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 |
|---|---|---|---|---|---|
Many important cellular membrane fission reactions are driven by ESCRT pathways , which culminate in disassembly of ESCRT-III polymers by the AAA ATPase Vps4 . We report a 4 . 3 Å resolution cryo-EM structure of the active Vps4 hexamer with its cofactor Vta1 , ADP·BeFx , and an ESCRT-III substrate peptide . Four Vps4 subunits form a helix whose interfaces are consistent with ATP binding , is stabilized by Vta1 , and binds the substrate peptide . The fifth subunit approximately continues this helix but appears to be dissociating . The final Vps4 subunit completes a notched-washer configuration as if transitioning between the ends of the helix . We propose that ATP binding propagates growth at one end of the helix while hydrolysis promotes disassembly at the other end , so that Vps4 ‘walks’ along ESCRT-III until it encounters the ordered N-terminal domain to destabilize the ESCRT-III lattice . This model may be generally applicable to other protein-translocating AAA ATPases .
The AAA ATPase Vps4 drives the ESCRT ( Endosomal Sorting Complexes Required for Transport ) pathways that mediate membrane deformation and fission in a wide range of cellular processes ( Monroe and Hill , 2016 ) . These include membrane severing during cytokinetic abscission , the formation of multivesicular bodies and exosomes , shedding of microvesicles and viruses , repair of lesions in the plasma membrane , pruning of neurons , removal of defective nuclear pore complex assembly intermediates , and nuclear envelope closure at mitotic exit ( Campsteijn et al . , 2016; Hurley , 2015; McCullough et al . , 2013; Olmos and Carlton , 2016 ) . The ESCRT pathways converge on the recruitment of ESCRT-III subunits , of which seven family members are recognized in yeast and 12 in human . The leading model is that the upstream factors recruit ESCRT-III subunits , which polymerize through their N-terminal domains to induce an inherently unstable membrane configuration that resolves by fission following Vps4-mediated disassembly or remodeling of the ESCRT-III polymer ( Henne et al . , 2013; McCullough et al . , 2013; Schöneberg et al . , 2017 ) . Multiple structures have been reported for domains of Vps4 ( Monroe and Hill , 2016 ) . The N-terminal MIT domain of Vps4 ( Scott et al . , 2005b ) binds ~20 residue MIT interacting motifs ( MIMs ) that are found at the C-termini of many ESCRT-III subunits ( Kieffer et al . , 2008; Obita et al . , 2007; Stuchell-Brereton et al . , 2007 ) . The MIT domain is followed by a flexible ~40 residue linker and an ~320 residue AAA ATPase cassette that comprises a large AAA ATPase domain and a small AAA ATPase domain ( Figures 1A and 2A ) , which contains an insertion known as the β domain that binds the dimeric C-terminal VSL domain of the Vta1 cofactor ( LIP5 in human ) in an interaction that promotes Vps4 assembly and ATPase activity ( Azmi et al . , 2006; Lottridge et al . , 2006; Scott et al . , 2005a ) . 10 . 7554/eLife . 24487 . 003Figure 1 . Vps4101-437-Hcp1 is an active hexamer . ( A ) Vps4 constructs and peptide-binding affinities assayed by fluorescence polarization . Peptide ‘20’ is a Vps2-derived 20-residue peptide C identified earlier ( Han et al . , 2015 ) . Peptide ‘8’ is an 8-residue fragment ( DEIVNKVL ) of peptide ‘20’ that retains essentially full binding affinity . The relatively weak binding of full-length Vps4 reflects autoinhibition mediated by the MIT domains ( Han et al . , 2015 ) . ( B ) Fluorescence polarization isotherms corresponding to values in panel A . Means and standard deviations are from four independent experiments . ( C ) Gel filtration of Vps4101-437-Hcp1 on a Superdex 200 column in 25 mM Tris/HCl pH 7 . 4 , 100 mM NaCl and 1 mM DTT . The protein elutes as a symmetric peak with an apparent molecular mass of 290 kDa , in good agreement with the calculated molecular mass of a hexamer ( 330 kDa ) . ( D ) ATPase activities for Vps4 constructs: 1 , Vps4 full-length; 2 , Vps481-437; 3 , Vps4101-437; 4 , Vps4101-437-Hcp1 . Vps4 subunit concentrations are indicated . Means and standard deviations from at least three independent measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 00310 . 7554/eLife . 24487 . 004Figure 1—source data 1 . Binding of fluorescently labeled ESCRT-III peptides to Vps4 , related to Figure 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 00410 . 7554/eLife . 24487 . 005Figure 1—source data 2 . ATPase activity of Vps4 constructs , related to Figure 1D . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 00510 . 7554/eLife . 24487 . 006Figure 1—figure supplement 1 . Vps4101-437-Hcp1 is a hexamer . Equilibrium sedimentation of Vps4101-437-Hcp1 . Absorbance is shown as a function of distance from the axis of rotation for three loading concentrations ( open circles , 10 µM; open squares , 5 µM; open diamonds , 2 . 5 µM Vps4 subunits ) . The data were fit to a single species model where the molecular weight was allowed to float or held constant as indicated ( black line ) , and residuals for each concentration are shown below . The molecular weight was fit to 329 , 402 Da in good agreement with the expected molecular weight for a hexamer of 334 , 223 Da . When the molecular weight was held constant , the data agree well with a hexamer model , whereas the residuals are clearly biased ( and in opposite directions ) when the molecular weight is set to a heptamer or pentamer . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 00610 . 7554/eLife . 24487 . 007Figure 2 . Structure of Vps4101-437:Vta1VSL:ESCRT-IIIpeptide:ADP·BeFx . ( A ) Structure of the complex . The Vps4 and Vta1 constructs used for cryo-EM structure determination are shown in color on the left , with excluded segments colored white . MIT , large AAA ATPase ( L ) , small AAA ATPase ( S ) and β domains of Vps4 are labeled , as are the t-MIT and VSL domains of the Vta1 dimer . L151 , a residue critical for hexamerization , is shown in gray spheres . ( B ) 4 . 3 Å map with the Vps4 model . ( C ) Side view of Vps4 hexamer , oriented with the subunit A-D helix axis vertical ( black line ) . ( D ) Same as panel D but with subunits E and F removed . The inset shows the position of pore loops 1 ( L1 , residues 203–210 , cyan ) and pore loops 2 ( L2 , 240–248 , dark blue ) relative to the ESCRT-III peptide ( dark green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 00710 . 7554/eLife . 24487 . 008Figure 2—figure supplement 1 . Vps4 3D reconstruction , refinement , and validation . ( A ) Representative cryo-EM image of Vps4101-437-Hcp1 particles . ( B ) Representative 2D class averages of Vps4101-437-Hcp1 particles . Red asterisks indicate classes with disordered Vps4 in which only the Hcp1 template is apparent . ( C ) ‘Gold standard’ FSC curves generated by RELION before ( blue ) and after ( orange ) Hcp1 signal subtraction . The FSC curve of the refined model ( comprising the large and small AAA ATPase domains of subunits A-E and substrate peptide ) against the final Hcp1-subtracted Vps4 map is shown in purple . ( D ) Cross-validation of the refined model . The refined model ( comprising large and small AAA ATPase domains of subunits A-E and the substrate peptide ) was randomly displaced by applying 0 . 5 Å shifts to all atoms and refined against one of the half maps generated by RELION . FSC curves are shown between the re-refined model against the half map used for re-refinement ( FSCwork , black ) and between the re-refined model and the other half map ( FSCtest , red ) . The agreement between the two FSC curves is an indicator that the model has not been overfit . ( E ) Local resolution estimates determined by ResMap . ( F ) The composite model indicating the refined portions of Vps4 ( colored ribbons ) and other regions limited to rigid body fitting ( Vps4 β domains , subunit F , and Vta1VSL , gray ribbons ) . Same orientation as panel ( E ) . Note that Vta1 densities are weak prior to 3D classification ( see Figure 2—figure supplement 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 00810 . 7554/eLife . 24487 . 009Figure 2—figure supplement 2 . 3D reconstruction workflow . Flow chart depicting classification and refinement of Vps4 particles . An initial model was generated from a gallery of non-CTF corrected 2D class averages , which was then used as a starting point for 3D classification . Particles from two classes showed ordered Vps4 features , which were then used to compute a 6 . 7 Å resolution consensus structure . Hcp1 densities were subtracted from raw images , followed by an additional round of RELION 3D classification and auto-refinement , which produced the final 4 . 3 Å resolution reconstruction . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 00910 . 7554/eLife . 24487 . 010Figure 2—figure supplement 3 . Additional validation of the 3D reconstruction . ( A ) Angular distribution plot based on RELION assignments and visualized in UCSF Chimera . ( B ) Comparison between reference-free 2D class averages and re-projections of the 3D reconstructions of the Vps4101-437-Hcp1 particle and Hcp1-subtracted Vps4 particle . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 01010 . 7554/eLife . 24487 . 011Figure 2—figure supplement 4 . Glutaraldehyde crosslinking improves the Vps4 density without distorting the structure . ( A ) Reference-free 2D class averages of non-crosslinked Vps4101-437-Hcp1 particles . Note that the Vps4 features are weaker and smeared out ( yellow arrows ) relative to the Hcp1 template and relative to the crosslinked sample ( Figure 2—figure supplement 1 ) . ( B ) 3D reconstruction of non-crosslinked Vps4101-437-Hcp1 particles reveals the Vps4 hexamer only at very low thresholds . ( C ) Comparison between non-crosslinked ( cyan ) versus crosslinked particles ( yellow ) reveal consistent features . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 01110 . 7554/eLife . 24487 . 012Figure 2—figure supplement 5 . Refined model and representative density . ( A ) Central β-sheet of subunit B with density . ( B ) Helices of subunit B . ( C ) Nucleotide density with ADP·BeFx and magnesium ion bound to subunit B . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 01210 . 7554/eLife . 24487 . 013Figure 2—figure supplement 6 . Identification and classification of Vta1 density . ( A ) Vta1VSL densities are visible at low threshold levels in an overall Hcp1-subtracted map . This observation prompted us to perform focused 3D classification with a mask over the expected Vta1 binding site . ( B ) Flow chart depicting 3D classification of the consensus structure with a focused mask ( yellow ) at the interface of subunits A and B . Classification revealed one distinct class with robust Vta1 features and the particles were isolated and subjected to an additional round of RELION auto-refinement ( light red ) . The same strategy was employed for each interface . ( C ) ‘Gold standard’ FSC plots of each of the six Vta1VSL datasets derived from focused 3D classification . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 01310 . 7554/eLife . 24487 . 014Figure 2—figure supplement 7 . Classification of subunit F Density . ( A ) Subunit F is poorly resolved but visible at low threshold levels in an overall Hcp1-subtracted map . This prompted us to perform focused 3D classification with a mask over subunit F . ( B ) Flow chart depicting 3D classification of the consensus structure with a focused mask ( yellow ) over subunit F . Classification revealed three distinct classes that could accommodate a rigid-body fit of the Vps4 crystal structure . See Figure 2—figure supplement 6 and Methods for details . ( C ) ‘Gold standard’ FSC plots of the three subunit F datasets derived from focused 3D classification . ( D ) Cut-away view depicting the local resolution of the F1 map determined by ResMap . Note that despite the overall ~7 Å resolution of the map , subunit F itself is less well resolved . ( E ) Fitting of Vps4 coordinates into the F1 structure confirms that our map resolves individual helices for Vps4 subunits A-E ( arrows ) , despite the lower resolution density for subunit F ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 01410 . 7554/eLife . 24487 . 015Figure 2—figure supplement 8 . Rigid-body fitting of Vps4 subunit F . Rigid-body fitting of Vps4 subunit F ( colored ribbon ) into three different density maps from focused 3D classification . The three models are related by pivoting of up to 16° about a point near the contact with Vta1VSL , close to the small AAA ATPase and β domains . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 015 Vps4 is monomeric or dimeric in the cytosol , and assembles to form an active hexamer upon concentration at the membrane/ESCRT-III surface ( Monroe et al . , 2014 ) . The central pore of this hexamer is thought to be lined by pore loop 1 and pore loop 2 ( residues 203–210 and 240–248 , respectively ) , which are highly conserved in AAA ATPases that act on protein substrates and play a critical role in substrate translocation ( Gonciarz et al . , 2008; Han et al . , 2015; Kieffer et al . , 2008; Monroe and Hill , 2016; Scott et al . , 2005a ) . Importantly , nucleotide-induced asymmetry is required for binding of a peptide from an ESCRT-III subunit to the central pore of the Vps4 hexamer , and the 1:1 stoichiometry of this interaction ( 1 peptide to 1 hexamer ) further confirms the asymmetric nature of the functional Vps4 complex ( Han et al . , 2015 ) . Although considerable effort has been devoted to visualizing this active state , several structures of Vps4 assembled in various configurations ( Caillat et al . , 2015; Hartmann et al . , 2008; Landsberg et al . , 2009; Yu et al . , 2008 ) have not provided a mechanistic model because they were determined for an inactive mutant , in the apo state , or in the presence of an inappropriate nucleotide . Guided by the insights that Vps4 is active as a hexamer ( Monroe et al . , 2014 ) that is stabilized by binding of ADP·BeFx ( Han et al . , 2015 ) and the VSL domain of the Vta1 cofactor ( Azmi et al . , 2006; Scott et al . , 2005a ) , we used S . cerevisiae proteins to prepare a Vps4-Vta1VSL-ESCRT-IIIpeptide-ADP·BeFx complex for structural studies . Determination of this structure by cryo-EM revealed a highly asymmetric configuration in which four of the six Vps4 subunits form a helix that is stabilized by ATP and Vta1 binding , and is fashioned to bind substrate peptide in a β-strand conformation approximately along the helix axis . The structure implies a helix propagation mechanism in which binding of ATP promotes growth at one end of the Vps4 helix and ATP hydrolysis promotes disassembly at the other end , such that the Vps4 hexamer ‘walks’ along the ESCRT-III polypeptide , thereby conveying the ESCRT-III substrate through the central Vps4 pore in an extended conformation .
The yeast Vps4 construct used in these studies ( residues 101–437 ) spans the AAA ATPase cassette . The active hexameric assembly was stabilized by expressing Vps4101-437 with a C-terminal 18-residue linker followed by the hexameric Pseudomonas aeruginosa Hcp1 protein ( Mougous et al . , 2006 ) . The active Vps4 conformation in the Hcp1 fusion protein was further stabilized by binding with ADP·BeFx , an ESCRT-III peptide , and the Vta1VSL domain . Importantly , this fusion protein binds an ESCRT-III substrate peptide with the same affinity as the Vps4 AAA ATPase cassette alone ( Figure 1AB ) . The 8-residue ESCRT-III peptide used in these studies was derived from the ESCRT-III subunit Vps2 ( residues 165–172 ) , and binds Vps4 with essentially the same ~200 nM KD as the 20-residue parent peptide that we characterized in an earlier study ( Han et al . , 2015 ) . As further controls , the fusion protein was found to elute from a size exclusion column as a single , symmetric peak ( Figure 1C ) , to form a stable hexamer as shown by equilibrium sedimentation ( Figure 1—figure supplement 1 ) , and to be a highly active ATPase ( Figure 1D ) . As shown below , other factors that indicate that Vps4 is not distorted by the Hcp1 fusion include the observation that Hcp1 has not imposed its 6-fold symmetry on the asymmetric Vps4 structure , the lack of contacts between Vps4 and Hcp1 in the overall consensus structure , and the short distance between Vps4 C-termini and Hcp1 N-termini ( 21–31 Å ) compared to the 60 Å that could be accommodated by the fully extended 18-residue linker sequence . The Vps4 complex structure was determined by cryo-EM at 4 . 3 Å overall resolution to reveal a highly asymmetric hexameric ring of Vps4 subunits that bind the ESCRT-III peptide in the central pore and six Vta1VSL dimers around the periphery ( Figure 2 , Table 1 , Figure 2—figure supplements 1–6 ) . The local resolution varies from 4 . 0 to 5 . 0 Å over much of the AAA ATPase cassettes of Vps4 subunits A-E , and to 7 Å or lower resolution at the β domains ( Figure 2—figure supplement 1 ) , Vta1VSL domains ( Figure 2—figure supplement 6 ) , and subunit F , which is distributed over at least three similar but distinct positions ( Figure 2—figure supplements 7–8 ) . The six Vps4 subunits adopt closely similar conformations but differ in the way that they contact each other . Although the limited resolution precludes detailed fitting of ADP·BeFx , the structure implies that the distinct Vps4 interfaces are coupled to binding of ATP , hydrolysis to ADP·Pi , and nucleotide exchange . Importantly , the ATP and ADP·Pi states can both be mimicked by ADP·BeFx . We propose that the distinct nucleotide states progress sequentially around the hexameric ring ( clockwise in Figure 2A ) , and that their step-wise conversion drives translocation of the ESCRT-III peptide , as discussed below . 10 . 7554/eLife . 24487 . 016Table 1 . Reconstruction , refinement , and model statistics of Vps4 . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 016Vps4101-437-Hcp1 , whole particleHcp1-subtracted Vps4ReconstructionParticle images58 , 15539 , 417Resolution ( unmasked , Å ) 6 . 75 . 7Resolution ( masked , Å ) 5 . 24 . 3Map sharpening B-factor ( Å2 ) -−208EM Databank Accession NumberEMD-8551EMD-8550Refinement and validation of Vps4 subunits A-EResolution used for refinement ( Å ) -4 . 3Number of atoms-10604R . M . S deviationBond length ( Å ) -0 . 01Bond angles ( ° ) -0 . 92RamachandranFavored ( % ) -91 . 13Allowed ( % ) -8 . 87Outlier ( % ) -0 . 00Molprobity score / percentile ( % ) -1 . 94/100thClashscore / percentile ( % ) -7 . 75/97thPDB-5UIE Vps4 subunits A-D form a right-handed helix that is created by the three very similar interfaces formed by the A-B , B-C , and C-D subunit pairs ( Figure 2 ) . These interfaces each bury ~2000 Å2 of surface area and appear fashioned to coordinate ATP ( Wendler et al . , 2012 ) , with R288 and R289 from the neighboring Vps4 subunit positioned to coordinate the ATP phosphates ( Figure 3 ) . The D-E interface ( ~1700 Å2 ) is similar to the A-B , B-C , and C-D interfaces at the central pore region of the hexamer but deviates at the nucleotide-binding site , where the large AAA ATPase domain of subunit E is rotated by ~15° so that R288 and R289 are displaced by ~2 Å and are no longer able to coordinate see the nucleotide phosphates . We have modeled the nucleotide at this site as ADP , but are open to the possibilities that it may represent either ADP or ADP·Pi . The displacement of subunit E increases toward the hexamer periphery , and is further exaggerated by an ~10° increase in the hinge angle between the large and small AAA ATPase domains that allows the small AAA ATPase domain of subunit E to maintain contact with subunit F . In contrast , the E-F and F-A interfaces , which bury only 500 and 900 Å2 , respectively , maintain contacts primarily near the hexamer periphery and appear open to allow nucleotide exchange . 10 . 7554/eLife . 24487 . 017Figure 3 . Interfaces in the asymmetric Vps4 hexamer . Vps4 subunit pairs superimposed on the large AAA ATPase domain of the first subunit , as indicated . A-B , B-C and C-D interfaces are equivalent . The nucleotide-binding site is slightly expanded at the D-E interface due to a 15° relative rotation of subunit E . The E-F and F-A sites are open for nucleotide exchange . Inset ( black rectangle ) , Close-up on the nucleotide binding site showing the nucleotide and coordinating P-loop for the first subunit , with the R288/R289-containing helix of the second subunits in color . These arginine finger Cα atoms shift by 2 Å at the D-E interface relative to the A-B , B-C , and C-D subunits . The E-F and F-A interfaces are shifted by 8 Å and 16 Å , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 017 The Vta1VSL dimers , which are pairs of helical hairpins that form a 4-helix bundle , stabilize the ring by forming struts between adjacent Vps4 subunits ( Figure 4A ) . Their density is clearly visible only after focused 3D classification ( Figure 2—figure supplement 6 ) , which reveals distinct density for all four helices of the Vta1VSL dimers adjacent to the β domains of Vps4 subunits A and B , while density for individual helices is less clearly defined but still apparent for Vta1VSL at subunits C and F ( Figure 4—figure supplement 1 ) . Subunits D and E do not show distinct VSL helices , but do show some overall density for the VSL bundle . This variation in the quality of Vta1VSL densities among the six interfaces likely reflects differences in occupancy and binding modes at each site , as discussed below . 10 . 7554/eLife . 24487 . 018Figure 4 . Vta1VSL contacts with Vps4 . ( A ) Density for the most clearly defined Vta1VSL ( bound to the Vps4 subunit A β domain ) . The Vta1VSL subunits are colored tan and teal . ( B ) Vta1VSL interaction with the first Vps4 subunit . This interface is modeled identically to a crystal structure of Vta1VSL in complex with a truncated Vps4 construct ( Yang and Hurley , 2010 ) . Additional N-terminal residues in the longer Vta1 construct used in this study are shown in white and their interaction with the small AAA ATPase domain of Vps4 is indicated with an asterisk . ( C ) Vta1VSL interaction with the second Vps4 subunit . Y303’ and Y310’ are labeled . ( D ) Overlap of subunit pairs on the small AAA ATPase domain of the first Vps4 ( residues 301–349 and 403–411 ) . Consequent RMSD values are shown for residues 300–311 and 320–331 of the second Vps4 subunit at the second Vta1 interface ( asterisk ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 01810 . 7554/eLife . 24487 . 019Figure 4—figure supplement 1 . Rigid-body fitting of Vps4 β domain-Vta1VSL complexes into each density map . ( A ) The A-B , B-C , C-D , and F-A Vta1VSL dimers show density for distinct helices . Density for D-E and E-F Vta1VSL dimers is poor and only S and β domain coordinates were used for rigid body fitting at these two interfaces . ( B ) Side view of the fitting at the A-B interface as in panel ( A ) showing distinct density for each of the four Vta1VSL helices . ( C ) Low threshold view of the A-B map reveals extensions of the Vta1VSL density that can accommodate the additional N-terminal residues included in our Vta1 construct ( black ribbon ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 019 In all cases , the Vta1VSL dimer contacts the β domain of one Vps4 subunit in an interface that we have modeled according to a previously reported crystal structure ( Yang and Hurley , 2010 ) in which Y303 and Y310 of the first Vta1 subunit contact two loops of the β domain ( residues 356–359 and 375–385 ) ( Figure 4B ) . Compared to the earlier crystal structure , the Vta1VSL construct in our structure is extended by 10 residues at the N-terminus , including residues of the ‘Vps4 stimulatory element’ ( Norgan et al . , 2013 ) . Some density for these residues is visible for the second subunit of the best-defined Vta1VSL dimers , where they make a small 70 Å2 contact with α7 and α9 in the small AAA ATPase domain of the same Vps4 subunit , as suggested previously ( Davies et al . , 2014 ) ( Figure 4B , Figure 4—figure supplement 1C ) . Consistent with their more clearly defined density , the Vta1VSL dimers bound to the β domains of subunits A , B , C , and F also contact the small AAA ATPase domain of the following Vps4 subunit ( Figure 4C ) , with Y310’ and surrounding residues at the hairpin end of the second Vta1 subunit contacting α6 and α7 ( residues 300–330 ) of the small AAA ATPase domain of the second Vps4 . This novel interaction shows an unusual use of the two-fold symmetry-related VSL dimer residues , Y310 and Y310’ , to contact different surfaces on neighboring Vps4 subunits , which is consistent with our biochemical finding that Vta1 stabilizes formation of the hexamer ( Monroe et al . , 2014; Scott et al . , 2005a ) rather than higher-order assemblies ( Xiao et al . , 2008; Yang and Hurley , 2010 ) . Vta1VSL dimers can bind in the same manner to the Vps4 subunit pairs F-A , A-B , B-C , C-D , but interactions at the D-E and E-F subunit pairs appear to be suboptimal . Superposition performed on the helices of the small AAA ATPase domains of the first Vps4 in the subunit pairs ( Figure 4D ) , which are relatively well defined , shows that Vta1 can make superimposable interactions with both the β domain of the first Vps4 subunit and with the small AAA ATPase domain of the second Vps4 subunit for the A-B , B-C , and C-D subunit pairs , and that F-A is quite similar . In contrast , the D-E and E-F interfaces are incompatible with Vta1 forming the same contacts between neighboring Vps4 subunits as seen at the A-B , B-C , C-D , and F-A interfaces ( Figure 4D ) . Instead , our preferred interpretation of the density is that Vta1VSL at the D-E and E-F interfaces remains bound to the β domain of the first Vps4 subunit but cannot form optimal contacts with its adjacent subunit . To test the importance of the interface seen between Vta1 and the small AAA ATPase domain of the second subunit , we quantified binding of Vta1VSL to the Vps4101-437-Hcp1 hexamer and to the Vps4101-437 L151D mutant , which is predominantly monomeric in the absence of Vta1VSL ( Gonciarz et al . , 2008 ) . The Vps4101-437-Hcp1 hexamer showed an ~30 fold tighter apparent KD than Vps4101-437 L151D ( Figure 5 ) , which supports our observation that Vta1VSL binds to two neighboring Vps4 subunits in the hexamer and is consistent with a role for Vta1 in stabilizing assembly of the active Vta1 hexamer ( Azmi et al . , 2006; Scott et al . , 2005a ) . Residues K321 , E322 and R325 of the small AAA ATPase domain of the second Vps4 subunit are located in the vicinity of residue Y310’ of the VSL domain ( Figure 5A ) . Consistent with their proximity to a region of the Vta1 surface that has negative electrostatic potential , mutating the lysine and arginine residues to alanine and aspartate , respectively , decreases binding to the Vps4 hexamer , whereas mutation of the glutamate to alanine increases binding affinity ( Figure 5B ) . 10 . 7554/eLife . 24487 . 020Figure 5 . Mutations at the Vta1 interface with the second Vps4 subunit alter binding affinity . ( A ) K321 , E322 and R325 of the Vps4 small AAA ATPase domain ( orange , labels in italic font ) contact Y310’ of Vta1VSL ( surface representation colored by electrostatic potential , kT/e ) in the interaction shown in Figure 4C . ( B ) Binding of fluorescently labeled Vta1VSL to the Vps4101-437-Hcp1 hexamer ( pink circles ) is ~24x tighter than binding to a monomeric Vps4 construct , Vps4101-437 L151D ( gold diamonds ) . Consistent with the Vta1 surface electrostatic potential , point mutations K321A and R325D weaken Vta1VSL binding 2-fold and 3-fold , respectively , while E322A strengthens binding 2-fold . Means and standard deviations are from at least three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 02010 . 7554/eLife . 24487 . 021Figure 5—source data 1 . Binding of Vta1VSL to Vps4 , related to Figure 5B . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 021 Density accommodates the 8-residue ESCRT-III peptide in an extended conformation . This density extends weakly at both ends and is not sufficiently clear to reliably build a model with side chains , consistent with the possibility that binding may occur in several overlapping positions ( Figure 6 ) . The primary contacts are with pore loop 1 ( residues 203–209 ) of the A , B , C , and D Vps4 subunits , whose closest Cα atoms are ~7 . 5 Å from the helix axis that is defined by these loops . There is also a contact with pore loop 1 of subunit E , although the loop density in that subunit is relatively weak . The peptide lies approximately along the helix axis , with Cα atoms modeled 1 . 0–2 . 7 Å ( average 1 . 6 Å ) from the axis , which is consistent with the model that substrates are translocated along or close to the helix axis , with some variation allowed to accommodate distinct amino acid sequences . The helical symmetry of pore loop 1 of Vps4 A-D is approximately continued by subunit E and has successive loops separated by a translation of 6 . 3 Å along the helix axis and a rotation of 60° ( Figure 7A ) . This matches the translation and rotation seen every two residues along a canonical β-strand , such that successive dipeptides of a β-strand that lies approximately along the helical axis could make equivalent interactions with pore loop 1 residues of successive Vps4 subunits . Hence , these four loops present a curved peptide-binding surface that extends into the hexamer pore . Pore loop 1 of subunit F is displaced ~14 Å from the helix axis and is therefore completely disengaged from the substrate ( Figure 7A ) . Although we do not observe a contact between the peptide density and residues of Vps4 pore loop 2 ( residues 241–251 ) and the density of pore loop 2 is generally quite poor , we note that these loops of the Vps4 A-D subunits are arrayed contiguously with the pore loop 1 residues through the hexamer pore and follow the same helical symmetry , which is consistent with the possibility that they continue the substrate binding surface used by Vps4 to translocate ESCRT-III subunits . 10 . 7554/eLife . 24487 . 022Figure 6 . Pore loops of Vps4 form a spiral staircase to coordinate the substrate peptide . Stereo view of the peptide and pore loop 1 ( residues 203–210 ) of subunits A-D with density . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 02210 . 7554/eLife . 24487 . 023Figure 7 . Peptide binding and mechanism of translocation . ( A ) The pore loop 1 residues of subunits A-D form a helix ( axis , black line ) that matches the symmetry of a canonical twisted β-strand , which rotates 60° and translates 6 . 3 Å every two residues . In white are the positions that subunits E and F would adopt if they continued this helix . The three positions seen for subunit F ( Figure 2—figure supplements 7–8 ) appear to be snapshots along the return path from the end of the helix at subunit E to the start of the helix at subunit A . ( B ) Steps along the translocation cycle inferred from the cryo-EM structure . The peptide shown is modeled as a β-strand along the helix axis of subunits A-D . Vps4 maintains a constant interaction with the peptide through steps 1 to 4 before dissociating at step 5 and rebinding 12 residues further up the peptide at step 7 , which is equivalent to step 1 . Nucleotides suggested by density and coordination geometry are labeled . Pore loop 1 contacts with the substrate peptide in steps 1–4 are indicated with an asterisk . The two subunits closest to the view direction are included with 50% transparency . The two horizontal lines are separated by 37 . 8 Å ( 12 residues ) and indicate points of substrate contact with pore loop 1 of the highlighted subunit . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 02310 . 7554/eLife . 24487 . 024Figure 7—figure supplement 1 . Comparison of the Vps4 hexamer with the ATPases of the 26S proteasome . ( A ) Superposition of secondary structure elements reveals a similar overall structure between Vps4 and the ATPase subunits of the 26S proteasome ( PDB 5GJQ , gray ) . ( B ) Top view and side view of the Vps4 pore loop 1 of Vps4 subunits ( rainbow ) and the proteasome ( gray ) shows close similarity about the helix axis ( black ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 02410 . 7554/eLife . 24487 . 025Figure 7—figure supplement 2 . Comparison of the Vps4 hexamer with the NSF D1 ring . ( A ) Superposition of secondary structure elements reveals a similar overall structure between Vps4 and the D1 ring of NSF ( PDB 3J94 , residues 215–489 , gray ) . ( B ) Top view and side view of the Vps4 ( rainbow ) and NSF D1 ( gray ) subunit pore loop one elements shows close similarity about the helix axis ( black ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 025 We propose that Vps4 translocates its ESCRT-III substrates through the hexamer pore by a helix propagation mechanism in which binding of ATP and Vta1 promotes growth at the A-end of the Vps4 helix while ATP hydrolysis and release promotes dissociation at the D/E-end of the helix ( Figure 7B , Figure 8 , Videos 1 and 2 ) . In repeated cycles , the propagating 4-subunit Vps4 helix will ‘walk’ along the substrate while binding it in an extended β-strand conformation and conveying it through the hexamer pore . Some variation in strand conformation may be tolerated but the overall effect is expected to be unfolding of ESCRT-III structure as the Vps4 helix advances . The structure indicates that Vta1 will promote this process by binding adjacent Vps4 subunits in the helical conformation . Moreover , as seen in the structure , Vta1VSL forms the same helix-promoting interaction between subunits F and A , as if pulling subunit F into an ATP-binding position at the leading end of the helix . At the other end of the helix , ATP hydrolysis correlates with expansion of the D-E nucleotide site to trigger disassembly of the helix , disengagement from the substrate by subunit F , opening of the interface to allow nucleotide exchange , and subsequent rebinding at the leading end of the propagating helix . An attractive feature of this model is that the symmetry match between a β-strand ESCRT-III substrate and the pore loop 1 residues of the Vps4 helix means that each of the Vps4 subunits can make identical interactions with consecutive ESCRT-III dipeptides , and that these interactions do not need to change during the translocation process . 10 . 7554/eLife . 24487 . 026Figure 8 . Schematic of one step in the translocation mechanism . Left , Subunits A-D form a helical surface of pore loop 1 residues that binds substrate in a β conformation along or close to the helix axis . The helix is stabilized by Vta1VSL binding to adjacent subunits and by ATP binding at subunit interfaces . Right , next step in the cycle where subunit F has bound ATP and assembled on the growing end of the Vps4 helix , ATP has been hydrolysed at the C-D interface , and the nucleotide-binding site of subunit E has been opened to allow ADP·Pi release and rebinding of ATP . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 02610 . 7554/eLife . 24487 . 027Video 1 . Top view of the proposed translocation mechanism . Vps4 reaction cycle modeled by linear interpolation between the six different states represented in the cryo-EM structure . The ESCRT-III substrate is modeled as a β-strand lying along the axis of the helix defined by Vps4 subunits A-D . Nucleotides are shown in pink ( ATP ) and gray ( ADP ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 02710 . 7554/eLife . 24487 . 028Video 2 . Side view of the proposed translocation mechanism . As for Video 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 028 The inherently weak hexamerization of Vps4 ( Monroe et al . , 2014 ) likely contributes to substrate specificity by coupling Vps4 recruitment to assembly of the active hexamer . As illustrated in Video 3 , we envision that Vps4 concentrates at ESCRT-III polymers by binding of its MIT domain with MIMs at the ESCRT-III C-termini . Vta1 possesses an N-terminal tandem MIT domain ( t-MIT ) ( Xiao et al . , 2008 ) that can also bind MIM sequences and is connected to the VSL domain by a flexible ~100 residue linker . Because Vps4 and Vta1 also bind to each other , the concentrating effect at ESCRT-III polymers will be synergistic , and will promote Vps4 binding to ATP and hexamerization around the 30–170 flexible residues that lie between the MIM and the folded N-terminal domain of various ESCRT-III subunits ( Han et al . , 2015 ) . The Vps4 complex can subsequently hydrolyze ATP to translocate along the polypeptide until the ESCRT-III N-terminal domain is destabilized and removed from the polymer . If the polymer remains intact after removal of one subunit , Vps4 and Vta1 would remain at high concentration due to the continuing MIM-MIT interactions and so could reassemble around another available ESCRT-III C-terminal sequence to repeat the process for as long as the ESCRT-III polymer persists to present an array of MIMs . 10 . 7554/eLife . 24487 . 029Video 3 . Model of Vps4 assembly and ESCRT-III disassembly . Vps4 ( purple ) is recruited to the ESCRT-III lattice ( green ) through binding of its N-terminal MIT domain to MIT Interacting Motifs ( MIMs ) ( Kieffer et al . , 2008; Obita et al . , 2007; Stuchell-Brereton et al . , 2007 ) , which are sequences at the ends of the long , flexible C-terminal tails of ESCRT-III subunits . The avidity effect of the ESCRT-III polymer promotes Vps4 hexamerization around flexible ESCRT-III sequences . The hexamer is further stabilized by the dimeric Vta1 protein ( tan ) , which also concentrates at the ESCRT-III polymer through its N-terminal t-MIT domain ( not shown ) ( Guo and Xu , 2015; Skalicky et al . , 2012; Vild et al . , 2015 ) . The Vps4 hexamer hydrolyzes ATP and translocates the substrate through the central pore , thereby destabilizing the ESCRT-III structure and removing the subunit from the lattice . We speculate that Vps4 and Vta1 remain bound to the ESCRT-III lattice via their MIT domains such that they are in position to process additional ESCRT-III subunits until the polymer is disassembled . This animation was created using Autodesk Maya ( Autodesk , Inc . ) from protein structural models exported from UCSF Chimera ( Pettersen et al . , 2004 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 029 The model that Vps4 translocates toward the N-terminal domain of ESCRT-III implies that substrate binds to the hexamer pore in a defined direction , which is a level of detail that is not resolved in our current structure . Also unresolved is the biological role of the 8-residue binding peptide that we identified from Vps2 , which based upon structural information available for other ESCRT-III proteins likely corresponds to a short helix that packs against the folded core in an isolated ESCRT-III subunit ( Bajorek et al . , 2009; Xiao et al . , 2009 ) and in an ESCRT-III polymer ( McCullough et al . , 2015 ) . These structures suggest that the peptide sequence disengages from the folded ESCRT-III core to bind Vps4 and raise the possibility that Vps4 destabilizes ESCRT-III structure simply by binding to this sequence , although our preferred model is that Vps4 initially hexamerizes around a more C-terminal segment and subsequently translocates beyond the site of the 8-residue peptide such that the entire ESCRT-III subunit is destabilized ( Yang et al . , 2015 ) . Structures of the substrate-bound DNA helicase E1 ( Enemark and Joshua-Tor , 2006 ) and RNA translocase Rho ( Thomsen and Berger , 2009; Thomsen et al . , 2016 ) suggested sequential mechanisms of polynucleotide translocation that are conceptually analogous to our proposal for polypeptide translocation by Vps4 . Thus , the structures indicate that Vps4 and the hexameric nucleic acid translocases function by forming a helical arrangement of subunits that matches the symmetry of their translocating substrate , with one or two transitioning subunits that are disengaged from the substrate , and with translocation achieved by sequential propagation of the ATPase helix . The structure of Vps4 superimposes with other hexameric AAA ATPases , including the ATPases of the 26S proteasome , which , like Vps4 , translocate protein substrates and adopt a right-handed helical notched-washer structure ( Förster et al . , 2013; Huang et al . , 2016; Lander et al . , 2013 ) . Multiple proteasome structures show four or five ATPase subunits forming a helix in which the pore loop 1 residues overlap Vps4 with RMSD values of ~2 . 3 Å ( Figure 7—figure supplement 1 ) . The ATP-bound conformation of the AAA ATPase NSF ( Zhao et al . , 2015 ) shows a similar overlap of pore loops ( Figure 7—figure supplement 2 ) , although the extent to which it may translocate substrate is unclear . It will be of considerable interest to determine the extent to which the geometry of substrate binding and the cycles of ATP-induced helix propagation envisioned here in light of the Vps4-substrate complex underlie the mechanisms of other protein-translocating AAA ATPases , and the extent to which variations on the idealized model , such as by variations in peptide binding geometry or in the sequence and timing of ATP hydrolysis , may apply .
Vps4 and Vta1 proteins were expressed in E . coli BL21 ( DE3 ) RIL ( Agilent Technologies , Santa Clara , CA ) from a pET151-based vector with a cleavable N-terminal 6xHis tag . Proteins and expression vectors used in this study are listed in Supplementary file 1 . Expression cultures were grown in ZY autoinduction media ( Studier , 2005 ) at 37°C for 6 hr and at 19°C for 16 hr . Cells were harvested by centrifugation and stored at −80°C . The same purification strategy was used for all Vps4 and Vta1 constructs . Cell pellets were thawed and resuspended in lysis buffer ( 50 mM Tris/HCl pH 8 . 0 , 300 mM NaCl , 5% ( v/v ) glycerol , 10 mM imidazole ) supplemented with protease inhibitors , 1 mg of DNAseI and 100 mg of lysozyme . After incubation on ice for 30 min , cells were lysed by sonication . Lysate was clarified by centrifugation and batch-bound to 10 ml of Ni-NTA agarose equilibrated in lysis buffer . Following a wash with 150 ml of lysis buffer , His-tagged protein was eluted with lysis buffer made up with 75 mM imidazole ( 500 mM for Vps4101-437-Hcp1 fusions ) . His tags were cleaved by incubation with 1 mg of TEV protease during dialysis against 25 mM Tris/HCl pH 8 . 0 , 150 mM NaCl , 1 mM EDTA , 1 mM DTT overnight at 4°C . Samples were then dialyzed into 25 mM Tris/HCl pH 8 . 0 , 100 mM NaCl for 4 hr with two buffer changes and incubated with 10 ml of Ni-NTA agarose equilibrated with dialysis buffer to remove the cleaved His tags and the His-tagged TEV protease . The sample was subsequently bound to a 5 ml HiTrap Q FF ion exchange column ( GE Healthcare ) equilibrated with Q buffer A ( 25 mM Tris/HCl pH 8 . 0 , 100 mM NaCl , 1 mM DTT ) , and eluted with a gradient of 0–50% Q buffer B ( 25 mM Tris/HCl pH 8 . 0 , 1 M NaCl , 1 mM DTT ) over 30 column volumes . Fractions containing the protein of interest were pooled , concentrated to ~5 ml and further purified by gel filtration into 25 mM Tris/HCl pH 7 . 4 , 100 mM NaCl , 1 mM DTT or 20 mM HEPES/NaOH pH 7 . 4 , 100 mM NaCl , 1 mM DTT using a 120 ml Superdex 200 column ( GE Healthcare , Chicago , IL ) . The yield per liter of expression culture was typically 20–30 mg for Vps4 proteins and 40–50 mg for Vta1VSL . Peptides DEIVNKVL ( Vps2 residues 165–172 ) and DEIVNKVLDEIGVDLNSQLQ ( Vps2 residues 165–184 ) were prepared by solid-phase synthesis on a Prelude X peptide synthesizer ( Protein Technologies , Inc . , Tucson , AZ ) using standard procedures and Fmoc chemistry ( Chan and White , 2000 ) . Unlabeled peptides were N-terminally acetylated and C-termini were produced as carboxyamides . For peptides used in fluorescence polarization assays , 5 ( 6 ) -carboxyfluorescein ( Acros Organics , Geel , Belgium ) was coupled to the N-terminal α-amine by standard coupling conditions . Following cleavage from resin by TFA , the peptides were precipitated with ice-cold ether , washed thoroughly with ether , and dried overnight under vacuum . Peptides were then HPLC purified on a Phenomenex 5 µm-C4 Jupiter column ( 10 × 250 mm , 300 Å ) at 5 ml/min over a 15 min gradient ( 20–80% ACN , 0 . 1% TFA ) . Peptide quality was verified by LC-MS using a FortisBIO 5 µm C4 column ( 4 . 6 × 150 mm , 300 Å ) coupled to an Agilent 6100 Series Single Quadrupole mass spectrometer . Binding of fluorescently labeled peptides to Vps4 was quantified in the presence of a 2:1 ratio of Vta1VSL:Vps4 subunits and 1 mM ADP·BeFx , as described previously ( Han et al . , 2015 ) . Briefly , peptides ( 1 nM ) were incubated at room temperature with Vps4:Vta1 complexes ( 0–180 µM Vps4 subunits ) in binding buffer ( 20 mM HEPES/NaOH , pH 7 . 4 , 100 mM NaCl , 1 mM ADP·BeFx , 10 mM magnesium chloride , 1 mM DTT ) in a total volume of 60 µl . When equilibrium was reached , parallel and perpendicular fluorescence intensities were measured on a Biotek Synergy Neo HTS microplate reader using an excitation wavelength of 485 nm and an emission wavelength of 528 nm . Because Vps4 binds a single peptide per hexamer ( Han et al . , 2015 ) , fluorescence polarization was plotted against the Vps4 hexamer concentration and dissociation constants were estimated by global fitting of the equation FP = [Vps4 hexamer]/ ( KD + [Vps4 hexamer] ) to data points from four independent experiment , where FP is the normalized fluorescence polarization or ‘fraction bound’ and independent experiments are defined as using different protein preparations , using GraphPad Prism 6 ( GraphPad Software , Inc . , La Jolla , CA ) . For binding studies , Vta1VSL S278 was replaced by cysteine . Fluorescent modification of the single cysteine was performed by incubating 8 μM Vta1VSL S278C with 200 μM fluorescein-5-maleimide ( Fisher Scientific ) in 25 mM Tris/HCl pH 7 . 4 , 100 mM NaCl , 5 mM EDTA overnight at 4°C . Excess label was removed using a PD-10 desalting column equilibrated in binding buffer ( 20 mM HEPES/NaOH pH 7 . 5 , 100 mM NaCl , 1 mM DTT ) . Vps4 constructs at monomer concentrations ranging from 0–60 µM were incubated with 3 nM of fluorescently labeled Vta1VSL in a total volume of 60 µl . Fluorescence polarization was read at equilibrium and apparent dissociation constants were estimated as described for the peptide binding studies ( above ) , except that in this case Vps4 subunit concentrations were used for both graphing and fitting because each subunit contains a potential binding site for the VSL dimer . The binding model does not account for potential differences in affinity to different binding sites in the context of the asymmetric Vps4 hexamer , and KD values are therefore referred to as apparent dissociation constants . The rate of ATP hydrolysis was determined using an end-point method modified from Merrill and Hanson ( Merrill and Hanson , 2010 ) . Vps4 at the indicated subunit concentration was incubated with 1 mM ATP at 37°C in 20 mM HEPES/NaOH pH 7 . 4 , 100 mM NaCl , 10 mM MgCl2 , 1 mM DTT in a total volume of 10 µl . The hydrolysis reaction was stopped after 5 min by the addition of 100 µl of malachite green color reagent ( 14 mM ammonium molybdate , 1 . 3 M HCl , 1 . 5 mM malachite green ) and 50 µl of 21% ( w/v ) citric acid . Absorbance at 650 nm was read using a Biotek Neo Synergy microplate reader and the amount of inorganic phosphate released by the reaction was determined using a sodium phosphate standard curve . Means and standard deviations reported in Figure 1D are from at least three independent experiments using different protein preparations with three or more technical replicates each . To confirm that Vps4101-437-Hcp1 is hexameric in solution , we performed equilibrium sedimentation analyses at 4°C using an XLI analytical ultracentrifuge ( Beckman Coulter , Indianapolis , IN ) with absorbance optics . Sample cells with 6-channel centerpieces were filled with 120 µl of Vps4101-437-Hcp1 in 25 mM Tris/HCl pH 7 . 4 , 100 mM NaCl at the indicated concentrations in the sample sectors and with 125 µl of buffer in the reference sectors . Absorbance scans at 280 nm were taken at equilibrium after centrifugation at 3000 rpm and 5000 rpm , respectively . Equilibrium sedimentation data were fit to a single species model in Heteroanalysis ( Cole , 2004 ) using a theoretical molecular mass of 55 , 703 . 8 Da per subunit , a partial specific volume of 0 . 732662 mL/g , and a buffer density of 1 . 0049 g/mL , as calculated in SEDNTERP ( Hayes et al . , 1995 ) . For crosslinking with glutaraldehyde , proteins were buffer-exchanged by extensive dialysis in 20 mM HEPES/NaOH pH 7 . 4 , 100 mM NaCl . Vps4101-437-Hcp1 ( final subunit concentration 18 µM ) , Vta1VSL ( final subunit concentration 36 µM ) and the 8-residue peptide ( from a 1 mM stock solution in water , final concentration 10 µM ) were combined in the presence of 1 mM ADP·BeFx and 5 mM magnesium chloride in a total volume of 4 . 8 ml , and incubated on ice for 30 min before equilibration to room temperature over 5 min . Crosslinking was initiated by addition of 50 µl of 2% glutaraldehyde solution ( diluted in dialysis buffer from an 8% stock , Fluka 49627 , final concentration 0 . 02% ) , and quenched after 30 min by adding 5 ml of 1 M glycine containing 1 mM ADP·BeFx and 5 mM MgCl2 . Following concentration to 0 . 5 ml , glutaraldehyde and glycine were removed by gel filtration into 25 mM Tris/HCl pH 7 . 4 , 100 mM NaCl , 1 mM ADP·BeFx , 5 mM MgCl2 , and 1 mM DTT using a Superdex-200 column with a bed volume of 24 ml . The extent of crosslinking was assessed by SDS PAGE analysis of the peak fraction . The elution volume was as expected for a hexameric complex based on protein standards . 3 . 5 µl of sample was applied to glow-discharged ( 25 mA , 25 s ) Quantifoil 1 . 2/1 . 3 holey carbon 400 mesh copper grids , which were plunge frozen in liquid ethane using a Vitrobot Mark III ( FEI , Hillsboro , OR ) set to 4°C , 80% relative humidity , 30 s wait time , −2 mm offset , and 8 s blotting time . Grids were stored in liquid nitrogen prior to data collection using SerialEM ( Mastronarde , 2005 ) on a Tecnai TF20 ( FEI ) operating at 200 kV using a Gatan 626 side entry cryo-holder . Movies were recorded using a K2 Summit direct detector ( Gatan , Pleasanton , CA ) in counting mode at a corrected magnification of 70 , 952× , corresponding to a physical pixel size of 0 . 7047 Å , and at a dose rate of ~5 e-/pixel/sec . Each movie was recorded as a stack of 40 subframes , each of which was accumulated for 0 . 2 s , totaling ~80 electrons per Å2 . Defocus values ranged between 0 . 8 to 2 . 0 µm . Movie frames were aligned , exposure filtered , and summed into a single micrograph using Unblur ( Grant and Grigorieff , 2015 ) ( Figure 2—figure supplement 1 ) . CTF parameters were determined using the program CTFFIND4 ( Rohou and Grigorieff , 2015 ) . Micrographs with poor CTF cross correlation scores were excluded from downstream analyses . 4059 particles were extracted from 41 micrographs after manual particle picking in EMAN2 using the e2boxer . py program ( Tang et al . , 2007 ) and used as input for non-CTF-corrected 2D class averaging in RELION ( Scheres , 2012 ) . The resulting 2D classes were used as templates for RELION auto-picking , which resulted in extraction of 180 , 172 particles from 703 micrographs for full CTF-corrected image processing . After four rounds of 2D classification , 108 , 733 particles were identified as having Vps4-like features and used for an initial round of 3D classification ( Figure 2—figure supplement 1 ) . The initial model for templated Vps4 was generated using a gallery of low-pass filtered ( 40 Å ) 2D classes in EMAN2 using the e2initialmodel . py program ( Tang et al . , 2007 ) , which yielded a double-layered 3D structure that was consistent with the dimensions of Hcp1 and Vps4 ( Figure 2—figure supplement 2 ) . After 3D classification , 58 , 155 particles were identified as having ordered Vps4 features and used for RELION auto-refinement to generate an overall structure at 6 . 7 Å resolution ( Table 1 ) . In order to optimize alignment on the Vps4 complex , we performed signal subtraction of Hcp1 densities in RELION using a previously described strategy ( Bai et al . , 2015 ) . Briefly , a soft-edged mask for Hcp1 was generated by subtracting a soft-edged Vps4 mask from the soft-edged mask of the entire Hcp1-Vps4 complex ( Figure 2—figure supplement 2 ) . This Hcp1 mask was applied to the 6 . 7 Å resolution map calculated from the consensus refinement of 58 , 155 particles , and the resulting masked map was used for Hcp1 signal subtraction from raw particles based on the particle orientations determined from the consensus refinement . This generated a new stack of particle images and a new STAR file with updated metadata that was used as input for a new round of RELION 3D classification and auto-refinement , which resulted in a Vps4 map at 4 . 3 Å resolution calculated from 39 , 417 particles ( Table 1 , Figure 2—figure supplements 1 and 2 ) . Local resolutions were estimated using ResMap ( Kucukelbir et al . , 2014 ) ( Figure 2—figure supplement 1 ) . Further quality control steps were taken by generating angular distribution plots , which confirmed a broad distribution of particle orientations , and comparisons between reference-free 2D class averages with 3D model reprojections of both the original consensus structure and the Hcp1-subtracted structure ( Figure 2—figure supplement 3 ) . To exclude the possibility that crosslinking with 0 . 02% glutaraldehyde might stabilize an artificial conformation of the Vps4 hexamer , we collected and processed a data set of non-crosslinked Vps4101-437-Hcp1 in complex with Vta1VSL , ESCRT-III peptide and ADP·BeFx . Samples were deposited on Quantifoil Graphene Oxide 2/4 200 mesh copper grids ( SPI Supplies ) glow-discharged for 25 s using a 10 mA current . Vitrification and data collection were performed as described above . 161 , 645 particles were extracted from 821 micrographs . After multiple rounds of 2D and 3D classification , particles were used for RELION auto-refinement , which yielded an ~13 Å resolution structure of the Vps4101-437-Hcp1 particle . Both 2D and 3D classes showed Vps4 features similar to those seen with the glutaraldehyde-crosslinked sample ( Figure 2—figure supplement 4 ) . However , Vps4 features are much better defined when the structure is stabilized by crosslinking . Some weak density was observed at the expected site for Vta1VSL at the Vps4 β domains ( Figure 2—figure supplement 6A ) . We therefore performed focused 3D classification using the Hcp1-subtracted dataset to identify particles that contain the Vta1 density ( Figure 2—figure supplement 6B ) . This was performed separately around each Vps4 subunit β domain , without particle re-alignment and by applying a generous soft-edged mask at the inter-subunit interface . The resulting classifications revealed Vta1 densities at each Vps4 subunit , and the corresponding particles were subjected to RELION auto-refinement . This strategy led to maps ranging between 5 . 3–7 . 2 Å resolution for the six sites ( Figure 2—figure supplement 6C , Figure 4—figure supplement 1 and Table 2 ) . Focused 3D classification that encompassed multiple Vta1 regions failed to enrich for a single class containing multiple Vta1 densities , presumably because the occupancy of Vta1 sites is low in the vitrified sample . 10 . 7554/eLife . 24487 . 030Table 2 . Reconstruction statistics of Vps4-Vta1 classes . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 030Vps4 Vta1VSL ( A-B ) Vps4 Vta1VSL ( B-C ) Vps4 Vta1VSL ( C-D ) Vps4 Vta1VSL ( D-E ) Vps4 Vta1VSL ( E-F ) Vps4 Vta1VSL ( F-A ) Particle images26 , 24313 , 06610 , 70011 , 68414 , 27426 , 964Resolution ( unmasked , Å ) 6 . 97 . 87 . 87 . 87 . 56 . 9Resolution ( masked , Å ) 5 . 46 . 77 . 26 . 96 . 55 . 3EMDB IDEMD-8552EMD-8553EMD-8554EMD-8555EMD-8556EMD-8557 The unsharpened Vps4 density map revealed reasonable density for subunit F at low contour levels ( Figure 2—figure supplement 7A ) . We therefore performed focused 3D classification with a custom mask over subunit F to identify particles that contain ordered subunit F density ( Figure 2—figure supplement 7B ) . The classification was performed without particle re-alignment ( i . e . , using the –skip_align flag in RELION ) and revealed three distinct classes with ordered subunit F density . A fourth class containing 47% of particles showed poor subunit F density . Particles from the three classes with ordered density were used for separate RELION auto-refinement calculations , which led to maps ranging between 6 . 9–7 . 2 Å resolution ( Figure 2—figure supplement 7C-E , Table 3 ) . The maps were used for rigid body fitting of Vps4 coordinates into each subunit F position ( Figure 2—figure supplement 8 ) . 10 . 7554/eLife . 24487 . 031Table 3 . Reconstruction statistics of Vps4-Subunit F classes . DOI: http://dx . doi . org/10 . 7554/eLife . 24487 . 031Vps4 F1Vps4 F2Vps4 F3Particle images690867947093Resolution ( unmasked , Å ) 7 . 88 . 17 . 8Resolution ( masked , Å ) 6 . 97 . 26 . 9EMDB IDEMD-8572EMD-8571EMD-8570 Model building was facilitated by the availability of a Vps4 AAA ATPase cassette crystal structure ( PDB 3EIE , [Gonciarz et al . , 2008] ) . The AAA ATPase cassettes for subunits A-E were fit to the 4 . 3 Å map as rigid bodies and subjected to real-space refinement using Phenix ( RRID:SCR_014224 ) ( Adams et al . , 2010 ) ( Figure 2—figure supplement 1F ) . Secondary structure restraints were applied during refinement . Guided by visual inspection of map similarity , NCS restraints were applied to Vps4 subunits A-E with the exception of residues 240–247 and 260–267 of subunit A and residues 204–207 ( pore loop 1 ) of subunit E . For Vps4 subunits A-E , residues 174–180 ( P-loop ) were restrained to a high resolution reference model ( PDB 5BQ5 , [Arias-Palomo and Berger , 2015] ) . For subunits A , B , and C , the distance between Be and the O3B of ADP was restrained to 1 . 6 Å , and the distance between Mg and F1 of BeF3 was restrained to 2 . 0 Å . For subunits D and E the nucleotide was refined as ADP , while the subunit F nucleotide site was empty . Residues 204–207 ( pore loop 1 ) , 240–247 ( pore loop 2 ) and 261–266 were absent in the previously reported structures and were built manually in Coot ( RRID:SCR_014222 ) ( Emsley et al . , 2010 ) . Because the 8-residue ESCRT-III peptide bound in the structure appears to occupy multiple sites , we did not attempt to build a detailed model but represented it as 8 Cα atoms in a low-energy extended conformation ( Figure 6 ) . To test for overfitting , all atoms in the refined model ( of the AAA ATPase cassettes of subunits A-E and the peptide substrate ) were randomly displaced by 0 . 5 Å and re-refined against one of the half maps derived from RELION auto-refinement . FSC curves for the re-refined model against the half map used for re-refinement ( FSCwork ) and against the other half map ( FSCtest ) showed close agreement ( Figure 2—figure supplement 1D ) , consistent with lack of overfitting . The refined model was assessed using MolProbity ( RRID:SCR_014226 ) ( Chen et al . , 2010 ) ( Table 1 ) . Models for Vta1VSL dimers and associated β domains were built by rigid body docking of a previously reported structure ( PDB 3MHV , [Yang and Hurley , 2010] ) . In cases where we observed additional density for N-terminal residues , the helix was extended accordingly . The β domains of subunits A , B and C and the corresponding VSL domains were subjected to rigid body refinement , whereas other β domains and VSL domains were positioned as docked by manual inspection . Subunit F was placed into the density by rigid-body fitting and not further refined . Finally , the model of the AAA ATPase cassettes for subunits A-E was combined with the models for Vta1VSL dimers and associated β domains and subunit F . In order to obtain reasonable geometry , the connecting residues were regularized in Coot . Figures of models and density maps were prepared using Chimera ( RRID:SCR_004097 ) ( Pettersen et al . , 2004 ) . Electrostatic potential was calculated using the Adaptive Poisson Boltzmann Solver ( APBS , RRID:SCR_008387 ) ( Baker et al . , 2001 ) implemented in Chimera . The complete model , including all 6 subunits of Vps4 AAA ATPase cassettes , 12 Vta1VSL domains , and the peptide , has been deposited into the PDB ( RRID:SCR_012820 ) together with the unsharpened Hcp-masked map . The unmasked map ( including both Hcp and Vps4 ) , sharpened Hcp-masked map , and the 6 maps for the Vta1VSL domain were deposited at the EMDB ( RRID:SCR_003207 ) . | Membranes surround multiple compartments within cells as well as the cell itself . In living cells , these membranes are remodeled continuously . This allows cells to divide , move molecules between different compartments and perform other essential activities . One important remodeling event is known as fission , which splits a membrane into separate parts . Large repeating structures ( or polymers ) of ESCRT-III proteins play a crucial role in membrane fission . Breaking apart ESCRT-III polymers triggers membrane fission and also recycles the ESCRT-III proteins so that they can be used again . An enzyme called Vps4 converts chemical energy ( stored in the form of a molecule called ATP ) into the mechanical force that breaks apart the ESCRT-III polymers . The active form of Vps4 consists of six Vps4 subunits working together to form a complex that includes a cofactor protein called Vta1 . Monroe et al . have now used a technique called cryo-electron microscopy to determine the structure of an active yeast Vps4-Vta1 complex while it is bound to a segment of an ESCRT-III protein . This revealed that four of the six Vps4 subunits form a helix ( which resembles a spiral staircase ) that binds ESCRT-III in its central pore . The structure implies that binding of ATP causes the Vps4 helix to grow at one end and that converting ATP into a molecule called ADP ( to release energy ) causes disassembly at the other end . The two additional Vps4 subunits move from the disassembling end to the growing end of the helix . In this manner , Vps4 ‘walks’ along ESCRT-III , thereby pulling it through the pore at the center of the Vps4 complex and triggering breakdown of the ESCRT-III polymer . Further work is now needed to understand exactly how this activity leads to membrane fission . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2017 | Structural basis of protein translocation by the Vps4-Vta1 AAA ATPase |
The Cdc45/Mcm2-7/GINS ( CMG ) helicase separates DNA strands during replication in eukaryotes . How the CMG is assembled and engages DNA substrates remains unclear . Using electron microscopy , we have determined the structure of the CMG in the presence of ATPγS and a DNA duplex bearing a 3′ single-stranded tail . The structure shows that the MCM subunits of the CMG bind preferentially to single-stranded DNA , establishes the polarity by which DNA enters into the Mcm2-7 pore , and explains how Cdc45 helps prevent DNA from dissociating from the helicase . The Mcm2-7 subcomplex forms a cracked-ring , right-handed spiral when DNA and nucleotide are bound , revealing unexpected congruencies between the CMG and both bacterial DnaB helicases and the AAA+ motor of the eukaryotic proteasome . The existence of a subpopulation of dimeric CMGs establishes the subunit register of Mcm2-7 double hexamers and together with the spiral form highlights how Mcm2-7 transitions through different conformational and assembly states as it matures into a functional helicase .
The faithful copying of DNA requires the correct spatial and temporal assembly of replication machineries at specific chromosomal loci known as origins . In eukaryotes , origins are licensed for replication by recruitment of the Mcm2-7 complex , a ring-shaped helicase that serves as the principal unwinding activity for separating parental DNA strands ( Blow , 1993; Bochman and Schwacha , 2008; Costa et al . , 2011; Lyubimov et al . , 2012 ) . Mcm2-7 is initially loaded around duplex DNA as an inactive double hexamer by the origin recognition complex ( ORC ) , Cdc6 , and Cdt1 in the G1 phase of the cell cycle ( Evrin et al . , 2009; Remus et al . , 2009 ) , forming a stable intermediate known as the pre-replicative complex ( pre-RC , Diffley et al . , 1994; Donovan et al . , 1997; Maiorano et al . , 2000; Sun et al . , 2013; Yanagi et al . , 2002 ) . Upon entry into S phase , Mcm2-7 associates with the GINS complex and Cdc45 generating an 11-member assembly termed the CMG ( Kanemaki et al . , 2003; Gambus et al . , 2006; Moyer et al . , 2006; Pacek et al . , 2006; Ilves et al . , 2010 ) . GINS/Cdc45 assembly is dependent on the CDK kinase ( Zegerman and Diffley , 2007 ) , while post-translational modification of Mcm subunits 2 , 4 , and 6 by the Cdc7/Dbf4 kinase ( DDK ) further contributes to CMG activation ( Labib , 2010; Sheu and Stillman , 2010 ) . Following the assembly of replicative polymerases and replisomal scaffolding factors ( Gambus et al . , 2009; Muramatsu et al . , 2010 ) , the two CMG particles split apart into discrete complexes that have been proposed to each encircle a single DNA strand during translocation ( Yardimci et al . , 2010; Boos et al . , 2012 ) . At present , multiple aspects of the Mcm2-7 loading and activation cycle remain poorly understood . Although the six homologous subunits of one Mcm2-7 complex are known to pair with a second Mcm2-7 complex through their N-terminal domains in the context of a double-hexamer ( Evrin et al . , 2009; Remus et al . , 2009 ) , the precise register by which these subunits interact with each other across the two rings is not known . How DDK phosphorylation of the Mcm2 , Mcm6 , and Mcm4 N-termini ( Labib , 2010 ) , or how a DDK-bypass mutation in the N-terminus of either Mcm4 ( Sheu and Stillman , 2010 ) or Mcm5 ( Jackson et al . , 1993 ) , might aid in the switch from an inactive Mcm2-7 double hexamer state to a functional CMG is similarly unclear , particularly as Mcm4 is spatially segregated from Mcm2 and Mcm5 ( Costa et al . , 2011 ) . During unwinding and fork progression , the CMG translocates 3′→5′ along DNA . How the various components of the CMG engage nucleic acid strands during this process has remained ill-defined . Cdc45 has recently been shown to contain a RecJ exonuclease domain that can bind DNA but that is catalytically inactive ( Petojevic et al . , unpublished data , as well as Sanchez-Pulido and Ponting , 2011; Krastanova et al . , 2012; Szambowska et al . , 2014 ) . Whether or how the Cdc45 RecJ fold might bind single DNA strands formed in the context of the CMG has not been established . Conflicting models likewise exist for how Mcm2-7 engages substrate DNAs as it moves 3′→5′ during strand separation , with biochemical data from archaeal MCMs and phylogenetic relationships to superfamily III ( SFIII ) helicases ( such as the SV40 Large T antigen and the papillomavirus E1 protein ) predicting mutually exclusive binding orientations ( McGeoch et al . , 2005; Enemark and Joshua-Tor , 2006; Rothenberg et al . , 2007; Lee et al . , 2014 ) . To begin to understand several extant questions surrounding how the CMG is formed and operates at molecular level , we have determined structure of the full-length complex from Drosophila melanogaster in the presence of a 3′-tailed DNA duplex and the non-hydrolyzable ATP analog , ATPγS , using negative-stain electron microscopy and single-particle reconstruction methods . The structure establishes that: 1 ) the CMG preferentially associates with single-stranded DNAs over double-stranded substrates , 2 ) the C-terminal ATPase domains of Mcm2-7 form the leading edge of the motor as it advances on a duplex , and 3 ) the RecJ domain of Cdc45 is oriented to favor the capture of DNA segments that might accidently escape the Mcm2-7 pore . Comparison of the new structure with a previously-determined apo CMG model ( Costa et al . , 2011 ) shows that the Mcm2-7 ATPase domains of the complex transition from a planar , open ring into a closed , right-handed spiral in the presence of both DNA and nucleotide . Analysis of this state alongside other ring-ATPases shows that the MCM spiral is most similar to that adopted by the bacterial DnaB helicase upon engaging single-stranded DNA ( Itsathitphaisarn et al . , 2012 ) , and that the GINS•Cdc45 complex bridges the junction between the ends of the spiral in a manner similar to that by which the Rpn1 accessory subunit spans a spiral Rpt1-6 ATPase assembly in the eukaryotic proteasome ( Lander et al . , 2012 ) . Interestingly , examination of a subpopulation of CMG dimers present in our EM data shows how two Mcm2-7 complexes associate within a double hexamer and suggests that this dimerized state persists during CMG formation , prior to separation during fork progression ( Ilves et al . , 2010; Yardimci et al . , 2010 ) . Collectively , our observations establish that Mcm2-7 unwinds DNA using an approach distinct from that of superfamily III helicases and highlight several new Mcm2-7 ring configurations and assembly states accessed by the motor during the initiation of DNA replication .
In a previous study , we determined the medium-resolution ( 28 Å ) structures of the Drosophila melanogaster CMG helicase in both an apo state and bound to a non-hydrolyzable ATP analog ( Costa et al . , 2011 ) . Though sufficient for mapping individual subunits within the CMG , both models revealed a planar structure for Mcm2-7 , with GINS and Cdc45 spanning a gap that appeared between Mcm2 and Mcm5 when nucleotide was omitted . Since insights into where DNA might bind to the CMG or how binding might potentially alter the structure of complex were unclear , we set out to trap and image a prospective translocation intermediate of the CMG using 3D single-particle electron microscopy . A purified solution of the CMG was first mixed with a 20 bp duplex DNA substrate bearing a single-stranded 3′-dT ( 40 ) tail and passed over a sizing column in the presence of the non-hydrolyzable ATP analog , ATPγS , to form a ternary complex . The complex did not behave well during cryo-preservation attempts using holey-carbon EM grids , so samples were instead deposited onto continuous carbon grids and exposed to uranyl formate for negative staining . A total of 29 , 913 particles were selected from EM micrographs acquired with JADAS automated data collection software ( JEOL , Zhang et al . , 2009 ) on a JEM2100 electron microscope . Following particle picking and 2D averaging , a 3D model was generated by projection matching using a low-pass filtered ( 60 Å ) , free-hand test-validated , nucleotide-bound structure of the CMG as a starting model ( ‘Materials and methods’; Rosenthal and Henderson , 2003; Lyubimov et al . , 2012 ) . CMG particles imaged with DNA and ATPγS turned out to be quite uniform , permitting structure determination to a higher resolution than that obtained previously ( 18 Å vs 28 Å resolution , Figure 1—figure supplement 1 ) . The resultant model ( Figure 1A ) in turn allowed for a more accurate fitting of the Mcm2-7 and GINS subunits ( Figure 1B , C ) , revealing several new features . For instance , the location of Psf1 C-terminus , which was previously not visible , was now clearly evident , and could be readily fit to a recently-published full-length structure of Psf1 from an archaeal ortholog ( Oyama et al . , 2011; Figure 1C ) . Flexing within the Mcm2-7 ring was also apparent with the C-terminal lobes of different MCM subunits displaying markedly distinct degrees of movement with respect to their associated N-terminal regions ( Figure 1D , E ) . Asymmetric positioning between the two tiers of an MCM ring has not been reported previously , demonstrating that these elements are conformationally independent of each other to some extent in the presence of DNA substrates . 10 . 7554/eLife . 03273 . 003Figure 1 . 18 Å resolution of a CMG–DNA–ATPγS complex . ( A ) Top-down view ( N-terminal MCM face ) of the CMG highlighting subunit positions . ( B ) Docking of homology models into the assembly . ( C ) Docked structures into segmented density for: top—a near-full-length , archaeal MCM monomer Mcm4 ( PDB ID 3F9V ) ; middle—the GINS complex ( PDB ID 2Q9Q and 3ANW , ‘Materials and methods’ ) ; bottom—the archaeal Mcm N-terminal domain hexamer ( PDB ID 1LTL and 2VL6 , ‘Materials and methods’ ) . ( D ) The N- and C-terminal domains of Mcm2-7 ( colors ) differentially flex around the helicase ring , with GINS–Cdc45 ( white ) wedging open Mcm5 in particular . ( E ) The N-terminal domains of Mcm2-7 are relatively planar , and are fit best by a hexameric , DNA-free structure of the archaeal MCM NTDs , indicating the observed intra-subunit flexing derives from ATPase domain movement . DOI: http://dx . doi . org/10 . 7554/eLife . 03273 . 00310 . 7554/eLife . 03273 . 004Figure 1—figure supplement 1 . Overview of EM data . ( A ) Representative micrographs . ( B ) Comparison of Euler plots obtained with either RELION ( left ) or SPARX ( right ) three-dimensional structure refinement . ( C ) Comparison between two-dimensional projections of the SPARX-refined , three-dimensional structure and reference-free , two-dimensional class averages . ( D ) Comparison of the independently determined , three-dimensional structures of DNA- and ATPγS-bound CMG particles generated with RELION and SPARX . ( E ) Fourier-shell correlation showing the calculated resolution for the 3D EM models . DOI: http://dx . doi . org/10 . 7554/eLife . 03273 . 004 The increase in resolution obtained for the CMG in the presence of DNA provided initial , support evidence for nucleic acid binding to the complex . More concrete evidence for DNA association was apparent in electron density maps generated for the CMG , which showed a rod-shaped feature jutting away from the C-terminal face of Mcm subunits 2 and 5 ( Figure 2A ) —this feature is absent in DNA-free 3D reconstructions of the CMG ( Costa et al . , 2011 ) . Because negative-stains are non-ideal for visualizing nucleic acids ( Grob et al . , 2012 ) , we further assessed DNA binding by biotin-labeling the duplex end of the oligonucleotide , mixing the CMG–DNA samples with streptavidin , and collecting new single-particle EM data . Inspection of the resultant 2D class averages from this approach revealed clear additional density compared to the unlabeled CMG–DNA particles ( Figure 2B ) , demonstrating that the tailed substrate indeed associates with the complex particles . Given the electron density features seen for the DNA and the distance the streptavidin ‘pointer’ resides from the complex , the EM data show that the CMG binds to the single-stranded end of the 3′-tailed DNA substrate , corroborating biochemical data indicating that the complex preferentially associates with and translocates along single-stranded DNA over duplex substrates ( Ilves et al . , 2010; Fu et al . , 2011 ) . 10 . 7554/eLife . 03273 . 005Figure 2 . Polarity of DNA binding by the CMG . ( A ) Observed experimental density seen at low contours reveals a rod-shaped extension ( green ) of comparable length to that expected for a 20mer DNA duplex that extends from the Mcm C-terminal motor domain . This feature is absent in DNA-free CMG reconstructions ( Costa et al . , 2011 ) . A schematic of the relative single- and double-stranded DNA regions of the substrate used for the present studies is shown at left . ( B ) Comparison of DNA–CMG class averages with and without streptavidin-labeling clearly marks the duplex end of the 3′-tailed duplex substrate . Note how the streptavidin density sits at a distance from the body of the CMG , indicating that the majority of the duplex region of the substrate is not bound by the CMG . DOI: http://dx . doi . org/10 . 7554/eLife . 03273 . 005 The ability to visualize not only DNA binding to the CMG but also the position of the duplex end with respect to the particle , resolves a key question concerning the polarity by which MCM helicases engage a presumptive translocation strand . MCMs and viral SF3 helicases , such as SV40 LTag and the papilloma virus E1 protein , are both AAA+ ATPases ( Neuwald et al . , 1999 ) . This relationship , coupled with shared ability of MCMs and SF3 enzymes to translocate along DNA in a 3′→5′ direction ( Kelman et al . , 1999; Chong et al . , 2000; Bochman and Schwacha , 2008; Moyer et al . , 2006 ) , has suggested that members of two helicase families might operate by a common translocation mechanism . However , studies of E1 and archaeal MCMs bound to DNA substrates have yielded conflicting data concerning the direction by which DNA threads through the helicase pore . In E1 , the 3′ end of DNA has been observed by X-ray crystallography to lie proximal to the C-terminal motor domains of the helicase ( Enemark and Joshua-Tor , 2006 ) . By contrast , based on FRET measurements between a dye-labeled DNA/MCM pair , the converse has been reported for Sulfolobus solfataricus MCM ( McGeoch et al . , 2005; Rothenberg et al . , 2007 ) . In the new CMG structure , the streptavidin appended to the duplex DNA end can be clearly seen to localize next to the C-terminal , AAA+ domain face of the particle ( Figure 2B ) . This finding not only demonstrates that a DNA segment bound by an MCM runs from the N-terminal collar to the ATPase motor region in a 3′ to 5′ direction , but also indicates that MCM and SF3 helicases bind substrates with opposing polarities . When the structure of the CMG was first reported , the fold of the associated Cdc45 subunit was unknown . As a consequence , although the general location of Cdc45 could be identified in both apo and ATP-bound forms of the CMG , the orientation and role of this subunit was left unresolved ( Costa et al . , 2011 ) . Recently , however , the N-terminus of Cdc45 was shown to belong to the RecJ family of ssDNA exonucleases ( Sanchez-Pulido and Ponting , 2011 ) . Interestingly , within this grouping , Cdc45 belongs to an offshoot branch that can still bind DNA , but that also possesses natural amino acid substitutions which would appear to inactivate any native hydrolase functions ( Krastanova et al . , 2012 ) . To understand how the RecJ fold of Cdc45 interfaces with Mcm2-7 and GINS , we built a homology model for DmCdc45 based on Thermus thermophilus RecJ and docked it into the higher-resolution , DNA-bound CMG reconstruction . The catalytic core and DNA tracking domain of the homology model fit unambiguously into only one region of the Cdc45 density ( Figure 3A ) , leaving only a single , unaccounted for region ( most likely corresponding to the C-terminal segment of Cdc45 outside the defunct exonuclease core , or possibly to the N-terminal extension present in Mcm2 ) that interdigitates between the N-terminal ‘A-domains’ of Mcm5 and Mcm2 ( Figure 3B , Figure 3—figure supplement 1 ) . Notably , in placing the Cdc45 RecJ domain , we found that this element appeared to contact the now-apparent C-terminal ‘B-domain’ of Psf1 ( Figure 3A ) . To test whether this interaction might be real or fortuitous , we subjected the Psf1 B-domain surface to site-directed mutagenesis and tested the ability of the mutant subunits to support binding to both Mcm2-7 and Cdc45 ( ‘Materials and methods’ ) . Ablation of either the Psf1 B-domain region ( residues 185–202 ) or Cdc45 N-terminal region ( residues 1–99 ) prevented CMG formation under the conditions used to purify the intact assembly ( Figure 3—figure supplement 2 ) . Likewise , while point mutations were unable to disrupt CMG formation as judged by co-immunoprecipitation , a quadruple Psf1 mutant ( E190A/L192A/V193A/R194A ) proved unable to interact with Cdc45 . Together , these data indicate that the Cdc45–Psf1 interaction evident from the EM data plays a critical role in CMG formation and/or stability . 10 . 7554/eLife . 03273 . 006Figure 3 . Cdc45 is positioned to permit trapping of single-stranded DNA . ( A ) Top—Segmented electron density corresponding to Cdc45 . A prominent horseshoe-shaped region fits well to the catalytic core of the homologous RecJ exonuclease ( PDB ID 1IR6 ) . Bottom—Docked models of RecJ and full-length GINS ( generated using an archaeal Psf1 homolog , PDB IDs 2Q9Q and 3ANW ) into DNA-bound CMG reconstructions highlight a previously unobserved interaction between the B-domain of Psf1 and the exonuclease-like domain of Cdc45 . ( B ) An extension of the Cdc45 RecJ-like region contacts and interdigitates between the A-domains of the Mcm2 and Mcm5 N-terminal regions . ( C ) The Mcm2-7 central channel ( black line ) and the Cdc45 DNA tracking groove ( red arrow ) are offset by ∼90° . ( D ) Schematic showing how the single-stranded DNA-binding groove of the Cdc45 RecJ-like domain could facilitate the capture of a leading strand segment if the Mcm5-2 DNA gate were to transiently open . DOI: http://dx . doi . org/10 . 7554/eLife . 03273 . 00610 . 7554/eLife . 03273 . 007Figure 3—figure supplement 1 . Comparison between the open N-terminal domain of Drosophila Mcm2-7 and the closed N-terminal region of the CMG . Docking of the Mcm2-7 DNA-interaction collar , GINS , and the RecJ exonuclease domain occupies most of the CMG density , with the exception of some unoccupied density shown in gray . We have tentatively assigned this region to the otherwise unaccounted for C-terminal domain of Cdc45; however , the density could also correspond to the disordered N-terminal tail of Mcm2 , which may become structured upon Cdc45 binding . DOI: http://dx . doi . org/10 . 7554/eLife . 03273 . 00710 . 7554/eLife . 03273 . 008Figure 3—figure supplement 2 . The Psf1 C-terminus and Cdc45 N-terminus are critical for the CMG formation . ( A ) ( Left panel ) Expression levels of individual subunits used for assessing the CMG formation were detected by antibodies against specific proteins as indicated ( asterisk indicates a degradation product of Psf2 ) . Loading was controlled by an α-tubulin antibody . ( Right panel ) Co-immunoprecipitation ( IP ) experiments analyzing CMG stability . Lanes are as follows: ( 1 ) wild-type CMG; ( 2 and 5–7 ) CMG with various C-terminal truncations in Psf1; ( 3 ) CMG without MCM5; ( 4 ) CMG without Cdc45 . Immunoprecipitation of FLAG-tagged Mcm3 subunit pulls down intact CMG complexes ( lane 1 ) . By contrast , removal of the entire Psf1 B-domain ( lane 2 , Psf11−139 ) , smaller truncations of Psf1 B-domain ( lanes 5–7 ) , the absence of Mcm5 ( lane 3 ) , or the absence of Cdc45 ( lane 4 ) all disrupt the CMG formation . ( B ) ( Left panel ) Expression levels of individual subunits used for assessing the CMG formation were detected by antibodies against specific proteins as indicated . ( Right panel ) Co-immunoprecipitation ( IP ) experiments analyzing the CMG stability . Lanes are as follows: ( 1 ) wild-type CMG; ( 2 ) CMG expression without Mcm5; ( 3 ) CMG expression without Mcm3; ( 4–9 ) CMG expression with various alanine substitutions in Psf1 C-terminal helix . Immunoprecipitation of Mcm3 via a FLAG-tag on this subunit pulls down the intact CMG complex ( lane 1 ) , and most point mutations in the C-terminal helix of Psf1 do not appear to compromise the CMG formation ( lanes 4 and 6–9 ) . By contrast , the alanine substitutions in four residues in the B-domain of Psf1 ( lane 5 ) , the absence of Mcm5 ( lane 2 ) , or a lack of Mcm3 ( the bait protein , lane 3 ) disrupt the CMG . The different alanine substitutions in Psf1 are listed below the gel at left . ( C ) Purification of the CMG with Cdc45Δ1−99 . ( Left ) Input ( 4 μl of 30 ml-total clarified whole cell lysate ) and IP ( 4 μl out of 5 ml-total eluate following anti-FLAG affinity purification ) of the CMG component utilizing Cdc45Δ1−99 are shown . The FLAG-tag on Mcm3 was used as bait for the IP . Co-immunoprecipitated proteins were separated by SDS-PAGE ( 10% ) and the CMG subunits were detected by antibodies against several CMG proteins as indicated on the right of the western blot . Mcm2 and Mcm5 represent the Mcm2-7 complex; Psf1 and Psf3 were used to follow the GINS complex . The experiment shows a near total loss of Psf1 and Psf3 from the CMG when the N-terminus of Cdc45 is truncated . ( Right ) MonoQ profile for the CMG purification , with the expected positions for various constituents labeled ( FLAG-Mcm3 alone , Mcm2-7 and the intact CMG complex ) . The CMG formation was abrogated when the N-terminus of Cdc45 is removed . DOI: http://dx . doi . org/10 . 7554/eLife . 03273 . 008 In previous apo and ATP-bound models of the CMG , the particle was seen to transition from a conformation in which the Mcm2/5 interface was open to one in which it was closed ( Costa et al . , 2011 ) . This transition in turn pinched off the large single channel that ran through the particle into two smaller channels , sealing the interior of Mcm2-7 away from the inner surface of GINS–Cdc45 . In the DNA-bound model , the CMG still exhibits two channels; however , docking of the Cdc45 RecJ domain shows that its exonuclease/DNA-tracking groove is offset by 90° with respect to the central axis of the Mcm2-7 pore ( Figure 3C ) . This orientation indicates that , were Cdc45 to bind DNA in a manner similar to RecJ , it would be poised to capture the leading DNA strand that might escape from Mcm5-2 gate ( Figure 3D ) . Consistent with this idea , cross-linking data in work to be published elsewhere ( Petojevic et al . ) show both that Cdc45 engages the leading strand of a fork substrate only in the absence of nucleotide and that this interaction is ablated by the mutation of residues suggested by the model to be important for DNA binding . How ATP-dependent physical movements within hexameric helicases are coupled to DNA binding and unwinding has long been a central question in the field ( Singleton et al . , 2007; Enemark and Joshua-Tor , 2008; Lyubimov et al . , 2011 ) . Notably , when comparing the new DNA-bound CMG model to the prior substrate-free state , we found that the AAA+ ATPase ring is no longer flat , but instead adopts a clear right-handed spiral ( Figure 4A , B ) . This change in conformation does not propagate into the N-terminal domains , which maintain a roughly planar character ( Figure 1E ) , but is instead offset by the variable flexing seen for the C-terminal domains in different positions around the ring ( Figure 1D ) . The observed asymmetry between the two MCM tiers indicates that the N-terminal domains form a relatively stable collar that likely helps to coordinate and restrain movements of the associated C-terminal ATPase regions . 10 . 7554/eLife . 03273 . 009Figure 4 . Global comparison of the DNA-bound Mcm2-7 region of the CMG with other hexameric helicases and ATPases . ( A ) Cut-away view ( removing Mcm5 ) of the Mcm2-7 central channel highlights a spiral organization for the Mcm2-6-4-7-3 AAA+ ATPase regions . Colored spheres demarcate the approximate center of mass for AAA+ pore loops as derived from the docking of MCM AAA+ domain as shown in Figure 1B . ( B ) Top-down view ( from the N-terminal face ) of MCM AAA+ domains docked into the DNA-bound CMG reconstruction showing the existence of a right-handed spiral . The CMG density has been removed for clarity . ( C ) In the presence of a single-stranded DNA , bacterial DnaB can adopt a right-handed spiral with a moderately-wide pore ( PDB ID 4ESV , Itsathitphaisarn et al . , 2012 ) . ( D ) The E1 helicase assembles into a right-handed spiral with a relatively narrow pore ( PDB ID 2GXA , Enemark and Joshua-Tor , 2006 ) . ( E ) Comparison of the AAA+ ring of the eukaryotic proteasome with Mcm2-7 region of the DNA- and ATPγS-bound CMG . The non-ATPase subunit Rpn1 binds to the side of the Rpt1-6 hetero-hexamer , wedging itself between the N-terminal and C-terminal tiers of the ATPase ring and helping to promote the formation of a right-handed ATPase domain spiral . Similar architectural features are apparent within the DNA–ATPγS–CMG complex , where GINS–Cdc45 occupy an analogous position . DOI: http://dx . doi . org/10 . 7554/eLife . 03273 . 009 Closer analysis of the AAA+ spiral reveals several features that have important implications for the action of MCM subunits during DNA unwinding . First , the largest inter-subunit shifts within the Mcm2-7 ring , which occur between Mcm subunits 2 and 5 , also correspond to the point where the GINS–Cdc45 complex docks against the helicase . Inspection of the DNA-bound model reveals that GINS–Cdc45 does not simply straddle the Mcm2/5 interface , but that portions of the accessory subunits actually wedge themselves between the N- and C-terminal tiers of the Mcm2-7 ring ( Figure 1D ) . This action widens the exterior groove between the MCM N- and C-terminal domains at their points-of-contact with GINS–Cdc45 , and is offset by a concomitant narrowing of the groove on the exterior MCM face opposite the GINS–Cdc45 binding site ( i . e . , Mcm4 , Figure 1D ) . The structural consequences resulting from GINS–Cdc45 binding suggests that these accessory subunits not only play a role in blocking access through the Mcm2/5 gate , as has been seen previously ( Costa et al . , 2011 ) , but that they also help stabilize a spiral configuration of ATPase centers when DNA is present . Since the mutation of active site residues at the Mcm2/5 interface ablates helicase activity ( Bochman et al . , 2008; Ilves et al . , 2010 ) , it is likely that the spiral state observed here , which positionally offsets the ATP-binding site of Mcm5 from the arginine-finger residue of Mcm2 , inter-converts with another conformation in which the Mcm2/5 interface is remodeled to form a catalytically functional ATPase center during the translocation cycle . Hence , a need for GINS–Cdc45 in preventing DNA from escaping Mcm2-7 would likely be infrequent and limited to instances when the Mcm2/5 gate accidentally opens for an extended period of time , such as at a roadblock created by other nucleoprotein complexes or DNA damage . A second unexpected feature of the DNA-bound CMG complex is that the spiral is more pronounced than that seen in SF3 helicases , and instead more closely approximating the spiral evinced by a RecA-family helicase , DnaB , in the presence of DNA ( Figure 4B–D ) . The width of the Mcm2-7 central channel ( as measured from homology models of the motor domains docked into the EM density ) is likewise significantly larger ( ∼30–35 Å ) compared to E1 ( ∼14 Å ) , and more closely approaches that of DnaB ( ∼22 Å ) . Interestingly , in E1 and DnaB the difference in channel diameter and subunit rise between the two proteins sculpts the DNA substrate bound by each helicase into a single-stranded helix whose relative pitches differ significantly; these geometric differences allow each subunit of DnaB to engage two nucleotides of DNA ( Itsathitphaisarn et al . , 2012 ) , whereas E1 binds only a single nucleotide per protomer ( Enemark and Joshua-Tor , 2006 ) . The similarity of the Mcm2-7 spiral to DnaB raises the interesting possibility that the helicase might translocate with a step-size greater than one nucleotide per ATP consumed; consistent with this notion , a recent study has shown that the MCM N-terminal DNA-binding collar of Pyrococcus furiosus binds four nucleotides per subunit ( Froelich et al . , 2014 ) . A third notable attribute of the ternary DNA–CMG–ATPγS model is that several structural features of the complex turn out to be most similar not to replicative helicases , but to a completely orthogonal system , namely , the regulatory subcomplex of the eukaryotic proteasome . The proteasome consists of several discrete subcomplexes including a heterohexameric unfoldase region , termed the ‘base’ , which ( like the CMG ) contains six homologous AAA+ ATPase subunits ( Rpt1-6 ) ( Forster et al . , 2013 ) . Recent cryo-EM studies have imaged the complete 26S yeast proteasome bound to ATP at ∼9 Å resolution showing that the AAA+ subunits of the base also form a right-handed spiral ( Lander et al . , 2012 ) . Comparison of proteasome spiral with that seen here for the CMG shows that these regions of the two systems exhibit a surprisingly similar global architecture ( Figure 4E ) . Moreover , the proteasome also contains an accessory subunit ( Rpn1 ) that—as observed here for GINS–Cdc45 in the context of the CMG—wedges itself between a subset of ATPase and OB-fold domains present in Rpt1-6 ( Lander et al . , 2012; Figure 4E ) . The structural congruencies exhibited between the CMG and proteasome ATPase subcomplexes suggest that , even though the substrates for the two systems differ greatly , both motors may share certain commonalities in how ATP turnover is coupled to movements that promote translocation . Such a similarity could underlie both the pronounced asymmetry of the CMG and proteasome ATPase rings , and the relatively high degree of tolerance shown by both systems toward active-site mutations within certain subunits ( Moreau et al . , 2007; Ilves et al . , 2010; Beckwith et al . , 2013 ) . Although the CMG has been observed to operate as a discrete single complex during replication ( Yardimci et al . , 2010 ) , the loading of the Mcm2-7 hexamer onto DNA by ORC , Cdc6 , and Cdt1 during initiation results in the transient formation of a catalytically inactive , head-to-head double hexamer intermediate ( Evrin et al . , 2009; Remus et al . , 2009 ) . The MCM N-terminal domains have been shown to comprise the dimer interface of the double hexamer ( Fletcher et al . , 2003; Costa et al . , 2006; Remus et al . , 2009 ) , and create a critical target site for the Dbf4-dependent Cdc7 protein kinase DDK ( Labib , 2010; Sheu and Stillman , 2006 , 2010 ) ; phosphorylation of the N-terminal tails of Mcm2/4/6 alters the configuration of the Mcm2-7 complex , but does not directly promote Mcm2-7 dissociation ( On et al . , 2014 ) . At present , it is unclear how the Mcm2 , Mcm4 , and Mcm6 subunits are aligned with respect to one another in the Mcm2-7 double hexamer . However , in the course of our studies , we found a new configuration of the CMG that sheds light on this issue . In particular , a small ( ∼5% ) but consistent population of CMG particles was seen to form a clear dimeric species that adopts a distinctive head-to-head configuration through its MCM N-terminal regions , and which consistently orients the GINS–Cdc45 subcomplex toward opposing sides of the two ring ( Figure 5A ) . The abundance and uniformity of CMG dimers , which were noted in several independent preparations , suggests that these particles represent a naturally occurring state of the assembly . 10 . 7554/eLife . 03273 . 010Figure 5 . The CMG can form head-to-head dimers that establish the interactions between Mcm2-7 double hexamers . ( A ) Experimentally observed 2D class averages showing that the CMG forms a defined double-hexamer in which the GINS-Cdc45 subcomplex is rotationally offset toward opposite sides of the Mcm2-7 rings . Arrowheads mark the position of GINS/Cdc45 . The structure of the archaeal N-terminal double hexamer is shown to depict the six possible Mcm2-7 registers that were tested . The class averages show that the interface between the N-terminal collars can partially crack open in some instances , suggesting that this region is somewhat unstable when the CMG is bound to ssDNA . ( B ) Computationally derived 2D projections of double-hexameric 3D CMG models in which Mcm2-7 N-terminal domain dimers ( modeled on an archaeal Mcm crystal structure of this region , PDB ID 1LTL ) have been manually offset by distinct rotational registers . The reference-free class averages best resemble a model in which Mcm5 from ring 1 juxtaposes with Mcm4 from ring 2 , although our data cannot formally rule out a configuration where Mcm4 might juxtapose with Mcm3 or Mcm2 ( although in these configurations the two Mcm2/5 gates would still remain misaligned ) . Arrowheads mark the position of GINS/Cdc45 . ( C ) Cartoon representation of the Mcm2-7 double ring register formed in head-to-head dimers of the CMG . DOI: http://dx . doi . org/10 . 7554/eLife . 03273 . 010 The observed organization of the CMG dimers has several implications for the formation and function of the helicase . For example , the 180° offset of GINS and Cdc45 present in the dimer places each Mcm2/5 gate on the opposite sides of the complex , facing away from each other ( Figure 5B , C ) . During origin melting , this configuration would allow a single DNA strand to escape each Mcm2-7 hexamer without steric interference from its partner CMG , enabling particle separation and the formation of two independent replication forks . The organization of the CMG dimer also indicates that the N-terminal regions of Mcm2 and Mcm5 of one hexamer associate in trans with the N-terminal regions of Mcm6 and Mcm4 of the partner hexamer . Such an interaction would help explain why Mcm4 and GINS have been seen to interact in pulldown studies ( Ilves et al . , 2010 ) , even though the two factors map to distal positions of the CMG in the context of a monomer ( Costa et al . , 2011 ) . Finally , the observed arrangement suggests that the ability of DDK to activate Mcm2-7 by phosphorylation of Mcm4 , 2 , and 6 ( Labib , 2010 ) ( as well as the ability of the Bob1 mutation in Mcm5 to bypass the requirement for DDK [Jackson et al . , 1993] ) could result from a destabilization of the CMG dimer contacts that are symmetrically apposed across the N-terminal collar ( Figure 5C ) . Altogether , our data establish that the two Mcm2/5 gates of a double hexamer are spatially segregated from each other and indicate that separation of Mcm2-7 double hexamers occurs subsequent to CMG formation . Such a mechanism is consistent with recent findings showing that the phosphorylation by DDK is insufficient to promote the separation of Mcm2-7 double hexamers on its own ( On et al . , 2014 ) . The existence of a dimeric CMG state , coupled with prior views of Mcm2-7 and the DNA-bound conformation of the Mcm2-7 ATPase domains seen here , highlights the innate plasticity of the MCM ring and the means by which different factors help remodel the helicase to support appropriate loading and activation during the initiation of DNA replication ( Figure 6 ) . Biochemical studies using Saccharomyces cerevisiae proteins first showed that Mcm2-7 on its own possesses a natural discontinuity—the ‘Mcm2/5 gate’—through which DNA can enter and exit the helicase pore ( Bochman and Schwacha , 2007 ) . Structural studies have corroborated this observation , additionally showing that metazoan Mcm2-7 rings preferentially assume a left-handed lock-washer shape and that ATP alone is incapable of fully inducing ring closure ( Costa et al . , 2011; Lyubimov et al . , 2012 ) . Following the action of ORC , Cdc6 , and Cdt1 , two Mcm2-7 hexamers become locked into a planarized co-joined ring ( Evrin et al . , 2009; Remus et al . , 2009 ) ; in the presence of GINS , Cdc45 , and single-stranded DNA , the Mcm2-7 ATPase domains shift again , but now into a right-handed spiral conformation ( Figure 4 ) . Thus , Mcm2-7 undergoes a chiral-flip in architectural state as it matures into a functional helicase with loading , activation , and DNA binding all appearing to participate in these transitions . 10 . 7554/eLife . 03273 . 011Figure 6 . Overview of Mcm2-7 organization , remodeling , and gate status during initiation and CMG formation . Following synthesis and assembly , the isolated metazoan Mcm2-7 motor forms an inactive left-handed spiral , irrespective of nucleotide state , with a discontinuity between Mcm5 and Mcm2 . The action of ORC , Cdc6 and Cdt1 results in the loading of a planar , head-to-head Mcm2-7 double hexamer onto a duplex DNA in which the Mcm2/5 gates are closed . Following loading , Dpb11-Sld2-Sld3 ( 11-2-3 ) chaperone GINS and Cdc45 onto the Mcm2-7 double hexamer and along with DDK promoted phosphorylation events , help promote both DNA melting , CMG formation and replication fork separation . Structural analysis of dimeric CMG particles ( Figure 5 ) indicates that the Mcm2/5 gates are localized on the opposing sides of the dodecameric Mcm2-7 complex and that CMG formation precedes separation of the double hexamer . DOI: http://dx . doi . org/10 . 7554/eLife . 03273 . 011 Why might Mcm2-7 traverse through such disparate intermediates ? One possibility is that Mcm2-7 initially assembles into an inactive conformation to prevent inadvertently triggering aberrant replication events . An alternative possibility , not necessarily incompatible with the first , is that Mcm2-7 might transition through different forms as a means to promote the melting of duplex replication origins . Interestingly , an important attribute of some of the states adopted by Mcm2-7 is the existence of a breach or subunit offset in the ATPase ring between the Mcm2 and Mcm5 subunits . Given the topological barrier that extended DNA segments present to encirclement by toroidal proteins , it is notable that two MCM-associated factors—Cdt1 and GINS–Cdc45—both appear to occlude the Mcm2/5 interface ( Costa et al . , 2011; Sun et al . , 2013 ) . Hence , controlled access through the Mcm2/5 gate appears useful not only for regulating DNA access into a preformed Mcm2-7 hexamer ( Bochman and Schwacha , 2007 ) , but also for preventing accidental DNA egress in the apo state and at stalled forks ( Petojevic et al . , unpublished data ) . The structure of the eukaryotic CMG complex bound to nucleotide and a 3′-tailed DNA substrate helps resolve many outstanding questions surrounding the mechanism by which this key replication assembly forms and operates . The structure reveals that the RecJ domain of Cdc45 , which recent data have found to bind to escaped leading strand substrates ( Petojevic et al . , unpublished data ) , takes up a position on the CMG that would aid in capturing a DNA segment that might escape from the Mcm2-7 pore . Our EM data show that the nucleotide-loaded CMG orients on single-stranded DNA such that the 3′ end enters first through the C-terminal AAA+ domains of Mcm2-7 , consistent with prior biochemical data indicating that archaeal MCMs move along DNA with the opposite polarity as superfamily III helicases ( McGeoch and Bell , 2005; Rothenberg et al . , 2007 ) . The binding of DNA to the CMG also induces the formation of a right-handed spiral in the MCM ATPase domains whose overall structure more closely mimics bacterial RecA-family replicative helicases than its viral AAA+ cousins , and whose relative arrangement of AAA+ domains and accessory subunits is mirrored by the regulatory subcomplex of the eukaryotic proteasome ( Lander et al . , 2012 ) . Finally , we find that the CMG can form head-to-head dimers in which the two Mcm2/5 gates are fully offset from each other , helping to fill in key gaps concerning the higher-order organization of the Mcm2-7 double hexamer and how the CMG matures from this state into two single particles that encircle complementary single strands . Future studies will be needed to definitively establish the full paths of the leading- and lagging-strand DNA bound to the CMG and the extent to which the disparate AAA+ and RecA-family ring-translocases share or diverge in coupling ATP turnover to specific subunit movements that drive substrate translocation .
Baculoviruses were constructed as previously described ( Ilves et al . , 2010 ) . Briefly , the MCM3 gene was tagged with FLAG epitope at the amino terminus through 5′ PCR oligonucleotides that inserted the epitope in frame . Mcm2 , Mcm3 , Mcm4 , Mcm5 , Psf1 , Psf2 , and Psf3 cDNAs were inserted between EcoRI and SpeI restriction sites of the pFastBac1 vector . The cDNA of Mcm6 and Mcm7 was inserted between BamHI and SpeI sites , and Cdc45 and Sld5 cDNA between EcoRI and Xba restriction sites . These vector templates were used for generation of CMG mutants through PCR based mutagenesis . Sequencing was used to verify the entire protein coding regions of all generated pFastBac1 constructs . Specific deletion mutants of either the Psf1 C-terminus or Cdc45 N-terminus included Psf11−139 , Psf11−170 , Psf11−176 , Psf11−184 , and Cdc45Δ1−99 . C-terminal deletions were constructed by introducing stop codons after the desired residue and by removing any remaining C-terminal sequences present from the original cDNA clone . N-terminal deletions were constructed by the removal of the pertinent cDNA regions and by introducing a start ( ATG ) codon in front of the desired residue . The same restriction enzyme sites as described above were used to subclone all truncation constructs into the pFastBac1 vector . To target the Psf1–Cdc45 interface , alanine substitutions in the C-terminus of Psf1 were introduced individually or in parallel into for Glu190 , Leu192 , Val193 , and Arg194 by site-directed mutagenesis . The CMG complex was purified as previously described ( Ilves et al . , 2010 ) , but with the following changes . Briefly , after co-infection with 11 distinct baculoviruses , proteins were expressed for 72 hr in Hi5 cells ( Invitrogen , Carlsbad , CA ) by culturing in 500-ml spinner flasks and co-infecting at 1 . 2^106 cells/ml density . Cells were lysed by hypotonic shock combined with one freeze–thaw cycle and Dounce homogenization . The lysate was clarified by centrifugation , and the protein complex was immunoaffinity purified over anti-flag ( M2 ) antibody-conjugated agarose beads ( Sigma , St . Louis , MO ) to bind a FLAG-tag on Mcm3 . To isolate the fully intact CMG from incomplete complexes , the protein preparation was passed over a Mono S HR 5/5 ion exchange column , followed by a Mono Q HR 5/5 ion exchange chromatography using an ÄKTA Purifier ( GE Healthcare , Piscataway , NJ ) . Peak fractions were collected and then both further purified and concentrated using a Mono Q PC 1 . 6/5 column coupled to a Pharmacia SMART system . The purified material was dialyzed into a buffer containing 25 mM HEPES ( pH 7 . 6 ) , 50 mM sodium acetate , 10 mM magnesium acetate , and 1 mM DTT . The final protein concentration as measured by the Bradford protein assay was 1 . 2 mg/ml . Oligonucleotides used for nucleoprotein complex reconstitution were synthesized by Integrated DNA Technology and shipped as lyophilized pellets . Oligo ‘LEAD60’ contained the sequence 5′–GGG-CAC-TTG-ATC-GGC-CAA-CCT-T39–3′ , while ‘3BTNLAG20’ contained the sequence 5′–GGT-TGG-CCG-ATC-AAG-TGC-CC–biotin–3′ . The oligonucleotides were dissolved in the CMG buffer and quantified by A260 . For annealing , LEAD60 and 3BTNLAG20 were mixed in equimolar amounts , briefly heated to 95°C and slow-cooled to 4°C . Previous work has shown that the ATPγS binds to a duplex-DNA substrate containing a 40mer 3′ single-stranded DNA tail ( Ilves et al . , 2010 ) . To form helicase/DNA complexes , 10 nmol of concentrated CMG were mixed with the annealed DNA with a 1 . 2 molar excess of nucleic acid in CMG buffer plus 0 . 1 mM ATPγS . After incubation at room temperature for 30 min , the sample was passed over a Superose 6 PC 3 . 2/30 gel filtration column using an ETTAN micropurification system ( GE Healthcare ) . A 50-μl fraction containing the center of the nucleoprotein elution peak was collected and immediately used for negative stain grid preparation , either with or without co-incubation with 1 . 2-fold molar excess of streptavidin . EM grids were prepared by floating a thin layer of continuous carbon over a 400-mesh copper grid ( Electron Microscopy Sciences , Hatfield , PA ) using a custom-made carbon-floating device . Four microliters ( ∼40 ng ) of the CMG–DNA or CMG–DNA–Streptavidin complex were then applied onto freshly glow-discharged grids for 30 s . The grids were laid on top of 75-μl drops of a fresh 2% ( wt/vol ) uranyl formate solution and stirred for five consecutive 10-s staining steps . The staining solution was then blotted dry and the grids were stored . Nucleoprotein particles were imaged using a JEM-2100 LaB6 electron microscope ( JEOL , Japan ) operated at 200 kV . Images were recorded at a nominal magnification of 50 , 000× on a Ultrascan 4k × 4k CCD camera ( Gatan , Pleasanton , CA ) , resulting in a 2 . 14 Å pixel size at the specimen level . The JEOL Automated Data Acquisition System ( JADAS , Zhang et al . , 2009 ) was used to automatically collect low-dose images with a 0 . 5 to 3 . 5 µm defocus at around 35 electrons per Å2 . In addition , manual data collection was performed under the same imaging conditions . In total , 579 micrographs were collected for the non-labelled and 436 micrographs for the labeled nucleoprotein complex . Particles were semi-automatically picked and phase flipped using the EMAN2 package , version 2 . 05 ( Tang et al . , 2007 ) . Reference free two-dimensional class-sums were obtained using RELION , version 1 . 2 ( Scheres , 2012 ) , except for the streptavidin-DNA-bound CMG complex that was processed using Imagic ( van Heel et al . , 1996 ) and the rotation and classification protocol implemented in Costa et al . ( 2011 ) . Working with the ADP ( BeF3 ) bound CMG as a starting model ( Costa et al . , 2011 ) , multi-model three-dimensional refinement was performed with an iterative projection-matching and back-projection protocol that employs libraries from the EMAN2 and SPARX software packages ( Hohn et al . , 2007; Tang et al . , 2007 ) or , in a parallel effort , using three-dimensional classification and refinement routines as implemented in RELION ( Scheres , 2012 ) . Refinement of the starting models began using an angular increment of 15° , progressing down to 2° with EMAN2/SPARX , and default parameters in RELION 1 . 2 ( 7 . 5° angular sampling , 5 pixel search range , and 1 pixel search step ) . The two approaches yielded virtually identical results ( Figure 1—figure supplement 1 ) with the best structure containing 7 , 409 particles ( obtained with EMAN2/SPARX and shown in all figures ) . The resolution was estimated by the ‘gold-standard’ Fourier Shell Correlation approach implemented in RELION ( Scheres , 2012 ) . 3D-maps were segmented using the Segger program ( Pintilie and Chiu , 2012 ) in UCSF Chimera ( Pettersen et al . , 2004 ) . The Chimera ‘Fit in Map’ option was used for rigid-body fitting of crystal structures and to generate all surface renderings included in the figures ( Pettersen et al . , 2004 ) . The full-length Psf1 protein was modelled by superposing the Thermococcus kodakaraensis GINS51 ( PDB entry 3ANW ) onto the human Psf1 structure . The N-terminal Mcm2-7 collar was modelled by superposing protomers of the Methanothermobacter thermautotrophicus MCM ( PDB entry 1LTL ) or the Sulfolobus solfataricus MCM ( PDB entry 2VL6 ) onto the isolated oligomerization subdomain of the M . thermautotrophicus MCM . The two structures exhibit markedly distinct configurations in their N-terminal α-helical region ( subdomain A ) , with the S . solfataricus MCM region matching best to the N-terminal domain configuration of Mcm4 and Mcm5 . The atomic model of the near full-length S . solfataricus MCM ( PDB entry 3F9V ) was found to fit the entire Mcm4 electron density region , whereas slight rigid body transformations of the MCM AAA+ domain relative to the N-terminal region were allowed to best fit the electron density of Mcms 2 , 3 , 5 , 6 , and 7 . A homology model based on the bacterial RecJ ( PDB entry 1IR6 ) was used for docking of Cdc45 . The EM map has been deposited in the 3D-EM database ( www . emdatabank . org ) with accession code EMD-2772 . | Before a cell divides , it must duplicate its DNA so that each new cell inherits its own copy of the genome . To do this , the DNA double helix must be unwound so that the two individual strands of DNA can serve as templates for making new DNA molecules . Unwinding begins when two helicase complexes , termed the Mcm2-7 rings , are loaded together onto the DNA . At first , the two Mcm2-7 rings encircle the double-stranded DNA and remain bound together in an inactive form . Activating the Mcm2-7 rings requires the binding of five other proteins to each ring , which forms two larger complexes called CMG helicases . When the CMG helicases form , the two DNA strands separate and an individual Mcm2-7 ring ends up encircling each of the single DNA strands . However , how an activated CMG complex is assembled , and how it binds to and unwinds DNA , is not fully understood . Now , Costa et al . have determined the three-dimensional structure of the fruit fly CMG helicase bound to a DNA double helix with a single-stranded overhang at one end . The activated Mcm2-7 ring binds to the overhang , which confirms previous findings indicating that the activated helicase prefers single-stranded over double-stranded DNA . The structure also shows that , as a CMG helicase slides along the single-stranded DNA towards the double-stranded DNA , it is the ring complex's ‘motor domains’ that lead the way , while its DNA-binding domains trail behind . Costa et al . also found that disrupting some of the interactions between two of the five proteins that bind to the Mcm2-7 ring either prevented the replicative helicase from forming or made it unstable . Furthermore , it was revealed that one of these two proteins—called Cdc45—was ideally placed to capture the strand of DNA that might be accidentally released from the Mcm2-7 ring . It was also discovered that when the complex is bound to DNA , the motor domains of the Mcm2-7 complex change shape from a flat ring to a spiral structure; the DNA-binding domains , however , remain in a flat ring . Costa et al . note that this structure is similar to that adopted by many viral and bacterial helicases , and that it even shares many features with the molecular machinery that breaks down unneeded or damaged proteins inside cells . Finally , Costa et al . were able to image a structure composed of two CMG complexes bound together . This reveals the relative orientation of the two Mcm2-7 rings before they separate and move in opposite directions to unravel the DNA . The findings of Costa et al . , combined with previous structural work in this field , demonstrate that the Mcm2-7 helicase complex can adopt many different shapes as it is assembled on DNA and activated to support DNA replication . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | DNA binding polarity, dimerization, and ATPase ring remodeling in the CMG helicase of the eukaryotic replisome |
Here , we document a collection of ∼7434 MiMIC ( Minos Mediated Integration Cassette ) insertions of which 2854 are inserted in coding introns . They allowed us to create a library of 400 GFP-tagged genes . We show that 72% of internally tagged proteins are functional , and that more than 90% can be imaged in unfixed tissues . Moreover , the tagged mRNAs can be knocked down by RNAi against GFP ( iGFPi ) , and the tagged proteins can be efficiently knocked down by deGradFP technology . The phenotypes associated with RNA and protein knockdown typically correspond to severe loss of function or null mutant phenotypes . Finally , we demonstrate reversible , spatial , and temporal knockdown of tagged proteins in larvae and adult flies . This new strategy and collection of strains allows unprecedented in vivo manipulations in flies for many genes . These strategies will likely extend to vertebrates .
Discoveries of gene function are most often driven by the integration of information related to cellular and subcellular protein expression patterns , genetic loss of function phenotypes , and protein–protein interactions . Antibodies to individual proteins can be used to localize their expression and for protein interaction studies . However , antibodies are not available for most proteins , and it is time consuming and expensive to generate them against a large number of individual proteins . Although antibodies have been documented against nearly 1900 fly proteins ( St Pierre et al . , 2014 ) , most have been lost or are not obtainable anymore . An extensive survey reveals that antibodies are currently available against ∼450 different proteins in Drosophila , a mere ∼3% of all protein coding genes ( Adams et al . , 2000 ) . An alternative to generating antibodies to individual proteins is to tag the genes with a common protein segment for which antibodies are commercially available . Endogenous tagging of proteins with a fluorescent marker not only provides useful localization data , but may also permit the conditional removal of the gene products . Standard epitope tags such as Flag , HA , V5 , GFP and other fluorescent tags allow the use of high-fidelity , commercially available antibodies against these tags , permitting biochemical experiments . Fluorescent tags also facilitate imaging in unfixed tissue or live animals . The most commonly used method of in vivo protein tagging is to introduce a transgene that contains a cDNA with a terminal epitope tag sequence that is expressed using an exogenous promoter ( Bischof et al . , 2013 ) . Although this may permit the determination of the subcellular localization of the protein , it does not allow assessment of the native expression pattern . It may also affect the distribution and function of the protein , because the transgene is typically not expressed at the endogenous level . An alternative is to tag the genomic locus using a genomic transgene , rather than a cDNA , because native regulatory elements then control the expression patterns and levels ( Venken et al . , 2008; Ejsmont et al . , 2009 ) . This genomic transgene can be integrated by a site-specific integrase into a defined docking site to minimize chromosomal position effects on the expression of the transgene . The tagged genomic transgene allows the analysis of the spatial and temporal patterns of expression of the protein , but does not provide a means to alter the expression of the endogenous copy of the gene . Finally , the most effective and informative strategy is to tag genes in their endogenous locations ( Venken et al . , 2011b ) . A large collection of such GFP tagged genes would permit numerous powerful genetic and biochemical applications in addition to determining expression patterns and protein distributions . These include efficient immunoprecipitations with anti-GFP nanoantibody followed by mass spectroscopy ( Neumüller et al . , 2012 ) , ChIP sequencing ( Nègre et al . , 2011 ) , iGFPi ( Neumüller et al . , 2012 ) and deGradFP mediated protein degradation ( Caussinus et al . , 2011; Urban et al . , 2014 ) . To tag numerous endogenous genes in Drosophila , protein trapping methods were previously developed based on screening untargeted insertions of transposable elements carrying a protein trap cassette . The first such protein trap vectors used P-elements , piggyBacs and piggyBacs with an internal P-element sequence ( Morin et al . , 2001; Buszczak et al . , 2007; Quiñones-Coello et al . , 2007; Aleksic et al . , 2009 ) . Only a small percentage of the transpositions function as protein traps , because the insertion must be within the intron between two protein-coding exons , as well as being in the right orientation and reading frame to create an in-frame protein fusion when the artificial exon is spliced into the mRNA of the inserted gene . The pooled results of three major efforts ( Buszczak et al . , 2007; Quiñones-Coello et al . , 2007; Aleksic et al . , 2009 ) yielded less than 600 unique genes containing protein traps ( Quiñones-Coello et al . , 2007 ) . Because of the insertion site biases of these transposons , it has been argued that this technology will not allow tagging more than about 5% of the Drosophila genes and that additional screening would yield very few novel tagged genes/proteins; hence , alternative approaches are needed ( Aleksic et al . , 2009 ) . We have previously shown that the Minos transposon-based MiMIC gene trap vector is much more efficient at generating intronic insertions in a much larger subset of Drosophila genes than either P-element or piggyBac vectors ( Venken et al . , 2011a ) . Moreover , MiMIC insertions in coding introns can be efficiently converted using RMCE to label the protein with a GFP or other epitope tag . We have vastly expanded the MiMIC collection , which now totals more than 7400 lines . We show that MiMIC is highly mutagenic and is an extremely efficient tool for gene/protein tagging . We created a new resource of 400 protein tagged genes and show that ∼72–77% of essential genes with internal GFP tags are functional . Importantly , iGFPi and deGradFP permit a temperature-dependent conditional knockdown of gene function that mimics a severe loss of function in specific cells or tissues in most instances . Finally , we document the reversible tissue-specific knockdown of proteins and reversible loss of function of the dunce gene . Hence , the MiMIC protein trap collection is a valuable resource as it allows numerous different applications . The resource and tools described here will allow researchers to address important biological questions , particularly in adult flies , as very limited tools are available to conditionally remove and restore protein function in the adult .
The goal of the Drosophila Gene Disruption Project ( GDP ) is to create resources to manipulate as many genes as possible ( Bellen et al . , 2011 ) . Currently , we use a Minos-based transposable element , because Minos has less insertion bias than the P-element and piggyBac transposable elements ( Thibault et al . , 2004; Metaxakis et al . , 2005; Bellen et al . , 2011; Spradling et al . , 2011 ) . We previously engineered the MiMIC gene trap vector , which contains a phiC31 attP site , a splice acceptor ( SA ) followed by stop codons in the three reading frames , a polyadenylation signal sequence , the yellow+ marker gene , and a second attP site in the opposite orientation ( Figure 1A ) . We previously generated and sequenced 4464 insertion lines and reported a curated collection of 1269 MiMIC insertions ( Figure 1—figure supplement 1 , [Venken et al . , 2011a] ) . 10 . 7554/eLife . 05338 . 003Figure 1 . Protein tagging with the MiMIC system . ( A ) Schematic of Recombinase-Mediated Cassette Exchange ( RMCE ) . The MiMIC transposable element consists of a splice acceptor ( SA ) followed by stop codons for all three reading frames , the EGFP coding sequence ( a readout for stop codon skipping ) , a polyadenylation signal ( PA ) and the yellow+ marker flanked by two inverted attP sites and two Minos inverted repeats . The gene trap cassette between the two attP sites can be replaced with the protein trap cassette containing a splice acceptor ( SA ) , EGFP-FlAsH-StrepII-TEV-3xFlag tag and splice donor ( SD ) flanked by two inverted attB sites . ( B ) Summary of relevant features of the MiMIC insertion collection based on the FlyBase 6 . 01 gene annotation . Note that the counts of insertions associated with features of the gene annotation do not sum to the total number of insertions . This is because some insertions are associated with more than one gene , some genes are associated with more than one insertion , and many genes have multiple annotated transcript isoforms . ( C ) Summary of viability of the sample of 200 lines with GFP-tagged genes . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 00310 . 7554/eLife . 05338 . 004Figure 1—source data 1 . Characterization of 200 unique protein trap lines: MI: MiMIC insertion , GT: Gene Trap , PT: Protein Trap , Y/N: Yes/No and L/V: Lethal/Viable . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 00410 . 7554/eLife . 05338 . 005Figure 1—source data 2 . List of fly strains used in the study . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 00510 . 7554/eLife . 05338 . 006Figure 1—figure supplement 1 . Generating MiMIC insertions . Males carrying the Mi{MIC} cassette inserted on the TM3 , Sb chromosome were crossed to females carrying a heat shock inducible source of Minos transposase . Progeny were heat shocked at 37°C for 1 hr for 5 consecutive days . The resulting progeny were crossed to y w flies . F2 flies were screened for wild type body color ( y ) + . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 006 To expand the MiMIC collection , we generated and screened an additional 11 , 196 single-insertion lines , mapped 10 , 504 additional insertions to unique sites in the genome sequence using inverse PCR , and selected 6131 additional strains for the GDP collection . Consistent with previous studies of Minos insertion sites ( Metaxakis et al . , 2005; Bellen et al . , 2011; Venken et al . , 2011a ) , a very significant fraction of unselected insertions ( 38 . 6% ) are in coding introns . As shown in Figure 1B we selected a total of 2854 MiMIC insertions in coding introns of 1862 distinct genes for inclusion in the GDP collection . Because many genes encode multiple protein isoforms , not all coding-intron insertions are equally useful . The collection includes 1732 insertions in constitutive coding introns that permit tagging of all annotated protein isoforms ( Gold set ) , 814 insertions in alternative coding introns that permit tagging of more than 50% of annotated protein isoforms ( Silver set ) , and 328 insertions in alternative coding that permit tagging of less than 50% of annotated protein isoforms ( Bronze set ) . Note that 78 of the coding intron insertions map within coding introns of two distinct , overlapping genes . The expanded MiMIC collection also includes insertions in coding exons , untranslated regions , non-coding introns , and putative control regions ( within 500 bp of the promoter ) of 2860 protein-coding genes and 359 non-coding RNA genes , as well as 1439 intergenic insertions . In total , the collection comprises 7434 insertions in 7400 lines associated with 4367 genes; 34 lines contain two insertions each . The project website ( http://flypush . imgen . bcm . tmc . edu/pscreen/ ) has a searchable database of all of the MiMIC lines that are currently available . MiMIC insertions in coding introns that are in the same orientation as the transcript ( ∼50% of coding intron insertions ) should function as gene traps ( GT ) . To further investigate the efficacy of MiMIC as a gene trap and other important properties of tagged proteins/genes , we characterized 200 genes with intronic MiMIC insertions by tagging with RMCE . To establish whether MiMIC functions as an efficient gene trap when inserted in the proper orientation in a coding intron , and to assess the efficiency of the splice acceptor site ( SA ) we focused on lethality . Of the 200 genes analyzed , 114 are essential genes based on FlyBase records ( St Pierre et al . , 2014 ) , 84 are non-essential , and 2 are unknown ( Figure 1C and Figure 1—source data 1 ) . The MiMIC insertions in coding introns of 63 of the 114 essential genes are inserted in the correct orientation to function as gene traps , and 58 of these 63 cause homozygous lethality and fail to complement null alleles or deficiencies ( Figure 1—source data 2 ) that delete the target gene ( Figure 1—source data 1 ) ; hence , MiMIC is highly mutagenic . The remaining 51 essential genes have a MiMIC insertion in the non-GT orientation . All these lines are homozygous viable or complement a null allele or a deletion , showing that an intronic MiMIC inserted in the wrong orientation does not significantly disrupt gene function . Note that; 11/51 of the MiMIC-bearing chromosomes are homozygous lethal , but this lethality is caused by second site mutations based on complementation data ( Figure 1—source data 1 ) . In summary , our data show that MiMIC functions as designed in 109/114 insertion lines , and is thus a highly reliable gene trap . The data also show that the SA in MiMIC is effective and that the integrated artificial exon is probably not frequently skipped by the pre-mRNA splicing machinery . MiMIC insertions in coding introns can be used to introduce an artificial exon encoding one or more protein tags via RMCE as described previously ( Venken et al . , 2011a ) . We selected an EGFP-FlAsH-StrepII-TEV-3xFlag tag ( hereafter abbreviated EGFP ) that is flanked on either side with a 4X ( GlyGlySer ) flexible linker ( Figure 1A ) . There are three versions of this protein trap cassette , one for each of the three intron reading frames . When the cassette with the proper reading frame is inserted into a coding intron in the proper orientation , the tag will be spliced into the mRNA of the target gene , and translation of this mRNA will result in a fusion protein with an internal tag inserted . We injected a plasmid DNA containing the donor tag cassette into embryos of ∼700 MiMIC strains ( intronic insertions ) expressing phiC31 integrase . We identified G1 progeny in which the RMCE event had occurred by the loss of the y+ marker of the MiMIC gene trap cassette and determined which RMCE events were in the proper orientation using a PCR assay ( Venken et al . , 2011a ) . We established 450 independent stocks in which ∼400 different genes are tagged . The success rate of obtaining at least one insertion in the correct orientation is currently 133/200 or 66% . Upon reinjection we obtained a 62% success rate . Hence , with two injections of about 500 embryos each we derived ∼175/200 properly tagged genes . A concern with protein tagging is that the tag might disrupt protein function . We were not able to assess what fraction of proteins retains normal function when tagged at terminal or internal locations , despite an extensive survey of the current literature . As a proxy , we characterized tagged alleles of essential genes . If a tag disrupts the function of an essential protein , it should cause lethality . After tagging 114 essential genes using RMCE as described above , we performed complementation tests with null mutations or deletions ( Figure 1—source data 2 ) to establish the fraction of tagged genes that failed to complement a severe loss of function or null allele . The lethality of 42 out of 58 lethal gene trap lines was reverted ( 72% ) upon tagging . Moreover , all five of the essential tagged genes that contain an insertion in the GT orientation , but are viable , and 41 of the 51 essential tagged genes that contain an insertion in the non-GT orientation are homozygous viable or complement a severe allele or deficiency uncovering the gene ( Figure 1C ) . Hence , 85 show no obvious phenotypes and three show subtle morphological phenotypes , suggesting a minimal impact on protein function . The remaining 26 tagged essential genes failed to complement null alleles or deficiencies . Hence , our data indicate that 72% ( 42/58 ) to 77% ( 88/114 ) of proteins tagged at various internal locations retain function or are not severely affected ( Figure 1C and Figure 1—source data 1 ) . However , a significant number of these chromosomes are homozygous lethal , and in almost all cases tested we have been able to eliminate the lethal mutation with backcrosses . To assess whether internal tagging of proteins within annotated protein domains is more disruptive than tagging within other regions of proteins , we mapped the tag insertion site in each of 200 tagged proteins and determined its position with respect to annotated protein domains . We observe a strong relationship between tagging location , annotated protein domains , and retention of protein function . Of the 88 tagged essential genes that retain protein function , 26 tags are inserted within annotated protein domains and 62 are not inserted within an annotated protein domain ( χ2 , p = 0 . 0001 ) . However , we found no such bias for internal tags that disrupt essential proteins: 14 tags inserted within annotated domains vs 12 inserted within unannotated sequences ( χ2 , p = 0 . 7 ) . In summary , insertion of a tag within known , conserved protein domains is more likely to disrupt protein function , and inserting an artificial exon that encodes a tag is effective and not disruptive in 77% of the proteins tested . The genes that were selected for tagging are a non-random set of genes whose expression is unknown in most cases . They were selected from requests made by members of the Drosophila neuroscience community . Although many were suspected to be expressed in the nervous system , quite a few are expressed in other tissues . As shown in Figure 2A , GFP expression can easily be detected in the brain ( Rab3 interacting molecule: Rim ) , muscles ( Myosin Heavy Chain: MHC ) , imaginal discs ( Abl tyrosine kinase: Abl ) , salivary gland nuclei ( CrebA ) , ovaries ( oo18 RNA-binding protein: orb ) , and testis ( Syncrip ) , in agreement with published expression data ( Lantz et al . , 1994; Fogerty et al . , 1999; Abrams and Andrew , 2005; Graf et al . , 2012; McDermott et al . , 2012 ) . Moreover , localization studies for Delta ( Dl ) and Ecdysone Receptor ( EcR ) show a precise colocalization of EGFP and the protein specific antibodies for Delta and EcR , respectively ( Figure 2—figure supplement 1 ) . Finally , proteins localize to the proper subcellular compartments as shown in Figure 2B for a cytoplasmic or organelle present in cytoplasm associated ( Catalase ) , nuclear ( H6-like-homeobox: Hmx ) and membrane associated protein ( dpr15 ) ( Figure 2B ) ( Beard and Holtzman , 1987; Hofmann et al . , 2010; Özkan et al . , 2013 ) . These examples indicate that the EGFP patterns faithfully report expression and protein distribution . 10 . 7554/eLife . 05338 . 007Figure 2 . Protein expression analysis after RMCE . ( A ) Examples of GFP expression patterns in different tissues: ( a ) larval brain ( Rab3 interacting molecule: Rim ) , ( b ) larval muscles ( Myosin Heavy Chain: MHC ) , ( c ) larval eye imaginal disc ( Abl tyrosine kinase: Abl ) , ( d ) larval salivary gland nuclei ( CrebA ) , ( e ) adult ovaries ( oo18 RNA-binding protein: orb ) , and ( f ) adult testis ( Syncrip ) , were detected using anti GFP antibody . Scale bars , 50 μm . ( B ) Subcellular localization of GFP tagged proteins: ( a ) cytoplasmic/organelle associated localization of the enzyme Catalase in larval gut tissue , Scale bar , 100 μm ( b ) nuclear localization of H6-like-homeobox ( Hmx ) in eye imaginal disc ( Green: Hmx-EGFP , Blue: DAPI ) and ( c ) membrane localization of Dpr15 in larval brain tissue ( Green: Dpr15-EGFP , Red: HRP ) . Scale bars , 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 00710 . 7554/eLife . 05338 . 008Figure 2—figure supplement 1 . Colocalization of protein trap GFP expression with specific corresponding antibodies . Signals from anti Delta ( Dl ) ( Red , top panel ) and anti Ecdysone Receptor ( EcR ) ( Red , bottom panel ) antibodies show colocalization with anti GFP ( green ) from the protein trap insertion in third instar larval wing imaginal disc and salivary gland respectively . Scale bars , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 008 To determine what fraction of our set of 200 proteins are expressed in the CNS of third instar larvae , we performed immunostaining with anti-GFP antibodies and detected EGFP expression in 168/200 lines ( 84% ) . The lines that did not show detectable expression in larval CNS were tested for expression in adult brain , and EGFP expression was observed in 11/32 lines ( 6% of total analyzed lines ) . To further establish what fraction of the expression patterns could be monitored using unfixed tissue , we compared anti-GFP stained third instar larval brains with unfixed brains in 40 randomly selected lines . We were able to detect expression of EGFP fluorescence in unfixed brains in 38 of 40 lines . Although the expression patterns in each case generally matched the pattern observed by GFP antibody staining ( Figure 3 ) , the fluorescence intensity is often lower in unfixed tissue . Note that all the images presented in Figure 2B were taken at the same laser intensity and gain for each pair of images . Increasing the gain and laser intensity revealed very similar if not identical expression patterns . These data show that most internally tagged proteins are present at levels that permit imaging in unfixed tissue . 10 . 7554/eLife . 05338 . 009Figure 3 . In vivo protein detection . Protein expression and distribution of GFP observed in unfixed third instar larval brains compared to those that were fixed and stained with an antibody against GFP . Each pair was imaged at the same confocal settings . Almost all pairs show very similar expression patterns but the gain or intensity needs to be adapted for genes that are expressed at low levels . Scale bar , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 00910 . 7554/eLife . 05338 . 010Figure 3—figure supplement 1 . A screenshot from the MiMIC protein database . A public website for the resource ( http://flypush . imgen . bcm . tmc . edu/pscreen/rmce ) containing all of the information about the MiMIC lines: insertion sites , associated genes , construct used for tagging , complementation data and images of brain expression patterns . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 010 As part of our objective to generate a useful set of gene and protein trap lines for the fly community , we created a public website for the resource ( http://flypush . imgen . bcm . tmc . edu/pscreen/rmce ) containing all of the information about the MiMIC lines , insertion sites , associated genes , construct used for tagging , complementation data and images of brain expression patterns for each of the 200 lines described here ( Figure 3—figure supplement 1 ) . We are in the process of documenting the expression patterns of the remaining tagged lines and these data will be added to the online project database soon . In addition to being very useful to determine expression pattern and subcellular localization of proteins , the GFP lines can be used to create conditional loss of function mutations via tag-mediated knockdown strategies . We tested two recently developed strategies to conditionally and reversibly knock down GFP tagged genes ( iGFPi ) ( Neumüller et al . , 2012 ) and proteins ( deGradFP ) ( Caussinus et al . , 2011 ) ( Figure 3A ) . There are three GFP RNAi transgenic lines available and each expresses a different short hairpin RNA ( shRNA ) against the EGFP tag under the control of the UAS/GAL4 system to knock down the expression of GFP-tagged genes ( Neumüller et al . , 2012 ) . There are two important advantages of GFP RNAi over gene-specific RNAi: no off-target effects have been documented and a single set of highly selective and efficient GFP shRNAs can be used to knock down the expression of any tagged gene . By using specific GAL4 drivers , knock down of a gene in almost any tissue or stage in developing animals or adults can be achieved ( Jenett et al . , 2012; Jory et al . , 2012; Manning et al . , 2012 ) . Alternatively , a GFP tagged protein can be degraded using the deGradFP system developed by Caussinus et al . ( 2011 ) . deGradFP is an ubiquitination-based system in which the F-Box domain of the Slmb protein N-terminus is fused to a single-chain nanoantibody fragment ( vhhGFP4 ) that recognizes the GFP tag . The system uses the host ubiquitination machinery to target GFP-tagged proteins for proteasomal degradation . To compare these two knockdown strategies , we tested four EGFP tagged genes . To test and compare knockdown strategies , we selected an intronic MiMIC insertion in alpha-Catenin ( α-Cat ) . This protein plays a critical role at the plasma membrane where it acts as an essential physical linker between the cadherin-β-catenin complex and the actin cytoskeleton ( Desai et al . , 2013 ) . We used RMCE to insert the EGFP tag ( Figure 3C ) . The tag is not inserted in a known protein domain . The homozygous EGFP tagged gene is viable and displays no obvious phenotypes . Because α-Cat null mutants are embryonic lethal ( Sarpal et al . , 2012 ) , and its loss specifically in the eye ( Seppa et al . , 2008 ) or ovaries ( Desai et al . , 2013 ) is associated with severe developmental or cellular defects , the tagged protein must be functional . To determine the subcellular localization of the tagged α-Cat protein ( α-Cat-EGFP-α-Cat ) , we performed co-immunostaining with anti-GFP antibody and anti-α-Cat antibodies ( Sarpal et al . , 2012 ) in the third instar larval eye disc . The tagged protein is ubiquitously expressed in the eye disc and localizes to adherens junctions as previously reported for α-Cat ( Sarpal et al . , 2012 ) . The signals from each antibody colocalize and the tagged α-Cat protein localizes to the proper intracellular domain , similar to the wild type protein ( Figure 3D-a; data not shown ) . As tagging with RMCE integrates an artificial exon , we determined whether exon skipping occurs . Given that MiMIC functions as an effective gene trap ( see above ) and since we used the same SA in the EGFP tagging construct ( Venken et al . , 2011a ) , we predicted that exon skipping would be rare . However , the integration of a splice donor ( SD ) in the inserted artificial exon cassette might alter the splicing properties and permit exon skipping . We performed Western blots of adult head extracts of y w and y w; α-Cat-EGFP-α-Cat with anti-GFP and anti-α-Cat antibodies ( Figure 4C ) ( Sarpal et al . , 2012 ) . A single anti-α-Cat reaction band of 100 kDa is detected in y w flies , whereas y w; α-Cat-EGFP-α-Cat extracts show a single band of about 135 kDa , in agreement with the estimated molecular weight ( MW ) of 35 kDa for the protein tag ( Figure 4C , left ) . The 135 kDa band was also observed in the anti-GFP immunoblot ( Figure 4C , right ) . Hence , we are not able to detect untagged protein in homozygous tagged animals , indicating that the vast majority protein is tagged , but we cannot exclude that some protein in not tagged . This observation is important , as knockdown with iGFPi or deGradFP may be ineffective if exon skipping occurs . 10 . 7554/eLife . 05338 . 011Figure 4 . Tagging and knock-down of α−Catenin . ( A ) Two UAS/GAL4 based knockdown strategies targeting GFP-sequence containing mRNA or GFP fusion protein . Left: Expression of a GAL4-inducible shRNA transgene against GFP ( UAS-GFP-RNAi ) will result in gene knockdown by degrading mRNA of fusion protein; Right: GFP fusion proteins can be targeted for ubiquitination-mediated degradation by a modified ubiquitination system called deGradFP ( UAS-NSlmbvhhGFP4 ) . ( B ) Schematic diagram of α-Cat locus ( based on FlyBase annotation release FB2014_05 ) . The coding regions of tagged isoform are shown as green , 5′ and 3′ UTRs are in blue . The insertion site of MI02577 is shown with a red triangle and the orientation is shown with a red arrow . The black bar at the bottom of third exon represents the Vinculin 1 domain . ( C ) Western blots of adult head extracts from control y w and y w; α-Cat-EGFP-α-Cat were probed with anti α-Cat ( on left ) and anti GFP ( on right ) . ( D ) Eye specific disruption of α-Cat expression causes eye related phenotypes . ( a ) An eye disc from a third instar larva expressing α-Cat-EGFP-α-Cat ( y w; α-Cat-EGFP-α-Cat ) is stained with anti GFP ( green ) and anti α-Cat ( red ) antibodies . Scale bar , 50 μm . On the right is a close up of the area boxed in eye disc on the left , which shows there is a strong co-localization of the GFP and α-Cat antibody signals . Scale bar , 5 μm . ( b ) Expression of deGradFP or GFP RNAi using ey-GAL4 at 28°C in the α-Cat-EGFP-α-Cat background result in rough eye phenotypes . Additionally , iGFPi knockdown causes a severe reduction in eye size . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 01110 . 7554/eLife . 05338 . 012Figure 4—figure supplement 1 . Temperature dependent Gal4 Expression . Western blots of extracts from act-GAL4 larvae raised at different temperatures were probed with anti GAL4 and anti Actin as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 01210 . 7554/eLife . 05338 . 013Figure 4—figure supplement 2 . α-Cat knockdown with RNAi in developing eyes . Expression of three different α-Cat RNAi lines using ey-GAL4 at 28°C results in a rough eye phenotype ( a ) , or pupal lethality with no head ( b ) or small head development ( c ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 013 Both iGFPi and deGradFP depend on the GAL4 binary system ( Brand et al . , 1994 ) to control spatial and temporal specificity . The GAL4 system exhibits some temperature dependency due to the presence of an hsp70 promoter upstream of GAL4 ( Duffy , 2002 ) . GAL4 is expressed at low levels at 18°C , while expression is elevated at 28°C . We tested this and show that with the act-GAL4 driver , the temperature sensitive expression of GAL4 is quite pronounced ( Figure 4—figure supplement 1 ) . We therefore wondered whether we could use temperature to modulate the GFP tag-mediated knockdown efficiency . To test this , we ubiquitously expressed deGradFP using act-GAL4 in flies expressing α-Cat-EGFP-α-Cat and maintained the cultures at 18°C or 28°C . Interestingly , y w; α-Cat-EGFP-α-Cat flies expressing deGradFP and kept at 18°C are viable , whereas animals kept at 28°C , to express deGradFP , are embryonic lethal ( Sarpal et al . , 2012 ) . To assess whether knockdown of α-Cat mediated by deGradFP or iGFPi phenocopies previously documented phenotypes , we expressed iGFPi or deGradFP using ey-GAL4 in y w; α-Cat-EGFP-α-Cat and y w control flies . Animals were raised at 28°C to enhance the GAL4 expression level . Both knockdown experiments cause a rough eye phenotype: however iGFPi results in a more severe phenotype ( Figure 4D-b ) . Knockdown with deGradFP causes rough eyes similar to one previously reported RNAi phenotype ( Seppa et al . , 2008 ) and knockdown using the α-Cat specific RNAi , FBst0033430 ( Figure 4—figure supplement 2-a ) , but is less severe than other two unpublished RNAi phenotypes ( FBst0038987 and FBst0038197 ) , both of which are pupal lethal when raised at 28°C ( Figure 4—figure supplement 2-b–c ) . Thus both tag-mediated knockdown strategies are effective at reducing gene function but , may not cause null phenotypes . Null mutations in discs large 1 ( dlg1 ) cause larval lethality , oversized imaginal discs and brains in third instar larvae , and defects in neuromuscular junction development ( Stewart et al . , 1972; Perrimon , 1988; Woods and Bryant , 1991; Woods et al . , 1996; Budnik et al . , 2006; Zhang et al . , 2007 ) . Using RMCE , we converted the MiMIC line MI06353 into a protein trap allele ( Figure 5A ) . The dlg1 locus contains 16 long isoforms and five approximately 24 kDa isoforms , the latter of which may not encode functional proteins ( Mendoza-Topaz et al . , 2008 ) . The MI06353 insertion is in an ideal location as it permits tagging of all the long , presumably functional Dlg1 isoforms . The EGFP insertion site is located between protein domains and is less likely to interfere with normal protein function ( Figure 5A; Figure 1C ) . Homozygous y w dlg1-EGFP-dlg1 flies are viable and do not exhibit obvious phenotypes , indicating that the tagged Dlg1 is able to function normally . 10 . 7554/eLife . 05338 . 014Figure 5 . Knockdown of Dlg1-EGFP-Dlg1 with ubiquitously expressed deGradFP causes characteristic embryonic and larval phenotypes . ( A ) A schematic of the discs-large1 ( dlg1 ) gene region ( based on FlyBase annotation release FB2014_05 ) . The site of MI06353 insertion is shown with the red triangle . Each isoform is labeled by name and molecular weight of the protein product . Isoforms tagged with EGFP at the MiMIC insertion site have black labels , green boxes ( coding exons ) and blue boxes ( 3′- , 5′-UTR exons ) ; while isoforms not tagged with EGFP have red labels with orange and brown exon boxes . The black bars represent each protein domain as they map on the genetic sequence . ( B ) Western blots of head extracts from control: y w and y w dlg1-EGFP-dlg1 were probed with anti Dlg1 ( upper ) , which recognizes the second PDZ domain , and anti GFP ( lower ) . ( C ) A diagram representing temperature conditions used in subsequent experiments to modify protein expression levels . The top bar indicates developmental stages: E ( embryo ) , L1 ( first instar larva ) , L2 ( second instar larva ) , L3 ( third instar larva ) , P ( pupa ) , A ( adult ) . The next bar is a time line ( in days ) of the developmental stages for animals kept at 18°C . The bars below indicate the time at which the animals were shifted to 28°C at the beginning of first , second or third instar , or kept continuously at 28°C . ( D ) Third instar larval brains stained with Dlg1 ( a and d ) and GFP ( b and e ) antibodies . y w dlg1-EGFP-dlg1;;UAS-NSlmbvhhGFP4/tub-GAL4 animals that were raised continuously at 18°C show robust larval brain expression of Dlg1 ( a–c ) with complete colocalization of Dlg1 and GFP expression ( c ) . However , animals that were shifted from 18°C to 28°C as second instar larvae ( see C ) have less Dlg1 expression as judged by both Dlg1 ( d ) and GFP ( e ) antibody staining , however colocalization of Dlg1 and GFP is still present in some areas ( f ) . Additionally , these brains ( d–f ) are significantly larger compared with controls ( a–c ) , which is a characteristic phenotype associated with loss of function alleles of dlg1 . The white arrows point to neuromuscular junctions . Scale bar , 100 μm . ( E ) Wing discs from third instar larvae labeled with GFP and E-cad antibodies . First instar larvae were shifted from 18°C to 28°C . The larvae that ubiquitously express deGradFP , y w dlg1-EGFP-dlg1;;UAS-NSlmbvhhGFP4/tub-GAL4 ( b ) have significantly larger wing discs compared with controls minimally or not expressing deGradFP , y w dlg1-EGFP-dlg1;;UAS-NSlmbvhhGFP4/+ ( a ) , another traditional dlg1 mutant phenotype . Scale bar , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 01410 . 7554/eLife . 05338 . 015Figure 5—figure supplement 1 . Dlg1 knockdown results in aberrant cellular morphology and organization in larval gut . Third instar larval midgut stained with GFP ( a and d ) and E-Cad ( b and e ) antibodies . y w dlg1-EGFP-dlg1;;UAS-NSlmbvhhGFP4/tub-GAL4 animals that were raised continuously at 18°C show robust larval gut expression of Dlg1 ( a–c ) and consistent cellular organization ( a–d ) . However , animals that were shifted from 18°C to 28°C as first instar larvae have less gut Dlg1-EGFP expression ( d ) and aberrant cellular formation and organization with abnormally shaped cells and disruption of the expression pattern for the cellular marker , E-Cad ( d–f ) . Scale bars , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 015 To assess whether exon skipping is occurring , we probed Western blots with anti Dlg1 antibody ( raised against the second PDZ domain; see Figure 4A; [Woods and Bryant , 1991] ) and noted a shift of ∼35 kDa for the various long EGFP tagged Dlg1 isoforms ( Figure 5B , top blot ) . Westerns with anti GFP show a nearly identical banding pattern in these flies , demonstrating that exon skipping is not occurring at a detectable level when assessed with western blots ( Figure 5B , bottom blot ) . Given that the previous GAL4 experiments indicate that there is significant temperature sensitivity we developed a simple diagram to explain how the temperature shifts are performed to temporally regulate expression of GAL4 and therefore deGradFP or iGFPi in subsequent experiments ( Figure 5C ) . Note that development takes about twice as long at 18°C and hence embryonic development lasts for about 2 full days . The Dlg1-EGFP-Dlg1 expression pattern and subcellular distribution as revealed with the anti Dlg1 and anti GFP antibodies also correspond to the published protein distribution ( Figure 4C-a–c ) , further confirming that the GFP tag does not disrupt Dlg1 localization . The y w dlg1-EGFP-dlg1;UAS-deGradFP/tub-GAL4 flies raised at 18°C develop normally and eclose without obvious phenotype . Moreover , GFP expression is not affected in third instar larvae ( data not shown ) . In contrast , embryos that are raised continuously at 28°C are embryonic lethal , an earlier lethal phase than dlg1 null mutants , which are third instar lethal . However , in zygotic null mutants , there is a large maternal component of Dlg1 ( Woods and Bryant , 1991 ) , and germ line clones of dlg1 null alleles are embryonic lethal ( Perrimon , 1988 ) , suggesting that the maternally deposited protein is degraded in y w dlg1-EGFP-dlg1; UAS-deGradFP/tub-GAL4 animals grown at 28°C . Furthermore , animals that are shifted to 28°C as either first or second instar larvae display hyperproliferative imaginal disc and brain phenotypes as third instar larvae ( Figure 5D–E ) . In these animals , we also observe a significant decrease in Dlg1-EGFP-Dlg1 expression in the brain ( Figure 5D-c–e ) and in the wing disc ( Figure 5E ) compared with controls . We also detect a mild decrease in NMJ growth in terms of bouton number and synapse complexity in the temperature shifted animals compared to unshifted controls ( Figure 5D-a , d-white arrows , data not shown ) , consistent with dlg1 loss of function ( Zhang et al . , 2007 ) . However , the decrease in Dlg1-EGFP-Dlg1 in the muscle is far less robust compared with other tissues , such as the brain , because the Tubulin driver leads to stronger expression in the brain compared to third instar larval muscles ( data not shown ) . In addition to hyperproliferation , y w dlg1-EGFP-dlg1; UAS-deGradFP/tub-GAL4 animals shifted to 28°C as first instar larvae , much like dlg1 zygotic mutants , also have defects in cell morphology and organization as observed in the larval midgut compared to animals raised at 18°C ( Figure 5—figure supplement 1 ) . Taken together , these data show that deGradFP not only prevents zygotic expression of Dlg1 , but also eliminates maternal deposited Dlg1 , thus recapitulating phenotypes associated with zygotic null mutations and the lethality associated with mutant germ line clones . Bruchpilot ( Brp ) is a functional homolog of the human presynaptic protein ELKS/CAST/ERC ( Wagh et al . , 2006 ) . It is essential for assembly of active zones , Ca2+ channel clustering at NMJs , and release of neurotransmitters . brp null mutants show loss of presynaptic dense projections and severe defects in synaptic transmission ( Kittel et al . , 2006 ) . We created an EGFP tagged brp allele using MI02987 , which tags all but one annotated transcript ( Figure 6A ) . The y w; brp-EGFP-brp flies are homozygous viable and healthy , whereas severe loss of function brp alleles cause third instar lethality ( Kittel et al . , 2006 ) . To determine whether exon skipping occurs , we performed Western blotting on adult head extracts of y w and y w; brp-EGFP-brp animals with anti Brp ( mAb nc82 ) ( Wagh et al . , 2006 ) and anti GFP antibodies The nc82 antibody recognizes two proteins of 165–175 and 200–210 kDa in y w controls ( Figure 6B , left ) and two bands of 200–210 and 235–245 kDa in brp-EGFP-brp flies ( Figure 6B , right ) . These bands correspond to the 165–175 kDa and 200–210 kDa isoforms that are tagged with EGFP , as the western blots probed with anti GFP show two bands of 200–210 and 230–240 kDa . Hence , both isoforms are present in the tagged flies and we observe no evidence of exon skipping . The spatial protein distributions of the tagged isoforms are not affected in Brp-EGFP-Brp as a comparison of anti Brp and anti GFP staining patterns in the L3 brain and NMJs shows that the signals co-localize in the third instar neuropil and at the active zones in the NMJs , and the localization is similar or identical to published data ( Figure 6C ) . 10 . 7554/eLife . 05338 . 016Figure 6 . Conditional knockdown of Brp-EGFP-Brp phenocopies loss of function alleles . ( A ) A schematic of the bruchpilot ( brp ) locus ( based on FlyBase release FB2014_05 ) , coding exons in green and , 5′ and 3′ UTR exons are in blue . The insertion site of MI02987 is shown with a red triangle and the orientation is shown with a red arrow . On left , tagged isoforms are indicated in black and untagged isoform is indicated in red . The black bars represent the CASK protein domains mapped onto genomic sequence . ( B ) Western blots of head extracts from y w; brp-EGFP-brp were probed with anti Brp ( on left ) and anti GFP ( on right ) . ( C ) y w; brp-EGFP-brp third instar larval brain stained with antibody to GFP ( a ) and neuromuscular junction ( NMJ ) ( b–d ) , was immunostained with antibodies to GFP ( green ) and Brp ( red ) . Scale bars 50 μm and 7 μm , respectively . ( D ) Knockdown of Brp with deGradFP or iGFPi at 28°C using ey-GAL4 driver results in altered physiology in the eye . ( a ) Western blot of adult head extracts probed with GFP antibody ( and Tubulin as a loading control ) . Brp-EGFP-Brp levels are reduced when deGradFP or iGFPi are expressed using ey-GAL4 driver ( at 28°C ) . ( b ) Quantification of ERG amplitudes and on-and off-transients for each genotype shown in ( c ) ( n = 6 ) . ERG amplitudes and on- and off-transients were normalized with respect to controls . Error bars represent SD . ( c ) ERG traces of flies ey-GAL4>deGradFP: y w;UAS-NSlmbvhhGFP4/ey-GAL4 , brp-EGFP-brp; ey-GAL4>deGradFP: y w; brp-EGFP-brp;UAS-NSlmbvhhGFP4/ey-GAL4 , ey-GAL4>GFP-RNAi: y w;UAS-GFP RNAi/ey-GAL4 and brp-EGFP-brp; ey-GAL4>GFP-RNAi: y w; brp-EGFP-brp;UAS-GFP RNAi/ey-GAL4 . Normal on- and off- transients , as shown in the controls , are indicated by red arrows . When either deGradFP or iGFPi is expressed with brp-EGFP-brp , on- and off- transients are lost ( indicated with the red circles ) and ERG amplitude is reduced . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 016 To compare the knockdown efficiencies in the adult eye , we expressed iGFPi and deGradFP using the ey-GAL4 driver in y w; brp-GFP-brp flies and assessed knockdown efficiencies by western blotting of single adult head extracts using anti GFP antibody , which typically reveals only the abundant larger Brp-EGFP-Brp isoforms ( upper band ) . ey-GAL4 drives expression in pupal and adult photoreceptors , lamina and medulla neurons , mushroom body neurons and other adult neurons ( Sheng et al . , 1997 ) . Hence , a decrease of 50% in Brp-EGFP-Brp levels in whole head extracts must correspond to a very substantial loss of the protein in the cells where ey-GAL4 is expressed , which accounts for an estimated 40–60% of brain mass ( Figure 6D-a ) . In agreement with the phenotype of a brp severe loss of function mutation in photoreceptors , we observe a complete loss of on- and off-transients ( Figure 6D-b , c ) in electroretinograms ( ERGs ) using both iGFPi and deGradFP knockdown , indicating severe synaptic transmission defects of photoreceptors . We also observe a 50% reduction in ERG amplitude in both experiments , suggesting that Brp might play a role in eye development or the visual transduction pathway ( Figure 6D-b , c ) . Similar reductions in ERG amplitude were previous described by Wagh et al . ( 2006 ) using RNAi against brp driven by GMR-GAL4 , another eye-specific driver . We performed similar experiments using GMR-GAL4 to drive UAS-deGradFP and UAS-GFP-RNAi and observed virtually identical phenotypes to those described here for ey-GAL4 ( data not shown ) . In summary , our data show that both knockdown systems are effective when the drivers are continuously expressed during development and in the adult . To determine the time needed to remove the Brp-EGFP-Brp protein and cause synaptic transmission defects at third instar NMJ , we expressed deGradFP with n-Syb-GAL4 , which drives expression in the nervous system . Homozygous y w; brp-EGFP-brp; deGradFP/n-Syb-GAL4 flies are viable and healthy at 18°C . We therefore raised y w; brp-EGFP-brp; deGradFP/n-Syb-GAL4 larvae at 18°C until the early third instar larval stage and then shifted them to 28°C for 6–9 hr , 12–16 hr and 18–24 hr to assess the time needed to knock down Brp and cause a loss of synaptic transmission at the neuromuscular junction ( Figure 7A-a ) . deGradFP knockdown efficiency is pronounced at NMJs in larvae that were kept at 28°C for 18–22 hr , as gauged by immunostaining with anti Brp ( nc82 ) ( Figure 7A-b ) and anti GFP ( data not shown ) . 10 . 7554/eLife . 05338 . 017Figure 7 . Neuronal expression of deGradFP in brp-EGFP-brp flies causes defects in synaptic transmission . ( A ) Disruption of brp function with deGradFP at 28°C , using n-Syb-GAL4 driver causes synaptic transmission defect . ( a ) Schematic diagram of temperature shift experimental parameters . y w; brp-EGFP-brp;UAS-NSlmbvhhGFP4/n-Syb-GAL4 larvae were shifted to 28°C as Late L2 or L3 larvae for the time indicated on right . ( b ) NMJ6/7 from third instar larvae that were raised at 18°C and shifted to 28°C for 18–22 hr were stained with an antibody to Brp ( nc82 ) . Brp expression is reduced in y w; brp-EGFP-brp; n-Syb>deGradFP compared with either y w; brp-EGFP-brp or n-Syb>deGradFP . Scale bar; 2 μm . ( c–f ) Electrophysiology was performed in y w; brp-EGFP-brp; UAS-NSlmbvhhGFP4/n-Syb-GAL4 larvae that were shifted to 28°C at the time indicated in ( a ) . EJP amplitudes ( c ) , mEJP amplitudes ( d ) , quantal content ( e ) , of control and knockdown are measured . Both EJP amplitudes and quantal content in knockdowns show a ∼76% reduction when larvae were raised at 18°C and shifted to 28°C for 18–22 hr . ( f ) Representative EJP traces obtained from controls ( black ) and 18–24 hr knockdowns ( red ) . Each electrophysiology recording is performed at 0 . 2 Hz in 0 . 5 mM [Ca2+] HL-3 solution . p value: **p < 0 . 01; ***p < 0 . 001 by Student's t-test . NS , not significant . Error bars indicate SEM . ( B ) A table summarizing lethality caused by disruption of Brp-EGFP-Brp with deGradFP or iGFPi using ubiquitous ( da-GAL4 ) or neuronal ( n-Syb-GAL4 ) GAL4 drivers after shifting animals to 28°C at different developmental stages . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 017 To assess the physiological consequences of Brp loss , we performed electrophysiological recordings on third instar larval NMJs in 0 . 5 mM extracellular Ca2+ . We used y w; deGradFP/n-Syb-GAL4 larvae with an untagged brp gene as control . There are no significant differences in excitatory junction potential ( EJP ) amplitudes and miniature junctional potentials ( mEJPs ) in controls ( EJP: 21 . 9 ± 3 . 4 mV , and mEJP: 1 . 79 ± 0 . 18 , n = 6 ) and 0 hr experimental animals ( EJP: 22 . 9 ± 4 . 5 mV , and mEJP: 1 . 8 ± 0 . 13 , n = 6 ) ( Figure 7A-c–f ) . However when the temperature is shifted to 28°C during third instar larval development , there is a loss in amplitude of the EJPs that becomes progressively more severe the earlier that the shift is made . Larvae kept at 28°C for 18–24 hr have EJP amplitudes that are reduced by 76% ( EJP: 5 . 5 ± 0 . 37 and mEJP: 1 . 87 mV ± 0 . 17 , n = 5 ) compared to animals kept at 18°C ( 0 hr ) ( Figure 7A-c–f ) . Moreover , larvae kept at 18°C ( 0 hr ) show no locomotion defects , whereas larvae kept at 28°C for 18–24 hr show severe locomotion defects ( data not shown ) . In summary , an 18 hr period of deGradFP expression at 28°C during third larval instar development is sufficient to remove most of the Brp-EGFP-Brp protein and cause severe loss of function phenotypes . To assess whether knockdown of Brp-EGFP-Brp affects adult behavior and to compare deGradFP and iGFPi efficiency in adult flies , animals expressing n-Syb-GAL4 were shifted from 18°C to 28°C at progressively later stages ( Figure 7B ) . The majority of animals raised continuously at 28°C and expressing deGradFP or iGFPi die as embryos or first instar larvae . Similarly , animals shifted to 28°C as first instar larvae ( L1 ) die in late second instar ( L2 ) . However , later shifts show major differences between animals expressing deGradFP vs iGFPi: iGFPi animals shifted to 28°C as L2 larvae or adults show no obvious phenotypes , whereas deGradFP animals shifted as L2 larvae or adults show locomotor defects and die as pupae ( L2 shifted ) or adults ( L2 or adult shifted ) . Moreover , temperature shifts to 28°C in adults allow flies to live for approximately 4 weeks after the shift . These data suggest that iGFPi is not as efficient as deGradFP at some stages of development or in some tissues , possibly because the Brp protein has a very long half-life and is less affected by depletion of mRNA levels by iGFPi than by degradation of the protein by deGradFP . The temperature sensitivity of the GAL4-UAS system in combination with deGradFP should allow the reversible removal of proteins . A reversible knockdown has numerous advantages: it is quick , simple and only requires temperature shifts; it can be performed during development and in adult flies of any age; it allows the unambiguous assignment of the phenotype to the protein knockdown; it obviates the need of control genotypes; and it determines if damage to cells upon loss of a protein is reversible or permanent . We therefore tested if we were able to reversibly knock down and regenerate the proper expression pattern and protein distribution for a few tagged genes . We selected two proteins with a large extacellular domain , Roughest ( rst ) and NetrinA ( NetA ) . The EGFP is inserted in the large extracellular domain of Rst and NetA . We were able to remove and re-establish protein expression and proper distribution of Roughest and NetrinA in third instar larval brains ( Figure 8 ) . Both Rst and NetA have distinct expression patterns in the larval brain which are significantly reduced after 24 hr at the restrictive temperature , 28°C . However , when larvae are returned to the permissive temperature , 18°C , for 24 hr , protein expression is restored to a level comparable to larvae that were never exposed to the restrictive temperature ( Figure 8 ) . Therefore , through temperature manipulation , GAL4 , and subsequently deGradFP expression can be controlled to reversibly disrupt target protein expression level . Given that GFP is in the extracellular domain , the degradation of these proteins must occur efficiently during protein synthesis . 10 . 7554/eLife . 05338 . 018Figure 8 . Knockdown and restoration of Rst and NetA protein expression in third instar larval brains . ( A ) ( a ) Roughest ( rst ) gene map ( chromosome X , based on FlyBase release FB2014_05 ) with the position of the MI04842 insertion shown by the red triangle . ( b ) Expression pattern of Rst-GFP-Rst in third instar larval brain when raised constitutively at 18°C . Expression is barely detectable after animals have been shifted to 28°C for 24 hr ( c ) . Expression is then restored by returning the animals to 18°C for 24 hr ( d ) . ( B ) ( a ) NetrinA ( NetA ) gene map ( chromosome X , based on FlyBase release FB2014_05 ) with the position of the MI04563 insertion shown by the red triangle . ( b ) Expression pattern of NetA-GFP-NetA in third instar larval brain when raised constitutively at 18°C . Expression is barely detectable after animals have been shifted to 28°C for 24 hr ( c ) . Expression is then restored by returning the animals to 18°C for 24 hr ( d ) . Scale bars , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 01810 . 7554/eLife . 05338 . 019Figure 8—figure supplement 1 . Variable knockdown efficiency of deGradFP . Western blots of adult head extracts probed with GFP antibody ( and Tubulin as a loading control ) . ( A ) y w Frq1-EGFP-Frq1;UAS-NSlmbvhhGFP4/+;n-Syb-GAL4/+ flies raised at 18°C until eclosion and shifted to 28°C as 1–3 day old adults for 24 hr , show a 96% reduction in GFP expression on Western blot compared to animals kept at 18°C . ( B ) y w CG14207-EGFP-CG14207;act-GAL4/+;UAS-NSlmbvhhGFP4/+ flies raised at 18°C until eclosion and shifted to 28°C as 1–3 day old adults for 3 days , show a 80% reduction in GFP expression on Western blot compared to animals kept at 18°C . ( C ) y w CG1632-EGFP-CG1632;act-GAL4/+;UAS-NSlmbvhhGFP4/+ flies raised at 18°C until eclosion and shifted to 28°C as 1–3 day old adults for 3 days , show a 45% reduction in GFP expression on western blot compared to animals kept at 18°C . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 019 Finally , to further assess the efficiency of protein knockdown through temperature regulation in combination with the deGradFP system , we preformed western blots against GFP for three other genes . Animals ( Frq1-EGFP-Frq1/Y;n-Syb-GAL4>deGradFP and CG14207-EGFP-CG14207/Y;act>GAL4 and CG1632-EGFP-CG1632/Y;act>deGradFP ) kept at 18°C through eclosion were shifted as 1–3 day old adult animals to 28°C for either 1 or 3 days . We observe a range of protein loss with western blots against GFP from ∼96% ( Frq1-EGFP with n-Syb-GAL4 ) to 45% ( CG1632-EGFP with act-GAL4 ) ( Figure 8—figure supplement 1 ) . Hence , knockdown efficiency is likely to vary significantly and should be assessed phenotypically and with western blots . To correlate the removal of a protein with a behavioral assay , we tagged the Dunce protein via RMCE using MI03415 ( Figure 9A ) . dunce ( dnc ) encodes a cAMP phosphodiesterase ( PDE ) that is required for learning and memory ( Davis and Kiger , 1981; Qiu et al . , 1991 ) . The EGFP cassette exon is positioned to tag all but one protein isoform ( Figure 9A ) . The Dnc-EGFP-Dnc protein is expressed in the adult brain and is mostly restricted to mushroom bodies ( MB ) , the center for learning and memory in Drosophila ( Figure 9B ) . A more detailed analysis reveals that the fusion protein is expressed in all MB lobes ( α , α′ , β , β′ and γ ) in the adult brain ( Figure 9B ) , and the GFP staining reveals a much crisper expression pattern than the anti Dunce antibody ( Figure 9—figure supplement 1A ) ( Nighorn et al . , 1991 ) . To knock down Dnc we used the 117y-GAL4 driver ( Armstrong et al . , 1998 ) which is expressed in MBs ( Figure 9—figure supplement 1 ) . The dnc-EGFP-dnc;117y-GAL4/+;UAS-deGradFP/+ flies were raised at 18°C . 1 day after eclosion , the flies were transferred to 28°C for 3 days , and then returned to 18°C for 2 days . We observe a progressive loss of Dnc-EGFP-Dnc , and after 48 hr at 28°C we could not or barely detect the tagged protein ( Figure 9C ) . Protein expression is restored 48 hr after a shift back to 18°C ( Figure 9C ) . To determine whether the decrease in Dnc-EGFP-Dnc expression levels correlates with a known behavioral phenotype associated with loss of dnc , we performed the aversive olfactory learning assay ( Tully and Quinn , 1985 ) . As expected , the dnc mutant flies , dnc1/dnc1 , dnc1/dncML and dnc1/dncM14 ( all partial loss of function combinations of dnc alleles ) exhibit a 50% reduction in performance index when compared to dnc1/+ , y w , Canton-S , and dnc-EGFP-dnc flies ( Figure 9D–E ) . In contrast , dnc-EGFP-dnc flies expressing deGradFP under the control of the MB specific driver and maintained at 28°C for 3 days exhibit a performance index that is decreased by 70% , similar to the most severe dunce alleles ( Davis and Kiger , 1981 ) . The same flies were then returned to 18°C , Dnc-EGFP-Dnc expression in MB was re-established within 24 hr , and the learning phenotype was recovered fully within 48 hr . Hence , deGradFP is able to reversibly alter protein expression levels . In addition to reestablishing Dnc-EGFP-Dnc expression , the learning score is reestablished after 2 days at 18°C ( Figure 9F ) . These data indicate that combining MiMIC-based EGFP tagging , UAS-deGradFP and GAL4 drivers allows us to achieve not only spatial and temporal control of protein expression , but to do so in a reversible manner , in order to reduce and then restore gene function in live flies . 10 . 7554/eLife . 05338 . 020Figure 9 . MiMIC mediated intronic tagging with EGFP permits a reversible spatial and temporal removal of proteins in flies . ( A ) dunce ( dnc ) gene map ( chromosome X 3070 . 4 kb-3237 . 8 kb , based on FlyBase release FB2014_05 ) with the position of the MI03415 insertion shown by the red triangle . ( B ) dnc-EGFP-dnc expression pattern in adult brain ( a ) . The α/β , α′/β′ and γ lobes of mushroom body ( MB ) are shown below ( b–g ) stained with anti GFP ( b and e ) and anti Dlg1 ( c and f ) antibodies . ( C ) dnc-EGFP-dnc can be spatially and temporally knocked-down and re-expressed with temperature shifts modulating expression of UAS-deGradFP under the control of the MB-GAL4 driver , 117y-GAL4 . Adult brains are stained with anti GFP . ( D ) dnc-EGFP-dnc flies show a normal learning score similar to Canton-S wild-type and y w flies . ( E ) Learning is impaired by temporal knockdown of dnc in MB caused by expression of UAS-deGradFP at 28°C for 3 days . ( F ) The learning deficit can be reversed with renewed dnc expression by shifting the animals back to 18°C for 2 days . The mean ± SEM is plotted for each treatment; n = 8 values for each group . ***p < 0 . 001 . Scale bars , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 02010 . 7554/eLife . 05338 . 021Figure 9—figure supplement 1 . Dnc expression pattern in mushroom bodies in the adult head . ( A ) Dnc-EGFP-Dnc expression in adult mushroom body: α/β and α′/β′ lobe ( a–c ) , γ lobe ( d–f ) . Adult heads are immuostained with ani GFP ( a and d ) and anti Dnc ( b and e ) antibodies . Scale bar , 50 μm . ( B ) Expression pattern of MB driver 117y-GAL4 . Scale bar , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05338 . 021
Here we describe a collection of 7434 MiMIC insertions that are inserted in or near ∼4500 different genes . Of these , 2800 insertions are in a coding intron and can be used for protein tagging of more than 1862 different genes . We introduced an artificial exon that encodes a SA-Linker-EGFP-FlAsH-StrepII-TEV-3xFlag-Linker-SD tag via RMCE ( Venken et al . , 2011a ) and created 450 protein trap alleles that tag 400 different genes . We present a detailed analysis of 200 tagged genes/proteins , and we performed a series of knockdown experiments on four tagged genes to assess the key features of this collection and demonstrate its power for functional analysis . We previously described MiMIC insertions in a coding intron of three genes that had recessive lethal phenotypes and showed that the lethality could be reverted by removing the gene trap cassette by RMCE ( Venken et al . , 2011a ) . Our data show that 92% of MiMICs inserted in a coding intron in the gene trap configuration are highly mutagenic and function as gene traps , while 100% of those that are in the opposite configuration are not mutagenic . These data demonstrate that the selected SA ( Splice Acceptor ) in the MiMIC element , also used in the tagging cassette , is highly efficient and minimizes exon skipping . This is also corroborated by the fact that we only observe the predicted tagged isoforms in three genes that were tested by Western blotting: α-Cat , dlg1 , and brp . This minimal level of exon skipping permits robust knockdown of the transcripts and proteins encoded by these genes using iGFPi and deGradFP ( see below ) . The analysis of the expression patterns and genetic tests of the 200 tagged genes reveals three highly valuable and surprising features of the library . First , EGFP expression can be detected in CNS of 90% of tagged genes . This is a surprisingly high number as it shows that most genes are expressed at sufficiently high levels to be detected with EGFP in their endogenous context . Second , intronic tagging does not obviously disrupt protein function in 77% of the 114 cases examined . This is also a surprisingly high portion of genes and is possibly due to the presence of unstructured linkers on either side of the tag . Third , expression of EGFP can be imaged in unfixed tissues in more than 90% of the samples tested . Note that this percentage has steadily risen during our project and the use of better confocal scopes has dramatically improved these numbers . The latter data are important as they document that most genes are expressed at levels that should permit direct imaging of GFP fluorescence in live tissue . These features make the collection very valuable and useful for many different applications . Comparison of two knockdown strategies based on iGFPi and deGradFP using four different genes reveals some differences . Both strategies appear to be efficient during development . RNAi expressed in the zygote is unlikely to remove the maternal proteins deposited in the egg , but they are removed or strongly reduced by deGradFP in the case of dlg1 . deGradFP is more efficient than the iGFPi system , except with knockdown of α-Cat in eyes where iGFPi exhibits a more robust phenotype . In contrast , for Brp-EGFP-Brp , RNAi knockdown causes similar phenotypes to deGradFP in embryogenesis and first instar larvae , but iGFPi leads to milder or no phenotypes at later stages , especially when the temperature shifts are performed in adults . These results suggest that the Brp protein is stable in adults and that the deGradFP system is able to target and disrupt Brp function , although it takes much longer to achieve than in embryos or larvae . In summary , both knockdown strategies should be explored when possible . It was previously shown that during embryogenesis deGradFP requires less than 3 hr to remove a protein encoded by a transgenic His2Av construct ( Caussinus et al . , 2011 ) . However , the temperature induced protein knockdowns that we observe take significantly longer . In the case of Brp-EGFP-Brp it takes at least 12 hr to observe a significant reduction in protein expression level in third instar larvae . Strong reductions in expression and a severe loss of synaptic transmission are observed 18 hr after a temperature shift in third instar larvae . For Dnc-EGFP-Dnc protein , we observe a severe knock down after 24 hr in adults , but it takes 48 hr at 28°C to observe a near complete loss of the protein in the mushroom bodies . In contrast to Brp-EFP-Brp and Dlg1-EGFP-Dlg1 where knockdown in adults can take up to 3 weeks or longer at 28°C to exhibit a severe phenotype . Yet , a ubiquitous adult knockdown with actin-GAL4 driving UAS-deGradFP for α-Cat , WASp and Bifid ( optomotorblind ) results in lethality within a few days ( data not shown ) . These observations suggest that differences in developmental stage , tissue , GAL4 strength and temperature will alter the dynamics of protein degradation with the deGradFP system leading to variation from protein to protein . As an additional note , we have observed that knockdown efficiency during developmental experiments is greatly enhanced when GAL4 is maternally deposited , by using GAL4 driver virgins . The ability to tailor conditions will permit analyses of different knock down levels at different temperatures with different drivers . Given that allelic series have been shown to be extremely valuable in genetic studies , important biological information can be derived by varying the knockdown conditions , providing tremendous flexibility for in vivo analyses . While Drosophila is recognized as being one of the premier model organisms for the genetic analysis of gene function , one limitation has been the difficulty of conditionally knocking down genes or proteins in adults , especially in the adult nervous system . To achieve this goal , typically a complex set of reagents needs to be created , even with the simplest designs . These include a null or severe loss of function mutation , a UAS-cDNA rescue construct , a ubiquitously expressed GAL4 driver to rescue the mutation , and a GAL80ts construct to conditionally inactivate the GAL4 ( McGuire et al . , 2003 ) . Using the library described here , we can now design large genetic screens in which we knock down proteins systematically in specific adult neurons or other tissues . The reversibility of the phenotype associated with the loss ( 28°C ) and restoration ( 18°C ) of the protein provide an important advantage as they permit us to directly pinpoint the cause of the phenotype , as well as the identity of the tissue that underlies the phenotype , as illustrated by the reversible MB-specific knockdown of the Dnc-EGFP-Dnc protein . While we have focused on the uses of MiMIC insertions in coding introns in this paper , MiMIC insertions in other parts of genes and in intergenic regions can be used for many other genomic manipulation applications ( Wesolowska and Rong , 2010; Venken et al . , 2011a; Venken and Bellen , 2012; Bateman et al . , 2013; Chen et al . , 2014; Vilain et al . , 2014 ) . Moreover , MiMICS in coding introns have recently been used to create gene specific GAL4 lines that permit one to reveal where genes are expressed when endogenous protein levels are low ( Diao et al . , 2014 , in press Cell Reports; Gnerer et al . , 2015 , in press Nucleic Acids Research ) . In summary , the resource that we have developed opens the door to numerous different applications and permits unprecedented manipulations of fly genes in vivo . Given the breath of applications associated with in vivo tagged genes and proteins , a genome wide collection of MiMIC insertions and tagged genes would be extremely useful . We are therefore using RMCE to insert a GFP tag into an additional 1200 genes in which there is a MiMIC insertion in a coding intron . Finally , the GDP has recently embarked on a large-scale project to tag 5000 selected genes by inserting a MiMIC-like cassette that can be targeted precisely using CRISPR/Cas9 technology .
New MiMIC insertions were generated by mobilizing the MI00827 insertion from the X chromosome to autosomal sites or by mobilizing the MI00000A or MI00000B insertions from a TM3 , Sb balancer chromosome ( Figure 1—figure supplement 1 ) . DNA segments flanking the new MiMIC insertions were amplified by inverse PCR , sequenced , and mapped by alignment to the reference genome sequence ( Venken et al . , 2011a ) . A detailed protocol is available on the GDP website ( http://flypush . imgen . bcm . tmc . edu/pscreen/ ) . Lines were selected for inclusion in the GDP collection by reviewing each insertion site with respect to annotated genes and the sites of other MiMIC insertions that were already part of the selected collection . Any insertion that was the first MiMIC hit in a gene was selected . When there were multiple insertions within a gene , priority was given to insertions within coding introns , coding exons , and introns in the 5′ UTR . Insertions within a coding intron or exon shared by the most annotated transcript isoforms were selected when possible . Multiple insertions were often selected for a gene if they hit different parts of the gene ( e . g . , one insertion in a coding intron and one in a coding exon ) or if they provided a way of differentially tagging multiple annotated protein isoforms . We selected one insertion line in each of several tandemly repeated gene families ( e . g . , the histone and 5S rRNA genes ) , even though we could not localize the insertion to a unique copy within the tandem array . One of the goals of the MiMIC screen was to create an array of MiMIC insertions throughout the genome , spaced no more than 40 kb apart ( Venken et al . , 2011a ) . Toward this end , we selected intergenic MiMIC insertions if they were inserted >20 kb from other MiMICs already in the collection . The attP sites of intergenic MiMICs can facilitate targeted mutagenesis of nearby genes ( Wesolowska and Rong , 2010 , 2013 ) and other genome engineering applications ( Venken and Bellen , 2012; Bateman et al . , 2013 ) . Lines that were selected for the GDP collection were balanced , and their insertion sites were verified by resequencing , before delivery to the BDSC . The flanking sequences of all MiMIC insertions sent to the BDSC are available from the project website and have been submitted to GenBank . Approximately 150 males of a specific MiMIC line were crossed to 400 virgins females expressing ΦC31 integrase using the vasa promoter inserted on the X chromosome or on the IV chromosome ( y1M{vas-int . B}ZH-2A w1118; snaSco/SM6a or y1M{vas-int . B}ZH-2A w*; Sb/TM6b , Hu , Tb or FM7j , B[1]; M{vas-int . B}ZH-102D ) . The resultant embryos were microinjected with pBS-KS-attB1-2-PT-SA-SD-[phase 0 , 1 or 2]-EGFP-FlAsH-StrepII-TEV-3xFlag . The F0 flies were crossed to balancer virgins or males of y1w67c23; In ( 2LR ) Gla , wgGla−1/SM6a or y* w*; D/TM6b , Hu , Tb or FM7j , B[1] . Transgenic F1 flies were scored for the loss of yellow+ ( yellow− phenotype ) and subsequently crossed to balancer virgins of y1w67c23; In ( 2LR ) Gla , wgGla−1/SM6a or y* w*; D/TM3 , Sb , Tb . Transgenic F2 flies were intercrossed to establish the final stock . Correct RMCE events were verified by PCR on genomic DNA obtained from 10–15 adult flies by using PureLink Genomic DNA Midi Kit ( Invitrogen , Life Technologies , Grand Island , NY ) . PCR was performed using two tag specific primers Tag-F and Tag-R , and two MiMIC specific primers Orientation-MiLF and Orientation-MiLR , in four different combinations . First PCR reaction was performed with Orientation-MiL-F and Tag-R , a second PCR reaction with primers Orientation-MiL-F and Tag-F , a third PCR reaction with primers Orientation-MiL-R and Tag-R , and a fourth PCR reaction was performed with primers Orientation-MiL-R and Tag-F . Third instar larvae were dissected for larval brains , imaginal discs , salivary gland , gut or NMJs in 1× PBS and fixed in 3 . 7% formaldehyde for 20 min ( NMJ ) or 30 min ( brain , discs , salivary gland , gut ) at room temperature and washed in 0 . 2% Triton X-100 . They were then incubated for 1 hr at RT in 10% NGS-PBS-0 . 2% Triton X-100 and stained with primary antibodies diluted in 10% NGS-PBS-0 . 2% Triton X-100 for 2 hr at RT or overnight at 4°C . The samples were washed and incubated with secondary antibodies and HRP ( where indicated ) for 2 hr at RT . The samples were then washed , stained with DAPI ( Invitrogen , Life Technologies , Grand Island , NY ) for 20 min ( where indicated ) and mounted in Vectashield ( Vector Labs , Burlingame , CA ) and imaged with a Zeiss LSM710 or a Leica SP8 confocal microscope and processed using Adobe Photoshop ( Adobe Systems Inc . , San Jose , CA , USA ) . Whole-mount immunolabeling of the adult brain was performed as previously described ( Lee et al . , 2011 ) . Briefly , brains were dissected in 1× PBS and fixed overnight in 4% paraformaldehyde in PBS on ice , transferred to 4% paraformaldehyde in PBS with 2% Triton X-100 at room temperature and vacuumed for 1 hr to remove the air sacs and left overnight in PBS with 2% Triton X-100 at 4°C . Following staining , brains were cleared and mounted in RapiClear ( SunJin Lab Co . , Taiwan ) and imaged with a Zeiss LSM710 confocal microscope under a 20× or 40× C-Apochromat water immersion objective lens and processed using Adobe Photoshop ( Adobe Systems Inc . , San Jose , CA , USA ) . Primary antibodies used: rabbit anti GFP 1:1000 ( LifeTechologies A11122 ) , guinea pig anti α-Cat 1:1000 ( Sarpal et al . , 2012 ) , mouse anti Dlg1 1:500 ( DSHB 4F3; [Parnas et al . , 2001] ) , rat anti E-Cadherin 1:100 ( DSHB DCAD2; [Oda et al . , 1994] ) , mouse anti Brp 1:30 ( DSHB nc82; [Wagh et al . , 2006] ) , rabbit anti Dlg1 1:500 ( Santa Cruz Biotechnology , Dallas , Texas ) , rabbit anti Dnc 1:10 ( gift from Ron Davis ) , FITC conjugated rat anti GFP 1:500 ( Santa Cruz Biotechnology , Dallas , Texas ) . Secondary antibodies used: Alexa 488 ( Invitrogen , Life Technologies , Grand Island , NY ) , and Cy3 or Cy5 conjugated secondary antibodies ( Jackson ImmunoResearch , West Grove , PA ) were used at 1:500 . Adult heads were homogenized in lysis buffer with an appropriate volume of 4 × Laemmli buffer and protease inhibitors . Samples were boiled at 95°C for 5 min and run on Mini-Protean TGX 4–20% gradient gels ( cat#456-1034; Bio-Rad , Hercules , California ) at 120 volts in a Bio-Rad Mini-PROTEIN NTM system followed by blotting onto a nitrocellulose membrane ( Bio-Rad Trans-Blot Turbo RTA transfer kit ) using the Trans-Blot Turbo Transfer system ( Bio-Rad , Hercules , California ) . The membrane was blocked with blocking buffer ( 5% milk in PBS-Tw-0 . 2% ) for 1 hr at room temperature and then probed with primary antibody diluted in 0 . 01% NaN3 containing blocking buffer overnight at 4°C . After washing , the membrane was incubated with secondary antibody diluted in blocking buffer for 2 hr at room temperature and further developed using SuperSignal West Dura Extended Duration Substrate ( #34075; Thermo Scientific , Waltham , MA ) and Odyssey CLx Infrared Imaging System ( LI-COR Biosciences , Lincoln , Nebraska ) . Signals were directly recorded and digitized by LAS 4000 machine ( FUJIFILM Corporation , Europe ) . Images were processed using Adobe Photoshop ( Adobe Systems Inc . , San Jose , CA , USA ) . For ERG recording , y*w*; brp-EGFP-brp; ey-GAL4 flies were crossed to y*w*; brp-EGFP-brp; UAS-NSlmbvhhGFP4 and y*w*; brp-EGFP-brp;UAS-GFP RNAi . For controls y*w*; ey-GAL4 flies were crossed to y*w*; UAS-NSlmbvhhGFP4 and y*w*; UAS-GFP RNAi . Crosses were consistently kept at 28°C . Desired progeny y*w*; brp-EGFP-brp; UAS-NSlmbvhhGFP4/ey-GAL4 , and y*w*; brp-EGFP-brp; UAS-GFP RNAi/ey-GAL4 , and respective control y*w*; UAS-NSlmbvhhGFP4/ey-GAL4 , and y*w*; UAS-GFP RNAi/ey-GAL4 were collected and ERGs were performed as previously described ( Ly et al . , 2008 ) . Briefly , adult flies were glued to a glass slide and a recording probe was placed on the surface of the eye , and a reference probe was inserted in the thorax . A flash of white light was given for 1 s , and the response was recorded and analyzed using AXON-pCLAMP 8 software . NMJ electrophysiology was performed as described previously ( Yao et al . , 2009 ) . Briefly , wandering third instar larvae were dissected in ice-cold , 0 . 25 mM calcium HL-3 ( 70 mM NaCl , 5 mM KCl , 20 mM MgCl2 , 10 mM NaHCO3 , 115 mM sucrose , 5 mM trehalose , and 5 mM HEPES; pH 7 . 2 ) , and rinsed with HL-3 containing 0 . 5 mM Ca2+ concentration . The fillet was incubated in the latter solution for at least 3 min before recording . Recordings were made from body-wall muscles 6 ( abdominal segment 3 ) with sharp electrodes filled with a 2:1 mixture of 2 M potassium acetate and 2 M potassium chloride . Data were collected only when resting membrane potential was below −65 mV . EJPs were evoked by directly stimulating the hemisegmental nerve through a glass capillary electrode at 0 . 2 Hz . Stimulus pulses were generated by pClamp 10 software ( Molecular Devices , Sunnyvale , CA ) , and the applied currents were 6 μA ± 3 with fixed stimulus duration at 0 . 3 ms . 30 evoked EJPs were recorded from each muscle for analysis . Miniature EJP ( mEJP ) events were collected for 2 min . Both EJPs and mEJPs were amplified with an Axoclamp 900A amplifier ( Molecular Devices ) and digitized by Digidata 1550 Digitizer ( Molecular Devices , Sunnyvale , CA ) . Experiments were performed at room temperature ( 20°C–22°C ) . EJPs were analyzed with pClamp 10 , and mEJPs were analyzed using the Mini Analysis Program ( Synaptosoft 29 Inc . , Decatur , GA ) . The EJPs amplitudes were corrected by nonlinear summation ( Feeney et al . , 1998 ) . The quantal content of evoked release was calculated from individual muscles by the ratio of the average EJP amplitude over the average mEJP amplitude . All flies are in y w background for conditional knock-down . For conditional protein knock-down , 1 day after eclosion flies are transferred from 18°C to 28°C for 3 days . For testing the reversibility , these flies are returned from the 28°C back to 18°C for 2 days . Aversive olfactory learning was performed by T-maze apparatus with a Pavlovian conditioning procedure as previously described ( Tully and Quinn , 1985 ) . Briefly , one training session consists of approximately 100 flies . During each session electrical shock is paired with the presence of one of the two odors ( 3-octanol and 4-methylcyclohexanol ) . Learning was measured 3 min after a single training session . Eight sessions were conducted for this odor-shock pairing ( ∼800 flies ) and then eight additional sessions were conducted with the other odor paired with the shock ( 1600 flies in total ) . The score is calculated as the number of flies avoiding the conditioned odor minus the number of flies avoiding the unconditioned odor divided by the total number of flies . The Performance Index ( PI ) is calculated as the average score of the two performances . Statistical analyses used KaleidaGraph 4 . 1 ( Synergy Software , Reading , PA ) . Performance Indexes were evaluated via one-way ANOVA followed by planned comparisons among the relevant groups with a Tukey Honestly Significant Difference test . All data are presented as mean ± SEM . *p < 0 . 05 . | In the last few decades , technical advances in altering the genes of organisms have led to many discoveries about how genes work . For example , it is now possible to add a specific DNA sequence to a gene so that the protein it makes will carry a ‘tag’ that enables us to track it in cells . One such tag is called green fluorescent protein ( GFP ) and it is often used to study other proteins in living cells because it produces green fluorescence that can be detected under a microscope . It is labor intensive to add tags to individual genes , so this limits the number of proteins that can be studied in this way . In 2011 , researchers developed a new method that can easily tag many genes in fruit flies . It makes use of small sections of DNA called transposons , which are able to move around the genome by ‘cutting’ themselves out of one location and ‘pasting’ themselves in somewhere else . The researchers used a transposon called Minos , which is naturally found in fruit flies . When Minos inserts into a gene , it often disrupts the gene and stops it from working . However , the researchers could swap the inserted transposon for a gene encoding GFP by making use of a natural process that rearranges DNA in cells . This resulted in the protein encoded by the gene containing GFP and so it can be detected under a microscope . This method allowed the researchers to create a collection of fly lines that have the GFP tag on many different proteins . Now , Nagarkar-Jaiswal et al . have greatly expanded this initial collection . More than 75% of GFP-tagged proteins worked normally and the flies producing these altered proteins remain healthy . It is possible to use a technique called RNA interference against the GFP to lower the production of the tagged proteins . Moreover , Nagarkar-Jaiswal et al . show that it is also possible to degrade the tagged proteins so that less protein is present . The removal of proteins is reversible and can be done in specific tissues during any phase in fly development . These techniques allow researchers to directly associate the loss of the protein with the consequences for the fly . This collection of fruit fly lines is a useful resource that can help us understand how genes work . The method for tagging the proteins could also be modified to work in other animals . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"tools",
"and",
"resources",
"neuroscience"
] | 2015 | A library of MiMICs allows tagging of genes and reversible, spatial and temporal knockdown of proteins in Drosophila |
The aminopeptidase DPP9 removes dipeptides from N-termini of substrates having a proline or alanine in second position . Although linked to several pathways including cell survival and metabolism , the molecular mechanisms underlying these outcomes are poorly understood . We identified a novel interaction of DPP9 with Filamin A , which recruits DPP9 to Syk , a central kinase in B-cell signalling . Syk signalling can be terminated by degradation , requiring the ubiquitin E3 ligase Cbl . We show that DPP9 cleaves Syk to produce a neo N-terminus with serine in position 1 . Pulse-chases combined with mutagenesis studies reveal that Ser1 strongly influences Syk stability . Furthermore , DPP9 silencing reduces Cbl interaction with Syk , suggesting that DPP9 processing is a prerequisite for Syk ubiquitination . Consistently , DPP9 inhibition stabilizes Syk , thereby modulating Syk signalling . Taken together , we demonstrate DPP9 as a negative regulator of Syk and conclude that DPP9 is a novel integral aminopeptidase of the N-end rule pathway .
Proteases of the DPPIV family are serine aminopeptidases that remove a dipeptide from the N-terminus of substrates having a Pro or an Ala residue in the second position ( NH2-XaaPro or NH2-XaaAla ) ( Yu et al . , 2010; Waumans et al . , 2015 ) . DPP8 and DPP9 are the only known intracellular members of this family , sharing 60% homology with a higher conservation ( >90% identity ) in their active site ( Zhang et al . , 2013 ) . Not surprisingly , DPP8 and DPP9 are similar in their biochemical properties ( Geiss-Friedlander et al . , 2009; Connolly et al . , 2008 ) . However , DPP9 , but not DPP8 , is rate-limiting for the hydrolysis of proline-containing peptides in the cytoplasm , and plays a role in maturation of antigenic peptides for presentation on MHC class I alleles ( Geiss-Friedlander et al . , 2009 ) . Knock-in mice expressing an inactive variant of DPP9 die 8–24 hr after birth , demonstrating its importance for neonatal survival which is not compensated by DPP8 ( Gall et al . , 2013 ) . DPP9 shows a broad tissue distribution ( Qi et al . , 2003; Ajami et al . , 2004 ) , localizes to the cytosol ( Ajami et al . , 2004 ) , nucleus ( Justa-Schuch et al . , 2014 ) , and to the leading edge of migrating cells ( Zhang et al . , 2015 ) . DPP9 is linked to several pathways including Akt signalling ( Yao et al . , 2011; Pilla et al . , 2013 ) , activation of pro-inflammatory M1 macrophages ( Matheeussen et al . , 2013 ) , cell migration ( Zhang et al . , 2015 ) and apoptosis of specific cell lines ( Matheeussen et al . , 2013; Spagnuolo et al . , 2013 ) . The molecular mechanisms leading to these different outcomes are poorly understood . For a profounder understanding of its molecular functions , we screened for interacting partners of DPP9 , reasoning that these will include substrates and regulators of this peptidase . Previously , we identified SUMO1 as a novel DPP9 associating protein , and demonstrated that SUMO1 acts as an allosteric activator of DPP9 ( Pilla et al . , 2013 , 2012 ) . Now we report filamin A ( FLNA ) as a novel interacting partner of DPP9 , which we identified in a yeast two-hybrid assay . FLNA is an actin binding protein that cross-links actin filaments into orthogonal networks , and is important for cell stiffening , cell adhesion and cell migration ( Nakamura et al . , 2011 ) . An actin-binding domain is located at the amino terminus of FLNA , followed by 24 Ig-like repeats . Repeat 24 is involved in FLNA dimerization , whereas the other repeats are important for the rod-like structure of FLNA ( Figure 1A ) ( Nakamura et al . , 2007 ) , and participate in numerous interactions with membrane , cytoplasmic and nuclear proteins ( Zhou et al . , 2010 ) . Consequently , FLNA influences several cellular signalling events . Most proteins identified so far bind preferentially to FLNA repeats 16–24 ( Yue et al . , 2013 ) . Surprisingly , we mapped the interaction between DPP9 and FLNA to repeats 5–7 . Apart from DPP9 , so far only the non-receptor tyrosine kinase Syk is known to bind to FLNA repeat 5 ( Falet et al . , 2010 ) . 10 . 7554/eLife . 16370 . 003Figure 1 . Filamin A ( FLNA ) is a novel DPP9 interacting protein . ( A ) Schematic representation of FLNA structure including numbering of the Ig-like domain repeats , and labelling of the actin-binding domain ( ABD ) . The asterisks mark the repeats lacking in the FLNA variant form used in ( B ) . ( B ) Pull-down assays showing direct interaction between recombinant DPP9 and recombinant FLAG tagged wt FLNA or a mutated form of FLAG-FLNA ( lacking repeats 4 , 9 , 12 , 17 , 19 , 21 , and 23 ) . Shown is a representative result of at least three independent experiments . ( C ) Recombinant DPP9 binds directly to GST- FLNA construct containing repeats 5–7 but not to GST-FLNA construct containing repeats 6–7 . Shown is a representative result of at least three independent experiments . ( D ) Co-immunoprecipitation of endogenous FLNA with endogenous DPP9 from HeLa cells treated with different cross-linkers . Binding was observed in the presence of the sulfhydryl cross-linker DPDPB . Shown is a representative result of at least three independent experiments . To control for the specificity of the cross link , we blotted for DPP8 , which did not bind to DPP9 in the presence of DPDPB ( E ) Quantification of the proximity ligation assay ( in situ PLA ) visualizing DPP9-FLNA interaction in HeLa cells treated with FLNA silencing oligos or non-targeting ( NT ) siRNAs for control shown in ( F ) . The number of PLA signals per cell were quantified in a blinded manner using the Duolink ImageTool software ( SIGMA ) . Data are represented as mean ± SEM . Signals of more than 130 cells were quantified for each condition respectively . Statistical analysis was carried out by an unpaired two-tailed t test ( ***p<0 . 0005 ) . ( F ) PLA showing interaction of DPP9 with FLNA in HeLa cells . Each red dot represents a single FLNA-DPP9 interaction . The number of PLA signals is significantly decreased in cells silenced for FLNA compared to cells treated with NT siRNA . Actin filaments are stained in green , nuclei were visualized by using HOECHST . Shown are representative images of at least three independent PLA experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 00310 . 7554/eLife . 16370 . 004Figure 1—figure supplement 1 . Co localization of DPP9 and FLNA . ( A ) Immunofluorescence microscopy images of HeLa cells decorated with goat anti-DPP9 antibody . To test for the specificity of the antibody , cells were treated with DPP9 targeting siRNA , control cells were treated with non-targeting ( NT ) siRNA oligonucleotides . Nuclei were stained with HOECHST . Indirect immunofluorescence images show that DPP9 signals are clearly reduced in DPP9-silenced cells . These experiments were repeated at least three times . ( B ) Immunofluorescence images showing an overlap in the localization of FLNA and DPP9 . HeLa cells were decorated with goat anti-DPP9 and a commercial anti-FLNA antibody . Nuclei were visualized by HOECHST staining . These experiments were repeated at least three times . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 004 Although it is also expressed in non-haematopoietic cells , Syk is best characterized as an essential component in B cell receptor ( BCR ) mediated signalling . Syk is activated upon BCR engagement , and consequently initiates a cascade of events by phosphorylating its downstream effectors ( Mócsai et al . , 2010; Geahlen , 2009 ) . The interaction between FLNA and Syk was identified in platelets , where loss of FLNA resulted in reduced signalling and mislocalization of Syk ( Falet et al . , 2010 ) . Here we show that DPP9 forms a complex with FLNA and Syk . In this complex FLNA acts as a recruiting factor linking DPP9 to Syk , resulting in cleavage of the Syk N-terminus by DPP9 , thus affecting Syk stability and Syk dependent signal transduction .
To identify novel proteins that interact with DPP9 , a human placenta library was screened in a yeast two-hybrid assay with full-length DPP9 as bait . FLNA was one of the most promising candidates , which was identified in 12 of 142 processed clones . The binding surface was mapped to residues 748–907 of FLNA , corresponding to FLNA repeats 5–7 . We verified direct interaction between FLNA and DPP9 by performing pull-down assays with the recombinant proteins ( Figure 1B ) . Only background binding of DPP9 was observed in the absence of FLNA . In line with the yeast two-hybrid assay showing the importance of repeats 5–7 , DPP9 bound to an FLNA deletion construct lacking repeats 4 , 9 , 12 , 17 , 19 , 21 , 23 ( Figure 1A and B ‘FLNA variant’ ) , which were previously mapped for several other FLNA interactions ( Nakamura et al . , 2011; Yue et al . , 2013 ) . DPP9 also associated with a truncated version of FLNA expressing only repeats 5–7 , but not with a shorter construct lacking repeat 5 ( Figure 1C ) . The interaction between DPP9 and FLNA was further analysed by performing co-immunoprecipitation assays against the endogenous proteins . To stabilize transient interactions , HeLa cells were incubated with various cross-linkers . Specific co-immunoprecipitation of FLNA with DPP9 was detected in cells treated with the sulfhydryl ( -SH ) cross-linker DPDPB , containing a spacer of 19 . 9 Å ( Figure 1D ) . Co-immunofluorescence microscopy images taken from HeLa cells decorated with antibodies targeting DPP9 and FLNA showed an overlap in the cellular localization of these two proteins ( Figure 1—figure supplement 1A and B ) . Next , in situ proximity ligation assays ( PLA ) were performed to visualize the association of endogenous DPP9 and FLNA in cells . Notably , we detected several distinct PLA dots in HeLa cells , each dot representing a single DPP9-FLNA interaction event . The number of PLA signals in control cells silenced for FLNA was strongly reduced compared to non-silenced cells ( Figure 1E and F ) . Taken together , these results describe a novel and direct interaction between DPP9 and FLNA , which requires FLNA repeat 5 and is readily detected in cells . Apart from DPP9 , only Syk is known so far to interact with FLNA repeat 5 ( Falet et al . , 2010 ) . By PLA we show that the association between FLNA and Syk , which was first identified in platelets , is conserved in HeLa cells ( Figure 2—figure supplement 1 ) . Strikingly , clear PLA signals were also detected in HeLa cells decorated with antibodies against endogenous DPP9 and Syk , suggesting an interaction between these proteins in cells ( Figure 2A and B ) . 10 . 7554/eLife . 16370 . 005Figure 2 . DPP9 interacts with the amino terminus of the tyrosine kinase Syk . ( A ) PLA in HeLa cells showing interaction of DPP9 with Syk . The number of PLA signals ( red ) representing DPP9-Syk interaction is significantly reduced in cells silenced for DPP9 compared to cells treated with NT siRNAs . Shown are representative images of at least three independent PLA experiments . Actin filaments are stained in green , nuclei are visualized with HOECHST staining . ( B ) Quantification of the PLA DPP9-Syk shown in ( A ) . Data are represented as mean ± SEM . Signals of more than 150 cells for each condition were quantified respectively in a blinded manner using the Duolink ImageTool software ( SIGMA ) . Statistical analysis was carried out by an unpaired two-tailed t test ( ***p<0 . 0005 ) . ( C ) Surface Plasmon Resonance ( SPR ) assays showing direct interaction between DPP9 wild type and a synthetic peptide covering the first 31 amino acids of Syk ( 1–31 ) . The binding affinity of the Syk ( 1–31 ) peptide is lower towards the inactive DPP9 variant ( DPP9 S730G ) . Depicted are equilibrium binding isotherms obtained from at least three repetitions for respective interaction pairs of recombinant DPP9 and DPP9 S730G with Syk ( 1–31 ) peptides . DPP9 was immobilized at the chip surface and the Syk ( 1–31 ) peptide was injected over the surface with concentrations varying from 16 µM to 0 . 125 µM . Binding affinities were calculated using Graph Pad Prism 6 . 0 . The Error is displayed as SEM . ( D ) Surface Plasmon Resonance ( SPR ) assays showing that the interaction of DPP9 with the peptide corresponding to Syk N-terminus requires the first two residues in Syk N-terminus . Depicted are equilibrium binding isotherms obtained from at least three repetitions for respective interaction pairs of recombinant DPP9 with Syk ( 1–31 ) or Syk ( 3–31 ) peptides . Recombinant His-tagged DPP9 was immobilized on the chip surface and the Syk ( 1–31 ) or Syk ( 3–31 ) peptide was injected over the surface with concentrations varying from 16 µM to 0 . 125 µM for Syk ( 1–31 ) and 32 µM to 1 µM for Syk ( 3–31 ) . Binding affinities were calculated as described in ( C ) . ( E ) Table summarizing the KD values . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 00510 . 7554/eLife . 16370 . 006Figure 2—figure supplement 1 . Interaction of FLNA with Syk is conserved in HeLa cells . ( A ) Syk is expressed in HeLa cells as shown by indirect immunofluorescence using a Syk-specific antibody . To test for specificity of the antibody cells were treated with Syk targeting siRNA , control cells were treated with non-targeting ( NT ) siRNA oligonucleotides . Syk signals were clearly decreased in Syk-silenced cells . HOECHST staining was applied to stain nuclei . ( B ) Interaction of Syk with FLNA in HeLa cells shown by PLA using FLNA- and Syk-specific antibodies . The number of PLA signals was markedly decreased in cells treated with siRNA against FLNA compared to control cells treated with non-targeting ( NT ) siRNA . Actin filaments are stained in green , nuclei were visualized by using HOECHST . Shown are representative images of at least three independent PLA experiments . ( C ) Quantification of the PLA FLNA-Syk in HeLa cells shown in ( B ) . Data are represented as mean ± SEM . Signals of more than 150 cells for each condition were quantified in a blinded manner using the Duolink ImageTool ( SIGMA ) . Statistical analysis was carried out by an unpaired two-tailed t test ( ***p<0 . 0005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 00610 . 7554/eLife . 16370 . 007Figure 2—figure supplement 2 . DPP9 cleaves after a Xaa-Pro/Ala . ( A ) Hydrolysis of MP-AMC by 12 , 5 nM recombinant DPP9 . An experiment was performed at least three times , each time in triplicates . Shown is a representative Michaelis-Menten analysis , data are represented as mean ± SEM . ( B ) Hydrolysis of MA-AMC by 12 , 5 nM recombinant DPP9 . An experiment was performed at least three times , each time in triplicates . Shown is a representative Michaelis-Menten analysis , data are represented as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 007 Interestingly , the amino terminus of Syk contains a DPP9 consensus cleavage site Xaa-Pro/Ala ( Connolly et al . , 2008; Ajami et al . , 2004 ) ( Figure 2—figure supplement 2A and B ) with an alanine in second position ( Met-Ala ) , suggesting that Syk may be a DPP9 substrate . Therefore , we tested whether DPP9 interacts with the amino terminus of Syk by conducting Surface Plasmon Resonance ( SPR ) assays . Recombinant DPP9 was immobilized on a chip and probed for interaction with a synthetic Syk ( 1–31 ) peptide covering the first 31 amino acids of full-length Syk ( MA↓SSGMADSANHLPFFFGNITREEAEDYLVQ ) . Based on published solved structures , the amino terminus of Syk includes an unstructured region followed by an α-helix ( residues 22–31 ) ( Grädler et al . , 2013; Fütterer et al . , 1998 ) . Of note , amino acids 1–7 are not resolved in any determined structure of Syk , suggesting that this region is flexible and thus accessible for interactions . We observed direct binding of DPP9 to the Syk ( 1–31 ) peptide with a KD of 52 ± 4 µM ( Figure 2C and E ) , suggesting a dynamic and transient interaction . Furthermore , an inactive DPP9 variant in which the serine in the active site was mutated to a glycine residue ( DPP9 S730G ) showed more than a five-fold reduction in its affinity to the Syk ( 1–31 ) peptide ( KD of 284 ± 27 µM ) compared to the wild type protein ( Figure 2C and E ) . A lower affinity ( KD of 156 ± 60 µM ) was also observed between wild type DPP9 and a Syk△MA ( 3–31 ) peptide , which lacks the first two amino acids Met-Ala ( Figure 2D and E ) . In summary , these results show a direct interaction between DPP9 and Syk N-terminus , which involves the active site of DPP9 , suggesting that Syk is a DPP9 substrate . Next in vitro cleavage assays were performed to test for processing of the Syk N-terminus peptide ( 1–31 ) by recombinant DPP9 . Mass spectrometry analysis of these reactions revealed that after six hours incubation , DPP9 had cleaved more than 99% of this peptide to remove the first two residues ( Figure 3A ) . Processing was strongly reduced to only 6 . 2% in control samples treated with the allosteric DPP8/9 peptide inhibitor SLRFLYEG ( Pilla et al . , 2013 ) . No cleavage was observed in the presence of the inactive variant DPP9 S730G ( Figure 3A ) . 10 . 7554/eLife . 16370 . 008Figure 3 . Syk is a novel substrate of DPP9 . ( A ) In vitro cleavage of a synthetic Syk peptide corresponding to the N-terminus of Syk ( 1–31 ) by recombinant DPP9 . 50 µM of a synthetic Syk ( 1–31 ) peptide was incubated for 6 hr , either alone or with 130 nM DPP9 . For control 10 µM allosteric DPP9 inhibitor SLRFLYEG was added in addition to 130 nM DPP9 and ( 6 hr ) . An additional control included the peptide and the inactive DPP9 S730G variant . Samples were analysed by high resolution liquid chromatography/tandem mass spectrometry in triplicate . Quantitation was achieved by extracting ion chromatograms and integrating peak areas for the most abundant 3+ charge state of the intact 1–31 ( [M+3H]3+m/z 1149 . 8589 ) and the cleaved 3–31 ( [M+3H]3+m/z 1082 . 4997 ) peptides . The identities and retention times of the peptides were established by accurate mass measurement and product ion spectra ( data not shown ) . ( B–G ) PLA assays showing that the interaction between DPP9 and Syk requires the active site of DPP9 . Shown are representative images with the corresponding quantifications of at least three independent PLA experiments . Actin filaments are stained in green , and nuclei were visualized by using HOECHST . The number of PLA signals ( red dots ) per cell were quantified in a blinded manner using the Duolink ImageTool software ( SIGMA ) . Signals of more than 300 cells were quantified for each condition respectively . Statistical analysis was carried out by an unpaired two-tailed t test ( **p<0 . 005; ***p<0 . 0005; n . s = not significant ) . ( B ) The interaction between DPP9 and Syk is markedly decreased in HeLa cells treated with 10 µM SLRFLYEG compared to control cells treated with DMSO . ( C ) Quantification of the PLA DPP9-Syk shown in ( B ) . Data are represented as mean ± SEM . ( D ) The number of PLA signals representing DPP9-Syk interactions per cell is reduced upon treatment of HeLa cells with the competitive DPP8/9 inhibitor 1G244 ( 10 µM , for 5 min ) compared to control cells treated with DMSO . ( E ) Quantification of the PLA DPP9-Syk shown in ( D ) . Data are represented as mean ± SEM . ( F ) The interaction of DPP9 with FLNA is not significantly altered upon treatment of HeLa cells with 1G244 ( 10 µM , 30 min ) compared to control cells treated with DMSO . ( G ) Quantification of the PLA DPP9- FLNA shown in ( F ) . Data are represented as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 00810 . 7554/eLife . 16370 . 009Figure 3—figure supplement 1 . Inhibition of DPP activity in HeLa cells with 1G244 . HeLa cells were treated with 10 µM DPP8/9 inhibitor 1G244 or DMSO for control ( 0 , 5 and 30 min ) . Cells were lysed and extracts ( 5 µg ) of were analysed for DPP activity in the presence of the artificial DPP substrate GP-AMC ( 250 µM ) or the unrelated substrate R-AMC ( 50 µM ) . Fluorescence was measured over time . Experiment was performed at least three times , each time in triplicates . Shown is a representative , data are represented as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 009 To further test whether DPP9 activity affects its interaction with Syk , HeLa cells were treated with SLRFLYEG . Previously we demonstrated that this inhibitor can be delivered into cells if it is pre-incubated with cell penetrating peptides ( Pep1 ) to form a non-covalent Pep-1-SLRFLYEG complex . Once in cells this complex dissociates leading to inhibition of DPP9 by SLRFLYEG ( Pilla et al . , 2013 ) . Consistently , exposure of cells to SLRFLYEG resulted in a significant reduction in PLA signals corresponding to DPP9-Syk interaction events , compared to the control cells treated with the carrier peptide only ( Figure 3B and C ) . Likewise , treatment of cells with the competitive DPP9 inhibitor 1G244 ( Wu et al . , 2009 ) also led to a clear decrease in the number of Syk-DPP9 PLA signals ( Figures 3D and E , Figure 3—figure supplement 1 ) . Of note 1G244 and all other available DPP9 inhibitors also target DPP8 due to the high conservation in the active site of both enzymes ( Van Goethem et al . , 2011 ) . For control , we measured the association of DPP9 with FLNA , which was not significantly altered by the 1G244 treatment ( Figure 3F and G ) . These results demonstrate that Syk , but not FLNA , requires access to the active site of DPP9 for interaction . Taken together , we conclude that Syk is a novel DPP9 substrate . What is the role of FLNA for the DPP9-Syk interaction ? Strikingly , immunofluorescence microscopy images show a drastic change in the cellular localization of DPP9 in FLNA silenced cells compared to control cells treated with non-targeting siRNA ( Figure 4A and B ) . In particular , upon FLNA silencing , DPP9 was no longer observed at the plasma membrane and was detected less in the cytosol , showing elevated levels in the nucleus ( Figure 4A and B ) . These results demonstrate that FLNA has a strong impact on the cellular localization of DPP9 . We further tested the importance of FLNA for the DPP9-Syk interaction by performing PLAs of DPP9-Syk in FLNA silenced cells , and compared to cells treated with non-targeting siRNA . We detected a significant reduction in the number of DPP9-Syk interaction events in the FLNA silenced cells as indicated by a lower number of PLA dots ( Figure 4C and D ) . Taken together , we conclude that the presence of FLNA is important for the ability of DPP9 and Syk to interact in cells , which eventually leads to Syk processing . 10 . 7554/eLife . 16370 . 010Figure 4 . FLNA - a scaffold linking DPP9 to Syk . ( A ) Immunofluorescence images of HeLa cells showing that the cellular localization of DPP9 is altered in FLNA silenced cells . DPP9 is shown in red , FLNA in green and nuclei are visualized with HOECHST staining ( blue ) . Shown are representative images of at least three independent experiments . ( B ) Quantification of the immunofluorescence shown in ( A ) . n = c: DPP9 is equally distributed in the cytosol and nucleus; n > c stronger DPP9 staining in the nucleus compared to the cytosol . Data are represented as mean ± SEM . Signals of more than 100 cells for each condition were quantified respectively . Statistical analysis was carried out by an unpaired two-tailed t test ( **p<0 . 005; ***p<0 . 0005 ) . ( C ) The interaction of DPP9 with Syk depends on the presence of FLNA in HeLa cells . The number of PLA signals ( red dots ) per cell is markedly decreased in cells treated with FLNA silencing oligonucleotides compared to control cells treated with NT siRNA oligonucleotides . Shown are representative images of at least three independent PLA experiments . ( D ) Quantification of the PLA DPP9-Syk shown in ( C ) . Data are represented as mean ± SEM . The number of PLA signals ( red dots ) per cell were quantified in a blinded manner using the Duolink ImageTool software ( SIGMA ) . Signals of more than 115 cells were quantified for each condition respectively . Actin filaments are stained in green , nuclei were visualized with HOECHST staining . Statistical analysis was carried out by an unpaired two-tailed t test ( ***p<0 . 0005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 010 To study the outcome of Syk cleavage by DPP9 we turned to a stable HeLa cell line for DPP9 silencing ( DPP9-kd ) . This cell line shows a 40% reduction in its capacity to cleave the artificial substrate Gly-Pro-AMC ( GP-AMC ) , which can be cleaved by DPP8 , DPP9 and DPPIV . Additionally , we detected at least a 50% reduction in DPP9 protein levels in these cells compared to the control parental cell line ( Figure 5A and B ) . Surprisingly , although we transfected wild type and DPP9-kd cells with equal amounts of plasmid DNA encoding C-terminal FLAG tagged Syk ( Syk-FLAG ) , higher Syk protein levels were observed in the DPP9-kd cell line ( Figure 5B ) . These unequal protein levels of the transiently transfected Syk constructs suggested that Syk may be more stable in the DPP9-kd cell line than in the wild type parental cell line . To test this hypothesis , we performed cycloheximide ( CHX ) chase experiments of DPP9-kd and the parental cells transfected with the Syk-FLAG construct . Cells were treated with CHX to inhibit protein synthesis ( ‘pulse’ , time 0 ) , and samples were taken at different time points for the chase . We found that Syk-FLAG is degraded in HeLa cells with a half-life of 6 hr . Remarkably , Syk-FLAG was more stable in the DPP9-kd cells compared to the control parental HeLa cells ( Figures 5C and D ) . These results suggest that processing of Syk by DPP9 and exposure of a neo Syk N-terminus with serine in position 1 ( MA↓S ) leads to Syk degradation , possibly by the N-end rule pathway . 10 . 7554/eLife . 16370 . 011Figure 5 . DPP9 determines Syk stability by exposing an N-terminal serine . ( A ) Reduced DPP activity in cytosolic extracts of HeLa cells with a stable silencing of DPP9 ( DPP9-kd ) . Cell extracts either from HeLa wt or DPP9-kd cells ( 5 µg ) were tested for DPP activity using the artificial DPP substrate GP-AMC ( 250 µM ) or the unrelated substrate R-AMC ( 50 µM ) . Fluorescence was measured over time . The experiments were performed at least three times , each time in triplicates . Shown is a representative , data are represented as mean ± SEM . ( B ) Higher steady-state levels of Syk in DPP9-kd compared to HeLa wt cells . Western blot analysis of cell lysates from HeLa wt or DPP9-kd cells transfected with 1 µg C-terminally FLAG-tagged full-length Syk . Shown is a representative result of at least three independent experiments . Tubulin was analysed as a loading control . ( C ) Syk stability is determined by DPP9 . HeLa wt or DPP9-kd cells were transfected with C-terminally FLAG-tagged full-length Syk and subjected to Cycloheximide ( CHX ) chase assays . GFP was analysed as a transfection and loading control . Shown is one representative result of at least three independent experiments . ( D ) Quantification of the Western blot results shown in ( C ) . The ratio of Syk/GFP at time 0 hr was normalized to 100% . For signal quantification GelQuant . NET software provided by biochemlabsolutions . com was used . ( E ) Syk stability is determined by the serine that is exposed after cleavage by DPP9 . HeLa wt cells were transfected with different Syk constructs: Syk-FLAG , SykS3G-FLAG or SykS3V-FLAG and subjected to CHX chase assays . Shown is one representative result of at least three independent experiments . ( F ) Quantification of the Western blot results shown in ( E ) , as described in ( D ) . ( G ) Syk stability is increased by mutating the A at position 2 in the DPP9 cleavage site to a D ( SykA2D ) . HeLa wt cells transfected with Syk-FLAG or SykA2D-FLAG were subjected to CHX chase assays . Shown is one representative result of at least three independent experiments . ( H ) Quantification of the Western blot results shown in ( G ) , as described in ( D ) . ( I ) Reduced interaction events between Syk and Cbl in the absence of DPP9 . Interaction of Syk with Cbl was visualized by PLA in HeLa wt and DPP9-kd cells . The number of PLA signals representing Syk-Cbl interactions per cell was reduced in DPP9-kd cells . For control , cells were treated with only one primary antibody ( anti Syk ) . ( J ) Quantification of the PLA Syk-Cbl in HeLa cells shown in ( I ) . Data are represented as mean ± SEM . The number of PLA signals per cell were quantified in a blinded manner using the Duolink ImageTool software ( SIGMA ) . Signals of more than 100 cells were quantified for each condition respectively using the Duolink ImageTool ( SIGMA ) . Statistical analysis was carried out by an unpaired two-tailed t test ( *p<0 . 05; n . s = not significant ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 011 The N-end rule pathway degrades proteins based on their first N-terminal residue ( N-degron ) ( Varshavsky , 2011 ) . N-end rule substrates are ubiquitinated by specific E3 ubiquitin ligases ( N-Recognins ) that recognize the N-degrons in proteins leading to their proteasome degradation ( Tasaki et al . , 2005 , 2009; Hwang et al . , 2010a ) . To test whether the serine that is exposed upon DPP9 cleavage indeed determines Syk stability , we mutated this residue to a glycine ( SykS3G ) or a valine ( SykS3V ) . Following processing by DPP9 these Syk constructs would differ only in the first amino acid in their N–terminus , which would be either Ser in wild type Syk , Gly or Val in the variant proteins . We chose these residues since substrates with Ser , Gly or Val in the P1’ position ( after the cleavage site ) are processed by DPP9 with a similar efficiency ( Geiss-Friedlander et al . , 2009 ) . Additionally , we mutated the alanine in position 2 to an aspartic acid ( SykA2D ) . This mutation destroys the DPP9 cleavage site , and thus should prevent Syk processing by DPP9 ( Connolly et al . , 2008 ) . The different Syk constructs were transfected into HeLa cells and their stability was compared by CHX chase experiments . Strikingly , the single mutation of Ser 3 to a Gly or a Val clearly increased the half-life of the SykS3G and SykS3V variants ( Figure 5E and F ) , compared to the wild-type protein . Furthermore , an increased half-life of Syk was also measured for the SykA2D variant , which cannot be processed by DPP9 ( Figure 5G and H ) . These results highlight the importance of Syk cleavage by DPP9 and the destabilizing effect of the serine that is exposed in the N-terminus of Syk upon cleavage . In B cells Syk signalling is initiated by the binding of an antigen to the BCR . This engagement of the BCR leads to receptor aggregation and phosphorylation of tyrosine residues in transmembrane ITAM containing proteins ( immunoreceptor tyrosine based activation motifs ) . Syk is recruited to phosphorylated ITAMS via two SH2 domains in its amino terminus , resulting in Syk phosphorylation and activation ( Mócsai et al . , 2010 ) . Syk is negatively regulated by the E3 ubiquitin ligase Cbl ( Ota et al . , 2000; Paolini et al . , 2001; Joazeiro et al . , 1999; Rao et al . , 2001a; Sohn et al . , 2003 ) . Proteasome inhibition in natural killer cells results in the accumulation of ubiquitinated Syk species , suggesting that ubiquitination targets the kinase for proteasomal degradation ( Paolini et al . , 2001 ) . Performing PLAs we established an interaction between endogenous Cbl and Syk in HeLa cells ( Figure 5I ) . Strikingly , the number of PLA signals corresponding to a Syk-Cbl interaction were significantly lower in the DPP9-kd cells compared to the control parental HeLa cells ( Figure 5I and J ) , suggesting a functional cross-talk between DPP9 and Cbl . We further tested the outcomes of Syk destabilization by DPP9 in B cells . Western blot analysis , activity assays and indirect immunofluorescence verify the expression of DPP9 in these cells ( Figure 6—figure supplement 1 ) . Next , PLA analysis shows that the binding of FLNA to Syk and DPP9 is conserved in DG-75 cells ( Figure 6A and B ) . Importantly , clear PLA signals corresponding to an interaction between DPP9 and Syk were also observed in these cells ( Figure 6C and D and Figure 6—figure supplement 2 ) . Of note , DPP8 , which shares 60% homology to DPP9 , did not appear to interact with Syk , although it is expressed in these cells . Similarly , no interaction above back-ground level was observed between Syk and the membrane protease DPPIV , which shares 26% homology with DPP9 ( Figure 6C and D ) . 10 . 7554/eLife . 16370 . 012Figure 6 . DPP9 regulates the stability of Syk in human Burkitt’s lymphoma B cells . ( A and B ) PLA showing that the interactions of FLNA with Syk and DPP9 are conserved in human DG-75 B cells . Each PLA interaction is shown here as a white dot , nuclei were visualized by using HOECHST . Control reactions ( NCtrl ) were performed with only one primary antibody ( αSyk , αFLNA or αDPP9 ) . Shown are representative images and quantifications of at least three independent PLA experiments . The number of PLA signals per cell were quantified in a blinded manner using the Duolink ImageTool software ( SIGMA ) . Signals of more than 80 cells were quantified for each condition respectively . Data are represented as mean ± SEM . Statistical analysis was carried out by an unpaired two-tailed t test ( ***p<0 . 0001 ) . ( C and D ) PLA in DG-75 cells showing that Syk interacts specifically with DPP9 but not with its homologs DPP8 and DPPIV . Control reactions ( NCtrl ) cells were treated with one primary antibody only: αDPP9 , αDPP8 or αDPPIV . Shown are quantifications of the PLA DPP9-Syk , DPP8-Syk and DPPIV-Syk in DG-75 cells from three independent experiments . Data are represented as mean ± SEM . The number of PLA signals per cell were quantified in a blinded manner using the Duolink ImageTool software ( SIGMA ) . Signals of more than 100 cells were quantified for each condition respectively . Statistical analysis was carried out by an unpaired two-tailed t test ( ***p<0 . 0001; n . s = not significant ) . ( E ) CHX chase experiment showing reduced stability of endogenous Syk upon stimulation of the BCR . Human DG-75 cells were stimulated with 12 µg/ml F ( ab’ ) 2 fragment goat-anti-human IgG+IgM ( + stim ) , or left untreated ( - stim ) , and simultaneously subjected to CHX chase . DPP9 was analysed as a loading control . Shown is one representative result of at least three independent experiments . ( F ) CHX chase experiments showing that the stability of endogenous Syk in stimulated DG-75 cells , is determined by the proteasome and by DPP9 . DG-75 cells were treated either with the DPP8/9 inhibitor 1G244 ( 10 µM ) , with the proteasome inhibitor MG132 ( 100 µM ) or with DMSO for control ( MOCK ) . Cell lysates were analysed for protein levels of Syk and of DPP9 for loading control by Western blotting . Shown is one representative result of at least three independent experiments . ( G ) Quantification of the Western blot results shown in ( F ) . The ratio of Syk/DPP9 at time 0 hr was normalized to 100% . For signal quantification GelQuant . NET software provided by biochemlabsolutions . com was used . ( H ) CHX chase experiment assaying the stability of endogenous phosphorylated Syk ( p-Y323 ) in stimulated DG-75 cells upon treatment with the DPP8/9 inhibitor 1G244 ( 10 µM ) or with DMSO for control ( MOCK ) . Tubulin was assayed as loading control . Shown is one representative result of at least three independent experiments . ( I ) Quantification of the Western blot results shown in ( H ) . The ratio of Syk p-Y323/tubulin at time 10 min was normalized to 100% . For signal quantification GelQuant . NET software provided by biochemlabsolutions . com was used . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 01210 . 7554/eLife . 16370 . 013Figure 6—figure supplement 1 . DPP9 expression , activity and interaction with Syk in the human DG-75 cells . ( A ) Total cell lysates ( 10 µg per lane ) of DG-75 cells stimulated with 12 µg/ml F ( ab’ ) 2 fragment goat-anti-human IgG+IgM ( 0 , 1 and 4 min ) were analyzed for DPP9 protein levels by Western blotting . Tubulin was used as loading control . Shown is a representative blot , an experiment was performed more than five times . ( B ) DG-75 cells were treated with 10 µM DPP8/9 inhibitor 1G244 or DMSO for control ( 0 , 5 and 30 min ) . Cell lysates ( 10 µg ) of were analysed for DPP activity in the presence of the artificial DPP substrate GP-AMC ( 250 µM ) or the unrelated substrate R-AMC ( 50 µM ) . Fluorescence was measured over time . An experiment was performed at least three times , each time in triplicates . Shown is a representative , data are represented as mean ± SEM . ( C ) Indirect immunofluorescence images of DG-75 cells decorated with antibodies against DPP9 , DPP8 and DPPIV . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 01310 . 7554/eLife . 16370 . 014Figure 6—figure supplement 2 . Controls for the DPP9-Syk PLA in DG-75 cells . PLA showing the interaction between Syk and DPP9 in DG-75 cells . Nuclei were visualized by using HOECHST . Shown are representative images with the corresponding quantifications of at least three independent PLA experiments . Signals were quantified in a blinded manner using the Duolink ImageTool ( SIGMA ) . ( A ) Interaction of DPP9 and Syk visualized by PLA in DG-75 cells . The number of PLA signals representing DPP9-Syk interactions per cell was reduced in control samples , in which Syk antibodies were pre-treated with a blocking peptide recognized by the Syk antibody . As an additional control , cells were treated with only one primary antibody ( here Syk ) . ( B ) Quantification of the PLA DPP9-Syk shown in ( A ) . Data are represented as mean ± SEM . Signals of more than 90 cells were quantified for each condition respectively . Statistical analysis was carried out by an unpaired two-tailed t test ( **p<0 . 005; n . s = not significant ) . ( C ) Interaction of DPP9 with Syk visualized by PLA in DG-75 cells as in ( A ) . The number of PLA signals per cell was markedly decreased in a control reaction using an unrelated anti-HA antibody instead of the anti-Syk antibody as a second primary antibody for the PLA . As a second control , cells were treated only with Syk antibodies . ( D ) Quantification of the PLA DPP9-Syk in DG-75 cells shown in ( C ) . Data are represented as mean ± SEM . Signals of more than 77 cells were quantified for each condition respectively . Statistical analysis was carried out by an unpaired two-tailed t test ( *p<0 . 05; **p<0 . 005; n . s = not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 014 Next , CHX chase experiments were performed to analyse the half-life of endogenous Syk in stimulated ( BCR-engaged ) and unstimulated DG-75 cells . BCR signalling was elicited by treating cells with anti-BCR F ( ab ) 2 when adding the CHX . DPP9 and Tubulin remained unaltered over the respective chase period and were assessed as a loading control ( Figure 6E ) . Whereas Syk was stable in unstimulated B cells over a time period of 6 hr , upon stimulation of the BCR , Syk was degraded with a half-life of 2 hr ( + stimulation , Figure 6E ) . Syk half-life in stimulated DG-75 cells is shorter than the previously published half-life of Syk in stimulated natural killer cells where it is estimated to be 8 hr ( Paolini et al . , 2001 ) . As expected , a clear stabilization of Syk was observed in stimulated DG-75 cells which were treated with the proteasome inhibitor MG132 ( Figure 6F and G ) . Notably , consistent with our findings showing stabilization of Syk in DPP9-kd HeLa cells , treatment of stimulated DG-75 cells with the DPP8/9 inhibitor 1G244 resulted in a clear stabilization of Syk ( Figure 6F and G ) . Next , we applied specific antibodies recognizing Syk phosphorylated on Y323 , which was previously shown to be important for binding to Cbl ( Yankee et al . , 1999; Lupher et al . , 1998 ) . CHX chase experiments on BCR stimulated cells show clear signals corresponding to the phosphorylation of Syk on Y323 upon stimulation of DG-75 cells ( Figure 6H and I ) , which clearly dropped in intensity over time . Notably , the intensities of signals corresponding to Syk pY323 did not drop as rapidly in 1G244 treated cells . Instead , we observed an increase in signal intensity up to 20 min post BCR stimulation , suggesting that during this time more Syk is phosphorylated on Y323 , but degradation is prevented due to inhibition of DPP9 . Taken together , these results show a central role for DPP9 in determining the stability of Syk after BCR-engagement , and suggest that DPP9 functions upstream of Cbl . In addition to Y323 , engagement of the BCR initiates the phosphorylation of Syk on multiple sites within a short time ( Furlong et al . , 1997; Bohnenberger et al . , 2011; Satpathy et al . , 2015 ) . Phosphorylation serves both to modulate Syk catalytic activity and to modify its interactions with other proteins ( Mócsai et al . , 2010; Geahlen , 2009; Krisenko and Geahlen , 2015 ) . Remarkably , inhibition of DPP9 in stimulated B-cells resulted also in a prolonged phosphorylation of Syk on Y352 ( Figure 7A and B ) , which is important both for the catalytic activity of Syk and for the activation of the proximal target Phospholipase C-γ2 ( PLCγ2 ) ( Tsang et al . , 2008; Law et al . , 1996 ) . Similar to Syk pY323 ( Figure 6H and I ) , signals for Syk-pY352 continued to increase up to 20 min in 1G244 treated cells , but not in the control cells ( Figure 7A and B ) . In total , the reduction in the signal intensities for Syk p-Y352 was slower in DG-75 cells treated with 1G244 compared to the mock treated control cells ( Figure 7A and B ) , suggesting that DPP9 cleaves active phosphorylated Syk . 10 . 7554/eLife . 16370 . 015Figure 7 . DPP9 targets phosphorylated Syk for degradation thus influencing Syk signalling in B cells , ( A ) Higher levels of endogenous active Syk ( phosphorylated on Y352 ) in stimulated DG-75 cells treated with the DPP8/9 inhibitor 1G244 compared to the mock ( DMSO ) treated cells . 1G244 ( 10 µM ) was added at the same time of BCR stimulation ( time 0 ) . Tubulin was assayed for loading control . Shown is a representative result of at least three independent pulse chase experiments . ( B ) Quantification of the Western blot results shown in ( A ) . The ratio of Syk p-Y352/tubulin at time 10 min was normalized to 100% . For signal quantification GelQuant . NET software provided by biochemlabsolutions . com was used . ( C ) Inhibition of DPP9 in DG-75 cells leads to increased Ca2+ mobilization , which is not dependant on BCR stimulation . Shown are flow cytometric Ca2+ profiles after the addition of 10 µM DPP8/9 inhibitor 1G244 or DMSO for control ( marked by an arrow ) . To monitor Ca2+ mobilization upon BCR stimulation in either 1G244-treated or control cells , cells were treated with 10 µg/ml F ( ab ) 2 goat-anti-human IgM . ( D ) Same as in ( C ) using Ramos cells as a second B cell line . ( E–F ) In the absence of BCR stimulation , DPP9 inhibition leads to higher basal levels of phosphorylated Syk and its down stream effector protein PLCγ2 . ( E ) Western blotting analysis of DG-75 cells treated with 1G244 in the absence of BCR stimulation . Lysates were analysed with antibodies specific against phosphorylated Syk ( p-Y352 ) or phosphorylated PLCγ2 . For loading control lysates were analysed with antibodies recognizing unmodified Syk and PLCγ2 . ( F ) Cells were treated for 20 min with 1G244 ( 10 µM ) or DMSO for control . Alternatively , cells were treated for 30 min with the allosteric DPP9 inhibitor SLRFLYEG peptide complexed with the carrier peptide ( pep-1 ) . Control cells were treated with the carrier peptide only . Following inhibitor treatment , cells were lysed and subjected to immunoprecipitation assays against Phospho-Y . Eluted proteins were analysed for Syk and PLCγ2 levels by Western blotting . Total protein levels in cell lysates were monitored for control . ( G ) Lower levels of phosphorylated ERK1/2 ( both bands ) are detected in the 1G244 treated DG-75 cells compared to mock ( DMSO ) treated cells . 1G244 ( 10 µM ) was added prior to BCR stimulation ( time 0 ) . DPP9 was assayed for loading control . Shown is a representative result of at least three independent pulse chase experiments . ( H ) Quantification of the Western blot results shown in ( G ) as described in ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 015 Next , we analysed the activity of PLCγ2 , which is phosphorylated by Syk upon BCR stimulation resulting in Ca2+ release from the endoplasmic reticulum ( Engelke et al . , 2007 ) . Remarkably , application of 1G244 to DG-75 cells triggered a rapid increase in Ca2+ fluxes , which occurred before the stimulation of the BCR ( Figure 7C ) . The levels of Ca2+ reached a plateau within less than 3 min after application of 1G244 , in DG-75 cells , and were less elevated upon BCR engagement compared to control cells . Similar results were also observed in Ramos cells , a second Burkitt’s lymphoma cell line ( Figure 7D ) . The rapid increase in calcium fluxes in response to DPP9 inhibition independently of BCR stimulation was surprising , since Syk appears to be stable in unstimulated B cells ( Figure 6E ) . Furthermore , in the absence of BCR stimulation , the steady state levels of Syk were comparable in cells treated with 1G244 and in mock treated cells ( no 1G244 ) ( Figure 7E ) . Similar results were observed for PLCγ2 . However , as we tested for the levels of active syk and PLCγ2 , we found that the phosphorylation levels of Syk ( Syk-pY352 ) and PLCγ2 ( PLCγ2-pY753 ) were higher in the 1G244 treated cells ( Figure 7E ) . These observations are in line with the increased Ca2+ fluxes under these conditions ( Figure 7C and D ) . Immunoprecipitation assays of DG-75 cell lysates with antibodies against phospho-tyrosine ( p-Y ) ( Figure 7F ) also revealed higher levels of phosphorylated Syk and PLCγ2 upon DPP9 inhibition with 1G244 ( Figure 7F ) , independently of BCR stimulation . Consistently , an increase in the steady state phosphorylation levels of PLCγ2 on pY753 was also observed in DG75 cells treated with the allosteric inhibitor of DPP9 SLRFLYEG in the absence of BCR engagement ( Figure 7F ) . To further test the biological consequences of DPP9 activity , we analysed the effect of DPP9 inhibition on the extracellular signal-related kinase ( ERK ) 1/2 which is a distal downstream target of Syk ( Slack et al , 2007 ) . We therefore incubated DG-75 cells with 1G244 prior to BCR stimulation and subsequently analysed ERK1/2 phosphorylation . While BCR engagement under DPP9 inhibition revealed a more transient ERK1/2 phosphorylation profile compared to control cells , it is important to note that the levels of ERK1/2 phosphorylation in resting cells were slightly augmented in cells that were treated with 1G244 ( Figure 7 G and H ) . These results are in line with the finding that DPP9 inhibition in unstimulated cells stabilizes an active phosphorylated form of Syk , resulting in ‘signal leakage’ as measured by higher levels of phosphorylated PLCγ2 forms and increased Ca2+ fluxes leading to reduced sensitivity to BCR engagement . Taken together , these results suggest that DPP9 is an integral negative regulator of Syk signalling by terminating Syk activity through processing the amino terminus of the kinase .
Here , we report the identification of Syk as a novel endogenous substrate of DPP9 . Being a central component of the B cell signalling , Syk initiates the phosphorylation of a cascade of downstream components , ultimately modulating cellular metabolism , cell migration , cell proliferation , apoptosis and gene transcription events ( Mócsai et al . , 2010; Lowell , 2011 ) . Syk is activated following BCR engagement , and is negatively regulated by phosphatases such as Shp-1 and PTPROt ( Dustin et al . , 1999; Alsadeq et al . , 2014; Chen et al . , 2006 ) , and by the E3 ubiquitin ligase Cbl ( Paolini et al . , 2001; Joazeiro et al . , 1999; Rao et al . , 2001a; Sohn et al . , 2003; Lupher et al . , 1998; Ota and Samelson , 1997 ) . Here , we present evidence for the role of DPP9 as a novel negative regulator of Syk that determines Syk stability . DPP9 inhibition results in increased half-life of Syk and higher levels of active phosphorylated Syk pY352 , suggesting that DPP9 targets the active form of Syk for degradation . Furthermore DPP9 inhibition results in a prolonged accumulation of Syk phosphorylated on Y323 , which is a critical residue for interaction with Cbl ( Yankee et al . , 1999; Lupher et al . , 1998 ) . Moreover , the interaction between Cbl and Syk is reduced to background level upon DPP9 silencing . Taken together , these results strongly suggest a cross talk between DPP9 and Cbl , in which cleavage of Syk by DPP9 is an up stream event to Syk ubiquitination by Cbl . Surprisingly , also in the absence of BCR-engagement , DPP9 inhibition leads to higher steady state levels of phosphorylated Syk ( pY352 ) and PLCγ2 ( pY753 ) as well as to a rapid mobilization of Ca2+ from the ER . These results show that inhibition of DPP9 prior to BCR engagement leads to a ‘leakage’ in Syk signalling , and consequently reduces the response of these cells to BCR-stimulation as measured by lower levels of Ca2+ release upon BCR-engagement probably due to the depletion of Ca2+ in the resting stage . Consistently , we observed reduced BCR mediated phosphorylation of ERK1/2 upon DPP9 inhibition in these cells . In addition to activated BCR signalling , which relays on the binding of antigens to the BCR , B cells also possess the so-called ‘tonic’ signalling , which maintains a constitutive baseline signal in B cells . Tonic signalling occurs in the absence of antigen binding , but nonetheless is dependent on the presence and activity of the BCR and some BCR components such as Syk , PLCγ2 and ERK1/2 ( Monroe , 2006; Wienands et al . , 1996; Shaffer and Schlissel , 1997 ) . The molecular details regarding this pathway are only emerging but it is clear that tonic signalling is essential for survival of B cells , including Burkitt lymphomas ( Corso et al . , 2016 ) . We suggest that by targeting Syk pY352 for degradation , DPP9 maintains low levels of activated Syk sufficient for tonic signalling , and thus can influence the strength of the BCR signalling in response to antigen engagement . By targeting Syk for degradation after BCR engagement , DPP9 can also influence the duration of the response to antigen BCR binding ( Figure 8A and B ) . 10 . 7554/eLife . 16370 . 016Figure 8 . Model - DPP9 is a negative regulator for Syk signalling , targeting active Syk for degradation by the N-end rule pathway . ( A ) Model: DPP9 regulates Syk signalling . By maintaining low levels of activated Syk in resting B cells DPP9 controls tonic signalling and preserves signalling capacity for the processes induced by BCR engagement . By targeting Syk for degradation after BCR engagement , DPP9 can also influence the duration of the response to antigen BCR binding . ( B ) Reduced DPP9 activity e . g by inhibition results in higher levels of active Syk in non-stimulated cells . This results in elevated Tonic signalling ( ‘signal leakage’ ) , consequently leading to a lower response upon BCR engagement . ( C ) Model: DPP9 targets Syk for degradation by the N-end rule pathway . In this model , FLNA acts as a recruitment platform for binding of Syk and DPP9 . FLNA then supports the cleavage of Syk by DPP9 , by increasing the local concentration of Syk and DPP9 , stabilizing the interaction between Syk and DPP9 , or by optimizing the orientation of DPP9 and Syk . The interaction between DPP and Syk leads to the processing of Syk N-terminus which removes the dipeptide MA , and exposes a neo Syk N-terminus with serine in position 1 is exposed . Subsequently the E3 ligase Cbl binds to Syk p-Y323 initiating its ubiquitintaion and degradation by the proteasome . DOI: http://dx . doi . org/10 . 7554/eLife . 16370 . 016 Several reports link DPP9 to various cellular processes , including cell migration , metabolism , cell proliferation and apoptosis ( Zhang et al . , 2015; Yao et al . , 2011; Matheeussen et al . , 2013; Spagnuolo et al . , 2013; Chen et al . , 2016; Han et al . , 2015 ) . It remains to be shown whether some of the effects caused by DPP9 silencing are due to stabilization of Syk in these cells . Since Syk appears to act as a negative regulator of specific tumours ( Coopman and Mueller , 2006 ) , inhibition of DPP9 to stabilize Syk in these cells may reflect a future approach for tumour therapy . Our data show that processing of Syk by DPP9 exposes a neo Syk N-terminus with a serine residue in position 1 ( Figure 8C ) . Silencing ( DPP9-kd ) or inhibition of DPP9 , or mutation of the DPP9 cleavage site in Syk ( SykD2A ) result in an increased stability of Syk . The significance of the exposed Ser for Syk stability was demonstrated by mutagenesis of this residue to a Gly or Val , resulting in a longer half-life of these variants . Therefore , we conclude that Syk is a novel substrate of the N-end rule pathway and DPP9 a novel peptidase in this pathway ( Figure 8C , and see more below ) . Of note classical primary N-degron are positively charged or bulky hydrophobic amino acids . According to this definition , Ser , Gly and Val are not destabilizing amino acids per se ( Bachmair and Varshavsky , 1989 ) , however they can form an N-degron upon N-terminal acetylation ( Ac/N-degrons ) ( Hwang et al . , 2010b; Shemorry et al . , 2013; Park et al . , 2015 ) . Interestingly , the difference in the stability of Syk wild-type , SykS3V and SykS3G mirrors the N-acetylation efficiency of these residues ( Arnesen et al . , 2009 ) . DPP9 inhibition results in a prolonged accumulation of Syk phosphorylated on Y323 , which is a critical residue for interaction with the E3 ligase Cbl ( Yankee et al . , 1999; Lupher et al . , 1998 ) . Moreover , the interaction between Cbl and Syk is reduced to background level upon DPP9 silencing . Taken together , these results strongly suggest a cross talk between DPP9 and Cbl , in which cleavage of Syk by DPP9 is an up stream event to Syk ubiquitination by Cbl . The functional link between Cbl and DPP9 suggests a novel role for Cbl in the N-end rule pathway , further expanding the growing list of ubiquitin E3 ligases which includes Doa10/Teb4 and Not4 that recognize non-primary N-degrons ( Hwang et al . , 2010b; Shemorry et al . , 2013; Park et al . , 2015 ) . Proteases such as calpains , separins , metalloproteases and caspases are linked to the N-end rule pathway , since they produce protein fragments that are degraded by this pathway ( Rao et al . , 2001b; Brower et al . , 2013; Piatkov et al . , 2014; Liu et al . , 2016 ) . Knowledge regarding the participation of aminopeptidases in the N-end rule pathway is limited to methionine aminopeptidases ( MetAPs ) ( Varshavsky , 2011 ) . These remove a single methionine from the amino terminus of a protein substrate leaving the protein otherwise intact . Of note , cleavage by MetAPs does not directly form an N-degron , since MetAPs remove the initiator methionine from substrates where the adjacent residue ( P1’ ) is not destabilizing: Ser , Pro , Gly , Thr , Ala , Val or Cys ( Bachmair and Varshavsky , 1989; Xiao et al . , 2010; Addlagatta et al . , 2005; Frottin et al . , 2006 ) . Nonetheless , MetAPs are considered an integral part of the N-end rule pathway , because these residues may be converted to Ac/N-degrons ( Hwang et al . , 2010b ) , albeit to different efficiencies . Our data place DPP9 as a second integral aminopeptidase of the N-end rule pathway . In contrast to MetAPs , DPP9 cleavage is permissive towards more amino acids in the P1’ position ( Xaa-Ala/Pro-Zaa ) , and is only inhibited if the residue after the cleavage site ( Zaa ) is Pro ( Geiss-Friedlander et al . , 2009 ) . Therefore DPP9 may directly produce an N-degron not requiring further modification ( Figure 8C ) . Additionally , DPP9 can cleave substrates with any amino acid in the P2 position ( Xaa ) apart from Asp or Glu , greatly increasing the number of potential substrates , including proteins undergoing progressive processing ( Geiss-Friedlander et al . , 2009; Connolly et al . , 2008 ) . Whereas MetAPs function largely in a co-translational manner , our data strongly suggest that DPP9 acts in a post-translational manner . Consequently , processing by DPP9 and MetAPs and funnelling of their products to the N-end rule pathway may be regulated by different signals . MetAP substrates , which were processed and acetylated at the N-terminus co-translationally , can be protected from degradation by masking of the N-degron , e . g . via protein interactions , and exposed upon changes in protein stoichiometry ( Hwang et al . , 2010b; Shemorry et al . , 2013 ) . On the other hand , DPP9 substrates may be processed only upon a specific post-translational cue . For Syk , our data suggest increased cleavage by DPP9 upon BCR stimulation indicated by a shorter half-life of Syk in stimulated cells , and clear stabilization by 1G244 treatment comparable to that observed with the proteasome inhibitor MG132 . This increased processing of Syk upon BCR stimulation may depend for example on Syk phosphorylation . Additionally , we found that although DPP9 interacts directly with Syk , in cells this interaction requires FLNA , suggesting that FLNA may serve as an interaction platform linking DPP9 and Syk ( Figure 8C ) . Consequently , FLNA may support Syk processing e . g . by increasing the local concentration of Syk and DPP9 , stabilizing the complex formation or by optimizing the orientation of DPP9 and Syk for efficient cleavage . In this sense , the function of FLNA may parallel that of F-box proteins in the ubiquitin ligase SCF complexes , which bind to substrates and thus support their ubiquitination by ubiquitin conjugating enzymes ( Edward and Kipreos , 2000 ) . Since FLNA interacts with multiple proteins , it is tempting to speculate that it acts as a general recruitment factor targeting DPP9 to additional potential substrates for cleavage . In conclusion , here we identify DPP9 as a novel upstream component of the N-end rule pathway . We suggest that depending on the substrate and the exposed residue in the neo N-terminus , DPP9 may cooperate with Cbl or other E3 ligases that then ubiquitinate the substrate resulting in proteasomal degradation .
HeLa DPP9 silenced cells ( DPP9-kd ) were custom made for us by GenScript . Culturing of HeLa DPP9 silenced cells ( DPP9-kd ) and the corresponding HeLa parental wt cells were cultured at 37°C and 5% CO2 in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum , 2 mM L-glutamine , 100 U/ml penicillin and 100 μg/ml streptomycin . To maintain the selection pressure on the DPP9-kd cells , 1 . 5 µg/ml puromycin ( Sigma-Aldrich , Germany ) was added to the growth medium . Human DG-75 and Ramos B cells ( purchased ATCC , Germany ) were cultured in VLE RPMI medium ( Biochrom , Germany ) supplemented with 3 mM glutamine , 10% heat-inactivated FCS , 100 mM sodium pyruvate and 100 U/ml penicillin and 100 μg/ml streptomycin . All cells are routinely tested for mycoplasma ( by GATC Biotech , Germany ) . C-terminally tagged full-length Syk was generated via PCR using a Syk-containing plasmid and was subsequently cloned into pcDNA3 . 1+ ( using the BamHI and NotI restriction sites ) . Single amino acid exchanges at the N-terminus were introduced using primers for site-directed mutagenesis . All plasmids were sequenced before usage . For overexpression experiments , HeLa ( P4 ) cells were transfected using the calcium-phosphate method at a cell confluency of 50–60% and analysed 48 hr after transfection . Control cells were transfected with a GFP plasmid ( MOCK ) . For silencing of DPP9 HeLa cells were transfected essentially as previously described in ( Geiss-Friedlander et al . , 2009 ) . SiRNA oligonucleotides for Syk were purchased from Santa Cruz Biotechnology Inc . Germany , siRNA oligonucleotides targeting FLNA were purchased from Invitrogen . For control , cells were transfected with a non-targeting siRNA oligonucleotide ( Invitrogen ) . Cells were analysed 72 hr after transfection . For Western blotting rabbit anti-DPP9 ( #ab42080 , 1:1000 ) was purchased from Abcam , England . We produced the goat anti-DPP9 antibodies ( 1:1000 ) by injecting recombinant DPP9-short protein to a goat . Antibody specificity is shown in Figure 1—figure supplement 1A . Rabbit anti-Syk ( #sc-1077 , 1:1000 ) and the corresponding blocking peptide rabbit anti-Syk ( #sc-929 ) , mouse anti-Tubulin ( #sc-32293 , 1:1000 ) , rabbit anti-GFP ( #sc-8334 , 1:1000 ) , and mouse anti-GST ( #sc138 , 1:1000 ) antibodies were obtained from Santa Cruz Biotechnology Inc . Mouse anti-FLAG ( #f1804 , 1:1000 ) was purchased from SIGMA and mouse anti-FLNA ( #MAB1680 , 1:1000 ) from Millipore , Germany . Mouse anti-pTyr 100 ( #9411 ) , rabbit anti-phospho-ERK1/2 ( #4370 1:2000 ) , rabbit anti PLCγ2 ( 1:2000 , Santa Cruz Biotechnology , #sc-9015 ) , rabbit anti p-PLCγ2 ( Y753 ) ( 1:500 , Santa Cruz Biotechnology , #sc-101785 ) , rabbit anti Syk ( #2712 , 1:1000 ) and rabbit anti-phospho-Syk ( Y323 , #2715 , 1:1000; Y352 , #2701 , 1:1000 ) were obtained from Cell Signalling Technology , The Netherlands . The enhanced chemiluminescence system ( Millipore ) was used for visualization of proteins on the membranes . For signal quantification GelQuant . NET software provided by biochemlabsolutions . com was used . In immunofluorescence studies and in situ Proximity ligation assays ( PLAs ) the following antibodies were used: self-generated goat anti-DPP9 ( 1:10–20 ) , mouse anti DPPIV ( 1:50–100; Santa Cruz Biotechnology , #sc-19607 ) , mouse anti DPP8 ( 1:50–100 , Santa Cruz Biotechnology , #sc-37699 ) , rabbit anti-Syk ( 1:100–1:200 ) , goat anti-Cbl ( C-15 ) ( 1:60; Santa Cruz Biotechnology , #sc-170-G ) , mouse anti-FLNA ( 1:100 ) , rabbit anti-FLNA ( 1:100 , NB100-58812; Novus Biologicals , Germany ) and mouse anti-HA ( 1:400 , , MMS-101P; Covance Germany ) . Secondary antibodies for immunofluorescence: donkey anti-mouse Alexa-Fluor-488 , donkey anti-goat Alexa-Fluor-594 and donkey αnti-rabbit Alexa-Fluor-594 , all at a dilution of 1:500; purchased from Molecular Probes . For PLAs the Duolink In situ PLA kit was purchased from Sigma-Aldrich . The PLA Probes ( oligonucleotide-conjugated secondary antibodies ) were used in the combinations: anti-Rabbit PLUS ( #DUO92002 ) with either a Goat MINUS ( #DUO92006 ) or a Mouse MINUS ( #DUO92004 ) , and a Goat MINUS ( #DUO92006 ) with a Mouse PLUS ( #DUO92001 ) ; the probes were diluted 1:5 . For staining of actin filaments in fixed cells CytoPainter Phalloidin-iFluor 488 ( 1:650 , #ab176753 , Abcam ) was used . For staining of actin filaments in fixed cells CytoPainter Phalloidin-iFluor 488 ( 1:650 , #ab176753 , Abcam ) was used . The DPP8/9-specific inhibitor 1G244 was purchased from AK Scientific , Inc . ( Union City , CA ) , the proteasome inhibitor MG132 was purchased from Enzo Life Sciences . All peptides including the SLRFLYEG peptide were custom-made by GenScript ( Hong Kong ) , to > 80% purity . A human placenta library was screened using full-length DPP9-short as bait . The assay was performed by Hybrigenics ( Paris , France ) . For immunofluorescence , HeLa cells grown on coverslips were fixed with 4% formaldehyde in PBS containing 10 μg/ml Hoechst 33 , 258 ( Molecular Probes ) for staining of the nuclei . Subsequently , cells were permeabilized with 0 . 2% Triton-X-100 in PBS , washed and blocked for 10 min in blocking buffer ( 2% BSA in PBS ) . For IF of DG-75 cells , 3*105 cells per slide were resuspended in PBS , added to the cover slip and incubated for 30 min at RT to allow attachment of the cells to the surface . Cells were fixed with 1% formaldehyde in PBS containing 10 μg/ml Hoechst 33 , 258 for staining of the nuclei . Subsequently , cells were permeabilized with 0 . 02% Triton-X-100 in PBS , washed and blocked for 30 min in blocking buffer . Both cell lines were incubated with primary antibodies for 90 min at 37°C . Following a PBS wash , cells were then incubated for 45 min at room temperature with the respective secondary antibodies . Cells were washed with PBS and water , and mounted in fluorescent mounting medium ( DAKO ) . Control samples were treated with secondary antibody only , to estimate background staining . Cells were analysed and images were taken using a LSM 510-Meta confocal microscope , oil immersion objective 63x/1 . 3 ( Carl Zeiss MicroImaging , Inc ) . Images were processed using the LSM image Browser ( Carl Zeiss MicroImaging , Inc ) and Adobe Photoshop . PLA was performed using the DUOLINK In Situ PLA Kit from Sigma Aldrich , according to the manufacturer's protocol . In short , HeLa cells and DG-75 cells were grown on coverslips and fixed as described above . To investigate consequences of DPP9 inhibition prior to fixation cells were treated either with 10 µM 1G244 or DMSO as MOCK for 5 or 30 min or alternatively with the allosteric peptide inhibitor SLRFLYEG . For entry into the cells the SLRFLYEG peptide was incubated with the carrier peptide ( pep1 ) to form the SLRFLYEG-pep1 complexes for 30 min or 1 hr at 37°C ( as described in Pilla et al . ( 2013 ) ) . Control assays contained pep1 alone . Both cell lines were then incubated with primary antibodies , 90 min at 37°C and actin filaments were simultaneously counterstained with CytoPainter Phalloidin-iFluor 488 Reagent ( Abcam - #ab176753 , 1:650 dilution ) . In case of a control for antibody specificity , the antibody was pre-incubated with the corresponding blocking peptide ( 1 µg/ml ) for 30 min at RT before addition to the cover slips . To estimate background staining in each experiment control slides were treated with one primary antibody only . After careful wash with PBS , cells were treated with the PLA probes according to the manufacturer's protocol . Cells were mounted in DAKO fluorescent mounting medium and analysed using a LSM 510-Meta confocal microscope , oil immersion objective 63x/1 . 3 ( Carl Zeiss MicroImaging , Inc ) . Taken images were processed using the LSM image Browser ( Carl Zeiss MicroImaging , Inc ) and subsequently analysed using the Duolink ImageTool ( SIGMA ) . Binding experiments were performed on a Reichert SPR Biosensor ( SR 7500 C , Reichert Instruments ) , using Ni2+ chelator chips ( NiHC 1000 m , Xantec Bioanalytics , Duesseldorf ) . The surface was initially purged of contaminants injecting 0 . 5 M EDTA , pH 8 . 5 , followed by equilibration steps washing the surface with immobilization buffer ( 20 mM Hepes , 150 mM NaCl , 0 . 005% Tween 20 , pH 7 . 4 ) at a flow rate of 30 µl/min . The left channel of the chip surface ( ligand channel ) was charged with Ni2+ using 0 . 3 M NiSO4 . A 200 nM solution of His-tagged , affinity purified DPP9 was subsequently injected at a flow rate of 30 µl/min . The ligand ( DPP9 ) was immobilized to a response of 2500 to 3500 µRIU . The right channel was kept empty and used as the first reference . Synthesized peptides correlating the first 31 amino acids of the Syk N-terminus were used as the analyte . A serial dilution of the peptide was injected over the chip surface at concentrations of: 16 µM , 8 µM , 4 µM , 2 µM , 1 µM , 0 . 5 µM , 0 . 25 µM and 0 . 125 µM . For each analyte sample twice as many buffer injections were performed , which were later on used as a second reference . After each binding experiment , containing two buffer injections and one analyte injection , the chip surface was washed with EDTA ( 0 . 5 M pH 8 . 5 ) and new ligand ( DPP9 ) was immobilized as described above . All binding experiments were carried out at a flow rate of 40 µl/min in SPR running buffer ( 20 mM Hepes , 150 mM NaCl , 0 . 005% Tween 20 , 50 µM EDTA , pH 7 . 4 ) . The response of each analyte sample was doubly referenced with the response obtained from the reference channel ( right channel ) and the response obtained by injecting buffer using Scrubber version 2 . 0c . Equilibrium binding analysis was performed using Graph Pad Prism 6 . 0 . Recombinant DPP9-short was expressed in SF9 insect cells and purified essentially as described in ( Nakamura et al . , 2007 ) . Full-length FLNA wild-type and variants were purified as described in ( Yue et al . , 2013 ) . For expression of GST-FLNA fragments 5–7: FLNA repeats 5–7 or 6–7 in pGEX4T1 were transformed into Escherichia coli BL21 ( Stratagene ) . Cells were grown to A600 0 . 6 and induced with 0 . 1 mM isopropyl 1-thio-β-D-galactopyranoside for 3 hr at 30°C . All following buffers were supplemented with protease inhibitors ( 1 µg/ml each of leupeptin , pepstatin , and aprotinin ) , and 1 mM dithiothreitol ( DTT ) . Cells were harvested and resuspended in lysis buffer ( 50 mM Tris-HCl , pH 8 . 0 , 100 mM NaCl , 1 mM EDTA , 1 mM EGTA ) . Cells were disrupted using an EmulsiFlex ( Avestin ) and centrifuged for 1 hr at 100 , 000 ×g . The supernatant was incubated with 1 ml Glutathion-Sepharose ( Macherey-Nagel ) for 1 hr at 4°C . Beads were washed at 4°C with binding buffer ( 50 mM Tris-HCl , pH 8 . 0 , 300 mM NaCl , 1 mM EDTA , 1 mM EGTA ) , supplemented with protease inhibitors and 1 mM DTT . Proteins were eluted with elution buffer ( 20 mM glutathione in 50 mM Tris-HCl , pH 8 . 0 , 300 mM NaCl , 1 mM EDTA , 1 mM EGTA ) supplemented with protease inhibitors and 1 mM DTT and further purified using an Äkta Purifier ( GE Healthcare ) equipped with a Superdex 75 size exclusion column ( GE Healthcare ) in Transport buffer ( 20 mM Hepes , pH 7 . 3 , 110 mM potassium acetate , 2 mM Mg acetate , 1 mM EGTA ) supplemented with protease inhibitors and 1 mM DTT . To measure DPP activity in DG-75 cells , 2*107 cells were resuspended in 2 ml of RPMI complete medium containing either 10 µM 1G244 or DMSO ( MOCK ) and incubated for the corresponding times ( 5 min , 30 min ) at 37°C . The reaction was stopped with 20 ml ice-cold PBS and cells were pelleted for 5 min at 500 g . Subsequently , cells were washed with 10 ml ice-cold PBS and were shock-frozen in liquid N2 . For activity measurements , cell pellets of the respective cell line were lysed in TB buffer ( 20 mM HEPES/KOH , pH 7 . 3 , 110 mM potassium acetate , 2 mM magnesium acetate , 0 . 5 mM EGTA ) supplemented with 0 . 02% Tween 20 and 1 mM DTT , centrifuged for 20 min at 55 , 000 rpm , 4°C . Next , 5 µg cell lysate was incubated with either 250 µM Gly-Pro-AMC ( GP-AMC ) or 50 µM Arg-AMC ( R-AMC ) , fluorescence release was measured using the Appliskan microplate fluorimeter ( Thermo Scientific ) with 380 nm ( excitation ) and 480 nm ( emission ) filters and SkanIt software . For subsequent analysis of the activity measurements Prism 5 . 0 ( GraphPad Software ) was used . For Michaelis-Menten analysis of Met-Ala-AMC ( MA-AMC ) or Met-Pro-AMC ( MP-AMC ) hydrolysis , 12 , 5 nM purified recombinant DPP9-short was incubated with various concentrations of MA-AMC or MP-AMC and fluorescence release was measured as described above . Each assay was performed at least three times , each time in triplicates ( technical repetitions ) . 50 µM of the Syk amino terminus peptide 1–31 ( MASSGMADSANHLPFFFGNITREEAEDYLVQ ) was incubated alone , in the presence of 130 nM DPP9 wt or its inactive variant DPP9 S730G . To test for inhibition , 10 µM peptide inhibitor ( SLRFLYEG ) was added . All reactions were performed in TB buffer ( 20 mM HEPES/KOH , pH 7 . 3 , 110 mM potassium acetate , 2 mM magnesium acetate , 0 . 5 mM EGTA ) supplemented with 0 . 2% Tween 20 . Reactions were stopped after 6 hr by dilution and acidification in aqueous 0 . 1% formic acid , 2% acetonitrile ( 1/500 , v:v ) . The resulting samples were analysed on a nanoLC425 nanoflow chromatography system coupled to a TripleToF 5600+ Plus mass spectrometer of QqToF geometry ( both SCIEX ) . In short , 5 µl of sample were pre-concentrated on a self-packed Reversed Phase-C18 precolumn ( Reprosil C18-AQ , Pore Size 120 Å , Particle Size 5 µm , 4 cm length , 0 . 15 cm I . D . , Dr . Maisch ) and separated on a self-packed Reversed Phase-C18 microcolumn ( Reprosil C18-AQ , 120 Å , 3 µm , 15 cm , 0 . 075 cm ) using a 15 min linear gradient ( 5 to 50% acetonitrile , 0 . 1% formic acid modifier , flow rate 300 nl/min , column temperature 50°C ) followed by a 5 min high organic cleaning step and a 15 min column re-equilibration . The eluent was introduced to the mass spectrometer using a Nanospray III ion source with Desolvation Chamber Interface ( SCIEX ) via a commercial Fused Silica tip ( FS360-20-10-N , New Objective ) at a spray voltage of 2 . 4 kV , a sheath gas setting of 12 and an interface heater temperature of 150°C . The MS acquisition cycle consisted of a 500 ms TOF MS survey scan that was used for profiling of substrate and product concentrations followed by data-dependent triggering of up to five 100 ms TOF product ion spectra to confirm the identity of detected analytes . Data analysis was performed using Analyst TF 1 . 7 and PeakView 2 . 1 softwares ( SCIEX ) . Analyses were performed in triplicates . 0 . 5–1 × 106 DG-75 cells/ml were seeded in 24 wells . 24 hr later cells were treated with CHX ( 100 µg/ml ) and , where stated , in parallel stimulated with 12 µg/ml F ( ab’ ) 2 fragment goat anti-human IgG + IgM ( H+L ) ( Dianova ) in the presence of either 10 µM 1G244 or 100 µM MG132 or DMSO as MOCK control . Cells were harvested at different time points , counted and shock-frozen in liquid N2 . Cell pellets were resuspended in the respective amount of sample buffer according to cell number . To analyse the stability of phosphorylated Syk cells were incubated in serum-free RPMI for 20 min before they were seeded in 24 wells ( 0 . 8 × 106 cells per well ) , treated with CHX and in parallel stimulated in the presence or absence of 10 µM 1G244 as described above . Cells were harvested at different time points in 250 µl sample buffer . For assays with HeLa wt and DPP9-kd cells 1 . 4 × 105 cells were seeded in 24 wells and transfected with various Syk-FLAG constructs together with a GFP plasmid on the next day . 48 hr after transfection , cells were treated with CHX and harvested at the respective time in sample buffer . 3∙107 DG-75 cells were washed with PBS , resuspended in serum-free RPMI in the presence or absence of 1G244 ( 10 µM ) . Unless otherwise stated , cells were incubated at 37°C with the inhibitor for 20 min . Cells were resuspended in cold lysis buffer ( 20 mM Tris/HCl ( pH 7 . 5 ) , 150 mM NaCl , 0 . 5 mM EDTA , 10 mM NaF , 10 µM MoO4 , 1 mM Na3VO4 , 1% IGEPAL CA-630 , PMSF , Aprotinin , Leupeptin ) , and incubated at 4°C for 1 hr with constant agitation . Cell lysates were centrifuged at 50000 g , at 4°C for 15 min . The supernatant was incubated for 30 min at 4°C with protein A beads to remove precipitating proteins . The pre-cleared supernatant was then subjected to IP with 5 µg/ml mouse anti-pTyr 100 ( #9411 ) for 2 hr at 4°C with constant agitation . This was followed by the addition of 20 µl protein A/G-magnetic beads for 30 min ( Thermo Fisher Scientific ) . Following careful washing , proteins were eluted with reducing 2 × sample buffer at 65°C for 5 min . 106 DG-75 or Ramos B cells were loaded in 700 µl RPMI containing 5% FCS , 1 µM Indo1-AM ( Molecular Probes ) , and 0 . 015% Pluronic F127 ( Molecular Probes ) at 30°C for 25 min . Subsequently , the cell suspension was diluted two-fold with RPMI 10% FCS and incubated for 10 min at 37°C . Cells were washed , treated with 10 µM 1G244 ( or mock-treated ) and prepared for measurements as described earlier ( Stork et al . , 2007 ) . For BCR-induced Ca2+ mobilization , cells were treated with 10 µg/ml F ( ab ) 2 goat-anti-human IgM ( Jackson Immunoresearch ) . Changes in the ratio of fluorescence intensities at 405 nm and 510 nm were monitored on a LSRII flow cytometer ( Becton Dickinson ) and analysed with FlowJo ( TriStar ) . | Proteins are made up of building blocks called amino acids bonded together to form chain-like molecules . Around twenty different amino acids are used to make proteins , and enzymes called proteases can recognize specific pairs of amino acids in proteins and cut the bonds between them . Dipeptidylpeptidase 9 ( or DPP9 for short ) is a protease that removes two amino acids from the end of a protein , just as long the second amino acid is one of two specific kinds ( namely , an alanine or a proline ) . The DPP9 protease influences a range of processes in the cell including cell death , signaling and survival . Indeed , mice born with an inactive version of DPP9 die shortly after birth , but it is not known why this happens . Justa-Schuch et al . investigated how the protease DPP9 controls processes inside cells and found an unexpected connection between DPP9 and another protein called Syk . The Syk protein is found in immune cells called B cells , and becomes highly activated whenever these cells are stimulated . Once activated Syk changes the activity of many proteins , affecting which genes are switched on and how the B cell moves and divides . By using DPP9 as a kind of bait , Justa-Schuch et al . found human proteins that bind to the protease . This search identified a protein called Filamin A that interacted with DPP9 , placing DPP9 close to Syk , which also binds to Filamin A . Further experiments showed that when DPP9 was located close to Syk , it cut the end of Syk . This cut left the Syk protein with a different amino acid exposed at its end , which in turn made it susceptible to being broken down inside the cell . Justa-Schuch et al . went on to show that DPP9 preferentially cleaved the active form of Syk . Since cleaved Syk was subsequently broken down , DPP9 acts as a shut-off mechanism for Syk after the B cell has been stimulated . The findings show that DPP9 can influence how much and how long the B cell responds to stimulation . Inhibitors of DPP9 may therefore be useful for stabilizing Syk , which is known to stop specific tumors from growing . Future work will investigate the mechanisms that control how Filamin A , DPP9 and Syk interact . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2016 | DPP9 is a novel component of the N-end rule pathway targeting the tyrosine kinase Syk |
Responses of midbrain dopamine ( DA ) neurons reflecting expected reward from sensory cues are critical for reward-based associative learning . However , critical pathways by which reward-related visual information is relayed to DA neurons remain unclear . To address this question , we investigated Pavlovian conditioning in macaque monkeys with unilateral primary visual cortex ( V1 ) lesions ( an animal model of ‘blindsight’ ) . Anticipatory licking responses to obtain juice drops were elicited in response to visual conditioned stimuli ( CS ) in the affected visual field . Subsequent pharmacological inactivation of the superior colliculus ( SC ) suppressed the anticipatory licking . Concurrent single unit recordings indicated that DA responses reflecting the reward expectation could be recorded in the absence of V1 , and that these responses were also suppressed by SC inactivation . These results indicate that the subcortical visual circuit can relay reward-predicting visual information to DA neurons and integrity of the SC is necessary for visually-elicited classically conditioned responses after V1 lesion .
Adaptive behaviour in a changing environment requires that we have to learn and update associations between unconditioned rewards and punishments , and the sensory stimuli that predict them . This form of associative learning is called classical or Pavlovian conditioning ( Pavlov , 1927 ) . The Pavlovian paradigm has been used widely to investigate the role of midbrain dopamine ( DA ) neurons in associative learning ( Schultz , 1998 ) . Much evidence indicates that the activity of DA neurons in the substantia nigra pars compacta ( SNc ) makes a key contribution in associative learning , in part , by encoding reward prediction errors . A reward prediction error is a scalar signal that signifies a current event is better or worse than predicted . In a series of pioneering experiments Schultz and colleagues ( Schultz et al . , 1992 , 1997; Mirenowicz and Schultz , 1994 ) showed that DA responses to an unpredicted reward ( unconditioned stimulus; UCS ) , gradually transferred to an unexpected predicting conditioned stimulus ( CS ) . If a predicting CS was presented but subsequent reward delivery was withheld , DA neurons would pause briefly at the expected time of reward delivery ( Schultz et al . , 1997 ) . These bidirectional sensory responses of DA neurons to events that were better or worse than expected led to the formulation of the reward prediction error hypothesis of DA signaling ( Montague et al . , 1996; Schultz , 1998 ) . Subsequent experiments have confirmed that phasic DA responses are sensitive to reward magnitude ( Tobler et al . , 2005 ) , reward probability ( Fiorillo et al . , 2003; Nakahara et al . , 2004; Matsumoto and Hikosaka , 2009 ) and reward delay ( Kobayashi and Schultz , 2008; Fiorillo et al . , 2008 ) . It has been shown that short latency phasic responses can be elicited in DA neurons by unexpected rewards ( Schultz , 1998; Fiorillo , 2013 ) or conditioned stimuli that predict future reward ( Matsumoto and Hikosaka , 2009 , 2007; Eshel et al . , 2015 ) . A critical feature of these early experiments was that the latency of sensory ( usually visually ) elicited DA responses was typically 100 ms or less following stimulus onset . This raised the question of by which route ( s ) is the visual information for reward expectation relayed to DA neurons in the ventral midbrain ( Redgrave et al . , 1999 ) . In a series of investigations , a novel projection from the subcortical midbrain superior colliculus ( SC ) directly to the midbrain DA neurons was demonstrated in rat ( Comoli et al . , 2003 ) , cat ( McHaffie et al . , 2006 ) and monkey ( May et al . , 2009 ) . The SC is an evolutionary archaic visual structure in the vertebrate brain that receives direct input from retinal ganglion cells ( Perry et al . , 1984 ) , and is especially sensitive to unexpected luminance changes ( Boehnke and Munoz , 2008 ) . A later study ( Dommett et al . , 2005 ) confirmed that the retino-tecto-nigral projections were involved in the short-latency phasic activation and release of DA in the basal ganglia following a transient light-flash . However , this investigation was conducted in anaesthetized rodents , and it remains to be determined whether the SC can play a critical role in the short-latency CS-elicited activation of DA neurons and conditioned responses in awake behaving non-human primates . During the evolutionary expansion of the cerebral cortex , the relative importance of the geniculo-striate projection to primary visual cortex ( V1 ) for visual perception increased ( Livingstone and Hubel , 1988 ) . This development offered a further potential route via V1 , by which visual information for reward expectation might be relayed to ventral midbrain DA neurons . Therefore , the specific purpose of the present study was to investigate whether the subcortical visual pathway via the SC can mediate the afferent visual CS signal in the Pavlovian conditioning paradigm and activate DA neurons at short-latency in primates . To do this , we used monkeys that had a unilateral lesion of cortical area , V1 . This preparation in which primary cortical visual processing was disabled was used to isolate the contribution of the SC that remained intact on the V1 lesioned side . After V1 damage , visual awareness is impaired in the lesion-affected visual field ( Cowey and Stoerig , 1995; Yoshida and Isa , 2015 ) . However , from both human ( Poppel et al . , 1973 ) and animal studies ( Cowey and Stoerig , 1995; Yoshida and Isa , 2015 ) it is known that a transient visual stimulus presented in the lesion-affected visual field can trigger a range of behavioural responses , in the apparent absence of subjective awareness . This phenomenon has been called ‘blindsight’ , where many of the residual visual competences are thought to be mediated by the SC ( Mohler and Wurtz , 1977 ) . Consequently , we have made use of animals that were used previously to characterize the phenomenon of ‘blindsight’; they have abilities to make saccadic eye movements to a visual target presented in the lesion-affected visual field ( Yoshida et al . , 2008 ) , despite their awareness to the visual target was impaired like human blindsight patient ( Yoshida and Isa , 2015 ) . This animal model enabled us to test whether the intact subcortical visual circuitry in this preparation can support visual Pavlovian conditioning and short-latency activation of DA neurons ( Schultz , 1998 ) . The purpose of this study was , therefore , to test whether unilaterally V1-lesioned monkeys could associate reward-predicting visual cues with subsequent reward ( Pavlovian conditioning ) , and whether visual CSs could activate midbrain DA neurons . To verify the role of subcortical visual processing , neural activity in the SC was suppressed with local injections of a pharmacological agent .
The right V1 of monkey K and U , and left V1 of monkey T was surgically removed by aspiration , 46 , 44 and 6 months before the present experiments , respectively . The lesion area was confirmed by MR images and the range of the lesion-affected visual field was confirmed by increased threshold for detecting saccadic targets at the beginning of the present experiments ( Figure 1A , Figure 1—figure supplement 1 ) . We presented targets at possible positions which covered the whole contralesional visual field ( monkey K; 3 directions × 3 eccentricities , monkey U; 5 directions × 4 eccentricities , monkey U; 5 directions × 3 eccentricities ) and luminance contrast sensitivity of all targets to induce saccadic eye movements clearly decreased in affected visual field ( Figure 1—figure supplement 1C ) . The visual deficits caused by these lesions was similar to the animals which were reported previously ( Yoshida et al . , 2008 ) . These results indicated that the V1 lesion affected most of the contralesional visual field , at least from 5° to 15° eccentricities . Visual input pathways from retina can be classified into two major pathways; one is cortical pathways via LGN and V1 , the other is subcortical pathways via the SC . The monkeys with unilateral V1 lesion were used to investigate abilities of the subcortical visual pathways through the SC ( Mohler and Wurtz , 1977; Kato et al . , 2011; Takaura et al . , 2011 ) . In this study , the V1 lesion allowed us to assess contribution of visual information via the SC to support visual classical conditioning and to evoke phasic DA responses following the presentation of conditioned stimuli . 10 . 7554/eLife . 24459 . 003Figure 1 . Pavlovian conditioning in V1 lesioned monkeys . ( A ) Left: lesion area ( depicted in gray ) on the whole brain image . Red lines ( 1 - 3 ) indicate dorso-ventral levels of horizontal slices shown on the right . Right: lesion area in monkey K ( depicted in gray ) is overlaid as black areas on axial slices traced from MR images . ( B ) Design of Pavlovian conditioning task in this study . Monkeys were required to fixate a central fixation point ( FP ) until CS offset . LR ( large reward ) and SR ( small reward ) trials were given at random order . In this task , LR was delivered during CS presentation , and SR was delivered after 1 . 5 s from CS offset . Abbreviations; RW ( reward ) . ( C ) Licking rates aligned at the CS onset ( monkey K ) . CSs were presented to intact visual field ( left panel ) and to lesion-affected visual field ( right panel ) . Red and blue lines indicate licking rates during LR and SR trials , respectively . Gray hatched area indicates CS presentation period . Red and blue vertical dashed lines indicate time of reward delivery in the LR and SR trials , respectively . ( D ) Licking rates during CS presentation were compared between LR and SR trials in monkey K ( left ) and U ( right ) . The CSs were presented either to the intact ( int , blue lines ) or lesion-affected ( aff , red lines ) hemifield . * = significant difference ( monkey K: p=6 . 1 × 10−5 ( aff ) , p=4 . 3 × 10−4 ( int ) , monkey U: p=4 . 3 × 10−4 ( aff ) , p=1 . 2 × 10−4 ( int ) , Wilcoxon signed-ranks test , α <0 . 05 ) . ( E ) Licking rates during CS presentation were compared between CS presented to lesion-affected and that to intact visual field in monkey K ( left ) and U ( right ) . There was no significant difference in the licking rates both in LR and SR trials . monkey K: p=0 . 33 ( LR ) , p=0 . 63 ( SL ) , monkey U: p=0 . 16 ( LR ) , p=0 . 084 ( SL ) , two sample t-test with Welch’s correction , α <0 . 05 ) . ( F ) Reversal learning; the effect of switching the CS assignment on licking rates in the intact and affected fields in monkey K . Licking rates during CS presentation to upper ( magenta ) or lower ( green ) visual field were plotted for individual days . CS positions were switched on the day indicated by the vertical red dashed lines . DOI: http://dx . doi . org/10 . 7554/eLife . 24459 . 00310 . 7554/eLife . 24459 . 004Figure 1—figure supplement 1 . Unilateral V1 lesion . ( A ) Locations of the V1 are shown as red area on the horizontal section traces of monkey K . ( B ) Traces of horizontal sections of the three monkeys' brain from their MR images . Their lesion areas are indicated by gray areas on the traces . Right V1 was lesioned in monkey K and U , whereas left V1 was lesioned in monkey T . ( C ) Deficit maps for the three monkeys ( K , U and T ) . Thresholds for detecting luminance contrast ( Michelson contrast ) are plotted over the whole visual field in each monkey with unilateral V1 lesion . The thresholds at individual target positions are displayed with a gray scale . Their sensitivity to luminance contrast was clearly reduced in the lesion-affected visual field . DOI: http://dx . doi . org/10 . 7554/eLife . 24459 . 00410 . 7554/eLife . 24459 . 005Figure 1—figure supplement 2 . Pavlovian conditioning in monkey U . Monkey U also provided a confirmatory dataset in the Pavlovian conditioning task . Arrangement of these figures was the same as Figure 1C and F . ( A ) Licking rates aligned at the CS onset ( monkey U ) . CSs were presented to intact visual field ( left panel ) and to lesion-affected visual field ( right panel ) . Red and blue lines indicate licking rates during LR and SR trials , respectively . Gray hatched area indicates CS presentation period . Red and blue vertical dashed lines indicate time of reward delivery in the LR and SR trials , respectively . ( B ) Reversal learning; the effect of switching the CS assignment on licking rates in the intact and affected fields in monkey U . Licking rates during CS presentation to upper ( magenta ) or lower ( green ) visual field were plotted for individual days . CS positions were switched on the day indicated by the vertical red dashed lines . DOI: http://dx . doi . org/10 . 7554/eLife . 24459 . 005 As a first step we investigated whether monkeys K and U , both with unilateral lesions of V1 , could learn the association between a visual CS and subsequent reward when the CS was presented in the lesion-affected ‘blind’ field ( Figure 1A ) . In this part of the study we presented two visual CSs; one predicted a large reward ( LR = 0 . 17 ml ) delivered during the CS presentation ( 1 . 3 s from CS onset ) , whereas the other predicted a small reward ( SR = 0 . 06 ml ) delivered 1 . 5 s after the CS offset . The two CSs could be discriminated by their location relative to central fixation point ( upper or lower visual field , Figure 1B ) . On separate days the CSs were presented to the lesion-affected or intact visual fields . After 12 days of having the CSs predict juice delivery ( approximately 200 trials/day ) , conditioned anticipatory licking was induced by both LR-CS and SR-CS ( Figure 1C , Figure 1—figure supplement 2A ) . The conditioned licking rate during the CS presentation was significantly higher in LR trials than in SR trials ( 15 sessions in monkey K and 16 sessions in monkey U , Figure 1D , α <0 . 05 , Wilcoxon signed-ranks test ) . In addition , the conditioned responses elicited by CSs presented to either the intact or lesion-affected visual fields were not reliably different ( Figure 1E ) , ( α <0 . 05 , two sample t-test with Welch’s correction ) . These results show that a visual cue presented in the V1 lesion-affected hemi-field can act as an effective CS in a Pavlovian conditioning task . Moreover , the monkeys were able to discriminate successfully between the difference in the magnitude and timing of reward predicted by CSs according to where they were presented in the lesion-affected hemi-field . To test the flexibility of associative learning and to exclude the possibility that the discriminability of the LR- and SR-CSs was simply determined by their respective locations , the upper and lower positions on the screen where the LR-CS and SR-CS appeared were switched ( Figure 1F , Figure 1—figure supplement 2B ) . After the switching , the high conditioned licking rate gradually changed to follow the new LR-CS , again irrespective of whether the CSs were presented in both intact and lesion-affected visual fields ( Figure 1F ) . After the successful reversal , the LR- and SR-CSs were switched back to their original assignment . At which point the conditioned responses switched back to follow the newly assigned LR-CS . These results indicate that monkeys can flexibly associate the locations of the visual CSs and the reward predicted by them even without V1 . To investigate whether visual processing in the SC was responsible for the expression of visually-evoked conditioned responses when the CSs were presented to the V1 lesion affected side , the GABA agonist muscimol ( 0 . 5 μL; 1 μg/μL concentration at a rate of 1 μL/15 s ) was injected into the ipsi-lesional SC of monkeys K and T . Thus , before the muscimol injection , neural activity of the SC was recorded , and the location of neurons responsive to LR-CS was identified on SCs retinotopic map . Muscimol was then injected into this location ( Figure 2A ) . The suppressive effect of the muscimol injection was confirmed by showing that the monkey failed to make saccades to the LR-CS location as previously shown for the blindsight monkeys by Kato et al . ( 2011 ) ( Figure 2B; see disappearance of saccades to the left-upward target ) . 10 . 7554/eLife . 24459 . 006Figure 2 . Effect of SC inactivation on conditioned behaviors . ( A ) A scheme of the SC inactivation experiments . Muscimol was injected into the point on the ipsi-lesional SC map representing the location of LR-CS in the visual field . ( B ) End points of saccadic eye movements before and after the SC inactivation ( left and right panel ) . The position of central fixation point is indicated by a blue cross . Circles indicate end points of visually guided saccades , and their colors indicate location of saccadic targets in individual quadrants . Impairment of saccades toward the upper-left target ( green ) indicates that muscimol effectively suppressed the neuronal activity at the injection site . ( C ) Licking rates in a daily session before ( left panel ) and after SC inactivation ( right panel ) in monkey T . The licking rates are plotted in the same manner as Figure 1C . Red and blue lines indicate the licking rates during the LR and SR trials , respectively . Gray hatched area indicates the CS presentation period . ( D ) Licking rate during 0 . 7 s from the CS onset in the SR trials are subtracted from licking rate in LR trials in monkey K ( blue line , N = 9 ) and T ( red line , N = 4 ) . The vertical lines indicate the SEM . Bef . : before inactivation , Dur: during inactivation . ( p=2 . 4 × 10−4 , Wilcoxon signed-ranks test , α <0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24459 . 00610 . 7554/eLife . 24459 . 007Figure 2—figure supplement 1 . Effect of SC inactivation on conditioned behavior in monkey K . Licking rates in a daily session before ( left panel ) and after SC inactivation ( right panel ) in monkey K . Monkey K also provided a confirmatory dataset in the Pavlovian conditioning task before ( left panel ) and during ( right panel ) the SC inactivation . Arrangement of these figures was the same as Figure 2C . Red and blue lines indicate the licking rates during the LR and SR trials , respectively . Gray hatched area indicates the CS presentation period . DOI: http://dx . doi . org/10 . 7554/eLife . 24459 . 007 Also , before the muscimol injection , anticipatory licking evoked by the LR-CS presentation ( 0–0 . 7 or 1 . 3 ms ) served as a baseline control in our Pavlovian conditioning task ( Figure 2C left ) . Immediately following the muscimol injection the monkeys continued to perform the LR-CS evoked conditioned anticipatory licking . However , over the next 20–30 min the normal conditioned response ( anticipatory licking ) gradually disappeared ( Figure 2C right ) . At which point , two new patterns of behaviour were observed: ( i ) in the case of monkey T ( Figure 2C right ) , all anticipatory response was abolished and licking appeared only after the juice reward was delivered; and ( ii ) for monkey K anticipatory licking was evoked shortly after the onset of both the LR-CS and SR-CS ( Figure 2—figure supplement 1 . In other words , the animal’s ability to discriminate between the CSs on the basis of position within the visual field was lost . Muscimol injections were administered in 13 experiments ( monkey K: 9 experiments , monkey T: 4 experiments ) . To assess the effect of the SC inactivation the difference between the licking rate during CS presentation was compared for LR and SR trials . Before the SC inactivation ( control ) , monkeys licked a reward spout more frequently during CS period in LR trials than in SR trials in all sessions . The difference of the licking rate between in LR trials and in SR trials was diminished after SC inactivation ( Wilcoxon signed-ranks test; p<0 . 001 ) . During the SC inactivation , the difference of licking rate was not significantly different from zero ( one-sample t-test; p>0 . 005 ) . The results were consistent in all sessions of both monkeys ( Figure 2D ) . These results indicate that the visual processing signifying CS onset by the SC on the V1 lesion-affected side was essential for a previously established conditioned response to be expressed in our Pavlovian conditioning task . It has been reported widely that dopamine neurons are phasically activated by unpredicted conditioned stimuli in Pavlovian tasks ( Schultz , 1998 ) . The purpose of the next phase of our study was , therefore , to investigate whether a visual CS presented to the V1 lesion-affected visual field had the capacity to evoke a phasic response in ipsilateral DA neurons in the current Pavlovian conditioning task . Monkeys K and T were used for these experiments . Neurons conforming to the electrophysiological criteria established for identifying putative DA neurons were recorded in the ventral midbrain . The neurons included in our sample therefore had low baseline firing rates ( <10 Hz ) , and broad spike-widths ( >0 . 45 ms between the first negative peak and next positive peak ) ( Figure 3B , C ) . The location of recorded neurons was later confirmed by identifying the site of small lesions made at some of the recording sites in tissue immunostained for tyrosine hydroxylase ( Figure 3D ) . 10 . 7554/eLife . 24459 . 008Figure 3 . DA neuron responses during Pavlovian conditioning task . ( A ) Schematic drawing of the experimental design for recording DA neuron activity in the monkey with unilateral V1 lesion . ( B ) Averaged spike waveforms of a presumed DA neuron in SNc and a non-DA neuron in the SNr . Amplitude of these spikes are normalized . Spike width was defined as the time between the first negative peak and second positive peak . ( C ) Histogram of the spike width . Red bars indicate the DA neurons and blue bars indicates the SNr neurons . ( D ) Left; a low magnification view of the SNc and surrounding structures stained with anti-TH immunohistochemistry . Scale bar = 5 . 0 mm . Right; a high magnification view of the area indicated by a blue square . Red arrows indicate locations of electrolytic markings . Scale bar = 2 . 0 mm . ( E ) Time course of the Pavlovian conditioning task ( the same as Figure 1B ) . ( F ) A typical DA neuron activity in V1 lesioned monkeys . Raster plots of a DA neuron from LR ( red ) and SR ( blue ) trials were sorted and shown on the top , receptively . The first trial was plotted at the bottom of the raster plot and the last trial was plotted at the top . Red and blue lines indicate average firing rates during LR and SR trials , respectively . These plots were aligned at the FP onset , CS onset , and RW delivery ( left , middle and right panels , respectively ) . ( G ) Responses of all recorded DA neurons to FP , CS and RW ( left , middle and right panels ) are superimposed . A thick red line in each panel is the averaged firing rate of DA neurons in LR trials , and a thick blue line is the averaged firing rate in SR trials . Thin lines behind the averaged lines are the averaged responses of individual neurons in LR trials ( red ) and in SR trials ( blue ) , respectively . ( H ) Firing rates of individual DA neurons within the time windows ( 100–300 ms from FP and CS or 150–350 ms from RW; left , middle and right panels ) . Blue lines indicate the average of all the neurons and SD of the firing rate in LR trials and in SR trials . * = significant difference ( N = 24 , p=0 . 82 ( FP ) , p=1 . 1 × 10−7 ( CS ) , p=0 . 27 ( RW ) , Wilcoxon signed-ranks test , α <0 . 05 ) . ( I ) The yellow background in the figures shows the period during which the responses to LR-CS and SR-CS were significantly different more than 15 ms ( N = 24 in affected , N = 16 in intact , two-sided sign test , α <0 . 05 ) . The two panels show averaged DA responses to CSs presented to the lesion-affected visual field ( upper panel ) , and to the visual field ( lower panel ) . Arrows under each figure indicate the earliest points where the LR and SR responses can be reliably discriminated for more than 50 ms ( 122 ms in the lesion-affected visual field , and 112 ms in intact visual field ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24459 . 00810 . 7554/eLife . 24459 . 009Figure 3—figure supplement 1 . Comparing DA responses to CSs in lesion-affected and intact visual field . These figures show firing rate of DA response to CS presented into lesion-affected and into intact visual field . Responses to LR-CS were compaired in A , and to SR-CS were in B . Time windows size to calicurate the firing rate was 100–300 ms from CS onset . In both cases , there are no significant difference ( N = 16 , p=0 . 958 ( LR-CS ) , p=0 . 796 ( SR-CS ) , one sample t-test , α <0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24459 . 009 Typical responses of putative DA neurons in our Pavlovian task are shown for a single case ( Figure 3F ) , and for the population of recorded neurons ( n = 24 ) ( Figure 3G ) . First , because of its task-relevance and unpredictability , putative DA neurons were activated robustly by the onset of the fixation point . However , this response was similar in LR trials and SR trials ( left-hand panels of Figure 3F and G ) because at the time the fixation point was presented the magnitude and timing of reward predicted by the upcoming CS was unknown . Subsequently , when the temporally uncertain CSs were presented , a clear difference in the putative dopamine response was evident between the LR and SR trials – a reliably larger response was evoked by the LR-CS ( central panels of Figure 3F and G ) . In this case , responses to predicted presentations of the juice reward were unreliable and significantly weaker than responses evoked either by the FP or CSs ( right-hand panels of Figure 3F and G ) . Confirmation of the above findings for the population of DA neurons ( n = 24 ) is illustrated in Figure 3H . In these figures , firing rate of these responses in a selected time window ( FP , CS: 0 . 1 s – 0 . 3 s from the onset , RW: 0 . 15 s – 0 . 35 s from the delivery ) was compared between the LR and SR trials . The left-hand panel shows that there was no reliable difference between the putative dopamine responses evoked by FP presentation in LR and SR trials ( Wilcoxon signed-ranks test ) . However , the LR-CS elicited a significantly larger responses compared with those evoked by the SR-CS ( central panel Figure 3H ) . These responses were not strongly affected by the V1 lesion . Firing rate of the responses to CSs presented into lesion affected and intact visual fields were not significantly different ( Figure 3—figure supplement 1 ) . Finally , there were no reliable differences in the responses evoked by the onset of the predicted LR or SR ( right-hand panel Figure 3H ) . The overall mean response latency was 107 ms while the latencies of the individual neurons were distributed between 60 to 160 ms after the LR-CS onset ( latency = the time when the neural response rate exceeded 2SD of their baseline activity ) . We calculated the earliest time points when difference between responses to LR-CS and SR-CS was observed . The earliest time points when response differentiation lasting more than 15 ms started was 122 ms from the CS onset in lesion-affected visual field , and 112 ms in intact visual field ( Figure 3I ) . This result indicates that the latency of the reward discrimination by DA neurons was minimally affected by the absence of V1 . These results showed in the absence of V1 , that temporally unpredicted visual CSs were able to elicit typical short latency and short duration phasic responses in ventral midbrain neurons , presumed to be dopaminergic . These neurons could discriminate the LR-CS and SR-CS , based on the location of their presentation within the lesion-affected visual field . These results indicate that the residual early visual structures ( most likely the midbrain SC ) retained the capacity to evoke differential phasic DA responses informed by the reward expected from CS . The final phase of our study sought to test the contribution of the SC . To test whether the transmission of visual signals via the SC was responsible for CS-evoked phasic DA responses , muscimol was injected into the ipsi-lesional SC ( Figure 4E ) . Thus , after the collection of control data on visually guided saccadic task and on Pavlovian conditioning task , baseline records of the responses of the DA neurons to the presentation of the fixation point , CS and reward were recorded . When all was done , muscimol was injected into the appropriate location of the SC ( see above ) and DA responses to the same sensory events were reassessed . Thus , the activity of a single DA neuron was recorded both before and after the muscimol injection . To ensure that the same recorded neuron was maintained throughout the session ( that is , for approximately 1 . 5 hr ) , its waveform was carefully monitored . Only when the DA waveforms remained constant before , after muscimol injection were the data included in our sample ( Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 24459 . 010Figure 4 . Effect of SC inactivation on cue-responses in DA neurons . ( A ) Activity of DA neurons before SC inactivation . Raster plots and firing rates plotted in the same manner as Figure 3F . These plots were aligned at FP onset , at CS onset , and at RW delivery ( left , middle and right panels , respectively ) . ( B ) Activity of DA neurons during SC inactivation After the SC inactivation , the responses to the FP were unchanged ( left ) , those to the LR-CS ( middle ) disappeared and those to RW ( right ) increased . ( C ) Population average of DA neuron responses ( N = 5 ) in LR trials before ( green ) and during SC inactivation ( magenta ) . These activities were aligned at FP onset , at CS onset and at RW delivery , respectively ( left , middle and right panels ) . ( D ) Firing rates of DA neurons in LR trials within different time windows ( 100–300 ms from FP and CS or 150–350 ms from RW; left , middle and right panels , respectively ) before and during SC inactivation . These time windows are the same as those in Figure 3H . * = significant difference ( N = 5 , p=0 . 067 ( FP ) , p=0 . 0025 ( CS ) , p=0 . 043 ( RW ) , one sample t-test , α <0 . 05 ) . ( E ) A schematic drawing of the experimental setup for the DA neuron recording and SC inactivation . Ipsi-lesional SC was inactivated . The neural activity was recorded from the ipsi-lesional SNc . DOI: http://dx . doi . org/10 . 7554/eLife . 24459 . 01010 . 7554/eLife . 24459 . 011Figure 4—figure supplement 1 . Spike waveforms of a DA neuron during a daily session . Comparing the spike waveform of a presumed DA neuron ( 1 ) before ( black ) , and ( 2 ) soon after muscimol injection ( blue ) and ( 3 ) at the end of recording ( green ) . Averaged spike waveforms obtained from individual time periods indicated by the three dotted squares with corresponding colors on the top . The spike waveforms did not appear to significantly change through the recording session . DOI: http://dx . doi . org/10 . 7554/eLife . 24459 . 01110 . 7554/eLife . 24459 . 012Figure 4—figure supplement 2 . firing rate of responses to SR-CS . These figures show firing rate of response to SR-CS before and after muscimol injection ( A ) and difference of firing rate between responses to LR-CS and to SR-CS during the SC inactivation ( time windows: 100–300 ms from CS onset ) . In both cases , there are no significant difference ( N = 5 , p=0 . 608 ( SR-CS ) , p=0 . 625 ( SC inactivation ) , one sample t-test , α <0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24459 . 012 The responses of a typical DA neuron are illustrated in Figure 4A and B . Before collicular inactivation ( Figure 4A ) the DA responses to the task-related stimuli were similar to those observed in previous experiments ( see above – Figure 3F and G ) . After the injection of muscimol , when the relevant SC was inactivated , the robust response evoked by the FP was largely unaffected ( compare Figure 4A and B ( left-hand panels ) , Figure 4C left and 4D left , Wilcoxon test , not significantly different ) . After the muscimol injection the response of the recorded neuron to presentation of the LR-CS was retained for a short while ( central panels Figure 4B ) . However , after a few trials the drug action became apparent , and the CS-evoked response was almost completely abolished ( central panel Figure 4B and C ) . It is also significant that in these early trials , when the reward delivery was still predicted by visual input from the SC , reward presentation failed to evoke a phasic DA response . However , as the colliculus became fully inhibited , the now unpredicted presentation of the reward evoked a robust phasic response , which in this case was clearly dependent on the magnitude and timing of reward predicted by the CS . This pattern of response was consistent in all recorded neurons ( Figure 4D ) . Also , for most of the recorded neurons , reward responses emerged as the inactivation progressed ( right panels of Figure 4B , C and D ) . In SR trials , firing rate to CS was unaltered by the injection . During the SC inactivation , DA responses evoked by the CS were not significantly different between LR and SR trials ( Figure 4—figure supplement 2 ) . Together , these results confirm that , in the absence of V1 , visual signals signifying CS onset , with the capacity to elicit a short latency phasic response in presumed DA neurons , are most likely to be relayed via the direct retino-tecto-nigral projection ( Comoli et al . , 2003; Dommett et al . , 2005 ) , although an indirect contribution , possibly involving the pedunculopontine nucleus cannot be ruled out at present ( Harting , 1977; Redgrave et al . , 1987 ) ; Kobayashi and Okada , 2007 ) .
Many studies have indicated that midbrain DA neurons causally contribute to reinforcement learning . For example , when reward expectation signals from DA neurons were impaired by D1 receptor blocker or when NMDA receptors were knocked out in DA receptor expressing neurons in various brain areas , conditioned response was impaired in many kinds of behavioral learning tasks ( Di Ciano et al . , 2001; Flagel et al . , 2011; Parker et al . , 2010 , 2011; Puig and Miller , 2012; Berridge and Robinson , 1998 ) . Alternatively , when DA neurons or neurons expressing D1 receptors were activated by electrical or optogenetical stimulation , various forms of conditioned behaviour were induced ( Olds and Milner , 1954; Adamantidis et al . , 2011; Ilango et al . , 2014; Steinberg et al . , 2013; Kravitz et al . , 2012 ) . Thus , such involvement of dopaminergic transmission or DA neuron activity in learning has been well studied , however , it remains unclear how DA neurons are able to signal the value or salience of unpredicted objects or events at short-latency . It has been proposed that the early phasic responses of DA neurons have two separable components; an early non-selective sensory response that represents temporal salient-event prediction errors , and a second component that codes the object/event’s reward value ( Joshua et al . , 2009; Bromberg-Martin et al . , 2010; Schultz , 2016 ) . This view immediately provokes the question of what early afferent visual processing could allow the DA neurons to respond in this fashion to conditioned visual stimuli ( the sensory modality that is most frequently used ) ? Following the onset of a visual CS response latencies in V1 are typically in the range 40–60 ms , while in the inferotemporal cortex where objects/events are identified they are slower in the range 80–100 ms ( Thorpe and Fabre-Thorpe , 2001 ) . Moreover , since there are no obvious direct connections to the ventral midbrain , the results of cortical visual processing are likely to be relayed via additional time consuming indirect routes . On the other hand , response latencies in the retino-recipient midbrain SC are significantly less ( 40–50 ms ) and there is a direct tectonigral projection to substantia nigra pars compacta ( Comoli et al . , 2003; McHaffie et al . , 2006; May et al . , 2009 ) . It is probable , therefore , that the earliest sensory component of the phasic DA response ( 70–150 ms ) is mediated via subcortical visual processing involving the SC ( Comoli et al . , 2003; Dommett et al . , 2005 ) . Two versions of the two-visual system hypothesis as an explanation for the bimodal characteristic of short latency phasic DA responses to visual CSs have been presented ( Joshua et al . , 2009; Bromberg-Martin et al . , 2010; Schultz , 2016; Redgrave et al . , 2017 ) . The first is that the initial component of the phasic DA response is a non-selective salience signal that represents a temporal salient-event prediction error ( Joshua et al 2009 , Bromberg-Martin et al . , 2010; Schultz , 2016 ) . The second phasic component is value-coded and takes longer to compute because the unexpected event needs to be identified before its value is known . Stimulus identification frequently requires stimulus detection , foveation and cortical analysis of geometric form , colour , texture , and apparent motion , in various permutations and combinations ( Nomoto et al . , 2010 ) . However , in the case of simple stimuli ( for example , luminance change at different spatial locations ) it is suggested that the non-selective salience and value components can merge to a near unimodal response that , in some cases , can be separated by sophisticated mathematical analysis ( Fiorillo et al . , 2013 ) . This version suggests that for both subcortical salience and cortical stimulus identification the early sensory responses have to be relayed through an unspecified ‘value-decoder’ that communicates with DA neurons , thereby enabling them to report reward prediction errors ( Schultz , 2016 ) . What is the likely location of the hypothesized ‘value decoder’ ? Uchida and colleagues recently identified all the brain regions which project to DA neurons in rodents . They report afferent connections from the striatum , amygdala , subthalamic nucleus , pedunculopontine nucleus , rostromedial reticular nucleus , and GABAergic neurons in the substantia nigra pars reticulata ( Watabe-Uchida et al . , 2012 ) . Consequently , there are many possible locations that receive input from primary visual structures , compute stimulus value and communicate this to DA neurons in the ventral midbrain . These indirect routes of communication can offer a perfectly reasonable explanation for the value coding of the second delayed component of the early phasic DA response . However , it is important to note that the earliest component ( 70–150 ms ) of phasic DA response is not always best described as a value insensitive salience signal . Both the present results ( where cortical visual processing is impaired ) , and earlier studies of Schultz and his colleagues involving intact monkeys ( Tobler et al . , 2005; Fiorillo , 2013 ) report that when CSs can be discriminated on the basis of luminance change at different locations ( a subcortical collicular visual competence – Boehnke and Munoz , 2008 ) , the phasic DA response latencies are frequently around 100 ms ( pre-gaze shift ) , unimodal and clearly code the predictive value of the CS . So how is it possible for unimodal phasic DA responses ( for example , Figure 1B – Tobler et al . , 2005 ) to code value at such short latencies ? Visual response latencies in intermediary structures identified above are too long ( typically >100 ms ) to account for value coding of a unimodal phasic DA response that peaks at about 100 ms . A second , rather simpler version of two-visual system hypothesis can explain value-coding of both components of the early phasic DA response ( Redgrave et al . , 2017 ) . The proposal is that the predictive value of a visual CS may already be encoded in the early sensory response of both the cortical and subcortical early visual systems . For example , there are many papers that demonstrate that an association with , or an expectation of reward can dramatically influence the magnitude of the initial sensory response in early sensory areas throughout the brain ( Mogami and Tanaka , 2006; Serences and Saproo , 2010; Metzger et al . , 2006; Leathers and Olson , 2012 ) , including the SC ( Ikeda and Hikosaka , 2003 ) . The most parsimonious explanation of how the earliest responses of DA neurons can be value-coded is , therefore , that they receive input from the SC that has been already value-coded through a classically conditioned process of sensory pre-tuning of the CS value in early sensory structures ( Ikeda and Hikosaka , 2003 ) . Thus , in our study and those of others , stimuli are conditioned by Pavlovian association with different levels/probabilities of reward , prior to the recording of DA neurons ( Fiorillo et al . , 2003; Tobler et al . , 2005; Matsumoto and Hikosaka , 2009 ) . The likely effect of this process would be to tune the initial sensory responses in early visual structures to reflect the reward predicted by the CS . According to this suggestion , if the object/event prediction error detected in early visual structures has been value-coded by prior Pavlovian association , the event prediction error would also be a reward prediction error . In the case of the SC , if a value-coded signal evoked by a CS was relayed to the DA neurons via the tectonigral projection ( Comoli et al . , 2003; Dommett et al . , 2005; McHaffie et al . , 2006; May et al . , 2009 ) , it would explain how DA neurons can signal reward prediction errors with latencies in the range 70–150 ms ( present study and Fiorillo et al . , 2003; Tobler et al . , 2005 ) . On the other hand , in the case of complex CSs that are presented at the same location , or randomly at different locations , the SC would certainly detect the luminance change associated with CS onset , ( Boehnke and Munoz , 2008 ) . However , because subcortical sensory processing cannot perform complex CS discriminations ( Boehnke and Munoz , 2008 ) , this onset response will not be value-coded , which might explain why , with complex CSs , the initial sensory component of the DA phasic response is a non-selective salient-event prediction error . A possible explanation of the second value-coded component of the phasic DA response could be that the cortical processing responsible for object/event identification is equally subject to Pavlovian pre-tuning ( Mogami and Tanaka , 2006; Serences and Saproo , 2010; Weil et al . , 2010 ) . It is well known that there are two kinds of DA responses; one is sensitive to the value of future events , and the other is sensitive to their salience ( Matsumoto and Hikosaka , 2008; Lerner et al . , 2015; Menegas et al . , 2017 ) . In the context of the present study , we are unable to tell whether our DA responses reflected value or salience , because we used only reward associated CSs . To confirm which kinds of DA responses are elicited thorogh the subcortical visual processing , we have to conduct another experiments using aversive stimuli . However , at least , we could demonstrate that DA neurons could differentiate either reward value or salience with the visual information mediated by the SC .
Three adult Japanese monkeys ( Macaca fuscata; all female , body weight 5–7 kg , monkey K , U and T ) were used in this study . Details of the procedures for training and surgery of the monkeys have been described in previous reports ( Yoshida et al . , 2008; Kato et al . , 2011 ) . Briefly , under isoflurane anesthesia ( 1 . 0–1 . 5% ) , the monkeys were implanted with a holder with which the head was stabilized during the behavioural and electrophysiological experiments . The monkeys were allowed to recover for more than two weeks after surgery before pre-lesion training . All the experimental procedures were performed in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and approved by the Committee for Animal Experiment at the National Institute of Natural Sciences . The right V1 of monkey K and U , and left V1 of monkey T were surgically removed by aspiration under isoflurane anesthesia ( 1 . 0–1 . 5% ) ( see Yoshida et al . , 2008 ) . The surgical operation was conducted before 46 months ( monkey K ) , 44 months ( monkey U ) , and six months ( monkey T ) from days when their training in this study was started . The opercular surface of the striate cortex and medial area in the Calcarine Sulcus was removed , while the ventrolateral part of the opercular surface , which encodes foveal vision ( visual field for eccentricity 0 to 1 . 0° ) remained intact ( Figure 1A , Figure 1–figure supplement 1A and B ) . Prior to the surgery , animals were trained on a visually guided saccadic eye movement task . Their ability to respond to visual stimuli was assessed both before and after the V1 lesion . A monitor ( Diamondcrysta WIDE RDT272WX ( BK ) , MITSUBISHI ) was positioned 34 . 5 cm in front of the monkeys’ face . A real-time experimental control system ( Tempo for Windows , Reflective Computing; http://reflectivecomputing . com/ ) was used for stimulus presentation and data collection . In this task , fixation point ( FP ) initially appeared at the center of monitor screen . Monkeys were required to maintain fixation in a window centered on the FP ( size , 2 . 5° radius ) for 1 . 6–2 . 0 s . A second target visual stimulus ( 0 . 6° ) was then presented randomly at one of five possible locations in the hemi-visual field for two monkeys ( monkey U and T ) and one of three possible locations in visual hemifield for one monkey ( monkey K ) ( Figure 1—figure supplement 1C ) . When the target appeared , the FP was extinguished and monkeys were required to make a saccade to the peripheral visual target . A window surrounding the target was a circle with a radius of half the distance between each target location ( radius = eccentricity × sin ( direction angle between neighboring target positions ) ⁄ 2 ) . This arrangement prevented the targets to overlap with each other . Target luminance Michelson contrast was 0 . 87–0 . 94 ( 13 . 4–31 . 3 Weber contrast ) on a background of 1 . 0 cd ⁄ m2 . Reward was delivered if monkeys made a correct saccade to the target within 1 s after target presentation and maintaining fixation within the target window ( 3 . 2° radius ) for 600 ms . Eye movements were measured with a video-based eye tracker ( EYE-TRAC 6; Applied Science Laboratories , sampling rate: 240 Hz ) . All statistical analysis in this study were performed on Matlab ( RRID:SCR_001622 ) . Details of the methods for calculations to construct the deficit map in these animals have been described previously ( Yoshida et al . , 2008 ) . Luminance contrast of the targets was varied randomly trial-by-trial ( 0 . 02 to 0 . 9 as expressed in Michelson contrast ( Weber contrast 0 . 04–18 . 0 ) ) . For this test , saccades landing in an area within a circle with a radius of half the distance between each target location ( radius = eccentricity × sin ( direction angle between neighboring target positions/2 ) ; 15° for monkey U and T , and 22 . 5° for monkey K ) were counted as correct responses . The sensitivity of luminance contrast was defined as that representing the percentage of correct responses corresponding to the sensitivity value d’ = 2 ( threshold for luminance contrast ) and deficit maps of individual monkeys were constructed with these values ( Figure 1—figure supplement 1C ) . In general , the visual field disrupted by the lesion site extended from eccentricities about 5–20° in the monkeys used in this study . The luminance contrast and CS size were retained from previous studies that investigated visual responses of V1 neurons to stimuli presented in the natural blind spot . Our previous study also precluded the possibility of stray-light affecting the results in the present experimental environment by demonstrating the absence of a saccadic response to visual stimuli presented in the natural blind spot . The present Pavlovian conditioning experiments were initiated 46 , 44 and 6 months after the V1 lesions in monkey K , U and T , respectively . The task sequence of the Pavlovian conditioning paradigm used in the present study is illustrated in Figure 1b . Conditioned stimuli ( CS ) ( 2 . 2° red square , luminance contrast: Michelson contrast 0 . 87 ( Weber contrast 13 . 4 ) against the background of 1 . 0 cd ⁄ m2 ) were presented in either the upper ( eccentricity: 10° , direction: 45° relative to the horizontal axis from the FP ) or lower quadrant ( eccentricity: 10° , direction: −45° relative to the horizontal axis from the central FP ) of the lesion-affected or intact visual hemifield . Experiments involving CS presentation to either the lesion-affected or intact visual hemifield were conducted on separate days . At the beginning of each trial , a fixation point ( FP ) appeared at the center of monitor . After a 0 . 7 to 1 . 2 s fixation period , a CS predicting a large reward ( LR-CS ) or a CS predicting a small reward ( SR-CS ) was presented for 1 . 0 or 1 . 7 s . The two CSs were pseudo-randomly alternated within a daily session . Throughout the task , monkeys were required to maintain their gaze on the central FP to assure that CS presentation was either to the lesion-affected , or intact visual hemi-field . If fixation was broken , the trial was terminated immediately . The conditioned response ( CR ) in this task was the anticipatory licking elicited by the CS presentation that occurred prior to the juice delivery . The CR was measured by detecting electric contact between the monkey and the reward tube or by a photo-detector in experiments involving electrophysiological recording . A lick was recorded when the monkeys’ tongue was observed to approach the reward spout . To quantify the conditioned response elicited by the visual CS , the number of licking responses detected during the cue presentation ( 0 to 1 . 3 s ) was counted in 0 . 1 s time bins in 14–16 sessions for each hemifield of each monkey . The frequency of licking ( licking rate ) was compared to a baseline frequency during the 1 s period ( −1 to 0 s ) before the CS onset ( one-tailed paired t test , significant level at p<0 . 05 ) . A principal aim of the study was to record from single DA neurons while the monkeys were engaged in the Pavlovian conditioning task . This was achieved using epoxylite-coated tungsten microelectrode ( impedance: 9–10 MΩ at 1 kHz , FHC ) . Voltage recording were bandpass-filtered between 0 . 1 ( or 0 . 3 ) and 10 kHz . Standard criteria were used for identification of putative DA neurons ( Ungless et al . , 2004 ) . First , the location of SNc and the VTA were estimated from MR images taken in advance . After having isolated a single neuron in the appropriate region , we tested whether the presentation of an unpredicted reward would cause a response . Two criteria to confirm the likelihood that we were recording from a DA neuron; ( 1 ) it had a low baseline activity between 1 . 0–10 . 0 Hz ( Schultz and Romo , 1987; Matsumoto and Hikosaka , 2009 ) ; and ( 2 ) the neuron had a spike width , which was clearly longer than those of nearby neurons in the substantia nigra pars reticulata ( SNr ) that had rates of baseline firing in excess of 40 Hz ( Ungless et al . , 2004; Matsumoto and Takada , 2013 ) . To determine the role of the residual subcortical visual circuit in eliciting conditioned responses in the Pavlovian task and CS-evoked responses in DA neurons we conducted experiments in which the SC on the V1 lesion-affected side was inactivated . In a previous study with these subjects ( Kato et al . , 2011 ) reported that the monkeys were unable to make saccades to parts of the visual field injected locally with the gamma aminobutyric acid A ( GABAA ) receptor agonist , muscimol . In our experiments we used additional single unit electrophysiology to locate the response field of the SC neurons responsive to the LR-CS . At these sites muscimol ( 0 . 5 μg in 0 . 5 μL ) was pressure-injected ( 0 . 4 μL/min ) using a 10 μL Hamilton syringe ( Hamilton Company , Reno , Nevada , USA ) mounted in a syringe pump . Conditioned response was measured both before and during inactivation of the SC . In some experiments we recorded the activity of presumed DA neurons while the animals were performing the Pavlovian task . Then , the SC was injected with muscimol . After recording DA activity for about 60 CS presentations , muscimol was injected into the SC while recording from the same neuron was maintained . In some sessions , post-injection trials started immediately after the injection , while in others they started 10 to 20 min after the injection . After all behavioural testing and electrophysiological recording had been completed with monkey K , two small electrolytic lesions were made in each recording track ( 20 mA , 30 s ) . The animal was then euthanized and coronal sections ( 40 µm ) of tissue that included SNc were immunostained for tyrosine hydroxylase ( TH ) to reveal the location of DA neurons ( Figure 3D ) . ( RRID:AB_390204 for the antibody ) | To survive and thrive , animals must learn to approach cues in their environment that are likely to lead to a desirable outcome and avoid those that might lead them to harm . A group of brain regions known as the midbrain dopamine system helps many animals to achieve this . Dopamine is the brain’s reward signal . Cues that predict rewards , such as the sight or smell of food , activate midbrain dopamine neurons . However , the details of this process remained unclear . Takakuwa et al . have now examined how visual information that signals reward reaches the midbrain dopamine neurons . The anatomy of the visual system suggests two main possibilities . Information may travel directly from the eyes to an area of the midbrain called the superior colliculus , and then onto the dopamine neurons . Alternatively , information may travel to the midbrain indirectly via a pathway that includes additional processing in the brain’s outer layer , the visual cortex . To distinguish between these routes , Takakuwa et al . studied monkeys in which the indirect pathway via the visual cortex had been damaged . Some people with damage to this pathway have a disorder called blindsight . They are able to detect the movement or location of stimuli , but they cannot consciously see those stimuli . The monkeys with damage to visual cortex were able to learn that an image on a screen predicted the delivery of fruit juice . After repeated trials , the monkeys began to lick the spout dispensing the juice whenever the image appeared , even if no juice was delivered . The monkeys’ midbrain dopamine neurons also sent more signals in response to the images , and showed greater activity when the images predicted large rewards than small ones . Takakuwa et al . next inactivated the superior colliculus with a drug and showed that this prevented both the licking behavior and the increased signaling . Together the findings show that visual information about potential rewards can reach midbrain dopamine neurons via a direct route through the superior colliculus , without needing to pass via the visual cortex . The next step is to determine how and when the visual cortex may get involved in this process to help animals maximize rewards . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Emergence of visually-evoked reward expectation signals in dopamine neurons via the superior colliculus in V1 lesioned monkeys |
Anti-malarial pre-erythrocytic vaccines ( PEV ) target transmission by inhibiting human infection but are currently partially protective . It has been posited , but never demonstrated , that co-administering transmission-blocking vaccines ( TBV ) would enhance malaria control . We hypothesized a mechanism that TBV could reduce parasite density in the mosquito salivary glands , thereby enhancing PEV efficacy . This was tested using a multigenerational population assay , passaging Plasmodium berghei to Anopheles stephensi mosquitoes . A combined efficacy of 90 . 8% ( 86 . 7–94 . 2% ) was observed in the PEV +TBV antibody group , higher than the estimated efficacy of 83 . 3% ( 95% CrI 79 . 1–87 . 0% ) if the two antibodies acted independently . Higher PEV efficacy at lower mosquito parasite loads was observed , comprising the first direct evidence that co-administering anti-sporozoite and anti-transmission interventions act synergistically , enhancing PEV efficacy across a range of TBV doses and transmission intensities . Combining partially effective vaccines of differing anti-parasitic classes is a pragmatic , powerful way to accelerate malaria elimination efforts .
Malaria remains a major global health challenge with an estimated 216 million new cases and 445 , 000 deaths in 2016 ( World Health Organization , 2017 ) . Whilst current tools have substantially reduced the global burden of disease , new tools will be needed to achieve malaria elimination ( Walker et al . , 2016 ) . Early development of malaria vaccines focused on either the pre-erythrocytic stage vaccine ( PEV ) – eliciting an immune response to prevent incoming sporozoites from establishing patent infection – or blood-stage – boosting natural responses to surface proteins on the infected erythrocytes ( Schwartz et al . , 2012 ) . The first malaria vaccine RTS , S/AS01 to complete Phase III trials is a PEV vaccine and has been demonstrated to be partially effective , reducing clinical incidence in 5 – 17-month-old children by 36 . 3% ( 95%CI: 31 . 8 – 40 . 5% ) over 40 months follow-up ( RTSS Clinical Trials Partnership , 2015 ) . Further candidate PEV vaccines include those that achieve protective immunity through irradiated/chemo-attenuated Plasmodium falciparum sporozoites , ( e . g . PfSPZ vaccines [Seder et al . , 2013] ) , those that use viral vectors to induce T-cell responses to provide protection ( de Barra et al . , 2014; MVVC group et al . , 2015 ) and the promising next-generation RTS , S-like vaccine , R21 ( Collins et al . , 2017 ) . A range of vaccines that target human-to-mosquito transmission by attacking sexual , sporogonic , and/or mosquito antigens ( transmission-blocking vaccines , TBV ) are also under development ( Talaat et al . , 2016; Wu et al . , 2008 ) . Pre-clinical investigations have identified multiple antigens ( e . g . Pfs25 , P230 , P48/45 ) as targets for TBV candidates that , when administered , can reduce transmission to mosquitoes ( Hoffman et al . , 2015; Sauerwein and Richie , 2015 ) , but complete , or reproducible , translation to the clinic has not been achieved so far ( Talaat et al . , 2016; Hoffman et al . , 2015; Sauerwein and Richie , 2015 ) . One of the major challenges encountered in developing PEV malaria vaccines is the partial protection achieved against each exposure , despite high levels of induced antibody titres . It has been hypothesized that this may be in part due to the over-dispersed distribution of sporozoites in each infectious bite , such that despite inducing a high per-parasite killing efficacy , the probability that at least one parasite reaches the liver and progresses to blood-stage infection remains high ( White et al . , 2013; Bejon et al . , 2005 ) . The classical approach to overcoming this is to attempt to further increase either the quantity , breadth or quality of the immune response ( Remarque et al . , 2012; Courtin et al . , 2009; Chaudhury et al . , 2016 ) . We hypothesized that an alternative mechanism would be to combine a PEV with approaches that reduce the number of sporozoites in the mosquito salivary glands . TBVs have been demonstrated to act in this way , reducing ookinete and sporozoite density ( Bompard et al . , 2017; Blagborough et al . , 2013 ) . We sought therefore to identify whether this mechanism could result in synergistic interactions between PEVs and TBVs co-administered within a population . To test this hypothesis , we used an established murine population assay to investigate the clearance of malaria over multiple generations in a controlled laboratory environment ( Blagborough et al . , 2013 ) . Here , the rodent malaria parasite Plasmodium berghei is passed between populations of mice by the direct feeding of Anopheles stephensi mosquitoes . To simulate the antibody response to a PEV , a monoclonal antibody ( mAb-3D11 ) - which targets the same corresponding parasite circumsporozoite protein ( CSP ) as RTS , S – was passively transferred ( intravenously ) into mice . To act as a partially effective PEV ( broadly comparable to RTS , S ) , a mAb-3D11 dose was selected which reduces mosquito-to-mouse transmission probability by ~50% ( as evaluated in naive mouse populations , transfused with differing doses of 3D11 , challenged with five mosquito bites , see Materials and methods ) . Sterilizing immunity of 47 . 2% was titrated over multiple challenges . The complementary actions of a TBV antibody response were simulated using an anti-Pfs25 monoclonal antibody 4B7 ( mAb-4B7 ) , administered by passive transfer , which targets the same parasite stages as the most currently advanced human TBV candidate , Pfs25 ( Talaat et al . , 2016 ) . This well-established transmission-blocking monoclonal antibody was used in combination with a transgenic P . berghei ( PbPfs25DR3 ) parasite that expresses Pfs25 in place of its rodent homologue . PbPfs25DR3 is phenotypically indistinguishable to WT P . berghei , expresses Pfs25 on the surface of the zygote/ookinete , and has been used previously to assay a range of TBVs ( Goodman et al . , 2011; Kapulu et al . , 2015 ) . A series of transfused mAb-4B7 doses were tested in multiple ( n = 6 ) direct feeding assays to titrate the appropriate doses to generate a 50% , 65% and 85% reduction in transmission to the mosquito ( measured as reduction in oocyst prevalence ) ( see Materials and methods ) . We undertook a series of experiments with either PEV alone , TBV alone at three different efficacies ( 50% , 65% and 85% ) , or combinations of the two across four generations of transmission . For each experiment mice were exposed to either 1 , 2 , 5 or 10 infectious mosquito bites to allow estimation of combined transmission-blocking efficacies between 20% and 100% ( Materials and methods ) .
At lower TBV antibody dose exposure levels ( 50% and 65% ) , the probability of eliminating all parasites from the mouse/mosquito populations was greater when PEV and TBV antibodies were administered together compared to singly , irrespective of the dose of the TBV antibody administered ( Figure 1 ) . Using statistical methods that explicitly capture parasite density and account for the impact of the interventions on both the prevalence and density of infection ( Sherrard-Smith et al . , 2017 ) , we estimated PEV antibody alone to reduce the prevalence of infection by 48 . 0% ( 95% credible interval , CrI , 36 . 6–58 . 0% ) . Similarly , use of TBV antibody alone reduced the prevalence of infection by 33 . 9% ( 95% CrI: 18 . 2–47 . 4% ) , 74 . 3% ( 65 . 7%–82 . 4% ) and 95 . 8% ( 90 . 2%–100% ) for the 50% , 65% and 85% individual efficacy titres , respectively . If the actions of the two antibody types were to act independently , the predicted combined efficacy would be 83 . 3% ( 79 . 1–87 . 0% ) . A substantially higher efficacy of 90 . 8% ( 86 . 7–94 . 2% ) ( p=0 . 0035 ) was observed in the PEV +TBV antibody group , indicating a synergistic interaction ( Table 1 ) . Here , the 95% Credible Intervals did not overlap the median estimates , demonstrating a significant difference between the two treatments . The same relationship was observed when examining vaccine efficacy against parasite density ( p=0 . 0025 ) ( Table 1 ) . Sub-dividing the data by TBV antibody dose , the greatest synergistic enhancement in efficacy against parasite prevalence was seen at lower doses of functional TBV antibodies ( Figure 2 ) . A TBV antibody dose which reduces mouse-to-mosquito transmission by 50% increased the efficacy of PEV antibody prevalence to 82 . 2% ( 74 . 6–88 . 9% ) compared to an expected 65% ( 57 . 7–73 . 2% ) efficacy if the vaccines acted independently , a strong indication of synergy ( p=0 . 0015 ) . Similar synergistic effects against parasite density were observed ( Figure 2 , p <0 . 0001 ) . Weaker synergistic effects were observed against parasite prevalence and density for the PEV +TBV antibody at the 65% dose ( Figure 2 , p=0 . 0675 , 0 . 02 , respectively ) ( Table 1 ) . At the highest TBV antibody dose ( 85% ) , the TBV alone already reduced parasites in the population to low levels , hence there was insufficient power to detect further synergy between the interventions as all parasites were eliminated from the experimental population ( Figure 1 ) . The impact of circulating TBV antibody on the reduction of parasites within the mosquito explains the greater efficacy of the PEV antibody in the combination treatment groups . The presence of TBV antibody ( mAb-4B7 ) reduced oocyst counts in infected mosquitoes ( ANOVA: F1 , 1525 = 75 . 3 , p<0 . 0001 ) . Similarly , TBV antibody presence reduced sporozoite density in infected mosquitoes ( ANOVA: F1 , 707 = 163 . 9 , p<0 . 0001 ) . Figure 3A , B and C further illustrate the effect that TBVs have on sporozoite density distribution for the TBV-50% , 65% and 85% single treatment groups respectively . At each dose , the tail of the distribution of sporozoites is curtailed progressively across transmission cycles . Figure 3D illustrates that the efficacy of the PEV is density-dependent , with higher efficacy achieved when mosquitoes have lower sporozoite density .
The efficacy of multiple PEV and TBV candidates against rodent and human parasites have been shown to depend on parasite density; TBV ( anti-Pfs25 , anti-Pfs48/45 and immune blocking serum ) efficacy decreases with increased parasite dose ( Churcher et al . , 2012; Miura et al . , 2016 ) and a representative PEV ( RTS , S ) only provides sterilizing immunity in volunteers against lightly infected mosquitoes ( Churcher et al . , 2017 ) . This suggests that both types of vaccine can halt transmission against a defined ( but currently uncharacterized ) quantity of parasites , which is enough to prevent onward infection in lightly infected mosquitoes/humans , but that onward transmission is still possible from heavily infected mosquitoes or individuals . Here , a partially effective TBV antibody reduced the number of parasites in infected mosquitoes , ensuring that the PEV antibodies encounter fewer parasites than would otherwise be expected if only a single antibody/vaccine class was administered in isolation . Thus , the subsequently reduced parasite burden increased the efficacy of the PEV when co-administered with a TBV . Potentially , a synergistic response could be induced by an unspecified biochemical or immunological interaction between the two vaccines . This explanation can be discounted within this system as the experimental design results in mosquito-to-mouse transmission being measured prior to the administration of a transmission-blocking intervention ( passive transfer of mAb-4B7 1 hr prior to blood feed ) as a new , naive batch of mice were infected in each generation . The greatest synergy was observed in the lower TBV dose group , although it is likely to operate across all TBV and PEV doses when vaccine reduces parasite density but fails to clear infection from host or vector . The 85% TBV dose administered alone eliminated the parasite over a single generation without the action of the PEV so there was no opportunity to show synergy . A 100% efficacious PEV or TBV would not require augmenting with an alternative vaccine , although their efficacy is still likely to drop over time as antibodies decay so combining highly effective vaccine may still be advantageous depending on the relative rates of antibody loss . While the direct translatability of rodent experiments to human health is variable , this approach is invaluable to demonstrate unequivocally that the mechanism behind the observed synergy is the direct result of TBV antibody reducing parasite density in infected mice and thereby enabling the density-dependent PEV antibody to be optimal ( Figure 3 ) . This mechanism is very likely to mirror that for humans , which cannot be tested directly due to ethical considerations and complex environmental variation . Whilst the murine system uses P . berghei , the mechanism of action of the PEV antibodies administered is matched to the antibody-based mechanism of the RTS , S vaccine; that is sporozoite invasion of the liver is inhibited by the presence of CSP-targeted antibodies in mice both in vivo and ex vivo ( Grüner et al . , 2003 ) . The P . berghei strain used here is genetically modified to express the human TBV candidate , P . falciparum P25 ( Pfs25 ) in place of its P . berghei counterpart . Thus , a proven anti-falciparum TBV mAb ( 4B7 ) can be used directly within the model ( Goodman et al . , 2011 ) . The evidence that co-administering TBV and PEV antibodies can accelerate toward controlling malaria transmission is a first step toward trialing such combinations in more natural parasite-vector-host combinations and environments . There is good reason to believe that the population dynamics of parasites and partially effective vaccines may be similar in human malaria . Transmission of human malaria from human-to-mosquito and mosquito-to-human is also considered to depend on the density of the parasite ( Churcher et al . , 2013; Sinden et al . , 2007 ) . The average number of oocysts in wild caught mosquitoes is likely to be substantially lower than the numbers observed in the rodent system ( Rosenberg , 2008 ) but oocyst distribution is highly over-dispersed ( Medley et al . , 1993 ) meaning that some mosquitoes have very-high-density infections . These highly infected mosquitoes are likely to be more infectious , so reducing their frequency through adding a TBV , could have additional impact on overall transmission . The epidemiological importance of any synergistic interaction between vaccine types in the field is hard to predict and will depend on many confounding factors such as human immunity , drug treatment , vector susceptibility , antigen escape , amongst others . Transmission is likely to be highly heterogeneous , caused by factors such as vaccine non-responders and super-transmitting hosts . The impact of different types of heterogeneity can be investigated under controlled laboratory scenarios using the murine population assay , varying the vaccinated coverage , antibody dose and changing biting heterogeneity within a population . This could help understand the relative importance of these different heterogeneities and could be used to support the design of appropriately powered Phase III trials ( or alternative trial designs ) to fully assess the impact of combining vaccine components with alternative mechanisms of action . There is no ‘magic bullet’ intervention against malaria and the current global strategy is to combine vector control and drug treatment tools in a timely manner to move towards malaria elimination . Our results suggest that the same approach might be taken for the use of vaccines and comprises the first practical demonstration that combining TBVs and PEVs may have auxiliary benefits . Synergism between PEV and TBVs could potentially enhance the efficacy of the current PEV vaccines , resulting in reduced burden and potentially elimination in areas where it was not previously possible . The development of novel anti-malarial vaccines is both costly and time-consuming . Combining partially effective vaccines of differing anti-parasitic classes may be therefore a pragmatic and powerful way to accelerate malaria elimination efforts . Synergism between PEV , TBVs and potentially a blood-stage vaccine ( either administered separately or as a multi-component vaccine ) could potentially enhance the efficacy of single vaccines , resulting in reduced burden and potentially elimination in areas where it was not previously possible .
The TBV mAb-4B7 neutralises the protein Pfs25 in sexual stages of the human malaria P . falciparum and reduces transmission of the parasite from host-to-mosquito ( Stura et al . , 1994 ) . The transgenic murine malaria parasite , P . berghei PbPfs25DR3 , expressing native Pfs25 in place of its rodent homologue , was used so that the same TBV antibody candidate could be examined within a mouse model ( Goodman et al . , 2011 ) . MAb-4B7 was administered and examined at sub-optimal concentrations titrated to reduce oocyst prevalence in the mosquito midgut ( as assessed using a direct feeding assay ) by either 50 , 65 or 85% . Given the severity of malaria , the WHO , the Strategic Advisory Group of Experts ( SAGE ) on Immunization and the Malaria Policy Advisory Committee ( MPAC ) recommended the RTS , S vaccine could be implemented in pilot countries in October 2015 ( http://www . malariavaccine . org/malaria-and-vaccines/first-generation-vaccine/rtss , accessed 04/04/2018 ) when the vaccine was demonstrating relatively low efficacies of just 36 . 3% in children aged 5 to 17 months ( RTSS Clinical Trials Partnership , 2015 ) . These TBV doses were chosen to bridge a range of malaria vaccine efficacies that might be acceptable by WHO , SAGE and MPAC . Briefly , to titrate the appropriate dose , female Tuck Ordinary ( TO ) mice ( 6–8 weeks old , Harlan , UK ) were treated with phenylhydrazine , and three days later , infected with 106 P . berghei PbPfs25DR3 ( Goodman et al . , 2011 ) . Three days later , infected mice were injected intravenously ( i . v . ) with 200 µl of purified mAb-4B7 at a range of doses . Negative control mice were transfused with 200 µl of phosphate buffered saline ( PBS ) . After 1 hr , mice were anesthetised and 50 Anopheles stephensi mosquitoes ( line SD 500 , previously starved for 24 hr ) were allowed to feed on each individual mouse . Mosquitoes were maintained as described in ( Blagborough et al . , 2013 ) , and after 10 days , 50 mosquitoes were dissected and microscopically examined to measure oocyst intensity and prevalence . This was repeated five times , with i . v . administered doses of mAb-4B7 ranging from 0 µg to 750 µg . Prevalence efficacy was estimated as a function of mAb-4B7 concentration using a generalised linear model framework ( Bolker et al . , 2009 ) in which experimental replicate was treated as a random effect . A Gompertz function ( Churcher et al . , 2013 ) was fitted to the data using maximum likelihood methods . Mean concentrations were estimated using the best-fitting model ( determined by log-likelihood tests ) with 95% confidence intervals obtained from the profile likelihood . We estimated that a mAb-4B7 dose of 284 . 2 µg i . v ( 244 . 7–337 . 3 µg ) was required to achieve a reduction in prevalence of 50% , a dose of 371 . 8 µg i . v ( 319 . 8–442 . 8 µg ) for 65% reduction and a dose of 629 . 5 µg i . v ( 525–777 . 1 µg ) for an 85% reduction . These calculated doses were then used in the mosquito-mouse model system as described below . The anti-P . berghei CSP mAb-3D11 ( Mishra et al . , 2012 ) is mechanistically similar to the recently registered RTS , S vaccine for human malaria , in that the presence of CSP-targeted antibodies in mice inhibit sporozoite invasion of the liver both in vivo and ex vivo ( Grüner et al . , 2003 ) . An appropriate dose for mAb-3D11 was estimated from 40 individual passive transfers , administering a range of mAb-3D11 doses ( 0–150 µg i . v . ) to mice ( 5 to 10 mice per experiment ) and determining the prevalence efficacy at the given dose . A logistic function was fitted to these data using RStan ( Stan Development Team , 2017 ) to determine the mean dose and 95% credible intervals that produces a ~50% reduction in the probability of infection . Consequently , mAb-3D11 antibody was administered at a single sub-optimal dose ( 50 µg i . v . ) that prevented 47 . 2% ( 38 . 0–62 . 0% 95%CI ) of transmission to mice that each received 5 potentially infectious mosquito bites . This dose was selected to match the approximate observed protection by RTS , S in human clinical trials . The mouse-to-mouse transmission model has been described in detail previously ( Upton et al . , 2015; Blagborough et al . , 2013 ) ( Figure 1—figure supplement 1 ) . Briefly , five female ( TO ) mice ( 6–8 weeks old , Harlan , UK ) were treated with phenylhydrazine , and , 3 days later , were infected with 106 P . berghei PbPfs25DR3 ( Goodman et al . , 2011 ) . Three days later , groups of infected mice were treated with the TBV antibody . After 1 hr , the mice were anesthetised and 500 An . stephensi mosquitoes ( line SD 500 , starved for 24 hr ) fed randomly on the five infected mice within each group . Mosquitoes were maintained as described in Blagborough et al . , 2013 . After 10 days , a sub-sample of 50 mosquitoes were microscopically examined to measure oocyst intensity and prevalence . After 21 days post-feeding , sporozoites are present in the salivary glands and are maximally infectious to the vertebrate host ( Blagborough et al . , 2013 ) . At this point , pre-defined numbers of mosquitoes ( to simulate mosquito biting rates of 1 , 2 , 5 and 10 mosquito bites per mouse ) were then randomly selected from the remaining mosquitoes and fed , for 20 min , on anesthetized mice from a naïve cohort . The mosquito biting rate is an aspect of the experimental design that can be varied to be able to estimate the effect size more precisely . If the mosquito biting rate is small ( say 1 or 2 ) , the probability that the infection is eliminated rapidly in the intervention arm of the experiment is high ( >80% ) for TBD/TBV with efficacies above 40% . Thus , we cannot discriminate at a low mosquito biting rate between a TBD/TBV with 60% efficacy and one with 80% efficacy ( they both eliminate ) . However , we do obtain a high degree of discrimination between an efficacy of 20% and one of 40% . Thus , a small mosquito biting rate is needed to get a precise estimate of the effect size of a TBD/TBV with lower efficacy . The converse also holds , so that a high mosquito biting rate ( up to around 10 based on our initial experiment ) is needed to obtain a precise estimate of a TBV/TBD with >80% efficacy . Using multiple mosquito biting rates increases the overall precision of our estimate of prevalence efficacy . Each group of mice ( five mice per group ) either received the PEV antibody or no -intervention ( negative control ) . Engorged mosquitoes were microscopically examined immediately after feeding to determine the number of sporozoites in the salivary glands . After 10 days , blood smears from each mouse were microscopically examined to determine the percentage parasitemia . These five mice were then given either the TBV antibody at the desired dose to achieve a 50% , 65% or 85% reduction in oocyst prevalence as required , or no intervention/control ( in accordance with the respective treatment arm ) . A new cohort of 500 naive mosquitoes was then allowed to blood feed on the mice . This mouse-to-mouse transmission cycle was repeated to a maximum of four cycles after the seeding mouse population or until no parasites had been detected in the system for two successive transmission cycles . The PEV and TBV antibodies at each dose ( corresponding to a reduction in transmission to mosquitoes of 50% , 65% and 85% ) were tested singly and in combination . Initial parasite density was measured by counting the number of infected red blood cells ( out of a total subsample of 1200 erythrocytes ) . The number of sporozoites in the salivary glands following blood feeding was counted on the logarithmic scale ( scores of 0–4 representing 0 , 1–10 , 11–100 , 101–1000 , 1000 + sporozoites , respectively ) . The data are provided in Table 1—source data 1 . All animal procedures were performed in accordance with the terms of the UK Animals ( Scientific Procedures ) Act ( PPL 70/8788 ) and were approved by the Imperial College Animal Welfare and Ethical Review Body ( AWERB ) LASA guidelines were adhered to at all points . The Office of Laboratory Animal Welfare Assurance for Imperial College covers all Public Health Service supported activities involving live vertebrates in the US ( no . A5634-01 ) . The complexity of the population assay requires non-standard methods of statistical analyses that can account for the non-linear dynamics of transmission and stochastic fluctuations seen by the relatively small number of mice used in each generation and treatment arm . These methods need to be able to determine whether the interaction between the different antibodies , that simulate vaccine-triggered antibodies , is below what would be expected if vaccine effects were less strong than expected if effects were multiplied ( sub-multiplicative ) , independent of the presence of the other vaccine ( multiplicative ) or synergistic , in that effects are enhanced for one or both vaccine types ( super-multiplicative ) . A density model was developed specifically for this purpose ( described in full , [Sherrard-Smith et al . , 2017] , Figure 1—figure supplement 1 ) . The structure of the model captures the experimental set up and fits explicitly to parasite densities during successive life stages to generate more precise estimates of simulated vaccine efficacy than the direct comparison of raw data alone , which can fluctuate widely due to chance ( as each generation only has five mice ) ( Table 1—source data 1 ) . Stochastic elimination ( or resurgence ) is possible in the mouse system given the ethical necessity to keep the mouse populations in each generation small; as only five mice are used so transmission could be halted or enhanced by natural variability in the physiological response from each individual mouse . The statistical mode , however , explicitly addresses this variation by modeling the distribution of pathogens at each stage of transmission for each mouse individually . By explicitly incorporating the heterogeneity in the mouse pathogen load the model avoids bias in the inference of the transmission process itself . All parameters were fitted jointly using a Bayesian posterior distribution in RStan ( version 2 . 13 . 1 , [Stan Development Team , 2017] ) . To ensure robust fits , a non-centered parameterization method was employed ( Papaspiliopoulos et al . , 2007; Betancourt and Girolami , 2015 ) . The model parameter fitting was achieved using a Hamiltonian Monte Carlo method ( Stan Development Team , 2017 ) , warmup was 500 and the subsequent 500 samples from each chain ( n = 4 ) were used for the posterior predictive checks ( Sherrard-Smith et al . , 2017 ) . The model was validated by visualizing the observed raw data measurements of parasite density in mice , the oocyst counts and the logarithmic counts of sporozoites in mosquitoes against model predictions . The data were analyzed at different scales , first taking all data together before breaking down the impact by the dose of the TBV antibody . The prevalence efficacy against infections in mice is the percentage difference in the proportion of infected hosts between the control and treatment arms of the experiment . The parasite density efficacy against infections in mice is the percentage difference in the mean parasite density per host between the control and treatment arms of the experiment . ( Similarly , efficacies can be calculated for the reduction in oocysts or sporozoites in mosquito populations . ) Efficacy estimates were generated for each arm of the experiment , for each posterior draw of the model ( 2000 posterior draws ) . This allows mean and 95% credible intervals to be calculated for each treatment group ( c ) and across mosquito biting rates ( m ) and transmission cycles ( i ) . Let Pc , m , ijindicate the prevalence of infected mice ( j = 1 ) or the mean asexual parasite density in the mouse population ( j = 2 ) . Treatment arm 0 represents the control data , and c indicates treatments 1 to 7 ( TBV antibody at 50% , 65% , 85% dose singly , PEV antibody singly , PEV and TBV antibody at 50% , 65% and 85% dose together ) , such that , ( 1 ) Ec , m , ij=P0 , m , ij−Pc , m , ijP0 , m , ij×100where Ec , m , ij is either the parasite prevalence efficacy ( j = 1 ) or density efficacy ( j = 2 ) against infections in mice as estimated by the posterior predictions of the density model ( Table 1 ) . To statistically assess whether the interaction between PEV and TBV antibodies are antagonistic , independent or synergistic , efficacy estimates against infections in mice for combined antibody treatments were compared to the expected estimates if antibodies had an independent impact ( expected efficacy ) . To estimate this expected efficacy the single antibody treatment groups were combined , as follows ( VanderWeele and Knol , 2014 ) : ( 2 ) Expected Efficacy=Epev* ( 1−Etbv , d ) +Etbv , dwhere Etbv , d is the prevalence or density efficacy for the TBV antibody treatment alone ( at the specified TBV dose of d = 50% , 65% or 85% ) , and Epev is the prevalence or density efficacy for the PEV alone . The ratio between efficacies for the combined treatments and the expected efficacy for matched treatments was used to assess synergy . A synergistic interaction is indicated when this ratio is greater than 1 , an independent interaction when equal to one and an antagonistic impact if less than 1 ( Figure 4 ) . The 95% credible intervals were calculated to give statistical support . Statistical evidence of a difference in treatment and control experiments ( p-value ) was defined as one minus the proportion of iterations from the model simulations that were greater than 1 . The observed synergy can be explained because the presence of TBV antibody reduces the sporozoite score ( as a measure of parasite density in the mosquito population ) which allows the PEV antibodies to achieve a greater efficiency . To understand whether the combined TBV antibody treatments improved the action of the PEV at higher parasite densities , the raw data recording the sporozoite scores of each mouse in the single treatment TBV antibody groups were plotted for each dose and across transmission cycles to demonstrate that there are progressively more mosquitoes without infection and progressively fewer mosquitoes with heavy infections ( Figure 3 ) . This highlights that the PEV in the combined treatment groups is acting against progressively fewer parasites in each transmission cycle because of the initial action of the TBV antibody . Simple analysis of variance was performed to confirm that the presence of the respective antibody type ( PEV acting against infections in mice , TBV acting against sporozoites in mosquitoes ) , as a binary covariate , could explain reduced parasite counts in mosquitoes or mice . To demonstrate that the PEV antibody has a density-dependent impact , binomial logistic regression curves were fitted to determine the relationship between the prevalence efficacy of the PEV antibody and the parasite density in mosquitoes ( measured as mean sporozoite score for each mosquito biting rate and transmission cycle , n = 16 ) . Parameters describing the regression curves were fitted using a Bayesian posterior distribution in RStan ( version 2 . 13 . 1 , [Stan Development Team , 2017] ) . The model parameter fitting was achieved using a Hamiltonian Monte Carlo method ( Stan Development Team , 2017 ) , warmup was 1000 and the subsequent 1000 samples from each chain ( n = 4 ) were used for the posterior predictive checks ( Stan Development Team , 2017 ) . All data analysis were conducted using the statistical software R ( version 3 . 2 . 2; [Core Team , 2014] ) . | In 2016 , malaria caused an estimated 216 million illnesses and 445 , 000 confirmed deaths globally . The disease is caused by a parasite , and mosquitos infected with the parasite transmit them to humans when they bite . In humans , the parasites enter the body and head to the liver before spending part of their life cycle in red blood cells , which cause the symptoms of the disease . Prevention efforts have reduced the burden of malaria but eliminating the disease will require new tools . One option is to use vaccines . The world’s first malaria vaccine – a so-called pre-erythrocytic vaccine ( PEV ) – targets the stages preceding the parasite reaching the liver . This vaccine prevents malaria parasites from infecting people , but it is only partially effective . Scientists are also developing transmission-blocking vaccines ( TBVs ) . These TBVs block the development of malaria parasites in mosquitos that bite vaccinated humans . So far , the most promising TBV candidates are also only partly effective . It is possible that using PEV and TBV vaccines together could boost their effectiveness , since the TBV vaccines reduce the number of parasites that infect each mosquito . This means that fewer parasites are injected into the next person . Currently , the PEVs work better when there are fewer parasites infecting a person . Now , Sherrard-Smith et al . show that combining TBVs with PEVs enhances their antimalarial effects . In the experiments , Sherrard-Smith et al . treated mice with either TBV or PEV vaccines , or both . Then , the mice were exposed to mosquitos infected with the malaria parasite . As expected , the TBV and PEV treatments were only partially effective when used alone . But exposing the mice to both TBVs and PEVs eliminated the parasites from the mosquitos and the mice . The combined benefit of TBVs and PEVs were greater than would be expected if either vaccine was acting alone and the effects were simply multiplied , suggesting they enhance each other’s effects . More studies of TBVs in humans are needed to prove they are safe and effective in the real world . More studies also are needed to confirm what Sherrard-Smith et al . found in mice would happen in humans treated with a combination of TBV and PEV vaccines . But if such future studies prove this combination approach is effective , it could be a powerful tool in the fight against malaria . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"microbiology",
"and",
"infectious",
"disease"
] | 2018 | Synergy in anti-malarial pre-erythrocytic and transmission-blocking antibodies is achieved by reducing parasite density |
ATP-dependent chromatin remodelling proteins represent a diverse family of proteins that share ATPase domains that are adapted to regulate protein–DNA interactions . Here , we present structures of the Saccharomyces cerevisiae Chd1 protein engaged with nucleosomes in the presence of the transition state mimic ADP-beryllium fluoride . The path of DNA strands through the ATPase domains indicates the presence of contacts conserved with single strand translocases and additional contacts with both strands that are unique to Snf2 related proteins . The structure provides connectivity between rearrangement of ATPase lobes to a closed , nucleotide bound state and the sensing of linker DNA . Two turns of linker DNA are prised off the surface of the histone octamer as a result of Chd1 binding , and both the histone H3 tail and ubiquitin conjugated to lysine 120 are re-orientated towards the unravelled DNA . This indicates how changes to nucleosome structure can alter the way in which histone epitopes are presented .
The extended family of ATPases related to the yeast Snf2 protein acts to alter DNA-protein interactions ( Flaus et al . , 2006; Narlikar et al . , 2013 ) . They act on a diverse range of substrates . For example , while the Mot1 protein acts on complexes between the TATA box binding protein BP and DNA ( Wollmann et al . , 2011 ) , the Snf2 protein carries out ATP-dependent nucleosome disruption ( CoteCôté et al . , 1994 ) . At the heart of all these proteins are paired domains capable of rearranging during the ATP hydrolysis cycle to create a ratchet like motion along DNA in single base increments ( Clapier et al . , 2017; Gu and Rice , 2010; Velankar et al . , 1999 ) . The yeast Chd1 protein is a member of this protein family and acts to organise nucleosomes over coding regions ( Gkikopoulos et al . , 2011; Ocampo et al . , 2016; Pointner et al . , 2012; Tran et al . , 2000 ) . Consistent with this , Chd1 is known to interact with elongation factors including the Spt4-Spt5 proteins , Paf1 and FACT ( Kelley et al . , 1999; Krogan et al . , 2002; Simic et al . , 2003 ) . The partially redundant functions of Chd1 and Isw1 in organising nucleosomes over coding regions are in turn required to prevent histone exchange and non-coding transcription ( Hennig et al . , 2012; Radman-Livaja et al . , 2012; Smolle et al . , 2012 ) . In addition to the positioning of nucleosomes , the distribution of many histone modifications is ordered with respect to promoters ( Liu et al . , 2005; Mayer et al . , 2010 ) . For example , histone H3 K4 methylation is frequently observed at promoters , while histone H3 K79 and K36 trimethylation are detected in coding regions ( Kizer et al . , 2005; Li et al . , 2003; Pokholok et al . , 2005 ) . Histone H2B is also observed to be ubiquitinylated within coding regions ( Fleming et al . , 2008; Xiao et al . , 2005 ) . Ubiquitinylation of histone H2B at lysine 123 in budding yeast , H2B K120 ( H2BK120ub ) in mammals , is dependent on the E2 ligase Rad6 ( Robzyk et al . , 2000 ) and the E3 ligase Bre1 ( Hwang et al . , 2003; Wood et al . , 2003 ) and removed by the deubiquitinases Ubp8 and Ubp10 ( Bonnet et al . , 2014; Schulze et al . , 2011; Wyce et al . , 2007 ) . A specific reader of H2BK120ub has not been identified . However , H2BK120ub does assist the histone chaperone FACT in enabling transcription through chromatin ( Pavri et al . , 2006 ) , and has been found to be required for methylation of histone H3 K4 and K79 ( Sun and Allis , 2002 ) . An intriguing aspect of H2BK120ub is that while mutation of the writer enzymes or K120 itself disrupts nucleosome organisation , deletion of the deubiquitinylases increases chromatin organisation ( Batta et al . , 2011 ) . One way in which H2BK120ub may influence nucleosome organisation is via effects on enzymes responsible for chromatin organisation . Consistent with this H2BK120ub increases nucleosome repositioning mediated by Chd1 ( Levendosky et al . , 2016 ) . Yeast Chd1 serves as a useful paradigm in that it functions predominantly as a single polypeptide . In addition , the catalytic core of the enzyme has been crystallised in association with the adjacent tandem chromodomains ( Hauk et al . , 2010 ) . Similarly the C-terminal region of the protein has been crystalized revealing that this region includes SANT and SLIDE domains that comprise the DNA binding domain ( DNABD ) ( Ryan et al . , 2011; Sharma et al . , 2011 ) and are also present in ISWI proteins ( Grüne et al . , 2003 ) . Chd1 enzyme engages nucleosomes in a conformation in which the SANT and SLIDE domains bind linker DNA , while the ATPase domains engage DNA at super helical location ( SHL ) 2 ( Nodelman et al . , 2017; Sundaramoorthy et al . , 2017 ) . Higher resolution structures of Chd1 ( Farnung et al . , 2017 ) , Snf2 ( Liu et al . , 2017 ) and INO80 ( Ayala et al . , 2018; Eustermann et al . , 2018 ) show that the ATPase domains make contacts with DNA via residues that are conserved in ancestral single-stranded ATPases and some unique to Snf2-related ATPases . The binding of the Chd1 DNABD unravels two turns of DNA from the surface of nucleosomes in a nucleotide-stimulated reaction ( Farnung et al . , 2017; Sundaramoorthy et al . , 2017 ) . Here , we report a structure for the yeast Chd1 protein in association with a nucleosome , bearing modifications that are found to occur within coding regions , where Chd1 is known to act . Interestingly , nucleosomal epitopes are observed to be reconfigured specifically on the side of the nucleosome on which DNA is unwrapped . This indicates the potential for changes to nucleosome structure to reconfigure the way in which histone epitopes are presented .
As Chd1 functions on transcribed genes , it is of interest to understand the interplay between Chd1 and histone modifications observed in coding region chromatin . As a result , nucleosomes were prepared in which histone H3 K36 was alkylated to mimic trimethylation ( Figure 1—figure supplement 1 ) and H2B cross-linked to ubiquitin ( Figure 1—figure supplement 2 ) . Conditions were established to favour binding of a single Chd1 to modified nucleosomes that included an asymmetric linker DNA extension of 14 bp ( Figure 1—figure supplement 2B ) in the presence of ADP-BeF . Purified complexes were frozen onto EM grids . 2D classification of some 893000 particles revealed 16 classes in which nucleosomes with the Chd1 molecule attached could be identified ( Figure 1—figure supplement 3B ) . Initial 3D classification resulted in five related classes ( Figure 1—figure supplement 3C ) . Three of these were combined and reclassified as six classes , one of which was selected for refinement . This resulted in the generation of a map with an average resolution of 4 . 5 Å ( FSC 0 . 143 ) ( Figure 1—figure supplement 4A ) . The resolution varies within the map , with resolution highest in the region occupied by the nucleosome and ATPase lobes and lower resolution in the vicinity of the DNABD and ubiquitin peptides ( Figure 1—figure supplement 4B ) . The nucleosome particles exhibited a preferred orientation , which may limit the resolution ( Figure 1—figure supplement 4C ) . A structural model was generated to fit the density map making use of the structures of a nucleosome assembled on the 601 DNA sequence , Chd1 chromoATPase , and DNABD ( Figure 1 ) . The fit for individual components of the structure to the electron density is shown in Figure 1—figure supplement 5 . The overall organisation of Chd1 is similar to that observed previously by cryo EM ( Farnung et al . , 2017; Sundaramoorthy et al . , 2017 ) and directed cross-linking ( Nodelman et al . , 2017 ) . The ATPase domains are bound at the SHL-2 location . Of the two SHL2 locations within nucleosomes , the bound site is in closest proximity to SANT-SLIDE domain bound linker DNA in physical space , but distal on the unwrapped linear DNA sequence ( Figure 1 ) . Chd1 predominantly contacts the nucleosome via contacts with DNA , via the DNABD in the linker and ATPase lobes at SHL2; contacts with histones are limited to the histone H3 and H4 N-terminal regions discussed below . We previously showed that Chd1 binding results in nucleotide-dependent unwrapping of nucleosomal DNA resulting from the interaction of the DNABD with linker DNA ( Sundaramoorthy et al . , 2017 ) . The higher resolution of the current structure shows that precisely two turns of nucleosomal DNA are unravelled ( Figure 1 ) . The extent of DNA unwrapping observed here when Chd1 is bound to nucleosomes flanked by a 14 base pair linker DNA is identical to that observed when Chd1 is bound to the opposite surface of the 601 nucleosome positioning sequence with a 63 base pair linker ( Farnung et al . , 2017 ) . As the interaction of histones with the two sides of the 601 positioning sequence differ quite dramatically ( Chua et al . , 2012; Hall et al . , 2009; Levendosky et al . , 2016; Ngo et al . , 2015 ) , this suggests that the extent of unwrapping is dominated by the properties of Chd1 rather than the affinity of DNA for the octamer . The path of this unwrapped DNA is oriented away from the plane of the wrapped DNA gyre and is kinked at the location where contacts are made with the SANT-SLIDE domains ( Figure 1 ) . Other than DNA unwrapping , we do not detect additional changes in the organisation of DNA on Chd1 bound nucleosomes at this resolution . The orientation of the DNABD is critical in determining the extent of DNA unwrapping . The only contacts detected between the DNABD and the remainder of Chd1 are contacts with the chromodomains ( Figure 2 contact I and II ) . The first of these is the interaction between K329 of chromodomain II and D1201 P1202 in the SLIDE component of the DNABD and has been observed previously ( Farnung et al . , 2017; Nodelman et al . , 2017 ) ( Figure 2—figure supplement 1A ) . The second contact is between S344 and K345 in the linker helix between chromodomain II and ATPase lobe I with the SANT component of the DNA binding domain at D1033-D1038 ( Figure 2—figure supplement 1A ) . Given that chromodomains are present in Chd1 enzymes but not ISWI and Snf2 remodellers , it makes sense that the residues contacted in the SANT and SLIDE domains are most highly conserved in Chd1 proteins ( Figure 2—figure supplement 1B ) ( Hall et al . , 2009; Meng et al . , 2015; Sundaramoorthy et al . , 2017 ) . The position of the chromodomains is determined by each of the four contacts made with other components of the complex ( Figure 2 ) . When not bound to nucleosomes , the tandem chromodomains of Chd1 are observed to impede DNA binding to the ATPase domains ( Hauk et al . , 2010 ) . This gave rise to the prediction that these domains would be rearranged in the nucleosome-bound state ( Hauk et al . , 2010 ) . This is indeed the case as the chromodomains undergo an 18 degree rotation when compared to the orientation observed in the crystal structure of Chd1 in the open state ( Figure 3 ) . Following repositioning , chromodomain I interacts with nucleosomal DNA at SHL1 ( Figure 2—figure supplement 2 ) as observed previously ( Farnung et al . , 2017; Nodelman et al . , 2017 ) . Coincident with repositioning of the chromodomains , ATPase lobe II is repositioned closer to lobe I . This results in residues including those contributing to the conserved Walker box motifs ( K407 and R804 , R807 ) being brought into an arrangement compatible with ATP catalysis . Density for ADP-BeF within the pocket formed by conserved residues from ATPase domains I and II is well defined ( Figure 2—figure supplement 3 ) . The repositioning of ATPase lobe II enables contacts to be made with nucleosomal DNA ( see below ) , the histone H4 tail and the histone H3 alpha one helix ( Figure 2—figure supplement 4 ) . These are the only direct contacts with histone components of the nucleosome . The contact with the H4 tails is conserved in mtISWI and Snf2 ( Liu et al . , 2017; Yan et al . , 2016 ) . D729 and E669 are conserved across all classes of remodelling enzyme but D725 is not as well conserved in Snf2-related enzymes ( Figure 2—figure supplement 4B ) . The conservation of this contact in Chd1 enzymes is consistent with the H4 tail playing an important role in regulating Chd1 activity; deletion or mutation of the H4 tail has been shown to reduce nucleosome sliding and ATPase activity ( Ferreira et al . , 2007 ) . The additional helices that make up the protrusion 2 region of ATPase lobe two in Chd1 are conserved in chromatin remodeling ATPases , but not within all SF2 DNA translocases . In the crystal structure of the Chd1 chromoATPase and the previously published Chd1-nucleosome structure this region was not mapped ( Farnung et al . , 2017; Hauk et al . , 2010 ) . However , within the structure presented here residues ranging from 632 to 647 pack against the alpha 1 helix of histone H3 . In particular the two conserved residues K632 and K642 are located closer to H3 α1 D81 and E73 ( Figure 2—figure supplement 4A ) . Deletion of this lobe two loop K632-K646 abolishes nucleosome sliding consistent with a role for this region in Chd1 action ( Figure 2—figure supplement 5 ) . The residues participating in the interaction are progressively less well conserved in ISWI and SNF2 related proteins ( Figure 2—figure supplement 4C ) . A loop from Phe1033 to Leu 1045 in an equivalent region of the yeast Snf2 protein is not assigned in the Snf2-nucleosome structure , but this region is positioned such that a related contact with histone H3 could be made . The structure , also provides clues as to how these conformational changes are driven . A central event is likely to be the closure of the cleft between ATPase domains driven by ATP binding ( Figure 2—figure supplement 3 ) . The 40o rotation of ATPase lobe II required to form the ATP binding pocket results in a positively charged surface , observed to interact with an acidic surface on the long helix of chromodomain I ( Figure 3A ) ( Hauk et al . , 2010 ) , being replaced by an acidic surface likely to repel chromodomain I ( Figure 3B ) . As a result , closure of the ATPase domains is anticipated to drive nucleotide-dependent repositioning of the chromodomains . Pulsed EPR was used to directly measure repositioning of the chromodomains in the absence of nucleosomes ( Figure 4 ) . The distance between engineered labels at V256C in chromodomain I and S524C in ATPase lobe1 is 4 . 4 nm in the open state , consistent with that observed in the crystal structure of the Chd1 chromoATPase domains ( Hauk et al . , 2010 ) . In the presence of ADP-BeF the 4 . 4 nm distance predominates , but a shoulder is observed consistent with a proportion of molecules adopting a new conformation with a distance of 5 . 6 nm ( Figure 4 ) which is similar to that observed in the ADP-BeF bound nucleosome by cryo-EM . This indicates that ATP binding is a driving event for repositioning of the chromodomains . The partial repositioning of the chromodomains observed in free Chd1 is likely to be stabilised by additional favourable interactions formed when this repositioning occurs within the context of nucleosome bound Chd1 . These include the formation of contacts between chromodomain I and DNA at SHL1 , between ATPase lobe II and the H3 alpha one helix , between ATPase lobe II and the histone H4 tail and most significantly the formation of a substantial interaction interface between ATPase lobe II and nucleosomal DNA at SHL2 . The repositioning of the chromodomains in turn acts as a lever to reposition the DNA binding domain . In the context of nucleosomes this results in nucleotide-dependent unwrapping of two turns of nucleosomal DNA ( Sundaramoorthy et al . , 2017 ) . Conversely , the interaction of the DNABD requires linker DNA to be accessible . In order to investigate how the ability of the DNABD to interact with linker DNA is affected by the presence of an adjacent nucleosome , interactions between dinucleosomes with different separations were modelled . With a linker length of 19 bp Chd1 can be modelled binding the linker between adjacent nucleosomes ( Figure 5 ) . However , as the linker between nucleosomes is reduced , steric clashes become increasingly prohibitive . The requirement for a 19 bp linker is likely to provide a limit below which engagement of the DNABD will be less stable . As this lower limit is set by clashes between the DNABD and the adjacent nucleosome , it is different from the length of linker required to occupy the DNA-binding surface of the SANT and SLIDE domains on a mononucleosome with a free DNA linker . In this latter case , seven base pairs of DNA make contact with the DNABD ( Figure 1 ) . The c19 bp separation below which access of the DNABD to linker becomes progressively more difficult resonates with the average inter-nucleosome spacing of 19 bp observed in Saccharomyces cerevisiae ( Tsankov et al . , 2010 ) . As the conformation of the DNABD is connected via the chromodomains to the ATPase domains , the structure of Chd1 provides molecular connectivity between the availability of nucleosomal linker DNA in excess of 19 bp and the generation of closed nucleotide bound motor domains . This potentially provides a mechanism via which linker DNA length regulates the rate of nucleosome movement ( Yang et al . , 2006 ) . Nucleosome repositioning is likely to be driven by the ability of the ATPase domains to drive ATP- dependent DNA translocation . This has been observed directly for several Snf2 family proteins ( Deindl et al . , 2013; Lia et al . , 2006; Sirinakis et al . , 2011; Zhang et al . , 2006 ) and is conserved within a wider family of superfamily II ATPases ( Singleton et al . , 2007 ) . Structures of superfamily II single stranded translocases , such as herpes virus NS3 , in different NTP bound states illustrate how the ratcheting motion of the ATPase domains drives translocation ( Gu and Rice , 2010 ) . To date such a series of structures does not exist for a double strand specific translocase . This raises the question as to whether structures of NS3 can be used to inform key aspects of the mechanism of Chd1 such as identifying the tracking strand . To do this we first align the ATPase lobes of Chd1 individually with NS3 . The ATPase lobes of Chd1 like other Snf2 related proteins contain additional helices not conserved with NS3 ( Dürr et al . , 2005; Liu et al . , 2017; Thomä et al . , 2005 ) . As a result , the alignment is restricted to conserved helices . In the case of lobe I and II , the RMSD of the fit is 4 . 9 Å and 6 . 5 Å , respectively ( Figure 6—figure supplement 1A ) . In the closed state alignment of both domains with the structure of NS3 in the ADP . BeF bound state results in an RMSD of 9 . 8 Å . Using this alignment , the ssDNA bound by NS3 can be docked into the Chd1-Nucleosome structure ( Figure 6A ) . This ssDNA aligns with the top strand of nucleosomal DNA ( Figure 6 ) . Conserved motif Ia in ATPase lobe one and motifs IV and V from ATPase lobe two contact this strand . These residues undergo a ratcheting motion during the course of ATP hydrolysis that drives the ssDNA through NS3 ( Gu and Rice , 2010 ) . Similar motion between these residues would be anticipated to drive nucleosomal DNA across the nucleosome dyad in the direction of the longer linker ( Figure 6B ) . It is notable that within Chd1 additional DNA contacts are made that differ from those observed in NS3 . Firstly , motifs II and III within lobe one contact the opposite DNA strand ( Figure 6A ) . As these motifs are intimately associated with motif Ia they would be anticipated to undergo a similar ratcheting motion with respect to the contacts made by lobe 2 . Secondly , the ATPase lobes of Chd1 like those of Snf2 contain additional helices including the protrusions to the helical lobes and the brace helix that are unique to chromatin remodelling enzymes ( Figure 6A ) ( Farnung et al . , 2017; Flaus et al . , 2006; Liu et al . , 2017 ) . These extend the binding cleft between the ATPase lobes and make additional contacts with both DNA strands . The structure of a fragment of the yeast Snf2 protein bound to a nucleosome revealed contacts between ATPase lobe one with DNA at SHL2 and the adjacent DNA gyre at SHL 6 ( Liu et al . , 2017 ) ( Figure 6—figure supplement 2A ) . The basic surface of lobe one responsible for this interaction is not conserved in Chd1 , and the acidic residues D464 and E468 make a similar interaction with DNA unlikely . In addition , DNA is not present in this location as it is lifted off the surface of the octamer ( Figure 6—figure supplement 2B ) . In the case of the Snf2 protein the interaction with the adjacent DNA gyre is proposed to anchor the translocase preventing it from transiting around the octamer surface ( Liu et al . , 2017 ) . Chd1 has a relatively small interaction interface with histones , so there is a similar requirement for DNA interactions to constrain motion of the whole protein . In the case of Chd1 this could instead be provided through the interaction of the chromodomains with DNA at SHL1 and through the interaction of the DNA binding domain with linker DNA . Amino acids 476 to 480 of lobe one also interact with DNA in the unravelled state ( Figure 6—figure supplement 2B ) . These residues are not conserved even in Chd1 proteins so the significance of this contact is not clear . Chromatin organising motor proteins are capable of catalysing bidirectional nucleosome repositioning that can occur as a result of the binding of two or one enzyme ( Blosser et al . , 2009; Qiu et al . , 2017; Racki et al . , 2009; Willhoft et al . , 2017 ) . As Chd1 binds to one side of the nucleosome , no steric clashes are anticipated should a second Chd1 bind linker DNA on the opposite side of the nucleosome . To investigate this further , complexes consisting of two Chd1 molecules bound to one nucleosome were prepared using nucleosomal DNA with symmetrical linkers of 14 base pairs and the images processed as indicated ( Figure 7—figure supplement 1 ) . Most particles were assigned to 2D classes in which two bound Chd1 molecules are discernible , though one is often more dominant likely as a result of the projections of the dominant orientations observed . All 3D classes have two bound Chd1 molecules and the best refined class provides an envelope with 11 Å resolution ( FSC 0 . 143 ) ( Figure 7 ) . Two Chd1 molecules bound in the same mode observed in the 1:1 complex can be docked into this volume . There are no direct contacts between the two Chd1 proteins suggesting that the two bound enzymes are likely to function independently . Previously , negative stain EM of two Chd1 molecules bound to a single nucleosome indicated that the DNA binding domain interacted with linker DNA on only one side of the nucleosome ( Nodelman et al . , 2017 ) . Our envelope shows that both DNA-binding domains can bind to linker DNA simultaneously and that the extent of DNA unwrapping is similar on both sides of the nucleosome . We do however observe that within some 3D classes the path of the unwrapped DNA is less clear on one side of the nucleosome . This could relate to differences in the dynamic interactions of Chd1 with DNA on the two sides of the nucleosome ( Tokuda et al . , 2018 ) . On the fully wrapped side of the nucleosome the H3 tail can be traced to proline 38 , emerging between the DNA gyres at SHL1 . In contrast , on the unwrapped side of the nucleosome the H3 tail can be traced to alanine 26 indicating that on this side of the nucleosome the H3 tail is better ordered . In addition , the trajectory of the tail is different to that observed in previous structures ( Figure 8A ) . This altered trajectory was also not observed in a previous structure of a Chd1-bound nucleosome ( Farnung et al . , 2017 ) . This structure was made in the presence of PAF1 and FACT complexes which may have contributed to noise in this region that is not apparent in our structure ( Farnung et al . , 2017 ) . A potential explanation for the defined and altered trajectory of the histone H3 tail on the unwrapped side of the nucleosome is that amino acids within the extreme N-terminal region , that are not resolved in our structure , interact with the unravelled DNA . The 25 N-terminal residues include eight lysine and arginine residues that could interact with DNA at different locations along the unravelled linker . Interestingly , deletion of the H3 tail to H3K36 increases the initial rate at which nucleosomes are repositioned by Chd1 ( Figure 8—figure supplement 1 ) . This effect is not dependent on DNA binding via the extreme N-terminus as deletion to K26 does not stimulate Chd1 activity ( Figure 8—figure supplement 1 ) . The region of H3 that exerts the repressive effect occupies the density shown in Figure 8A . Surprisingly , no contacts between this region of H3 and Chd1 are observed . It is possible that this region interacts with regions of Chd1 such as the N-terminus that are not resolved , or that the repressive effect is exerted when Chd1 is bound in a conformation different to that observed by EM . The electron density for ubiquitin molecules is not as well defined as other components of the complex , and limiting for the overall resolution ( Figure 1 ) . This is likely to reflect mobility of the ubiquitin peptides . Consistent with this , the electron density determined from X-ray diffraction on crystallised nucleosomes with ubiquitin conjugated at this location resulted in no density attributable to ubiquitin ( Machida et al . , 2016 ) . On the wrapped side of the nucleosome , ubiquitin is located adjacent to the acidic patch that is widely used as an interface for nucleosome binding proteins ( McGinty and Tan , 2016 ) ( Figure 8B ) . This is also the location that ubiquitin conjugated to H2A K15 has been observed to occupy on unbound nucleosomes ( Wilson et al . , 2016 ) and likely represents a favourable conformation for ubiquitin when coupled at different sites within this locality ( Vlaming et al . , 2014 ) . On the unwrapped side of the nucleosome , the ubiquitin peptide is displaced across the lateral surface towards the DNA . The unwrapped DNA is oriented away from the lateral surface and this positions the DNA backbone at SHL6 in contact with the repositioned ubiquitin ( Figure 8B ) . In 2D classes , ubiquitin is more prominent on the unwrapped side suggesting it is more tightly constrained . The residues interacting with DNA include Lys 48 Arg 54 and Asp 60 . It is likely this interaction stabilises DNA in the unwrapped state . Using a fluorescence-based assay , H2BK120ub has previously been observed to stimulate the activity of Chd1 118–1274 approximately 2-fold ( Levendosky et al . , 2016 ) . Using a gel-based assay with Chd1 1–1305 , we observe a slightly greater 3-fold effect ( Figure 8—figure supplement 2 ) .
The Chd1 enzyme has the ability to organise spaced arrays of nucleosomes both in vitro and in vivo ( Gkikopoulos et al . , 2011; Lusser et al . , 2005; Robinson and Schultz , 2003 ) . Enzymes that exhibit this organising activity typically reposition nucleosomes away from the ends of short DNA fragments . This is also true for Chd1 ( McKnight et al . , 2011; Stockdale et al . , 2006 ) . Repositioning of nucleosomes with this directionality conflicts with the directionality of translocation inferred from docking the tracking strand of NS3 into Chd1 ( Farnung et al . , 2017 ) ( Figure 6 ) . Tracking along this strand with 3’−5’ directionality would instead be anticipated to draw DNA into the nucleosome from the side of the nucleosome that has no linker DNA . Inferring the mechanism of Chd1 from NS3 is complicated by the fact these enzymes are not so closely related . Conserved motifs are difficult to align based on sequence alone . In addition some aspects of nucleic acid binding by both Snf2 and Chd1 profoundly differ from NS3 . Notably , motifs II and III within lobe one contact the opposite , 3’−5’ stand , which is not present in NS3 . In addition , Snf2-related chromatin remodelling enzymes contain features that extend the nucleic acid binding cleft between the two ATPase lobes and make contacts with both strands ( Figure 6 ) . As a result of the extensive contacts with both strands , it is possible that the assignment of guide and tracking strands within remodelling ATPases is not absolute as tracking may be coupled to both strands . Consistent with this , experiments that have probed the action of remodelling enzymes using short gaps in either strand of nucleosomal DNA have found them to be sensitive to lesions in either strand ( Saha et al . , 2005; Zofall et al . , 2006 ) . The introduction of gaps in nucleosomal DNA has also been used to infer the directionality with which ATPases’s move along DNA . Introduction of gaps distal to the SHL two location closest to the entry linker DNA has been observed to impede the action of Snf2 , Iswi and Chd1 enzymes ( McKnight et al . , 2011; Saha et al . , 2005; Zofall et al . , 2006 ) . This has been used as evidence that the enzyme translocation that drives repositioning initiates from the SHL two located distal to the entry DNA . As a consequence it has been proposed that Chd1 bound in the cross gyres conformation , that we and others observe , represents an inactive state in which the interaction of the DNABD with ( exit ) linker DNA is inhibitory ( Nodelman et al . , 2017 ) . Confounding this , the Chd1-nucleosome structure bears the hallmarks of an active DNA translocase . Density for ADP . BeF is observed in the nucleotide binding pocket , and the ATPase domains are repositioned to a closed conformation with conserved residues positioned for catalysis ( Figure 2—figure supplement 3 ) . The closure of the ATPase domains in the Chd1-nucleosome complex is connected to repositioning of the chromodomains , which in turn levers the DNABD position and allows DNA to gain access to the cleft between the ATPase domains . Further study will be required to reconcile these observations . We nonetheless speculate that the initial stages of remodelling by Chd1 are most effective on nucleosomes lacking exit DNA . Consistent with this the initial rate of Chd1 ATPase activity has been observed to be greater on nucleosomes lacking exit linker DNA ( Nodelman et al . , 2017 ) . In cells , lack of linker could arise in close packed nucleosomes and represent a suitable substrate for a nucleosome spacing enzyme . The rapid action of Chd1 on close packed chromatin would generate new exit linker DNA which would then be available to be bound by the DNABD in the conformation observed by in the EM structures . In this case , Chd1 bound in the conformation shown in Figure 1 , could be active but represent a state in which the DNABD is bound to exit rather than entry DNA . This reconciles many of the biochemical observations , and is compatible with the directionality inferred from the interaction of DNA with the ATPase domains . However , it requires that the Chd1 ATPase domains exhibit a preference for initial binding to nucleosomes on the side that initially lacks linker DNA . The basis for this is not clear , but in the case of ISWI enzymes the initial and subsequent translocation events have been observed to differ ( Deindl et al . , 2013 ) , raising the possibility that they could be regulated independently . Following initial repositioning , the engagement of the DNABD with the new exit linker may provide an opportunity for large changes in enzyme conformation , perhaps related to those observed for SNF2H ( Leonard and Narlikar , 2015 ) and enabling individual Chd1 molecules to switch between different sides of the nucleosome enabling bidirectional repositioning ( Qiu et al . , 2017 ) . Both budding yeast Chd1 and human Chd2 are found to be enriched within coding regions ( de Dieuleveult et al . , 2016; Gkikopoulos et al . , 2011; Lee et al . , 2017 ) . Histone H3 K36me3 is a hallmark of coding region nucleosomes , so we prepared nucleosomes alkylated to mimic trimethylation at this position . Alkylation modestly stimulates Chd1 activity ( Figure 1—figure supplement 1 ) , raising the possibility that this modification is recognised by the enzyme , possibly via the chromodomains . However , we observe electron density for the histone H3 tail to residue 26 , indicating that H3K36 does not stably interact with the chromodomains or any other component of Chd1 in the structure reported here . Furthermore , for this interaction to occur , either the chromodomains would need to be repositioned , or the structure of the N-terminus of H3 reconfigured for example by unfolding of the alpha–N helix ( Elsässer et al . , 2012; Liu et al . , 2012 ) . The improved density for the H3 tail on the unwrapped side of the nucleosome is most likely to result from the interaction of the basic N-terminal region of the H3 tail , which is not resolved , with DNA . It is notable that in the fully wrapped state the H3 tail would need to follow a very different path in order to interact with DNA . Consistent with this the trajectory of the H3 tail on the unwrapped side of the nucleosome is different to that observed in structures of intact nucleosomes ( Wilson et al . , 2016 ) . This raises the possibility that changes to DNA wrapping could affect the way in which histone tail epitopes are displayed . In principle , such effects could be positive or negative . For example the tudor domain of PHF1 preferentially interacts with trimethylated H3K36 on partially unwrapped nucleosomes ( Gibson et al . , 2017 ) . The interaction of the PHD domains of Chd4 with DNA is also inhibited by nucleosomal DNA ( Gatchalian et al . , 2017 ) . As a result if Chd4 generates unwrapped structures similar to those observed with Chd1 the interaction of these domains would be enhanced . The reconfiguration of the H3 tail by Chd1 has the potential to affect the interaction of histone reader , writer and eraser enzymes with the tail and as a result the distribution of these modifications in chromatin . Such effects have been observed , as Chd1 contributes to the establishment of boundaries between H3K4me3 and H3K36me3 at most transcribed genes ( Lee et al . , 2017 ) . H2BK120ub is also enriched in coding region chromatin , and directly stimulates Chd1 activity ( Levendosky et al . , 2016 ) ( Figure 8—figure supplement 2 ) ( Sundaramoorthy et al . , 2017 ) . It has also been observed that H2BK120ub negatively affects the activity of some ISWI containing enzymes ( Fierz et al . , 2011 ) . As organisation of coding region nucleosomes involves these and other enzymes ( Krietenstein et al . , 2016; Ocampo et al . , 2016; Parnell et al . , 2015 ) , H2BK120Ub has the potential to regulate interplay between different enzymes . Ubiquitin on the unwrapped side of the nucleosome is repositioned such that it interacts directly with DNA . As in the case of the H3 tail , the repositioning of the ubiquitin resulting from Chd1-directed DNA unwrapping could potentially affect interactions with the factors involved in the placing , removal or recognition of H2BK120ub . The most striking functional evidence for this interplay is that H2BK120ub is greatly reduced in Chd1 mutants ( Lee et al . , 2012 ) . One possible explanation for this effect is that Chd1 sequesters ubiquitin in a conformation less accessible for removal . Consistent with this , the position of ubiquitin on the unwrapped side of Chd1 bound nucleosomes is incompatible with interaction with the SAGA DUB module ( Morgan et al . , 2016 ) . Interestingly , the paradigm for trans regulation between histone modifications stems from the interplay between H2BK120ub and H3 K4 methylation ( Sun and Allis , 2002 ) , both of which are influenced by Chd1 binding . While Chd1 is not required for H3K4me3 ( Lee et al . , 2012 ) , it does influence the distribution of this histone modification ( Lee et al . , 2017 ) . H2BK120ub has previously been observed to directly affect chromosome structure at the level of chromatin fibre formation ( Debelouchina et al . , 2017; Fierz et al . , 2011 ) . Our observations show a new role for H2BK120ub at the level of nucleosomal DNA wrapping . The specific relocation of ubiquitin on the unravelled side of the nucleosome , the local distortion of H2B at the site of attachment and the presence of lysine and arginine residues at the site of interaction with DNA all indicate this is a favourable interaction that stabilises DNA in the unwrapped state . The outer turns of nucleosomal DNA rapidly associate and dissociate on millisecond time scales , with occupancy of the unwrapped state estimated at 10% ( Li et al . , 2005 ) . The ubiquitin interaction we have observed would be anticipated to stabilise the transiently unwrapped state increasing its abundance . It is however , unlikely that the unwrapped state predominates in the absence of Chd1 or other factors that promote unwrapping as the structure of isolated ubiquitinylated nucleosomes is unchanged ( Machida et al . , 2016 ) . Nonetheless , increased occupancy of the transiently unwrapped state would be anticipated to facilitate access to nucleosomal DNA . Chromatin folding to form higher order structures is likely to be favoured by fully wrapped nucleosomes , and so an increase in the proportion of unwrapped nucleosomes could potentially contribute to the effects of H2BK120ub on chromatin fibre formation ( Fierz et al . , 2011 ) . Many other processes involving chromatin dynamics are linked to H2BK120ub including transcription ( Bonnet et al . , 2014 ) , DNA repair ( Moyal et al . , 2011; Nakamura et al . , 2011 ) and DNA replication ( Lin et al . , 2014 ) . A more stable unwrapped state could also provide an explanation for the association of factors that lack recognised ubiquitin interaction domains , with ubiquitinylated chromatin ( Shema-Yaacoby et al . , 2013 ) . Interestingly , H2BK120ub associating proteins include human Chd1 , SWI/SNF complex , pol II and the elongation factors NELF and DISF ( Shema-Yaacoby et al . , 2013 ) . The change in the position of ubiquitin also has the potential to indirectly affect the way in which other factors interact with ubiquitinylated nucleosomes . On the wrapped side of the nucleosome , ubiquitin is positioned such that it occludes access to the acidic patch formed by the cleft between histones H2A and H2B . This provides a surface via which many proteins including LANA peptides ( Barbera et al . , 2006 ) , RCC1 ( Makde et al . , 2010 ) , Sir3 ( Armache et al . , 2011 ) , PRC1 ( McGinty et al . , 2014 ) and the SAGA DUB module ( Morgan et al . , 2016 ) interact with nucleosomes . The repositioning of ubiquitin away from the acidic patch on the unwrapped side of the nucleosome improves access to the acidic patch . In this way , H2BK120ub may provide a means of regulating access to the acidic patch that is sensitive to changes in nucleosome structure . Although the repositioning of the H3 tail and ubiquitin were observed on Chd1-bound nucleosomes , the potential for reconfiguration of histone epitopes may be more general . All processes that generate local DNA unwrapping would be anticipated to result in similar repositioning of histone tail epitopes . In particular , where combinations of modifications are recognised bivalently , the spatial alignment of epitopes will be important for recognition by coupled reader domains . This potentially provides a means of tuning signalling via histone modifications to regions of transient histone dynamics .
ScChd1 C-terminal and N-terminal truncations were made from the full length clone described in Ryan et al , using an inverse PCR strategy ( Ryan et al . , 2011 ) . A similar approach was used to generate a chd1 lobe2 Δ632–646 deletion . Site-directed mutagenesis was used to introduce cysteine residues at strategic locations on ScChd1 1-1305ΔC using standard cloning procedure . All proteins were expressed in Rosetta2 ( DE3 ) pLysS Escherichia coli cells at 20° C in Auto-induction media , and the purification of the protein was carried out typically as described in Ryan et al . After the purification of the protein the GST tag was cleaved with precision protease and the tag cleaved proteins were subjected to size exclusion chromatography using Superdex S200 10/300 GL columns ( GE Healthcare ) . Expression and purification of Xenopus laevis histones were carried out as described previously ( Luger et al . , 1999 ) . Alkylation of cysteine-mutant histones to generate histones modified with methyl-lysine analogues was performed as in ( Simon et al . , 2007 ) . Approximately 10 mg of lyophilised cysteine mutant histone was resuspended in 800 µL ( me3 ) or 900 uL ( me0 ) degassed alkylation buffer ( 1M HEPES , 10 mM D , L-methionine , 4M Guanidine HCl , pH7 . 8 ) . Histones were reduced with fresh 30 mM DTT for 30 min at room temperature . For trimethyl-lysine analogues , the reduced histone was added to approximately 125 mg of ( 2-Bromoethyl ) trimethylammonium bromide ( Sigma 117196–25G ) in 200 µL of DMF and incubated in the dark at 50⁰C for 3 hr . An additional 10 µL of DTT was added , and the reaction was allowed to proceed overnight at room temperature . For generation of the unmethylated lysine analogue , 75 µL of 1M 2-Bromoethylamine hydrobromide ( Fluka 06670–100G ) was added to the reduced histone and was incubated at room temperature in the dark for 3 hr . An additional 10 µL of DTT was added for 30 min prior to the addition of an extra 75 µL of alkylating agent , and the reaction was allowed to proceed overnight at room temperature in the dark . The reaction was terminated with the addition of 50 µL 2-mercaptoethanol for 30 min and the alkylated histone was desalted either by dialysis into water with 2 mM 2-mercaptoethanol or on a PD-10 desalting column ( GE 52130800 ) . The shift in molecular weight associated was confirmed via MALDI-TOF mass spectrometry . Recombinant expression of xH2B K120C and His-TEV-Ubiquitin G76C mutant proteins was induced with IPTG for 4 hr in Rosetta 2 DE3 pLysS cells grown at 37⁰C . Inclusion body purification followed by cation exchange chromatography was performed to isolate the histone protein . Ubiquitin was purified using HisPur cobalt resin with 150 mM sodium chloride/20 mM Tris pH8 buffer and eluted with 350 mM imidazole . Histones and ubiquitin were desalted by dialysis into water with 2 mM 2-mercaptoethanol and lyophilised for storage . Proteins were re-suspended in 50 mM ammonium bicarbonate pH eight and treated with 2 mM TCEP for 1 hr . Ellman’s reagent was used to ascertain the concentration of free sulfhydryls , and xH2b and ubiquitin were combined at equimolar ratios , as defined by the Ellman’s assay , and diluted with 50 mM ammonium bicarbonate to 200–250 uM each protein . The proteins were crosslinked at room temperature with four hourly additions of ¾ molar ratio of 1 , 3 dichloroacetone ( freshly prepared in DMF ) . An equal volume of denaturing buffer ( 7M Guanidine HCl , 350 mM sodium chloride , 25 mM Tris pH7 . 5 ) was added to the reactions , which were purified using HisPur cobalt resin , pre-equilibrated in denaturing buffer . The His-TEV-Ub-xH2B crosslinked product was eluted with 350 mM imidazole and dialysed into SAUDE200 buffer ( 7M Urea , 20 mM sodium acetate , 200 mM sodium chloride , 1 mM EDTA , 5 mM 2-mercaptoethanol ) overnight . The ubiquitinated histone was further purified over a cation exchange column , as before , and fractions were dialysed into water with 2 mM 2-mercaptoethanol and lyophilised for storage . Xenopus H2B-K120 ubiquitinylated histones were refolded in equimolar ratios with H2A and similarly H3 K36 methyl analogue histones were refolded in equimolar ratios with histone H4 to obtain dimers and tetramers as described previously for wild type histones Dyer et al . , and purified on a size exclusion chromatography using S200 gel filtation column . The peak fractions were analysed with SDS-PAGE gel and pooled . 601 DNA fragments of respective lengths for recombinant nucleosome assembly were generated by PCR method as described previously ( Sundaramoorthy et al . , 2017 ) . Nucleosomes were generated by salt dialysis as described previously by combining H2A/H2B-K120 ubiquitin dimer , H3K36 methyl lysine analogue tetramer ( 2:1 ratio ) with DNA containing PCR-amplified Widom 601 DNA sequence . Nucleosomes were reconstituted on Cy3 ( me0 ) and Cy5 ( me3 ) labelled DNA , based on the 601 sequence , with a 47 bp extension . Repositioning by Chd1 was performed in 40 mM Tris pH7 . 4 , 50 mM KCl , 3 mM MgCl2 , 1 mM ATP , 100 nM each nucleosome , and 10 nM Chd1; 10 µL was removed at each time point ( T = 0 , 4 , 8 , 16 , 32 , and 64 min ) , placed on ice , and stopped with the addition of 100 ng/µL competitor DNA , 200 mM NaCl , and 1 . 6% sucrose . Repositioned nucleosomes were run on 6% PAGE/0 . 2X TBE gels in recirculating 0 . 2X TBE buffer for 3–4 hr at 300V . The percent of repositioned nucleosomes was analysed using Aida image analysis software . Data were fit to a hyperbola in Sigma Plot , to determine the initial rate of repositioning . Xenopus laevis nucleosomes ( 20 nM ) , reconstituted on Cy3 labelled 0W11 DNA , were bound to titrations of Chd1 enzymes ( concentration specified in figure legend ) in 50 mM Tris pH 7 . 5 , 50 mM sodium chloride , and 3 mM magnesium chloride supplemented with 100 µg/mL BSA . Unbound and bound nucleosomes were separated on a pre-run 6% polyacrylamide gel ( 49:1 acrylamide: bis-acrylamide ) in 0 . 5X TBE buffer for 1 hr at 150V . The gel shift was scanned on Fujifilm FLA-5100 imaging system at 532 nm . MTSL was conjugated to introduced cysteines immediately following size exclusion purification as described in Hammond et al . ( 2016 ) . Excess unreacted labels were removed from the sample by dialysis . PELDOR experiments were conducted at Q-band ( 34 GHz ) operating on a Bruker ELEXSYS E580 spectrometer with a probe head supporting a cylindrical resonator ER 5106QT-2w and a Bruker 400 U second microwave source unit as described previously ( Hammond et al . , 2016 ) . All measurements reported here were made at 50K . Data analysis was carried out using the DeerAnalysis 2013 package ( Jeschke and Polyhach , 2007 ) . The dipolar coupling evolution data were first corrected to remove background decay . Tikhonov regularisation was then used to determine distance distributions from each dataset . To model the distance distribution for the open conformation of Chd1 helicase lobes crystal structure of chromo helicase ( PDB Code: 3MWY ) ( Hauk et al . , 2010 ) was used . For the closed conformation refined cryoEM structure of Chd1 bound to nucleosome in the presence of ADP . BeFx described in this study was used as a model . For each structure , R1 spin labels were added and the distribution simulated for each position using MTSSL wizard in Pymol . Also the average distance from the distribution from a pair of spin labels were calculated using MTSSL wizard in Pymol . The appropriate ratio of ScChd1 ( 1-1305Δ57–88 ) to nucleosome for 1:1 and the 2:1 complex formation in the presence of 5-fold molar excess of ADP-BeFx was determined by titration and native PAGE analysis . The formed complex was then purified by size exclusion gel filtration using a PC 3 . 2/30 superdex 200 analytical column in 20 mM Tris , 50 mM NaCl and 250 µM ADP . BeFx . ( formed by mixing 1:1:3 molar ratio of ADP:BeCl2:NaF ) . In a typical run 50uLs of 20 µM of sample was injected using Dionex autoloader . 50uLs fractions were collected and analysed in 6% Native PAGE gel and appropriate fractions containing ScChd1-nucleosome complexes were pooled together . A 4 µl drop of sample was then applied to C-flat Holey carbon foil ( 400 mesh R1 . 2/1 . 3 µM ) pre-cleaned with glow discharge ( Quorum technologies ) . After 15 s incubation , grids were double side blotted for 4 s in a FEI cryo-plunger ( FEI Mark III ) at 90% humidity and plunge frozen into −172°C liquefied ethane . Standard vitrobot filter paper Ø 55/20 mm , Grade 595 was used for blotting . The prepared grids are initially checked for its ice quality and the particle distribution using a JEOL 2010 microscope with side-entry cryo-holder operated at 200 keV and equipped with a gatan 4k × 4 k CCD camera . Cryo-grids were then stored in liquid nitrogen and dry-shipped to respective centre for data collection . For the 1:1 complex the data were acquired on a FEI Titan Krios transmission electron microscope ( TEM ) operated at 300 keV , equipped with a K2 summit direct detector ( Gatan ) . The 2:1 complex data were collected with FEI Titan Krios microscope equipped with Falcon three detector . Automated data acquisition was carried out using FEI EPU software at a nominal magnification of 105 , 000 × for the 1:1 complex and 59 , 000X for the 2:1 complex . For both the datasets , the data were collected as a movie with 32 frames per movie for the 1:1 complex and 40 frames per movie for the 2:1 . Details of the data collection and processing parameters for both the datasets are included in Table 1 . The movie frames were subjected to frame wise motion correction using MotionCor2 ( Zheng et al . , 2017 ) . CTF correction was then performed on the motion corrected summed image using GCTF ( Zhang , 2016 ) . Subsequent image processing was performed with RELION 2 . 0 . 4 ( Scheres , 2012 ) . About 5000 particles from 50 micrographs were first handpicked in RELION , extracted and 2D classes were generated . These 2D classes were then used as a reference in RELION auto pick routine and particles were picked from respective number of micrographs from each dataset . The auto picked particles were subsequently extracted and sorted . An iterative round of two-dimensional classification was performed to discard poorly averaging particles , contamination and exploded particles . On the resultant cleaned up particle stack a hierarchical three-dimensional classification and refinement was performed as described in the results . A low pass filtered low resolution chd1 engaged nucleosome structure was used as an initial model in the 3D classification . At the three-dimensional refinement stages a soft mask that encompasses the entire Chd1-nucleosome complex was applied . Post processing of refined models was performed with automatic B factor determination in RELION . Resolution for the refined reconstruction are determined at 0 . 143 FSC cut off as 4 . 5 Å for the 1:1 complex and 10 Å for the 2:1 Complex . Local resolution estimates were determined for the 1:1 complex using Resmap-1 . 4 . For model building X . laevis nucleosome with Widom 601 sequence ( PDB 3LZ0 ) , the S . cerevisiae Chd1 DNA-binding domain ( PDB 3TED ) and the ATPase core with tandem chromo domain ( 3MWY ) were used . The domains were individually placed into the electron density using UCSF chimera and fitted as a rigid body . The path of the unwrapped DNA , H4 tail region and H3 tail region were manually built in Coot . Protein back bone restraints and DNA base pair , and parallel pair restraints were generated using ProSMART and LibG modules ( Nicholls et al . , 2012 ) . The generated restraints were then used as constraint in jelly body refinement with CCPEM REFMAC program . ADP·BeF3 was built by superpositioning ATP-gamma-S from the inactive Chd1 structure ( PDB code 3MWY ) onto our model , inspected in COOT and replacing the ATP analogue with ADP·BeFx . For the 2:1 complex the structural model of 2chd1 bound to one nucleosome with 14 bp symmetrical linker on both side was generated using one chd1 bound to 0W14 nucleosome presented in this study and the one chd1 bound to 63W0 nucleosome ( PDB-ID 5O9G ) . The model was then rigid body docked into the reconstructed 2chd1:1nucleosome volume in chimera . The quality of the fit is assessed by visual examination and the correlation coefficient between the model and the map . Sequence alignments were generated using JALVIEW ( Waterhouse et al . , 2009 ) . The EM volumes and the fitted model can be accessed from the EMDB database with EMDB accession number EMD-4318 and PDB code 6FTX for the 1:1 complex and EMD-4336 and EMD-4336 for the 2:1 complex . | The DNA inside cells contains all the information needed to build an organism . Human DNA measures about 2 metres . To condense it , DNA is wrapped around eight histone proteins to form disc-like structures , called nucleosomes . Nucleosomes are further compressed into chromatin fibres that make up our chromosomes . The way DNA is packaged and positioned into the nucleosomes can be variably controlled and affects how genes are switched on and off . Although all cells have the same DNA , the way specific genes are turned on and off gives rise to the different types of cells in our body . Specialised motor proteins , called ‘chromatin remodellers’ , control the positioning of nucleosomes inside cells . In yeast cells , for example , the protein Chd1 moves nucleosomes along the DNA so that they are evenly spaced . So far , it has been unclear how chromatin remodellers interact with nucleosomes . To investigate this further , Sundaramoorthy et al . studied the structure of Chd1 bound to a nucleosome , in which one histone protein was modified with a molecule , called ubiquitin , which is present on genes where Chd1 is known to be active . The structure revealed that both Chd1 and the nucleosome did not have their usual shape . Moreover , Chd1 partially unwrapped DNA from the nucleosomes . As a consequence , the ubiquitin moved to interact with the unwrapped DNA; as did a flexible area on one of the histones , known as ‘histone tail’ . Both ubiquitin and histone tails play important roles in signalling processes on chromatin . Therefore , such a rearrangement could affect the transmission of signals from chromatin . The organisation of nucleosomes affects the accessibility of the underlying DNA . As a result , any process that happens on DNA is affected – including controlling when genes are turned on and off under normal conditions , and when things go wrong during diseases . A better knowledge of how the organisation of nucleosomes is controlled will improve our understanding of gene regulation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2018 | Structure of the chromatin remodelling enzyme Chd1 bound to a ubiquitinylated nucleosome |
Malaria transmission is spatially heterogeneous . This reduces the efficacy of control strategies , but focusing control strategies on clusters or ‘hotspots’ of transmission may be highly effective . Among 1500 homesteads in coastal Kenya we calculated ( a ) the fraction of febrile children with positive malaria smears per homestead , and ( b ) the mean age of children with malaria per homestead . These two measures were inversely correlated , indicating that children in homesteads at higher transmission acquire immunity more rapidly . This inverse correlation increased gradually with increasing spatial scale of analysis , and hotspots of febrile malaria were identified at every scale . We found hotspots within hotspots , down to the level of an individual homestead . Febrile malaria hotspots were temporally unstable , but 4 km radius hotspots could be targeted for 1 month following 1 month periods of surveillance .
The transmission of infectious disease often shows substantial heterogeneity ( Woolhouse et al . , 1997 ) . Malaria transmission is determined by mosquito ecology and behavior , which is in turn determined by rainfall , hydrology , soils , human behavior and population distributions , and a range of other social , biotic and abiotic factors . Heterogeneity of malaria transmission is apparent at global scale ( Gething et al . , 2011 ) , regional scale ( Kleinschmidt et al . , 2001a; Noor et al . , 2009 ) , and at fine scale in , for instance , Mali ( Gaudart et al . , 2006 ) , Ghana ( Kreuels et al . , 2008 ) , Ethiopia ( Yeshiwondim et al . , 2009 ) Kenya ( Brooker et al . , 2004; Ernst et al . , 2006; Bejon et al . , 2010 ) , and Tanzania ( Bousema et al . , 2010 ) . This spatial heterogeneity makes transmission relatively resilient to indiscriminate control efforts , but also provides an opportunity to engage in targeted malaria control on clusters of transmission ( or ‘hotspots’ ) , a strategy that is predicted to be highly effective ( Dye and Hasibeder 1986; Woolhouse et al . , 1997 ) . We have previously identified hotspots of malaria using active surveillance ( Bejon et al . , 2010 ) . Others have identified hotspots using passive surveillance in health facilities linked to demographic surveillance systems ( Ernst et al . , 2006 ) . Passive surveillance is more readily scaled up , but may be biased by variations in access to health care facilities and socially-determined health-seeking behavior ( Sumba et al . , 2008; Franckel and Lalou 2009 ) . The incidence of febrile malaria presenting to health care is thus biased by access to care . This bias may be countered by using the malaria positive fraction ( MPF ) among children with fever ( also termed ‘slide positivity rate’ in some publications [Jensen et al . , 2009] ) . The MPF includes all febrile children presenting to the dispensary as the denominator , hence controlling for access to health care , in contrast to incidence for which all children in the community are included in the denominator . The MPF is less likely to show systematic spatial bias with distance from the health facility since parental accounts of illness have not been found to discriminate malaria from non-malarial fever ( Luxemburger et al . , 1998; Mwangi et al . , 2005 ) , and diagnostic testing is not available outside the dispensary . We present data from demographic surveillance linked to passive case detection in Pingilikani dispensary in Kilifi District , coastal Kenya . Data are collected from 1500 homesteads within an 8 km radius followed for 9 years . We analyse the spatial heterogeneity of malaria cases in order to determine the temporal and spatial scales of case clustering so as to inform targeting in malaria control programmes . We also excluded visits with specific symptoms such as skin infections or cutaneous abscesses , otitis media , and gastroenteritis ( >4 episodes diarrhoea per day ) that might have been the primary motivation for seeking health care rather than fever per se .
MPF was inversely correlated with the average age of children with malaria , Spearman's rank correlation ( rs ) = −0 . 16 , p<0 . 0001 ( Figure 1A–C ) . This suggests that greater exposure to malaria ( i . e . , high MPF ) leads to more rapid acquisition of immunity as children grow up , hence predominantly younger children visiting the dispensary with febrile malaria . There was no evidence that this relationship was confounded by spatial clustering of age: the average age of children with non-malarial fever did not show spatial clustering ( Moran's I = 0 . 01 , p=0 . 5 within 1 km and Moran's I = 0 . 02 , p=0 . 5 within 5 km ) and was not associated with MPF ( rs = −0 . 02 , p=0 . 4 ) . We examined the effect of spatial scale at which this correlation occurred by imposing grids of increasing cell size on the study area , calculating rs within each cell of the grid , and then estimating the mean rs at each scale of grid ( Figure 1D , blue lines ) . The mean rs trended gradually away from 0 as the grid divisions became larger in scale . This pattern suggests gradual differentiation in transmission characteristics as the distance between homesteads included within a cell of the grid increases . We then examined the patterns seen on applying this analysis to simulated data . In order to exclude that this trend was a result of cells at fine-scale containing fewer homesteads , we ran permutations of the data using after randomly re-assigning spatial coordinates to the homesteads . These permutations show that a consistent correlation at rs = −0 . 16 throughout the range of grid sizes , albeit with greater uncertainty with smaller cell size ( Figure 1D , red lines ) . Hence , the trend of a gradually increasing inverse correlation as the grid size increases does not appear to be explained simply by having fewer homesteads in each cell at fine scale . In order to determine the pattern that might be seen with specific spatial scales of clustering , we conducted further simulations by imposed patterns with specific scales on the spatial coordinates of the homesteads , in varying proportions with random noise using a gamma distribution . These simulations show that a specific scale of clustering produces ‘spikes’ in rs as the cell size varies , with the position of the spike coinciding with scale of the clustering ( Figure 1—figure supplement 1 ) . Reducing the Signal:Noise ratio eventually obscured the ‘spikes’ due to a characteristic pattern , but only at the point where the overall correlation was no longer discernible ( Figure 1—figure supplement 2 ) . Adding a gradient to the simulated characteristic scale attenuated but did not obscure the ‘spikes’ ( Figure 1—figure supplement 3 ) . 10 . 7554/eLife . 02130 . 005Figure 1 . Geographical distribution of malaria positive fraction and average age of febrile malaria . Each plotted point represents an individual homestead , where the colour shading indicates the malaria positive fraction ( MPF ) in panel A , or the average age of children who test positive for malaria in panel B . Panel C shows the scatter plot for MPF vs average age ( Spearman's rank correlation coefficient ( rs ) = −0 . 16 , p<0 . 0001 ) . Panel D shows rs ( y axis ) plotted against scale of analysis ( x axis ) , where a grid with varying cell size is imposed on the study area , rs is calculated within each cell and then the mean rs presented , with 95% confidence intervals produced by boot-strap ( blue solid and dashed lines , respectively ) , and the results of analysis of spatially-random permutations of the data with equivalent cell size are shown for comparison ( red solid and dashed lines , respectively ) . The analysis shown in panel D was compared on simulations with varying simulated characteristic scales , Signal:Noise ratios and with added gradients ( Figure 1—figure supplements 1–3 , respectively ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02130 . 00510 . 7554/eLife . 02130 . 006Figure 1—figure supplement 1 . Simulated data with varying imposed scales of clustering . Simulated data using imposed spatial clustering at specific scales are analysed to determine rs ( y axis ) plotted against scale of analysis ( x axis ) , where a grid with varying cell size is imposed on the study area , rs is calculated within each cell and then the mean rs presented , with 95% confidence intervals produced by boot-strap ( blue solid and dashed lines , respectively ) . The six panels show the appearances of different imposed scales as shown in the sub-titles . DOI: http://dx . doi . org/10 . 7554/eLife . 02130 . 00610 . 7554/eLife . 02130 . 007Figure 1—figure supplement 2 . Simulated data with varying signal to noise ratios . Simulated data using imposed spatial clustering at specific scales are analysed to determine rs ( y axis ) plotted against scale of analysis ( x axis ) , where a grid with varying cell size is imposed on the study area , rs is calculated within each cell and then the mean rs presented , with 95% confidence intervals produced by boot-strap ( blue solid and dashed lines , respectively ) . The six panels show the appearances using different Signal:Noise ratios . DOI: http://dx . doi . org/10 . 7554/eLife . 02130 . 00710 . 7554/eLife . 02130 . 008Figure 1—figure supplement 3 . Simulated data with varying gradients around imposed scales of clustering . Simulated data using imposed spatial clustering at specific scales are analysed to determine rs ( y axis ) plotted against scale of analysis ( x axis ) , where a grid with varying cell size is imposed on the study area , rs is calculated within each cell and then the mean rs presented , with 95% confidence intervals produced by boot-strap ( blue solid and dashed lines , respectively ) . The six panels show the appearances using gradients of varying spatial scales around the simulated clustering . DOI: http://dx . doi . org/10 . 7554/eLife . 02130 . 008 Using the Bernoulli model in SaTScan ( Kulldorff , 1997 ) , we identified a hotspot with a radius of 5 . 8 km at p<0 . 00001 ( Figure 2A ) using the full data set ( for which n = 20 , 702 ) . However , on re-analysis of the children within this hotspot ( in which n = 5300 ) , we identified a further hotspot ( with a radius of 0 . 76 km ) within the 5 . 8 km hotspot ( p<0 . 00001 , Figure 2B ) . Then on further re-analysis of the homesteads within that 0 . 76 km hotspot ( within which n = 1406 ) , we identified a third significant hotspot ( p=0 . 016 ) which comprised a single homestead , in which there were 36 episodes of malaria compared with 3 malaria negative fevers ( Figure 2D ) . When we selected a random 5-km square area outside the original 5 . 8 km radius hotspot , we identified a hotspot within this area a fourth hotspot with a 1 . 32 km radius ( p<0 . 00001 , Figure 2C ) . 10 . 7554/eLife . 02130 . 009Figure 2 . Hotspots within hotspots . Each plotted point represents an individual homestead , where the colour shading indicates the malaria positive fraction ( MPF ) . Hotspots are identified using SATScan , using the whole study area ( panel A ) , then repeated within the hotspot ( panel B ) , within the hotspot of panel B ( panel D ) , and then within a randomly chosen area outside the hotspot ( panel C ) . The semi-variogram and log–log semi-variogram plot are shown in Figure 2—figure supplements 1 and 2 , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 02130 . 00910 . 7554/eLife . 02130 . 010Figure 2—figure supplement 1 . Semi-variogram . The semi-variogram is shown for MPF . A lowess smoothed line is superimposed on the data points . DOI: http://dx . doi . org/10 . 7554/eLife . 02130 . 01010 . 7554/eLife . 02130 . 011Figure 2—figure supplement 2 . Log-log plot of semi-variogram . The log–log plot of the semi-variogram is shown for MPF . A lowess smoothed line is superimposed on the data points . DOI: http://dx . doi . org/10 . 7554/eLife . 02130 . 011 To further explore the scale of spatial clustering , we plotted the semivariogram ( Figure 2—figure supplement 1 ) and the log–log transformed semivariogram ( Figure 2—figure supplement 2 ) . These plots suggested linear fits for the semivariogram , suggesting that spatial clustering occurred over a range of spatial scales . We also examined temporal trends for individual homesteads ( Figure 3 ) . There was an inverse correlation between the mean MPF and the variance in MPF over the 10-year study period ( rs = −0 . 61 , p<0 . 0001 , Figure 3A ) . The temporal trends for two subsets of homestead can be seen in Figure 3B ( stable high MPF ) and Figure 3C ( unstable low MPF ) , suggesting that homesteads can be characterized as stable high transmission homesteads or unstable low transmission homesteads . Infant parasite rates have been proposed as a measure of transmission intensity that minimizes the offsetting of acquired immunity in macro-epidemiological studies ( Snow et al . , 1996 ) . We therefore hypothesized that the malaria positive fractions in children <1 year of age ( hereafter ‘MPF<1yr’ ) would measure transmission intensity without the offsetting of acquired immunity , and that unstable transmission would result in higher risk of malaria in older children . To test this hypothesis , we calculated the mean MPF<1yr and the variance in MPF<1yr for each homestead over the 9 years of follow up and tested the relationships between these metrics and risk of malaria in older children in multivariable linear regression models . 10 . 7554/eLife . 02130 . 012Figure 3 . Temporal variations in malaria positive fraction . ( Panel A ) shows the scatter plot of individual homesteads by mean malaria positive fraction ( MPF ) on the x axis vs variance in MPF on the y axis ( rs = −0 . 61 , p<0 . 0001 ) . A labelled blue circle indicates subset q ( homesteads with high variance but low mean MPF ) and subset p ( homesteads with low variance and high mean MPF ) . The temporal trends for these two subsets are shown on panels ( B and C ) , respectively . The median trend for the study area is shown in red . ( Panel D ) shows the regression coefficients ( y axis ) for the malaria positive fractions ( MPF ) in older children when regressed on; ( i ) the mean MPF in children <1 year of age ( MPF<1y ) and ( ii ) MPF in older children when regressed on the variance in MPF<1y over the 9 years of the study . Separate multivariable regression models ( i . e . , with mean MPF<1y and variance in MPF<1y as explanatory variables ) are fit for each age group as shown on the x axis ( excluding children <1 year of age , whose data are used to calculate MPF<1y ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02130 . 012 In multivariable linear regression models , MPF<1yr was strongly correlated with MPFs in children in the 1- to 2-year-old and 2- to 3-year-old age group , but progressively less strongly correlated with MPF in older children ( Figure 3Di ) . The regression coefficient was ∼0 . 4 for 1–2 year olds , meaning that each unit increase in MPF<1yr is associated with a 0 . 4 increase in the MPF for 1- to 2-year-old children . On the other hand , the variance in MPF<1yr was not correlated with MPFs in 1- to 2- or 2- to 3-year-old children , but was progressively more strongly correlated with MPF in older children ( Figure 3Dii ) . Hence there were high stable transmission homesteads , with predominantly younger children getting febrile malaria , and low unstable transmission homesteads , with increasing risk to older children . This pattern of high stable vs low unstable transmission also occurs between regions or countries , and demonstrates a similarity between the micro- and macro-epidemiology of malaria ( Hay et al . , 2008 ) . We then used our data set to simulate the accuracy of targeting cases that a malaria control programme might achieve on conducting surveillance over a defined period of time followed by targeted control . We assumed that malaria control programmes would need to define a priori the period of time to use for surveillance , and also to select a spatial scale at which to define hotspots . For varying time periods and spatial scales , we determined the % of excess malaria cases within the targeted hotspots compared with the surrounding area in the period of time immediately following the simulated surveillance . One week periods of surveillance ( top left panel of Figure 4 ) did not identify hotspots that are still present the following week at fine spatial scales ( i . e . , the plotted line indicates that the accuracy of targeting is 0% at scales of less than 1 km ) . On the other hand , at larger spatial scales we found that 1 week periods of surveillance were more accurate , resulting in the targeting of areas with a 60% excess of new malaria cases compared with the surrounding area at a scale of an 8 km diameter . A similar pattern was seen for monthly periods of surveillance . Longer surveillance periods ( e . g . , 6 months ) resulted in targeting areas with an excess of 20% malaria cases compared with the surrounding area over the range of spatial scales examined . 10 . 7554/eLife . 02130 . 013Figure 4 . Theoretical accuracy of targeted control undertaken at varying temporal and spatial scales . The accuracy of varying strategies of hotspot identification is shown . Each panel is labelled with the time period of surveillance data used . The x axis shows the diameter of hotspot defined . In each case hotspots were selected to account for 20% of the homesteads in the area . The y axis shows the increase that would have been present assuming that they were targeted in the time period following their identification . DOI: http://dx . doi . org/10 . 7554/eLife . 02130 . 013 Mass distributions of Insecticide Treated Nets ( ITNs ) in the area began in 2006 . ITN use was surveyed in 2009 and 2010 . We found that children using ITNs had a reduced risk of malaria by logistic regression ( i . e . , OR = 0 . 69 , 95%CI 0 . 67 to 0 . 8 , p<0 . 001 ) , in keeping with previous literature on the personal protection provided by ITN use ( Lim et al . , 2011 ) . On the other hand , we did not identify significant evidence that ITN use was clustered spatially ( Moran's I = 0 . 02 , p=0 . 5 ) . Furthermore , adding ITN use as a covariate in SaTScan analysis to locate hotspots had little effect on results; the addition of ITN use as a covariate changed the location of the hotspot by 120 m , and changed the predicted radius of the hotspot from 5 . 4 km to 5 . 2 km . On re-analysis of the homesteads within the 5 . 4 km hotspot , a further 0 . 87 km hotspot was identified the position and radius of which were not altered by the inclusion of ITN use as a covariate . Finally , within this 0 . 87 km hotspot the same 7 homesteads were identified as a hotspot irrespective of the inclusion of ITN use as a covariate . We did not identify significant evidence that ITN use correlated mean MPF<1yr ( rs = −0 . 04 , p=0 . 04 ) or with the variance in MPF<1yr ( rs = −0 . 01 , p=0 . 7 ) . Hence , ITNs provided personal protection from malaria , but we were unable to show that they explained the spatial micro-epidemiological patterns .
We found that malaria cases were spatially heterogeneous in an 8-km radius area of coastal Kenya . The strongly significant inverse correlation between the malaria positive fraction ( MPF ) and average age of children presenting with malaria suggests variable acquisition of immunity between homesteads . Homesteads at high transmission intensity have a high MPF and a young average age of malaria ( with older children becoming immune and therefore not presenting to the dispensary ) whereas homesteads at low transmission intensity have a low MPF but an older average age of malaria since older children are not becoming immune as rapidly . In theory , this inverse correlation might have arisen because of heterogeneity at various spatial scales . For instance , there might have been a block of homesteads all at high transmission in one half of the study area ( thus with high MPF and low average age ) and a second block of homesteads at low transmission in the other half ( with low MPF and high average age ) . On the other hand , the inverse correlation might have arisen because of a random distribution of ‘high’ and ‘low’ transmission intensity homesteads throughout the study area . To determine at which spatial scale transmission was heterogeneous , we conducted an analysis where correlation coefficient was recalculated within each cell of a grid superimposed on the study area . The mean correlation coefficient of all cells was then presented as the cell size of the grid used was increased ( Figure 1D ) . This analysis was done to identify the most influential geographical scale at which the inverse correlation was observed . In simulated data , we noted ‘spikes’ where the inverse correlation was abruptly lost when the size of cells in the grid coincides with the size of the geographical ‘blocks’ of homesteads that drove the inverse correlation , as seen in Figure 1—figure supplement 1 . Similar spikes were seen after adding simulated noise and gradients in space over which the correlation varied ( Figure 1—figure supplements 1 , 2 and 3 ) . Real-world data would contain more complex sources of variation than we have simulated , and hence may not produce distinct spikes . Nevertheless , the analysis of these simulations suggests that discontinuities in the correlation between MPF and average age of malaria over cell size might be expected when clustering is at a specific spatial scale . In fact there was no such discontinuity in the function shown in Figure 1D , indicating that the inverse correlation was present at every geographical scale examined within our study . It is likely that this pattern would extend at greater geographical scales , since a similar inverse correlation between the age distributions of malaria cases and transmission intensity can be seen on comparing countries and regions ( Okiro et al . , 2009 ) . The pattern of spatial heterogeneity is relevant to malaria control , since targeted disease control is predicted to be highly effective ( Woolhouse et al . , 1997 ) . Spatial targeting is particularly appropriate for malaria ‘hotspots’ ( Coleman et al . , 2009; Moonen et al . , 2010; Bousema et al . , 2012; Sturrock et al . , 2013 ) and many malaria control programmes are already engaged in spatially-targeted intervention ( Zhou et al . , 2010; Loha et al . , 2012 ) . Our data showing clustering at varying spatial scales suggest that malaria control programs can expect to identify hotspots at many different geographical scales . We demonstrate that hotspots occur within hotspots , down to the level of a single homestead , and also that hotspots can be identified on ‘zooming in’ on random areas outside the main hotspot ( Figure 2C ) . These hotspots were based on analysis of a large dataset with adequate power , and were strongly significant based on the multiple permutations run in SaTScan , suggesting that type I statistical error is an unlikely explanation for our findings . The complexity of presenting ‘hotspots within hotspots’ to a malaria control programme is further compounded by the temporal instability of the spatial pattern ( Figure 3 ) . We therefore simulated the accuracy with which hotspots could be targeted using varying spatial scales and varying time periods of surveillance . We found that using data aggregated over 1 month of surveillance to define 4 to 8 km diameter hotspots would provide greatest accuracy , but this information is only relevant for 1 month before temporal instability necessitates further surveillance . One might therefore consider a continuous programme of parallel surveillance and targeting , where the surveillance data are examined at the end of each month to determine the location to be targeted for the following month . Continuous surveillance would allow adaptive targeting of hotspots for the following month . Such a strategy might be employed all year round , or for a limited period of the year depending on local seasonality . ( Cairns et al . , 2012 ) Targeting at this spatial scale has the added practical advantage that it could be done with village-level location data and would not require fine-scale geo-positional data . There are some caveats to this recommendation . Our observations are from a single site . Other sites should examine their local data to determine whether a similar targeting strategy is appropriate . Furthermore , some hotspots did show temporal stability . For instance , we identified a 6 km diameter hotspot south east of the dispensary that maintained a 30–60% increase in MPF compared with the surrounding area throughout the 9-year surveillance . Children with positive microscopy slides for malaria presenting at the dispensary may have genuine febrile malaria , or alternatively may have chronic asymptomatic parasitaemia with co-incident non-malarial fever . Previous studies estimating malaria attributable fractions in the locality suggest 61% of the children in our analysis would have malaria as the proximate cause of their illness , with the other 39% having chronic asymptomatic parasitaemia with co-incident fever from another cause ( Olotu et al . , 2011 ) . We have previously demonstrated that spatial heterogeneity is more temporally stable when analysed for asymptomatic parasitaemia rather than febrile malaria ( Bejon et al . , 2010 ) . Targeting hotspots of asymptomatic parasitaemia would require community surveys rather than dispensary monitoring , which may need to be done less frequently than monitoring of febrile malaria episodes . Furthermore MPF is not a comprehensive indicator of transmission intensity . Homesteads with consistently low average ages of febrile malaria are likely to be stable high transmission homesteads ( such as those in subset p of Figure 3A ) which amplify transmission in the areas surrounding them . Targeting such high transmission homesteads to interrupt transmission may be highly effective ( Woolhouse et al . , 1997 ) . The stronger inverse correlation between MPF and average age of febrile malaria as spatial scale increases ( Figure 1 ) suggests that the spatial heterogeneity of transmission is progressively more stable at more coarse spatial scales . Malaria transmission is determined by mosquito ecology and behavior . Mosquito ecology may be determined by obvious geographical features such as altitude ( Reyburn et al . , 2005 ) , cultivation practices ( Lindsay et al . , 1991 ) , streams and dams ( Ghebreyesus et al . , 1999 ) , wind direction ( Midega et al . , 2012 ) and mosquito searching behaviour for hosts ( Smith et al . , 2004 ) . Ecological models based on such features have been developed using frequentist techniques ( Omumbo et al . , 2005 ) , Bayesian approaches ( Craig et al . , 2007 ) , and fuzzy logic ( Snow et al . , 1998 ) . However , the same ecological factor may act inconsistently in different geographical areas ( Kleinschmidt et al . , 2001b; Gemperli et al . , 2006; Noor et al . , 2008 ) , and the effect of ecological factors is modified by fine-scale vector and host movement ( Perkins et al . , 2013 ) . Our data suggests that the environmental factors determining malaria transmission operate at a range of spatial scales . We might speculate that mosquito breeding site density could be equally influenced by proximity to a large geographical feature such as a river , or to a micro-geographical feature such as a cow hoof-print ( Sattler et al . , 2005 ) . Hence ecological models of malaria transmission will need to include data at a range of spatial scales in order to accurately predict malaria risk .
Pingilikani Dispensary is 40 km to the North of Mombasa , in Kilifi Country , Coast Province , Kenya . The population relies mainly on subsistence farming and experiences all year round malaria transmission , with ‘long’ and ‘short’ rains each year causing two peaks in transmission . Estimates of the local EIR were 22–53 in 2003 ( 1 ) , and 21 . 7 infective bites per person per year in 2010 ( 2 ) . Between 2003 and 2011 , data were collected on all children ( i . e . , ≤15 years of age ) attending the dispensary . Demographic surveillance is conducted for the 240 , 000 people in a 900 square kilometre area in Kilifi County . Four-monthly enumeration rounds were conducted to identify births , deaths , and migration ( 3 ) . Each inhabitant is described by their family relationships and their homestead of residence , with geospatial coordinates , and assigned a unique personal identifier . These details were used to link children visiting Pingilikani dispensary to geospatial coordinates for the homestead of residence . During enumeration rounds in 2009–2011 ITN use per individual was established during visits to the homestead , as reported by a homestead representative . We restrict analysis to within an 8 km radius of the dispensary , which accounted for >96% of all visits to the dispensary and excluded visits with specific symptoms such as skin infections or cutaneous abscesses , otitis media , and gastroenteritis ( >4 episodes diarrhoea per day ) that might have been the primary motivation for seeking health care rather than fever per se . These latter exclusions combined accounted for 14% of all visits . All children presenting for assessment ( except those with trauma as their only concern ) had finger-prick blood samples examined for malaria parasites . Thick and thin blood smears were stained with 10% Giemsa and examined at x1000 magnification for asexual Plasmodium falciparum parasites . 100 fields were examined before slides could be considered negative . Amodiaquine was the first-line anti-malarial from 2003 to 2005 , when policy changed to Co-artemether . Fever was defined as either reported fever by the parents or measured fever , that is , axilliary temperature ≥37 . 5°C ( Mackowiak et al . , 1997 ) . The malaria positive fraction ( MPF ) was calculated as the fraction of febrile children attending the dispensary with fever who were positive for malaria parasites by blood smear examination . MPF was aggregated by homestead . Multiple identifications of fever and parasitaemia in the same child within 21 days were considered a single episode . The average age of febrile malaria was calculated as the arithmetic mean age at which children visited the dispensary with fever and malaria parasites . Correlations between average age of febrile malaria and MPF per homestead were calculated using spearman's rank correlation coefficient . Grids of gradually increasing cell size were calculated using longitude and latitude coordinates . Simulations were done using the distribution of homesteads identified in our study . We applied a factor to MPF ( positive ) and average age ( negative ) to the homesteads within a block of varying size to induce the appearance of clustering at a given spatial scale . Random noise was added to these simulations using a gamma distribution . In the first round of simulations we set the Signal:Noise ratio ( i . e . , the ratio between the factor applied to MPF and average age vs the mean amplitude of the noise ) to reproduce the rs seen in the real data . In the second round of simulations , we varied the Signal:Noise Ratio as shown in individual panels , and in the third round of simulations we introduced a gradient over which the correlation emerged , where the factor applied to MPF and average age was tapered in a uniform way towards 1 beginning at the edge of the simulated block . Hotspots were defined using SaTScan software to calculate the spatial scan statistic ( Kulldorff , 1997 ) . The software is freely available and can be downloaded from www . satscan . org . The version used in this analysis was downloaded in November 2012 , as v9 . 1 for a 64-bit system . The spatial scan statistic uses a scanning window that moves across space . The scanning windows are circles centred on each homestead , with a radius varied from inclusion of only the single homestead it is centred on through to 30% of the population size . When using the Bernoulli model , the software calculates the fraction of cases/controls inside vs outside the each possible scanning window , and selects the window giving the highest probability of a case within the scanning window compared with the probability of a case outside the window . In our application of the Bernoulli model , cases were febrile children with parasitaemia and controls were febrile children without parasitaemia . The test of significance needs to take into account the whole process of selecting the optimal window rather than simply the comparison of inside vs outside the optimal window . This is achieved by running random permutations of the case/control data over the spatial co-ordinates of homesteads and determining the log-likelihood statistic for the model fit by the optimal window for each random permutation . The log-likelihood statistic for the real data is then compared with the statistics on the random permutations to derive a p value . We used 9999 replications in our study . The maximum hotspot size was set at 30% of the population , and the inference level for significance was set at 0 . 05 . The main analysis was done without adjustment for covariates , and a secondary analysis was conducted for the 2009/2010 data with and without ITN use as a covariate . Kernel smoothing with a 1 km radius is used for spatial display graphs , but all analyses of correlation are conducted on raw data without smoothing . Semivariograms , Moran's I and linear regression models were run in Stata version 12 ( StataCorp , Texas ) . Semivariograms were constructed using 0 . 1 km intervals between 0 . 1 km and 10 km . Moran's I was assessed globally using cumulative bands of <0 . 1 , <0 . 5 , <1 and <2 and <5 kms . | Malaria remains a formidable threat to public health in tropical regions . The parasite that causes the disease is transmitted to humans by bites from infected mosquitoes , and the complicated lifecycle of the parasite makes developing vaccines difficult . However , preventive strategies are effective at reducing the spread of malaria . The two most widely used and effective strategies are the use of pesticide-treated bed nets to create a barrier between sleeping families and biting mosquitoes , and indoor residual spraying to reduce the numbers of mosquitoes biting sleeping families in homesteads . Other potential preventive strategies include killing mosquito larvae in breeding sites and mass anti-malarial drug treatment for infected humans . Targeting preventive efforts to malaria hotspots—the areas where the risk of malaria transmission is greatest—may help to eliminate malaria more efficiently . Unfortunately , identifying hotspots is complicated as there are many different factors that affect how malaria spreads . These factors range from ecological conditions such as rainfall and soil type , to human effects like population density and migration . Bejon et al . have examined the patterns of malaria transmission in Kenya over 9 years . Over this period , 54% of children who went to health clinics with a fever tested positive for the parasite that causes malaria . Infected children from areas with the highest rate of malaria infection were , on average , younger than those from less infected regions . This makes sense as in highly affected areas children have a greater chance of encountering the parasite at an early age . They are therefore more likely to get malaria when younger and , as exposure to the parasite can provide some immunity to a child , they are also less likely to get infected again when older . In addition , mapping the spread of malaria reveals hotspots at different geographical scales . Bejon et al . could see hotspots within hotspots , and in some cases could go as far as identifying the individual homesteads most at risk of malaria . Public health workers could potentially use these analyses to identify areas that are likely to be hotspots and then target preventive measures there for the next month . However , the constantly changing locations of the hotspots means workers would have to reanalyse the data and retarget their interventions at the end of each month . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"microbiology",
"and",
"infectious",
"disease"
] | 2014 | A micro-epidemiological analysis of febrile malaria in Coastal Kenya showing hotspots within hotspots |
More than 90% of lung cancers are caused by cigarette smoke and air pollution , with polycyclic aromatic hydrocarbons ( PAHs ) as key carcinogens . In Xuanwei City of Yunnan Province , the lung cancer incidence is among the highest in China , attributed to smoky coal combustion-generated PAH pollution . Here , we screened for abnormal inflammatory factors in non-small cell lung cancers ( NSCLCs ) from Xuanwei and control regions ( CR ) where smoky coal was not used , and found that a chemokine CXCL13 was overexpressed in 63/70 ( 90% ) of Xuanwei NSCLCs and 44/71 ( 62% ) of smoker and 27/60 ( 45% ) of non-smoker CR patients . CXCL13 overexpression was associated with the region Xuanwei and cigarette smoke . The key carcinogen benzo ( a ) pyrene ( BaP ) induced CXCL13 production in lung epithelial cells and in mice prior to development of detectable lung cancer . Deficiency in Cxcl13 or its receptor , Cxcr5 , significantly attenuated BaP-induced lung cancer in mice , demonstrating CXCL13’s critical role in PAH-induced lung carcinogenesis .
Air pollution is a diverse mixture of pollutants that originated from anthropogenic and natural sources , is comprised of particulate matter ( PM ) , gases ( e . g . , sulfur oxides , carbon monoxide , ozone ) , organic compounds ( e . g . , polycyclic aromatic hydrocarbons , PAHs ) , metals ( e . g . , lead , vanadium , and nickel ) , and others , such as microbes ( Akimoto , 2003; Huang et al . , 2014 ) . Air pollution is a global environmental health risk that affects the populations in developed and developing countries alike , and satellite observations suggest that 80% of the global population resides in locations where the ambient pollutant concentrations exceed the World Health Organization ( WHO ) Air Quality Guideline ( van Donkelaar et al . , 2010 ) . Outdoor air pollution in cities and rural areas was estimated to cause 3 . 7 million premature deaths annually worldwide in 2012 , including 220 , 000 deaths due to lung cancer ( WHO , 2014 ) . Recently , outdoor ( Loomis et al . , 2013 ) and indoor ( WHO , 2010 ) air pollution has been classified as a Group 1 carcinogen in humans by the International Agency for Research on Cancer ( IARC ) of WHO . Indeed , the risk of lung cancer rises by 18% for every increase of 5 μg/m3 of PM smaller than 2 . 5 μm in diameter ( PM2 . 5 ) in the environment; the risk increases by 22% for every increase of 10 μg/m3 in PM smaller than 10 μm ( PM10 ) ( Raaschou-Nielsen et al . , 2013 ) . There have been efforts to investigate how air pollution causes human cancers by exposing cells and animals to PM or related chemicals ( Straif et al . , 2012 ) , but the carcinogenic mechanism remains elusive , at least in part , due to the difficulty of identifying human lung cancers that are causally associated with air pollution . Xuanwei City of the Yunnan Province in China provides a significant association between air pollution and lung cancer ( Lan et al . , 2002; Sinton et al . , 1995; Mumford et al . , 1987 ) . Until the 1970s , residents of this region used smoky coal in unvented indoor fire pits for domestic cooking and heating , which are processes that release high concentrations of PM2 . 5/PM10 that contain high concentration PAHs ( Mumford et al . , 1987 ) . Indoor air pollution in Xuanwei is associated with a significant increase in the absolute lifetime risk of developing lung cancer ( Xiao et al . , 2012 ) , and the lung cancer incidence in this region is among the highest in China ( Mumford , et al . , 1987; Xiao et al . , 2012; Li et al . , 2011 ) . In particular , nearly all women in this region are non-smokers and cook food on the household stove , and the lung cancer incidence in women is high . In Xuanwei , the male-to-female ratio of lung cancer incidences is 1 . 09:1 , while that of China’s national average reaches 2 . 08:1 ( Xiao et al . , 2012 ) . Tobacco smoke was weakly and non-significantly associated with lung cancer risk in this region ( Kim et al . , 2014 ) . In the 1990s , a reduction in the lung cancer incidence was noted after stove improvement , supporting the association between indoor air pollution and lung cancer ( Lan et al . , 2002 ) . These findings had been cited by the IARC monograph to classify indoor emissions from household coal combustion as ‘carcinogenic to humans ( Group 1 ) ( WHO , 2010 ) . However , lung cancer incidence in Xuanwei is still increasing ( Li et al . , 2011 ) , possibly due to pollutants generated by coal-burning industrial plants that have moved into the area ( Cao and Gao , 2012 ) . Therefore , the population in this highly polluted region ( HPR ) provides a unique opportunity to dissect air pollution-related lung carcinogenesis . Furthermore , the key carcinogens of smoky coal combustion are PAHs ( Lv et al . , 2009 ) , which are also the key carcinogens of tobacco smoke that causes more than 85% of global lung cancer deaths ( Hecht , 2012 ) . Xuanwei lung cancer is , therefore , applicable to elucidate tobacco smoke-induced lung carcinogenesis . Chronic inflammation can promote cancer formation , progression and metastasis by inducing oncogenic mutations , genomic instability , and enhanced angiogenesis ( Grivennikov et al . , 2010 ) . Studies show that PAHs can cause immune suppression ( Zaccaria and McClure , 2013 ) and induce the secretion of cytokines/chemokines , including tumor necrosis factor ( TNF ) -α , interleukin ( IL ) -1β , IL-6 , IL-8 , CC chemokine ligand 1 ( CCL1 ) , chemokine ( C-X-C motif ) ligand 1 ( CXCL1 ) , and CXCL5 ( N'Diaye et al . , 2006; Umannová et al . , 2011; Chen et al . , 2012; Dreij et al . , 2010 ) . These factors may facilitate cancer initiation and progression . However , most of these factors were identified in cellular and animal PAH exposure models . Clinically relevant inflammation factors should be identified to shed insights into PAH-related carcinogenesis and provide novel therapeutic targets for lung cancer . To systematically investigate air pollution-induced lung carcinogenesis , we used Xuanwei lung cancers to analyze the abnormalities in the cancer genomes ( Yu et al . , 2015 ) , genome-wide DNA methylation , non-coding RNAs ( miRNAs [Pan et al . , 2015] and lncRNAs ) , and inflammation factors . The abnormalities found in the HPR lung cancers were tested in patients from control regions ( CRs ) where smoky coal was not used , to compare the difference between HPR and control region ( CR ) lung cancers . In this study , we explored the abnormal inflammatory factors in HPR non-small cell lung cancers ( NSCLCs ) .
Using a microarray analysis of 84 cytokines/chemokines in tumor samples and their adjacent normal lung tissues of eight HPR NSCLCs , we found that the expression of four cytokines ( IL-1F5 , IL-1F9 , MIF , and SPP1 ) and seven chemokines ( CXCL13 , CCL7 , CCL20 , CCL26 , CXCL6 , CXCL9 , and CXCL14 ) was increased in tumors compared with their counterpart normal lung tissues ( Figure 1A ) . Among them , CXCL13 was the most significantly up-regulated gene , with an average of 63-fold ( 10 . 48–173 . 65 ) higher expression in the tumor samples than in the normal controls . The expression of three cytokines ( IL-1α , IL-5 , and TNF-α ) and nine chemokines ( CCL4 , CCL17 , CCL18 , CCL21 , CCL23 , CXCL1 , CXCL2 , CXCL3 , and CXCL5 ) was lower in the tumor samples than in the normal controls ( Figure 1A ) . 10 . 7554/eLife . 09419 . 003Figure 1 . CXCL13 expression in lung cancer . ( A ) A PCR array was used to detect the expression of 84 cytokines/chemokines in eight highly polluted region ( HPR ) lung cancers . ( B ) The ratios of CXCL13 in tumor samples to their counterpart normal lung tissues from both the HPR and control region ( CR ) non-small cell lung cancers ( NSCLCs ) . ( C ) Comparison of the CXCL13tumor/CXCL13normal values of the HPR patients with the CR cases . ( D , E ) CXCL13 expression was detected by immunohistochemistry ( IHC ) in HPR and CR patients ( D ) , and the immunoreactivity score was calculated ( E ) . ( F ) Western blot analyses of lysates from the tumors and adjacent normal lung tissues harvested from CR NSCLCs . ( G ) The concentrations of CXCL13 in the blood samples from healthy donors ( HDs ) and HPR and CR patients were detected by ELISA . ( H , I ) CXCL13 expression in Oncomine reports . ( H ) CXCL13 expression was detected by microarrays in tumor samples and normal lung tissues . AD , adenocarcinoma; A+S , adenocarcinoma and squamous cell carcinoma; Ca , Canada; It , Italy; Jp , Japan; SC , squamous cell carcinoma; Tw , Taiwan , China . ( I ) The expression of CXCL13 was detected in tumor tissues of smokers and non-smokers . ( J ) In mouse Gene Expression Omnibus ( GEO ) data sets , the expression of cxcl13 in indicated mice was detected by microarray . ( K ) The relationship between the CXCL13 expression and the tumor stages of lung cancer patients . ( L ) Overall survival of 54 CR patients ( see Table 3 for their baseline demographic characteristics ) . The median follow-up was 1087 days ( range , 187–1845 days ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 00310 . 7554/eLife . 09419 . 004Figure 1—source data 1 . Sequences of primers for real-time PCR and ChIP , and siRNA . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 00410 . 7554/eLife . 09419 . 005Figure 1—figure supplement 1 . Kaplan–Meier estimates of survival of patients with non-small cell lung cancer ( NSCLC ) according to age , cancer stage , and histology . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 005 We expanded these observations , and found that the expression of CXCL13 was elevated in the tumor tissues from 63/70 ( 90% ) HPR patients ( Figure 1B ) and 71/131 ( 54 . 2% ) CR patients ( Table 1 ) compared with their normal controls . CXCL13 expression was much higher in the HPR patients than the CR cases ( p<0 . 002; Figure 1C ) . Using immunohistochemistry ( IHC ) and immunoreactivity scoring , we showed that the expression of the CXCL13 protein was significantly higher in the tumor samples than their adjacent normal controls ( Figure 1D , E ) . CXCL13 expression was higher in the HPR NSCLCs than the CR patients , and the CR smoker NSCLCs had higher CXCL13 levels than the CR non-smoker patients ( Figure 1B , D , E , and Table 1 ) . Using Western blot assays , we found that CXCL13 was much higher in the tumor samples than their adjacent normal lung tissues ( Figure 1F ) . ELISA showed that the CXCL13 serum concentration was higher in the HPR patients compared with the CR patients , while the latter was higher than the healthy donors ( Figure 1G ) . 10 . 7554/eLife . 09419 . 006Table 1 . Baseline demographic characteristics of the 201 patients who underwent CXCL13 analyses . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 006CharacteristicsTotalHighly polluted region ( HPR ) Control region ( CR ) p values ( HPR vs CR ) Case , nCXCL13 high , n ( % ) p valuesCase , nCXCL13 high , n ( % ) p valuesCase , nCXCL13 high , n ( % ) p valuesTotal201134 ( 66 . 7 ) 7063 ( 90 ) 13171 ( 54 . 2 ) 0 . 0000003G: Male13488 ( 65 . 7 ) 0 . 674741 ( 87 . 2 ) 0 . 278747 ( 54 ) 0 . 950 . 0001 Female6746 ( 68 . 7 ) 2322 ( 95 . 7 ) 4424 ( 54 . 5 ) 0 . 0006A: <65 y14098 ( 70 ) 0 . 215651 ( 91 . 1 ) 0 . 558447 ( 56 ) 0 . 70 . 000009 ≥65 y5634 ( 60 . 7 ) 1412 ( 85 . 7 ) 4222 ( 52 . 4 ) 0 . 03Unknown52 ( 40 ) 52 ( 40 ) S: Smoker10775 ( 70 . 1 ) 0 . 183631 ( 86 . 1 ) 0 . 177144 ( 62 ) 0 . 040 . 01 Non-smoker8753 ( 60 . 9 ) 2726 ( 96 . 3 ) 6027 ( 45 ) 0 . 000006 Unknown76 ( 85 . 7 ) 76 ( 85 . 7 ) H: Adenocarcinoma ( AD ) 13191 ( 69 . 5 ) 0 . 454844 ( 91 . 7 ) 0 . 378347 ( 56 . 6 ) 0 . 840 . 00003Squamous cell carcinoma ( SCC ) 6139 ( 63 . 9 ) 1916 ( 84 . 2 ) 4223 ( 54 . 8 ) 0 . 03Others94 ( 44 . 4 ) 33 ( 100 ) 61 ( 16 . 7 ) Tumor node metastasis ( TNM ) : I8851 ( 58 ) 0 . 0072923 ( 79 . 3 ) 0 . 025928 ( 47 . 5 ) 0 . 050 . 004 II2717 ( 63 ) 98 ( 88 . 9 ) 189 ( 50 ) 0 . 005 III5843 ( 74 . 1 ) 1717 ( 100 ) 4126 ( 63 . 4 ) 0 . 004 IV2219 ( 86 . 4 ) 1111 ( 100 ) 118 ( 72 . 7 ) 0 . 06 Unknown64 ( 66 . 7 ) 44 ( 100 ) 20 ( 0 ) G , gender; A , age; S , smoke; H , histology . CXCL13 expression was not significantly different in smoker and non-smoker HPR patients ( p=0 . 17; Table 1 ) , suggesting that severe air pollution had a carcinogenic effect on humans . In NSCLCs from CRs , however , the expression of CXCL13 was significantly higher in smokers ( 44/71 , 62% ) than in non-smokers ( 27/60 , 45% , p=0 . 04; Table 1 ) , suggesting a potential association between tobacco smoke and CXCL13 expression . The multivariate logistic analyses showed that among the 201 NSCLCs , CXCL13-high was associated with HPR ( p=4 . 6×10–6 ) and tobacco smoke ( p=0 . 032; Table 2 ) . 10 . 7554/eLife . 09419 . 007Table 2 . Multivariate logistic analyses of the association between CXCL13 high expression and clinical characteristics . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 007Highly polluted region ( HPR ) patients , n=70 Variable Odds ratio 95 . 0% confidence interval P value Age1 . 4830 . 177–12 . 4520 . 716Gender4 . 7110 . 488–45 . 510 . 18Smoking0 . 6520 . 115–3 . 6820 . 628Histology0 . 370 . 069–1 . 9920 . 247TNM stage3 . 0920 . 765–12 . 5020 . 113Control region ( CR ) patients , n=131 Variable Odds ratio 95 . 0% confidence interval P value Age0 . 9140 . 455–1 . 8320 . 799Gender2 . 150 . 772–5 . 9930 . 143Smoking0 . 5130 . 251–1 . 0520 . 06Histology1 . 1480 . 569–2 . 3130 . 7TNM stage2 . 3551 . 157–4 . 7930 . 018Total ( HPR and CR patients ) , n=201 Variable Odds ratio 95 . 0% confidence interval P value Region7 . 9083 . 272–19 . 1144 . 6×10-6Age0 . 9640 . 5–1 . 8610 . 914Gender2 . 2540 . 897–5 . 6620 . 084Smoking0 . 3940 . 168–0 . 9250 . 032Histology0 . 9960 . 52–1 . 9070 . 99TNM stage2 . 7071 . 401–5 . 2320 . 003 To investigate CXCL13 expression in NSCLCs of other cohorts , a cancer microarray database Oncomine ( Rhodes et al . , 2004 ) ( www . oncomine . org ) was applied . We found that in several works of this database ( Okayama et al . , 2012; Bhattacharjee et al . , 2001; Hou et al . , 2010; Landi et al . , 2008; Selamat et al . , 2012; Talbot et al . , 2005; Su et al . , 2007; Stearman et al . , 2005 ) , CXCL13 in tumor samples was elevated compared with their paired normal lung tissues or other normal controls ( Figure 1H ) . CXCL13 was also higher in smoker NSCLCs than non-smoker patients in some studies ( Okayama et al . , 2012; Landi et al . , 2008; Selamat et al . , 2012 ) ( Figure 1I ) . In microarray data sets GSE6135 ( Ji et al . , 2007 ) , GSE21581 ( Carretero et al . , 2010 ) , and GSE54353 ( Xu et al . , 2014 ) deposited in the Gene Expression Omnibus ( GEO; http://www . ncbi . nlm . nih . gov/geo/ ) from genetically engineered mouse models of lung cancer , Cxcl13 was increased in KrasG12DStk11-/- mice ( especially in metastatic tumors ) , as compared with KrasG12DStk11wt mice ( Figure 1J ) . These results suggest that CXCL13 overexpression was not specific to the Chinese cohorts , and Cxcl13 may play a role in Stk11-related lung tumorigenesis . We showed patients of both HPR and CR regions with relatively earlier disease ( stages I and II ) had lower CXCL13 , while those with advanced disease ( stages III and IV ) had higher CXCL13 ( Figure 1K ) , and multivariate logistic analyses showed that CXCL13-high was associated with TNM stage ( Table 2 , p=0 . 003 ) . In 54 CR patients whose survival information was available ( Table 3 ) , the median survival time of CXCL13-high patients ( 965 days ) was much shorter than the CXCL13-low cases ( 1193 days , p=0 . 03; Figure 1L ) . Kaplan-Meier estimates of survival of patients with NSCLC according to age ( Figure 1—figure supplement 1A ) , cancer stage ( Figure 1—figure supplement 1B ) , and histology ( Figure 1—figure supplement 1C ) confirmed that patients with stages III–IV lung cancer had shorter survival time than those with earlier stages of NSCLCs . 10 . 7554/eLife . 09419 . 008Table 3 . Baseline demographic characteristics of 54 control region ( CR ) lung cancer patients whose survival information was available . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 008CharacteristicsCase , n CXCL13-high , n ( % ) P* ValueTotal 5425 ( 46 . 3 ) Gender Male3816 ( 42 . 1 ) 0 . 34Female169 ( 56 . 3 ) Smoking Smoker3316 ( 48 . 5 ) 0 . 69Non-smoker219 ( 42 . 9 ) Age , years 0 . 72<653817 ( 44 . 7 ) ≥65168 ( 50 ) Histology Adenocarcinoma3215 ( 46 . 9 ) 0 . 83squamous cell carcinoma189 ( 50 ) small cell lung cancer10 ( 0 ) Others31 ( 33 . 3 ) TNM stage 0 . 65I2612 ( 46 . 2 ) II62 ( 33 . 3 ) III189 ( 50 ) IV42 ( 50 ) PAHs were reported to be the major carcinogens in the PM2 . 5/PM10 in HPR , as well as in the PM at urban locations in Beijing , Shanghai , Guangzhou and Xi’an in China during January 2013 ( Huang et al . , 2014 ) . Clinically , a long latency is required for individuals to develop lung cancer since they were first exposed to smoking or air pollution . To test the effects of PAHs on cytokine/chemokine production , the normal human lung epithelial 16HBE cells ( Cozens et al . , 1994 ) were exposed to a representative PAH compound benzo ( a ) pyrene ( BaP ) at 1 μM for a long period of time ( 30 days ) . We found that CXCL13 was the most significantly up-regulated gene among the 84 cytokines/chemokines ( Figure 2A ) . We confirmed that BaP up-regulated CXCL13 at both the mRNA and protein levels in a dose- and time-dependent manner in 16HBE and A549 lung cancer cells ( Figure 2B , C ) . 10 . 7554/eLife . 09419 . 009Figure 2 . Benzo ( a ) pyrene ( BaP ) induces CXCL13 in vitro and in vivo . ( A ) A PCR array analysis of the expression of 84 cytokines/chemokines in 16HBE normal lung epithelial cells treated with 1 μM BaP for 30 days . ( B ) The cells were treated with BaP at 10 μM for indicated time points or with the indicated concentrations for 72 hr , and CXCL13 expression was assessed by real-time RT-PCR . The experiments were conducted in triplicate and repeated three times . The error bars represent the SD . ( C ) The cells were treated with BaP as described in ( B ) , and the concentration of CXCL13 in the supernatants was evaluated by ELISA . ( D ) The A/J mice were treated with BaP and/or dexamethasone ( DEX ) for 5 weeks ( see also Figure 2—figure supplement 1A ) and sacrificed 6 months later . The lung tissues were isolated and analyzed by Hematoxylin and eosin ( HE ) staining or immunohistochemistry ( IHC ) using an anti-Cxcl13 antibody ( left panel ) . The immunoreactivity score was calculated ( right panel ) . ( E ) Cxcl13 expression was detected in the lung tissues by real-time PCR . ( F ) The concentration of Cxcl13 in mouse serum was assayed by ELISA . ( G ) IHC assays of mice’ lung tumor tissues using anti-Cd68 , anti-Ttf1 , and anti-Cxcl13 antibodies . ( H ) Immunofluorescence assay of mice’ lung tumor tissues using antibodies against Cxcl13 ( green ) , Cd68 ( red ) , and Ttf1 ( white ) . 4' , 6-diamidino-2-phenylindole ( DAPI ) was used to stain the nucleus ( blue ) . ( I ) The survival curves of the mice treated with BaP and/or DEX ( n=8 for each group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 00910 . 7554/eLife . 09419 . 010Figure 2—figure supplement 1 . Benzo ( a ) pyrene ( BaP ) induces lung cancer in A/J mice . ( A ) Schematic represents the protocols for administration of BaP and/or dexamethasone ( DEX ) in A/J mice . ( B ) MicroCT and 3D reconstruction of lung cancer in mice treated with BaP and/or DEX . Av No . T , average of numbers of tumors in the mice . ( C ) Tumor volume of mice treated with BaP and/or DEX . ( D ) One month after BaP treatment ( 100 mg/kg ) , the mice were scanned with microCT and then sacrificed , lung tissues were isolated and analyzed by HE staining . ( E ) Cxcl12 concentration in peripheral blood of the mice was assayed by ELISA . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 010 To test the in vitro observations in vivo , A/J mice were treated with vehicle control ( corn oil ) or 50–100 mg/kg BaP ( Figure 2—figure supplement 1A , modified from previous studies ( Wattenberg and Estensen , 1996 ) . All BaP-treated animals developed lung cancer within 6 months , as detected by a microCT scan ( Figure 2—figure supplement 1B , C ) and histopathology ( Figure 2D ) . We found that Cxcl13 was significantly up-regulated in the tumor samples compared with their normal counterparts at both the mRNA ( Figure 2E ) and protein ( Figure 2D ) levels , which paralleled the increase in Cxcl13 serum concentrations ( Figure 2F ) . Notably , BaP increased Cxcl13 concentrations in peripheral blood of the mice at 1 month after the beginning of the treatment ( Figure 2F ) , at which point no tumors were observed ( Figure 2—figure supplement 1D ) . To determine the source of Cxcl13 , we performed IHC and immunofluorescence assays in lung cancer tissues of the mice . To do this , thyroid transcription factor-1 ( Ttf-1 ) which is expressed in the lung epithelial cells ( Lazzaro et al . , 1991 ) , was used to mark lung cancer cells , and Cd68 stain was employed to mark macrophages which can produce Cxcl13 in inflammatory lesions with lymphoid neogenesis ( Carlsen et al . , 2004 ) . By IHC assay , we found that both the Ttf1-positive lung cancer cells and Cd68-positive macrophages were stained positive for Cxcl13 , but Ttf1-positive cells constituted more than 95% of the cellular component of the lung tumor tissues of the mice ( Figure 2G ) . This observation was confirmed by immunofluorescence assay ( Figure 2H ) using antibodies against Cxcl13 ( green ) , Cd68 ( red ) , and Ttf1 ( white ) . These results indicate that Ttf1 positive lung cancer cells were the main source of Cxcl13 in mice exposed to BaP . We further showed that the anti-inflammatory drug dexamethasone ( DEX ) ( Wattenberg and Estensen , 1996 ) down-regulated Cxcl13 mRNA and protein , and reduced the Cxcl13 serum concentration ( Figure 2D–F ) . DEX reduced the tumor burden ( Figure 2—figure supplement 1B , C , Figure 2D ) and prolonged the survival ( Figure 2I ) of the mice . Cxcl12 , which was shown to be associated with lung cancer ( Teicher and Fricker , 2010 ) , was not significantly changed in the BaP-treated mice ( Figure 2—figure supplement 1E ) . We screened for transcription factors that could regulate CXCL13 , and found potential aryl hydrocarbon receptor ( AhR ) ( Figure 3A ) and V-Rel Avian Reticuloendotheliosis Viral Oncogene Homolog A ( RelA ) ( data not shown ) binding sites in its promoter . AhR is a ligand-activated transcription factor that binds the xenobiotic-responsive element ( XRE ) or aryl hydrocarbon response element ( AHRE ) to regulate genes in response to the planar aromatic hydrocarbon 2 , 3 , 7 , 8-tetrachlorodibenzo-p-dioxin ( TCDD ) ( Fujisawa-Sehara et al . , 1987; Lo and Matthews , 2012; Shimizu et al . , 2000 ) . The potential XRE-like sequence , 5’-GCCCAGGCTGGAGTGCAG-3’ , is located at 1 . 7 kb downstream from the transcription start site ( TSS; Figure 3A ) . The potential interaction between AhR and CXCL13 was analyzed by the chromatin immunoprecipitation ( ChIP ) assay , and the results showed that AhR could not bind CXCL13 in A549 cells without BaP treatment . Interestingly , AhR-CXCL13 interaction was detected and CXCL13 was significantly up-regulated in the presence of BaP , revealed by quantitative PCR ( qPCR ) ( Figure 3B ) . However , significant interaction between AhR and CXCL12 , CXCL14 , CXCL11 , or CXCL2 was not detected ( Figure 3B ) . BaP significantly induced luciferase activity driven by the CXCL13 promoter containing XRE-like sequence; deletion of the XRE-like element , co-incubation with the AhR antagonist α-NF ( Wilhelmsson et al . , 1994 ) or transfection of an AhR-specific siRNA significantly reduced luciferase activity ( Figure 3C–E ) . 10 . 7554/eLife . 09419 . 011Figure 3 . CXCL13 is a target gene of aryl hydrocarbon receptor ( AhR ) . ( A ) The AhR binding site is located at 1 . 7 kb downstream of the CXCL13 transcription start site ( TSS ) . ( B ) A chromatin immunoprecipitation ( ChIP ) assay was performed in BaP-treated or untreated 16HBE cells . The enriched CXCL13 was detected by qPCR . ( C ) The A549 cells were transfected with the wild-type ( WT ) or mutant ( deletion mutation ( mut ) in the XRE-like sequence ) CXCL13 promoter-luciferase reporter construct , treated with BaP and/or α-NF for 48 hr , and assessed by the luciferase assays . ( D , E ) A549 cells were transfected with AhR-specific siRNAs , and western blot was performed to detect the expression of AhR . Three siRNAs were used , and the result of one was shown ( D ) . Luciferase assays were performed in A549 cells transfected with the WT CXCL13 promoter-luciferase reporter construct and siRNAs in the absence or presence of BaP ( E ) . ( F , G ) Western blot analyses of AhR in the cytoplasm and nucleus of 16HBE cells co-incubated with BaP , with or without α-NF treatment ( F ) or siRNA transfections ( G ) . ( H , I ) Immunofluorescence assays of AhR expression in 16HBE cells co-incubated with BaP , with or without α-NF treatment ( H ) or siRNA transfections ( I ) . ( J , K ) CXCL13 mRNA ( detected by qPCR; J ) and protein ( in supernatants of the cells detected by ELISA; K ) levels in the AhR-silenced 16HBE cells treated with BaP and/or α-NF . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 011 TCDD can activate AhR by inducing its translocation from the cytoplasm to nucleus ( Pollenz et al . , 1994 ) . We reported that α-NF and siAhR drastically inhibited BaP caused translocation of AhR to nucleus in 16HBE cells ( Figure 3F–I ) . These results might explain why α-NF and siAhR significantly decreased CXCL13 expression ( Figure 3J ) and concentration in the supernatant ( Figure 3K ) in BaP-treated 16HBE cells . To uncover the role of CXCL13 in BaP-induced lung cancer , Cxcl13 knockout mice ( Cyster et al . , 2000 ) were treated with 100 mg/kg BaP twice a week for 8 weeks ( Figure 4—figure supplement 1A , B ) , and the tumor burden of the mice was evaluated . First , tumor nodules in histologic sections of mice upon BaP treatment were analyzed as described ( Tan et al . , 2013 ) , and the results showed that at treatment time points of 120 days , 180 days , and 240 days , Cxcl13-/- mice had much fewer lesions than Cxcl13+/- and Cxcl13+/+ mice ( Figure 4A ) . Then , microCT was used to detect visible tumors in mice 240 days after BaP treatment . We found that Cxcl13-/- mice harbored much fewer lung tumors than Cxcl13+/- and Cxcl13+/+ mice ( Figure 4B ) . Consistent with this observation , the tumor volume of Cxcl13-/- mice was significantly smaller than Cxcl13+/- and Cxcl13+/+ mice ( Figure 4C ) . In Cxcl13+/+ and Cxcl13-/+ mice , the Cxcl13 serum concentrations were elevated 3 months after BaP treatment , at which time point no tumors were detected ( Figure 4D ) . Moreover , the life span of BaP-treated Cxcl13-/- mice was prolonged compared with the Cxcl13+/+ and Cxcl13-/+ mice ( Figure 4E ) . These results indicate that CXCL13 is required for BaP-induced lung cancer . 10 . 7554/eLife . 09419 . 012Figure 4 . Cxcl13 and Cxcr5 are critical to benzo ( a ) pyrene ( BaP ) -induced lung cancer . ( A ) Cxcl13 deficiency mice were treated with BaP , sacrificed 120 days , 180 days or 240 days later , and the tumor nodules in histologic sections were analyzed . See also ( Figure 4—figure supplement 1 ) . ( B ) MicroCT scanning images and HE staining of lung sections from the BaP-treated Cxcl13 wild-type ( WT ) or knockout mice . ( C ) Tumor volume of the microCT scanning of the mice . ( D ) Serum concentrations of Cxcl13 in the BaP-treated Cxcl13 WT or knockout mice . ( E ) Life span of the BaP-treated Cxcl13+/+ , Cxcl13+/- , and Cxcl13-/- mice . ( F , G ) Cxcr5 expression in non-small cell lung cancers ( NSCLCs , n=24; F ) and in A/J mice treated with BaP ( n=6 for each group; G ) . ( H ) Cxcr5 deficiency mice were treated with BaP , sacrificed 120 days , 180 days or 240 days later , and the tumor nodules in histologic sections were analyzed . See also Figure 4—figure supplement 1 . ( I ) MicroCT scanning images and HE staining of lung sections from BaP-treated Cxcr5 WT or knockout mice . ( J ) Tumor volume of the microCT scanning of the mice . ( K ) Life span of the BaP-treated Cxcr5 WT or knockout mice . *p<0 . 05; **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 01210 . 7554/eLife . 09419 . 013Figure 4—figure supplement 1 . Treatment of Cxcl13-/- and Cxcr5-/- mice with benzo ( a ) pyrene ( BaP ) . ( A ) The expression of Cxcl13 in the mice . ( B ) Schematic represents the protocols for administration of BaP in the mice . ( C ) The expression of Cxcr5 in the mice . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 013 CXCL13 primarily functions by binding to the G protein coupled receptor CXCR5 , and CXCL13 is currently the only known ligand of CXCR5 ( Förster et al . , 1996; Lazennec and Richmond , 2010 ) . We tested the expression of CXCR5 in 24 NSCLCs , and found that it was expressed by the tumor samples but was not significantly higher than in paired normal lung tissues ( p=0 . 089; Figure 4F ) . In A/J mice treated with BaP , Cxcr5 was slightly up-regulated ( p=0 . 07; Figure 4G ) . To further show the role CXCR5 plays in lung carcinogenesis , Cxcr5-/- mice ( Förster et al . , 1996 ) were treated with BaP at 100 mg/kg BaP twice a week for 8 weeks ( Figure 4—figure supplement 1B , C ) . We showed that BaP treatment induced lung cancer in Cxcr5+/+ mice ( Figure 4H , I ) . However , BaP-treated Cxcr5+/- mice developed fewer tumors , and BaP-treated Cxcr5-/- mice had much ewer lung tumors ( Figure 4H , I ) . Consistently , the tumor volume of Cxcr5-/- mice was significantly lower than Cxcr5+/- and Cxcr5+/+ mice ( Figure 4J ) . Furthermore , Cxcr5-/- mice had a prolonged life span compared with the wild-type and Cxcr5+/- mice ( Figure 4K ) . These results demonstrate that CXCR5 is also critical for BaP-induced lung cancer . CXCR5 is expressed by B cells , T cells and macrophages ( Förster et al . , 1996; Nigrovic and Lee , 2005 ) . We investigated which cells were involved in BaP-induced lung cancer by analyzing cell surface markers in the tumor microenvironment using antibodies against CD45 , CD19 , CD4 , and CD68 . Flow cytometry analysis showed a significant increase in Cd68+ macrophages in the lung tumor samples from BaP-treated A/J mice ( Figure 5A , B ) , which was confirmed by IHC ( Figure 5C ) . We found that 94 . 7% of these macrophages strongly expressed Cxcr5 ( Figure 5D ) . However , B cells and T cells were not enriched in the tumor microenvironment ( Figure 5—figure supplement 1 ) . Moreover , CD68+ macrophages were enriched in human lung tumor tissues revealed by IHC ( Figure 5C ) and immunofluorescence assays ( Figure 5E ) . The Tamm–Horsfall protein ( THP-1 ) macrophage cell line also expressed high levels of CXCR5 and CD68 ( Figure 5E ) . 10 . 7554/eLife . 09419 . 014Figure 5 . Tumor-associated macrophages in benzo ( a ) pyrene ( BaP ) -induced lung cancer . ( A , B ) Flow cytometry analysis of Cd68+ macrophages in BaP-induced tumors . A representative gating is shown . The numbers indicate the Cd68+ cells in the quadrant expressed as the percentage of the total Cd45+ leukocytes from the same tumor ( A ) . The means+SD of the Cd68+ cells from the mice ( n=10 for each group ) are shown ( B ) . See also Figure 5—figure supplement 1 . ( C ) IHC analysis of CD68+ macrophages in tumor samples from BaP-treated mice and highly polluted region ( HPR ) patients . ( D ) Flow cytometry analysis of Cd68+ macrophages isolated from tumor samples of mice treated with 50 mg/kg BaP using an anti-Cxcr5 antibody . ( E ) Immunofluorescence analysis of tumor-associated macrophages in tumor samples from HPR patients and THP-1 cells using anti-CD68 and anti-CXCR5 antibodies; DAPI was used to counterstain the nucleus . ( F ) A trans-well migration assay was performed by plating THP-1 cells in the lower chambers , and the indicated cells in the upper chambers , with or without anti-CXCR5 antibody . ( G , H ) Bioluminescent assays of mice that were inoculated with A549- Luciferase ( Luc ) or A549-Luc-CXCL13 cells ( 8×105 ) in the right lung . THP-1 cells ( 8×105 ) were injected via the tail vein . Representative images ( G ) and total luminous flux ( H ) were shown . ( I ) Lung sections of the mice were stained with HE . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 01410 . 7554/eLife . 09419 . 015Figure 5—figure supplement 1 . Analysis of B cells and T cells in benzo ( a ) pyrene ( BaP ) -induced lung cancer . ( A ) The expression of Cd19 in Cd45+ leukocytes sorted from lung cancer tumor samples was analyzed by flow cytometry . ( B ) The expression of Cd4 was analyzed in Cd45+ leukocytes by flow cytometry . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 015 We investigated the potential role for these macrophages in promoting lung cancer . In a trans-well migration assay , THP-1 cells induced the migration of A549-Luciferase ( Luc ) cells and H1975 cells from the upper chamber to the lower chamber ( Figure 5F ) . Stable transfection of CXCL13 into A549-Luc cells ( designated A549-Luc-CXCL13 cells ) significantly enhanced the migration activity , which was antagonized by an anti-CXCR5 antibody ( Figure 5F ) . Addition of CXCL13 to THP-1 cells enhanced the migration of H1975 cells , while an anti-CXCR5 antibody inhibited this effect ( Figure 5F ) . A549-Luc-CXCL13 cells were injected into the right lung of the non-obese diabetic/severe combined immunodeficiency ( NOD/SCID ) mice ( whose macrophages were functionally immature ) , and bioluminescence was recorded 30 days later . We showed that compared with the A549-Luc cells , the A549-Luc-CXCL13 cells significantly increased the tumor burden in the recipient mice ( Figure 5G , H ) . Furthermore , tail vein-injection of THP-1 cells increased the volume of tumors formed by the A549-Luc-CXCL13 cells ( Figure 5G , H ) and caused obvious metastases to the left lung of the mice ( Figure 5I ) . The above results suggested that CXCL13 may induce macrophages to produce critical factors that promote lung cancer progression/metastasis . To identify these factors , a whole genome gene expression assay was performed in the eight patients ( Figure 1A ) , and 436 genes were found to be associated with CXCL13 ( a coefficient >0 . 5 or <−0 . 5; Figure 6—source data 1 ) . These genes were enriched in cytokine-cytokine receptor interactions , the Wnt signaling pathway , the calcium signaling pathway , and others ( Figure 6A ) . Among the secretion-related genes , MMP12 , secreted phosphoprotein 1 ( SPP1 ) , which encodes a macrophage-secreted cytokine SPP1 or osteopontin ( Kohri et al . , 1992 ) , and MMP7 had the highest coefficient values for interacting with CXCL13 ( Figure 6B ) . We tested 11 secretion-related genes in THP-1 cells , and found that the expression of SPP1 was the highest ( Figure 6C ) . Therefore , it was chosen for further investigation . In THP-1 cells , CXCL13 treatment significantly increased SPP1 concentrations in the supernatant ( Figure 6D ) . In trans-well assays , the addition of CXCL13 to THP-1 cells in the lower chamber enhanced the migration of H1975 cells , while SPP1 silencing by siSPP1 in THP-1 cells attenuated this effect ( Figure 6E , left ) . Transfection of CXCL13 into A549 cells increased their migration activity , while SPP1 silencing in THP-1 cells inhibited A549-CXCL13 cell migration ( Figure 6E , right panel ) . In BaP-treated A/J mice , the Spp1 serum concentration was significantly increased ( Figure 6F ) , and the tumor samples showed significantly increased Spp1 IHC staining compared with the control mice ( Figure 6G ) . By IHC assay , we found that Cd68 positive macrophages were also strongly stained by Spp1 , whereas Ttf1-positive lung cancer cells were only weakly stained positive for Spp1 ( Figure 6H ) . This observation was confirmed by immunofluorescence assay ( Figure 6I ) using antibodies against Spp1 ( green ) , Cd68 ( red ) , and Ttf ( white ) . These results indicate that Cd68-positive macrophages were the main source of Spp1 in mice exposed to BaP . 10 . 7554/eLife . 09419 . 016Figure 6 . Identification of macrophage-secreted SPP1 as a downstream effector of CXCL13 . ( A ) The pathway analysis of CXCL13 associated genes . The data from the microarray data sets of the eight highly polluted region ( HPR ) lung cancers are shown . See also Figure 6—source data 1 . ( B ) Gene ranking according to the correlation with CXCL13 expression . The genes were filtered based on extracellular localization to identify paracrine mediators . The list on the right shows the genes that correlate most significantly with CXCL13 . ( C ) The mRNA expression of the candidate gene was detected in THP-1 cells by qPCR . ( D ) The concentration of SPP1 in supernatants of THP-1 cells in the absence or presence of CXCL13 . ( E ) Trans-well migration assays were performed by plating THP-1 cells ( transfected with siSPP1 or siNC ) in the lower chambers , and the indicated cells ( CXCL13-treated or untreated ) in the upper chambers . ( F ) Spp1 serum concentrations of mice treated with benzo ( a ) pyrene ( BaP ) and/or dexamethasone ( DEX ) were detected by ELISA . ( G ) Spp1 expression in lung section of mice treated with BaP and/or DEX was determined by IHC . ( H ) IHC assays using antibodies against Cd68 , Ttf1 , and Spp1 . ( I ) Immunofluorescence assay using antibodies against Spp1 ( green ) , Cd68 ( red ) , and Ttf1 ( white ) . DAPI was used to stain the nucleus ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 01610 . 7554/eLife . 09419 . 017Figure 6—source data 1 . CXCL13-associated genes in lung cancer . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 01710 . 7554/eLife . 09419 . 018Figure 6—figure supplement 1 . Activation of β-catenin and epithelial mesenchymal transition ( EMT ) in benzo ( a ) pyrene ( BaP ) -induced lung cancer . ( A ) The expression of indicated genes in lung tissues from A/J mice treated with BaP and/or dexamethasone ( DEX ) was detected by real-time PCR . ( B ) Western blot analyses of lysates of lung tissues from A/J mice treated with BaP and/or DEX . ( C ) Western blot analyses of cytoplasmic and nucleic protein fractions of lung tissues from A/J mice treated with BaP and/or DEX , using indicated antibodies . ( D ) Western blot analysis of lysates of lung tissues from NOD/SCID mice injected with indicated cells . ( E ) Western blot analyses of cytoplasmic and nucleic protein fractions in lung tissues from NOD/SCID mice injected with indicated cells . ( F ) Western blot analyses of lysates of lung tissues from Cxcr5 mutant mice treated with BaP . ( G ) IHC assays for the expression of β–catenin in Cxcr5 mutant mice upon BaP . ( H ) Immunofluorescence analyses of β-catenin in A549 cells transfected with control vector ( V ) or SPP1 , or treated with THP-1 supernatant ( Mϕs . ) or supernatant of THP-1 cells co-incubated with CXCL13 ( M13s . ) . ( I ) Western blot analyses of cytoplasmic and nucleic protein fractions of A549 cells transfected with control vector ( V ) or SPP1 , or treated with Mϕs . or M13s . ( J ) Real-time PCR assays of indicated genes in A549 cells transfected with control vector or SPP1 , or treated with Mϕs . or M13s . ( K ) Western blot analyses of lysates of A549 cells transfected with control vector ( V ) or SPP1 , or treated with Mϕs . or M13s , using indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 09419 . 018 At least two tumors were found in left and/or right lungs of each BaP-treated mouse ( Figure 2—figure supplement 1B ) , and CXCL13 induced cancer cell migration and metastasis ( Figure 5F–I , Figure 6E ) . We tested the epithelial mesenchymal transition ( EMT ) phenotypes that are associated with cancer progression and metastasis ( Kalluri and Weinberg , 2009 ) in tumor samples from the BaP-treated A/J mice , and found that the expression of E-Cadherin was down-regulated , while N-Cadherin , Vimentin , Slug and Snail were up-regulated ( Figure 6—figure supplement 1A ) . At protein level , E-Cadherin was down-regulated , while N-Cadherin and Vimentin were up-regulated in the tumors ( Figure 6—figure supplement 1B ) . β-catenin transport to the nucleus is critical for cells to enter into an EMT and acquire an invasive phenotype ( Kim , 2002 ) . We showed that the nuclear β-catenin levels were increased in the tumor samples from A/J mice ( Figure 6—figure supplement 1C ) . In the tumor samples from NOD/SCID mice injected with A549-Luc-CXCL13 cells ( Figure 6—figure supplement 1D , E ) and BaP-treated Cxcr5+/+ mice ( Figure 6—figure supplement 1F , G ) , the expression of E-Cadherin was low , while the N-Cadherin and nuclear β-catenin were high . SPP1 was shown to activate β-catenin ( Robertson and Chellaiah , 2010 ) . To study the role of SPP1 in promoting the progression of lung cancer , A549 cells were transfected with pcDNA3 . 1-flag-SPP1 . We found that SPP1 overexpression decreased the cytoplasmic expression but increased the nuclear expression of β-catenin ( Figure 6—figure supplement 1H , I ) . SPP1 overexpression in cancer cells led to the down-regulation of E-Cadherin and up-regulation of N-Cadherin and Slug at both the mRNA and protein levels ( Figure 6—figure supplement 1J , K ) , and up-regulation of Vimentin and Snail at the mRNA level ( Figure 6—figure supplement 1J ) . These findings were also observed in A549 cells supplemented with supernatants from THP-1 cells that were co-incubated with CXCL13 ( Figure 6—figure supplement 1H–K ) . These results suggest that SPP1 may transactivate β-catenin to induce an EMT in BaP-treated mice .
In this study , we took the advantage of the epidemiology profile of Xuanwei lung cancer to systematically explore the abnormalities in inflammatory factors that are critical to air pollution-induced lung cancer , and reported that CXCL13 was up-regulated in 63/70 ( 90% ) HPR NSCLCs ( Table 1 ) . BaP induced lung epithelial cells to secret CXCL13 ( Figure 2 ) , which recruited tumor-associated macrophages ( TAMs ) and induced SPP1 production ( Figure 5 , 6 ) . SPP1 activated β-catenin by facilitating its translocation to the nucleus and promoted an EMT , leading to the progression of lung cancer ( Figure 6—figure supplement 1 ) . CXCL13-CXCR5 signaling was required for PAH-induced lung cancer , because knockdown of Cxcl13 or Cxcr5 attenuated BaP-induced lung cancer in mice ( Figure 4 ) . These results suggest that the ubiquitous carcinogen BaP may induce tumor-promoting inflammation to facilitate the invasive growth and metastasis of lung cancer , shedding new insights into the carcinogenic mechanisms of air pollution . CXCL13 may be associated with lung cancer risk in smokers ( Shiels et al . , 2013 ) . We showed that in HPR , smoker and non-smoker NSCLCs had approximately equal expression levels of CXCL13 , suggesting a carcinogenic effect of air pollution . In CR , however , smoker patients had higher CXCL13 expression than non-smoker cases ( Figure 1 and Table 1 ) . Moreover , high CXCL13 expression was associated with advanced stage cancer and poor prognosis ( Figure 2 ) . These results indicate that CXCL13 is a common and critical factor in air pollution-induced and tobacco smoke-induced lung carcinogenesis . Furthermore , the serum concentration of CXCL13 was elevated in mice before the emergence of detectable lung tumor by microCT ( Figure 2 and Figure 2—figure supplement 1 ) , suggesting that this chemokine may represent a biomarker for early diagnosis , which warrants further investigation . PAHs are key carcinogens in air pollution and tobacco smoke ( Huang et al . , 2014; Hecht , 2012 ) , and cause mutations in TP53 and K-RAS , resulting in uncontrolled cell growth ( Hecht , 2012 ) . PAHs also perturb the immune system and induce production of the inflammatory factors to facilitate cancer progression in cellular and animal models ( Zaccaria and McClure , 2013; N'Diaye et al . , 2006; Umannová et al . , 2011; Chen et al . , 2012; Dreij et al . , 2010 ) . By treating 16HBE cells with a relatively low concentration ( 1 μM ) of BaP for a long time course ( 30 days ) to mimic BaP-induced lung cancer in humans , we showed that CXCL13 was the most significantly up-regulated gene among the 84 cytokines/chemokines ( Figure 2 ) . Furthermore , CXCL13 was increased in 134/201 ( 66 . 7% ) NSCLCs ( Table 1 ) , suggesting the clinical relevance of CXCL13 in BaP-induced lung cancer . AhR is critical for the carcinogenic action of BaP ( Shimizu et al . , 2000 ) . We found that CXCL13 was a direct target of AhR ( Figure 3 ) , and deficiency in CXCL13 or its receptor , CXCR5 , abrogated BaP-induced lung cancer ( Figure 4 ) , while an anti-CXCR5 antibody significantly inhibited lung cancer cell migration ( Figure 5 ) . These results further demonstrate CXCL13’s critical role in environmental pollutant-induced lung cancer , but its role in pre-malignancy and cell transformation remains to be addressed . Cross-talk between cancer cells and cells of the neoplastic stroma is involved in the acquired ability for invasive growth and metastasis ( Hanahan and Weinberg , 2011 ) . Using a whole -genome gene expression array ( Figure 6 ) , we discovered that TAMs secreted SPP1 , which activates β–catenin and is regulated by this transcription factor ( Robertson and Chellaiah , 2010; El-Tanani et al . , 2001 ) . SPP1 induced the nuclear localization and activation of β–catenin in epithelial and cancer cells , resulting in an EMT phenotype ( Figure 6—figure supplement 1H–K ) . At this stage , the detailed mechanisms of CXCL13-CXCR5-induced SPP1 production were unclear . However , our results suggested that SPP1-β–catenin could form a positive feedback loop to promote EMT and lung cancer progression . CXCL13 was shown to be able to induce an EMT in breast cancer via RANKL and Src activation ( Biswas et al . , 2014 ) . We found that BaP or CXCL13 treatment or SPP1 overexpression in lung epithelial or cancer cells did not up-regulate RANKL or Src ( data not shown ) , but did activate β–catenin ( Figure 6—figure supplement 1 ) , suggesting that CXCL13 may induce an EMT via different mechanisms in different settings . Chemokines can modulate tumor cell proliferation , survival , angiogenesis , senescence and metastasis ( Wang et al . , 2015 ) . For example , CCL20 is induced by another important carcinogen in tobacco smoke , nicotine-derived nitrosaminoketone , and is overexpressed in smoker lung cancers and inversely associated with prognosis . CCL20 promotes lung cancer cell proliferation and migration ( Wang et al . , 2015 ) . CXCL12 and CCR7 regulate the metastasis of breast cancers and NSCLCs ( Phillips et al . , 2003; Müller et al . , 2001; Takanami , 2003 ) , while CXCL1/2 enhance cell survival and promote chemoresistance ( Acharyya et al . , 2012 ) . CXCL13 is overexpressed in breast cancer ( Panse et al . , 2008 ) and mediates prostate cancer cell proliferation ( El-Haibi et al . , 2011 ) . Given their critical roles in cancer , pharmaceutical development pipelines are filled with new chemokine-targeting drugs to treat malignancies ( Klarenbeek et al . , 2012 ) . DEX is a synthetic glucocorticoid used to counteract certain side effects of chemotherapies or as a direct chemotherapeutic agent in certain malignancies . It reduces lung tumor multiplicities in the tobacco smoke-exposed and non-exposed mice ( Witschi et al . , 2005 ) . We showed that DEX inhibited CXCL13 production by epithelial cells and SPP1 production by TAMs ( Figure 3 and Figure 6—figure supplement 1 ) , inhibited the EMT ( Figure 6—figure supplement 1 ) , reduced the tumor burden and prolonged the life span of the mice ( Figure 2 ) , revealing a new mechanism for this existing drug . Because DEX is a wide-spectrum anti-inflammatory drug that may cause severe side effects , more specific , CXCL13/CXCR5-targeting antibodies or small molecules should be developed to combat lung cancer , the leading cause of cancer-related mortality accounting for 1 . 59 million deaths worldwide in 2012 ( WHO , 2012 ) .
The study was approved by the local research ethics committees of all participating sites; all lung cancer samples were collected with informed consent . A total of 201 previously untreated NSCLCs from HPRs or CRs were included ( Table 1 ) . The HPR patients were diagnosed in the last 5 years in participant hospitals in Yunnan Province , and the diagnosis of lung cancer was confirmed by at least three pathologists . Those who fulfilled the following criteria were selected for this study: ( 1 ) Residents of Xuanwei where the smoky coal was used . ( 2 ) Resided in their communities and never stayed in other regions for a long time ( 6 months or more ) . ( 3 ) Previously untreated primary lung cancer . ( 4 ) The tissue samples were taken at the time of surgery and quickly frozen in liquid nitrogen . The tumor samples contained a tumor cellularity of greater than 60% and the matched control samples had no tumor content . Serum samples were obtained from 80 untreated NSCLCs ( 40 from HPR and 40 from CR ) and 40 healthy donors . The antibodies used in this work were: recombinant human CXCL13 , human CXCL13 Quantikine ELISA Kit , mouse CXCL13 Quantikine ELISA Kit , mouse CXCL12 Quantikine ELISA Kit , mouse anti-human CXCR5-PE mAb , goat anti-human CXCL13 polyclonal Ab , goat anti-mouse CXCL13 polyclonal Ab ( R&D , Minneapolis , MN ) , rat anti-mouse CD68-PE , rat anti-mouse CD45-APC , rat anti-mouse CD4-PerCP-CY5 . 5 , rat anti-mouse CD19-FITC , rat anti-mouse CXCR5-PE/CY7 mAb ( Biolegend , San Diego , CA ) , mouse anti-human CD68 mAb ( Dako , Glostrup , Denmark ) , rabbit anti-mouse SPP1 polyclonal Ab ( Proteintech , Chicago , IL ) , anti-TTF1 ( Abcam , Cambridge , UK ) , rabbit anti-mouse E-Cadherin mAb , EMT Antibody Sampler Kit ( Cell Signaling Technology , Beverly , MA ) , rabbit anti-human AhR polyclonal Ab , goat anti-human Lamin B polyclonal Ab , mouse anti-human ɑ-tubulin mAb ( Santa Cruz Biotechnology , Santa Cruz , CA ) , mouse anti-human β-actin antibody ( Sigma , St . Louis , MO ) , Alexa Fluor 488 Donkey anti-Goat IgG ( H+L ) , Alexa Fluor 555 Donkey Anti-Mouse IgG ( H+L ) , Alexa Fluor 647 Donkey anti-Rabbit IgG ( H+L ) ( Life Technology , Thermo Fisher Scientific , Basingstoke , UK ) , and mouse SPP1 ELISA Kit ( Shanghai GenePharma , Shanghai , China ) . BaP and DEX were purchased from Sigma . The expression of inflammatory factors in eight HPR patients ( Figure 1A ) was determined using the Human Inflammatory Cytokines & Receptors RT2 Profiler PCR Array , which contains 84 cytokines/chemokines and their receptors . CXCL13 expression was detected in additional 193 HPR and CR patients , in BaP-treated cells , and in BaP-treated A/J , Cxcl13-/-48 or Cxcr5-/-50 mice by real-time PCR , Western blot analysis , ELISA , or IHC using anti-CXCL13 and anti-SPP1 antibodies . The immunoreactivity score ( IRS ) was calculated as IRS ( 0–12 ) =CP ( 0–4 ) ×SI ( 0–3 ) , where CP is the percentage of CXCL13-positive epithelial cells and SI is the staining intensity ( Remmele and Stegner , 1987 ) . The lung adenocarcinoma cancer cell lines A549 and NCI-H1975 were obtained from ATCC ( Manassas , VA , USA ) . Human normal bronchial epithelial cell line 16HBE was obtained from Clonetics ( Walkersville , MD ) and cultured according to standard protocols . The macrophages were obtained by incubation of THP-1 cells ( ATCC ) with 100 ng/mL phorbol 12-myristate 13-acetate ( PMA ) for 24 hr and then the culture medium was refreshed with serum-free Dulbecco’s Modified Eagle Medium ( DMEM ) for another 24 hr ( Tsuchiya et al . , 1982 ) . For preparation of conditioned medium , PMA-differentiated THP-1 cells were plated in six-well plates and co-cultured with CXCL13 for 48 hr . The medium was then centrifuged at 200×g for 10 min at 4°C to remove cell debris and stored at −80°C until use . For the analysis of EMT signaling by western blot analysis , supernatants were ultrafiltered for protein enrichment using a centrifugal filter device ( Millipore , Darmstadt , Germany ) . We screened for transcription factors that could regulate CXCL13 by website-based prediction ( http://www . sabiosciences . com/chipqpcrsearch . php ? species_id=0&factor=Over+200+TF&gene=CXCL13&nfactor=n&ninfo=n&ngene=n&B2=Search ) , and CXCL13 promoter with wild-type or deletion mutation XRE-like element was cloned into pGL3-Basic vector . SPP1 was cloned into pcDNA3 . 1-flag vector . Cells were transfected with plasmids or siRNA using the Lipofectamine 2000 ( Invitrogen , Frederick , MD ) or lentivirus according to manufacturer’s instruction . CXCL13 was cloned into the retroviral vector pGC-FU-GFP-IRES-Puromycin , and then transfected into the virus-packaging Phoenix cells . The supernatant was used to infect A549-Luc cells , and the infected cells were selected with 0 . 5 μg/mL Puromycin ( Gene Oparation , Ann Arbor , MI ) and 500 μg/mL G418 ( Calbiochem , San Diego , CA ) . The total RNA was isolated using the TRIZOL reagent ( Invitrogen ) and the phenol-chloroform extraction method according to the manufacturer’s instruction . Total RNA ( 2 μg ) was annealed with random primers at 65°C for 5 min . The cDNA was synthesized using a 1st-STRAND cDNA Synthesis Kit ( Fermentas , Pittsburgh , PA ) . Quantitative real-time PCR was carried out using SYBR Premix ExTaq ( Takara ) . The primer sequences for real-time PCR are listed in Figure 1—source data 1 . Each sample was analyzed in triplicate three times . Trans-wells with 8-μm pores were incubated at 37°C in a CO2 incubator for at least 1 hr . The differentiated THP-1 cells were seeded in the lower chamber , and A549-Luc-CXCL13 or H1975 cells ( 1×104 ) with serum-free DMEM/RPMI 1640 were added to the upper chamber and allowed to migrate for 24 hr . Cells on the inserts were fixed with 90% ethanol , stained with 0 . 0005% Gentian Violet Solution , and washed with PBS . Non-migrated cells on the upper side of the inserts were wiped off with a cotton swab . Migrated cells were counted in five microscopic fields at 4× magnification , and the counts were averaged ( Meijer et al . , 2006 ) . A549 cells ( 5×106 ) were seeded onto 10-cm dishes and treated with or without BaP ( 5 μM ) for 2 h , and ChIP assay was performed as described ( Zhang et al . , 2012 ) . Both input and immunoprecipitated DNA samples were analyzed by qPCR to determine the relative amounts of DNA from the CXCL13 gene promoter region present in the samples . The primer pairs used here were listed in Figure 1—source data 1 . Mouse lung cancer tissues were dissected into 2-mm fragments , followed by collagenase IV ( 0 . 2% , Sigma ) digestion for 40 min at 37°C . A single-cell suspension was generated through a 200-mm-stainless steel wire mesh . The dissociated cancer cells labeled with indicated cell surface markers were sorted by MoFlo XDP Cell Sorter ( Beckman Coulter , Brea , CA ) , and the data were analyzed on the Summit Software v5 . 0 ( Beckman Coulter ) . All FACS analyses and sorting were paired with matched isotype control . Dead cells were excluded based on scatter profile . Cells grown on coverslip ( 24 mm×24 mm ) were fixed with 4% paraformaldehyde for 15 min , washed with 150 mM glycine in PBS , and permeabilized with 0 . 3% Triton X-100 in PBS for 20 min at room temperature . After blocking with 5% BSA , the cell smears were incubated with the indicated primary antibodies overnight at 4°C , washed , and FITC/PE-labeled secondary antibody in PBS was added to the cell smears . Images were taken by a laser scanning confocal microscopy ( Zeiss , Oberkochen , Germany ) . IHC assay was performed using anti-CXCL13 and anti-SPP1 antibodies as previously described ( Ma et al . , 2011 ) . Briefly , formalin-fixed , paraffin-embedded human or mouse lung cancer tissue specimens ( 5 µm ) were deparaffinized through xylene and graded alcohol , and subjected to a heat-induced epitope retrieval step in citrate buffer solution . The sections were then blocked with 5% BSA for 30 min and incubated with indicated antibodies at 4°C overnight , followed by incubation with secondary antibodies for 90 min at 37°C . Detection was achieved with 3 , 3'-diaminobenzidine ( DAB , Zhongshan Golden Bridge Biotechnology , Beijing , China ) and counterstained with hematoxylin , dehydrated , cleared and mounted as in routine processing . The scoring of immunoreactivity was performed as described ( Remmele and Stegner , 1987 ) . Concentration of CXCL13 in serum and cell culture supernatant was determined by ELISA using a commercially available ELISA kit ( R&D ) . The absorbance of the plates was read at 450 nm using an automated microplate reader ( Bio-Tek , Winooski , VT , USA ) . Eight pairs of human lung cancer tissues and their adjacent normal lung tissues were homogenized by Biopulverizer ( Biospec , Bartlesville , OK ) . Total RNA was extracted from homogenized samples using the RNeasy Mini Kit ( Qiagen , Valencia , CA ) . Total RNA ( 1 µg ) was labeled and hybridized with a One-ColorQuick Amp Labeling Kit and Gene Expression Hybridization Kit ( Agilent Technologies , Santa Clara , CA ) . Hybridization signals were detected using a DNA microarray scanner G2565BA ( Agilent Technologies ) , and all scanned images were analyzed using Agilent Feature Extraction Software . Quantile normalization and subsequent data processing were performed using the GeneSpring GX v11 . 5 . 1 software package ( Agilent Technologies ) . Quality assessment of mRNA data after filtering was carried out by Box Plot and Scatter-Plot . To identify differentially expressed mRNAs with statistical significance , Volcano Plot filtering between the two groups ( fold change ≥ 1 . 5 , p values ≤ 0 . 05 ) was performed . Pathway analysis was performed using the KEGG ( Kyoto Encyclopedia of Genes and Genomes ) database . The genes correlated with CXCL13 were analyzed by Pearson’s correlation coefficient , and the results were listed in Figure 6—source data 1 . Cells were lysed on ice for 30 min in RIPA buffer ( 50 mM Tris-HCl pH 7 . 4 , 150 mM NaCl , 0 . 1% SDS , 1% deoxycholate , 1% Triton X-100 , 1 mM EDTA , 5 mM NaF , 1 mM sodium vanadate , and protease inhibitors cocktail ) , and protein extracts were quantitated . Proteins ( 20 μg ) were subjected to 10–15% sodium dodecyl sulfate-polyacrylamide gel electrophoresis ( SDS-PAGE ) , electrophoresed and transferred onto a nitrocellulose membrane . After blocking with 5% non-fat milk in Tris-buffered saline , the membrane was washed and incubated with the indicated primary and secondary antibodies and detected by Luminescent Image Analyzer LSA 4000 ( GE , Fairfield , CO , USA ) . The animal studies were approved by the Institutional Review Board of Institute of Zoology , Chinese Academy of Sciences . All animal studies were conducted according to protocols approved by the Animal Ethics Committee of the Institute of Zoology , Chinese Academy of Sciences . Female NOD/SCID mice ( 5–6 weeks old ) were purchased from Vital River Laboratory Animal Technology ( Beijing , China ) . A/J mice ( 5–6 weeks old , female ) , homozygous Cxcl13-deficient , and homozygous Cxcr5-deficient mice were purchased from the Jackson Laboratory ( Bar Harbor , Maine , USA ) . The exposure protocols were modified from previous studies ( Wattenberg and Estensen , 1996; Estensen and Wattenberg , 1993 ) . The A/J mice were exposed to BaP ( 0 mg/kg , 50 mg/kg , 100 mg/kg twice per week for 5 weeks ) in corn oil via oral gavage , and treated with or without DEX ( 0 . 5 mg/kg per day for 10 weeks in diet; Figure 2—figure supplement 1A ) . The Cxcl13-/- and Cxcr5-/- mice were treated with BaP at 100 mg/kg twice per week for 8 weeks ( Figure 4—figure supplement 1B ) . For microCT ( PerkinElmer , Waltham , MA ) and bioluminescence ( using an IVIS Spectrum In Vivo Imaging System , PerkinElmer ) analyses , mice were anesthetized by mixture of oxygen/isoflurane inhalation and positioned with legs fully extended , and assayed according to manufacturers’ instruction . Survival of the mice was evaluated from the first day of treatment until death or became moribund , at which time points the mice were sacrificed . All statistical analyses were conducted using GraphPad Prism 5 ( GraphPad Software , La Jolla , CA ) and SPSS 12 . 0 software for Windows ( Chicago , IL ) . Statistically significant differences were determined by Student’s t-test , Wilcoxon rank sum test , one-way analysis of variance , the chi-squared test , or multivariate logistic analysis . P values less than 0 . 05 were considered statistically significant in all cases . | Lung cancer causes the most cancer deaths worldwide . For decades , people have known that lung cancer is associated with environmental factors , and both cigarette smoke and air pollution are known to cause cancers in humans . Smoke and air pollution both contain chemicals called polycyclic aromatic hydrocarbons ( or PAHs ) . These chemicals cause chronic inflammation of the lung , which in turn is a major risk factor for developing lung cancer . However , it is unclear exactly how PAHs trigger inflammation and cancer . Xuanwei City in China is suited to the study of this question because until the 1970s its inhabitants used 'smoky coal' for cooking in unventilated indoor spaces; this produced high levels of small particles that contain high concentrations of PAHs . Women from this region , who traditionally do most of the cooking , have rates of lung cancer comparable to those of men . In other parts of China a woman’s chance of getting lung cancer is approximately half that of a man’s . Therefore , Xuanwei City provides a setting in which air pollution is a main contributor to lung cancer risk . Wang , Cheng et al . have now compared the levels of certain proteins ( which are linked to inflammation ) in lung cancer patients from Xuanwei City with those in control regions of China . The level of one such protein marker , called CXCL13 , was particularly high in almost all patients from Xuanwei City , but only highly expressed in half of the patients from the control regions . Moreover , there was also a clear link between cigarette smoke and CXCL13 expression because , in control regions , smokers were much more likely to have high levels of CXCL13 than non-smokers . To test whether PAHs cause CXCL13 expression , Wang , Cheng et al . first exposed normal lung epithelial cells , cancer cells and then mice to a PAH . These experiments showed that CXCL13 levels did indeed increase and the mice developed lung tumours . However , when the genes for CXCL13 or its binding partner were deleted , the mice no longer got cancer when exposed to the PAH . This shows that CXCL13 signalling is an important mechanism by which PAHs cause lung cancer . Lastly , further experiments showed that CXCL13’s binding partner is highly expressed on some immune cells that can promote lung cancer . Importantly , the over-expression of CXCL13 occurred before the lung tumours developed . This might provide a new treatment strategy in which CXCL13 signalling could be inhibited after the exposure to PAHs . Future studies may now focus on discovering new drugs , or modifying existing drugs , to achieve this goal . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cancer",
"biology"
] | 2015 | The chemokine CXCL13 in lung cancers associated with environmental polycyclic aromatic hydrocarbons pollution |
Previous studies have demonstrated the importance of the primary sensory cortex for the detection , discrimination , and awareness of visual stimuli , but it is unknown how neuronal populations in this area process detected and undetected stimuli differently . Critical differences may reside in the mean strength of responses to visual stimuli , as reflected in bulk signals detectable in functional magnetic resonance imaging , electro-encephalogram , or magnetoencephalography studies , or may be more subtly composed of differentiated activity of individual sensory neurons . Quantifying single-cell Ca2+ responses to visual stimuli recorded with in vivo two-photon imaging , we found that visual detection correlates more strongly with population response heterogeneity rather than overall response strength . Moreover , neuronal populations showed consistencies in activation patterns across temporally spaced trials in association with hit responses , but not during nondetections . Contrary to models relying on temporally stable networks or bulk signaling , these results suggest that detection depends on transient differentiation in neuronal activity within cortical populations .
Lesion studies in humans and animals indicate the causal importance of the primary visual cortex ( V1 ) in detection , discrimination , and awareness of visual stimuli ( Lashley , 1943; Weiskrantz et al . , 1974; Weiskrantz , 1996 ) , and this role has been recently confirmed by direct optogenetic inhibition of mouse V1 ( Glickfeld et al . , 2013 ) . Visual perception has been proposed to arise from interactions between stimulus-specific processing in V1 and neural activity in higher visual and frontoparietal areas , involving both feed-forward propagation of activity and recurrent , top-down feedback ( Shadlen and Newsome , 1996; Britten and van Wezel , 1998; Lamme et al . , 2000; Haynes et al . , 2005 ) . Critical in unraveling neural correlates of vision is how detected and undetected stimuli are processed differently , especially when these stimuli are physically identical . For instance , it has been suggested that the intensity , duration , and reproducibility of sensory neural activity may provide signatures critical for visual perception ( e . g . Moutoussis and Zeki , 2002; Schurger et al . , 2010 ) . In addition , it has been proposed that neural activity in V1 does not correlate with visual perception because stimuli that were seen or not seen evoked similar V1 blood-oxygenation-level-dependent signals ( Vuilleumier et al . , 2001; Rees , 2000 ) , but this remains an area of substantial controversy ( Ress and Heeger , 2003; Palmer et al . , 2007; Nienborg and Cumming , 2014 ) . In this context , it is important to recall that functional magnetic resonance imaging ( fMRI ) , electro-encephalogram ( EEG ) , and magnetoencephalography ( MEG ) rely on a mean-field approach , leaving open the possibility that neural correlates of perception may be coded in more subtle ways that take into account the local differentiation present in populations of sensory neurons . Such local , functional differentiation is supported by single- or multiunit recording studies in visual , auditory , and somatosensory areas of animals trained to make perceptual decisions ( Logothetis et al . , 1995; Britten et al . , 1996; Posner and Gilbert , 1999; Petersen , 2002; Luna et al . , 2005; Palmer et al . , 2007; Mitchell et al . , 2009; Cohen and Maunsell , 2009; Cohen and Maunsell , 2011; Sachidhanandam et al . , 2013; Chen et al . , 2013; Miyashita and Feldman , 2013; Doron et al . , 2014; McGinley et al . , 2015 ) . Over the last decade , it has become clear that the shared response variability between neurons ( i . e . noise correlation ) might be particularly important for sensory processing because noise correlations can influence the amount of information that can be extracted from neuronal population codes ( Averbeck et al . , 2006; Cafaro and Rieke , 2010 ) . Furthermore , it has been observed that these correlations can be reduced during stimulus presentation ( Gutnisky and Dragoi , 2008; Snyder et al . , 2014 ) and directed attention , which may aid in disentangling stimulus information from noisy population responses ( Mitchell et al . , 2009; Cohen and Maunsell , 2009; Herrero et al . , 2013 ) . Although noise correlations have been studied well , they have the drawback of not being an instantaneous measure—their computation requires integrating neural activity over multiple time points or stimulus repetitions . Instantaneous aspects of population activity in cortex , such as temporal spike co-occurrence and population sparseness , seem critical for efficient neural coding ( Olshausen and Field , 1997; Vinje , 2000; Benucci et al . , 2013; Harris and Mrsic-Flogel , 2013 ) . Some population-based measures have been proposed and tested in somatosensory and auditory cortex ( Romo et al . , 2003; Safaai et al . , 2013; Carnevale et al . , 2013; Buran et al . , 2014 ) . It has , for example , been shown that measures based on the variability and correlations between neurons correlate better with the animal’s decision than simpler approaches based on the mean spiking rate ( Safaai et al . , 2013; Carnevale et al . , 2013 ) . However , in the domain of visual perception the behavioral relevance of only few population measures has been experimentally tested in paradigms where animals report behaviorally whether they have seen a stimulus or not . Therefore , we investigated correlates of visual stimulus detection using two-photon calcium imaging of populations of ~100 neurons in V1 L2/3 of mice performing a detection task as superficial layers are easy to access with calcium imaging and have been reported to show neural correlates with stimulus detection ( van der Togt , 2006; Ito and Gilbert , 1999 ) . Our first aim was to examine whether visual detection correlates with the mean visual response strength of V1 neurons or rather with other metrics of population responses , such as noise correlation or variance . This led us to develop a novel population metric—response heterogeneity—that correlates better with stimulus detection performance , and particularly with the animal’s reaction time , than traditional measures by capturing the dissimilarity of neuronal responses within a population . Second , an assumption in many computational models of vision is that neurons in distributed cortical architectures have relatively fixed roles in encoding visual features , but modulate their activation in a temporally dynamic manner based on attentional needs that can influence perception ( e . g . Jones and Palmer , 1987; Itti et al . , 1998; Desimone , 1998; Dayan and Abbott , 2001; Deco and Rolls , 2004; Reynolds and Heeger , 2009 ) . To study whether modulations of neuronal activity that influence stimulus perception show temporally recurring patterns , we asked whether population activation patterns are more similar across trials that repeat the same stimulus presentation when the stimulus is successfully detected . We report that ( 1 ) visual stimulus detection does not correlate well with mean response strength , but is significantly correlated with population heterogeneity; ( 2 ) neuronal populations show consistencies in activation patterns across temporally spaced trials in association with hit responses , but not when the animal fails to report a stimulus; and ( 3 ) in addition to heterogeneity , multidimensional structures in neuronal population responses provide information on visual detection .
As a first approach to examine population correlates of visual detection , we investigated differences in mean activity levels between hit and miss trials ( Figure 2 ) . We defined each neuron’s response during a trial as the mean dF/F0 during the entire stimulus presentation ( Figure 2a , b ) . Because hit/miss would arguably be stronger in the population of neurons that prefer the features of the visual stimulus , we started out with investigating neural correlates of detection in the preferred population . We , therefore , calculated each neuron’s preferred stimulus orientation ( see ‘Materials and methods’ ) , and for the analysis in Figure 2c , d took for each trial the responses of only neurons that preferred the presented stimulus orientation ( henceforth ‘preferred population’ ) . In Figure 2c , all trials of a single animal were grouped by stimulus contrast and behavioral response [hit/false-alarm ( ‘response’ ) or miss/correct-rejection ( ‘no-response’ ) ] , and the average preferred population response was calculated for hits and misses . As expected , the mean response increased with higher stimulus contrasts ( Figure 2—figure supplement 1 for traces across time ) . However , for this animal we did not find a significant difference between hit and miss trials for any individual contrast [false-discovery rate ( FDR ) -corrected paired t-test , p>0 . 05 for all contrasts] ( note that for both false alarms and correct rejections V1 mean population response was indistinguishable from zero; Figure 2c , d , 0% contrast; Figure 2—figure supplement 1a ) . When grouping the test contrasts ( 0 . 5–32% ) , the data did show a modestly higher response for hit than miss trials for single animals as well as across animals ( p<0 . 05 ) . We , therefore , asked whether this increase in neuronal responses during stimulus detection was due to consistent response enhancements of specific neurons or due to a population-distributed process . 10 . 7554/eLife . 10163 . 005Figure 2 . The difference in neural activity between hit and miss trials can be partly explained by consistent hit-associated increases in activity of specific neurons , and somewhat better by trial-by-trial population-wide fluctuations , but these mean-based approaches fall short of being fully descriptive . ( a ) Recorded traces from 10 randomly selected neurons over four subsequent trials . For further analysis , we took the mean dF/F0 per neuron over the visual stimulation period ( thick colored lines ) as single mean neural activity measure per trial . ( b ) Data of one entire recording block consisting of 74 tuned and simultaneously recorded neurons over 336 trials . Blue rectangle shows the four trials depicted in panel ( a ) . ( c ) In an example animal , the detection of stimuli ( green ) with test contrasts ( 0 . 5–32% ) correlated with a modest increase in preferred population dF/F0 over undetected stimuli ( red ) ( two-sample t-test , p<0 . 05 ) but none of the individual contrasts reached statistical significance ( resp . vs . no resp . , two-sample t-tests , FDR-corrected p>0 . 05 ) . ( d ) As ( c ) , but for mean over all animals the graph shows a small , but consistent overall difference of dF/F0 with visual detection ( test contrasts 0 . 5–32% , n=8 animals , p<0 . 05 ) . ( c , d ) Shaded areas show the standard error of the mean . ( e ) The hit-associated increase in neural activity per neuron for all hit trials of test contrast stimuli ( panel I ) can be partly explained by specific neurons showing consistent dF/F0 increases or decreases across trials ( panel II ) partly by trial-by-trial population-wide fluctuations regardless of neuronal identity ( panel III ) and somewhat better by both ( panel IV ) . ( f ) Control analysis by shuffling neuronal identities ( IDs ) per trial ( n=1000 iterations , black distribution ) shows that the population activity is more predictable based on consistent hit modulations per neuron ( top panel ) and more neurons are significantly hit-modulated ( bottom panel ) than can be expected by chance . ( g ) Analyses as in ( f ) , but across animals; comparison versus shuffle-based R2-expectation showed above-chance ( at α=0 . 05 ) predictability of hit modulations using neuron ID , trial ID or both for respectively 7/8 , 8/8 , and 8/8 animals ( left panel ) . The fraction of significantly hit-modulated neurons was above chance ( at α=0 . 05 ) for 7/8 animals ( right panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 00510 . 7554/eLife . 10163 . 006Figure 2—figure supplement 1 . Several control analyses reveal no confounding effects of motor preparation , modulatory feedback , and eye movements . ( a–f ) Traces of population dF/F0 averaged over animals ( n=8 ) for different contrasts show the difference between response and no-response trials over time . ( a ) Probe trials ( 0% stimulus contrast ) . Note the absence of any neuronal activity during responses to 0% contrast probe trials ( green line , t-test vs . 0 , p=0 . 549 ) and ( b ) 0 . 5% contrast , ( c ) 2% contrast , ( d ) 8% contrast , ( e ) 32% contrast , and ( f ) 100% contrast . ( a–f ) Response trials ( green ) show quicker offsets because the stimulus turns off when the animal makes a licking response . Also note the absence of any motor-related neural activity in panel ( a ) supporting the interpretation that the observed correlates are unrelated to motor activity , preparation or reward expectancy . ( g , h ) Removal of locomotion trials does not qualitatively affect neural correlates of stimulus detection . ( g ) When computed only on trials where animals were not moving ( 89 . 9% of trials ) , the mean population dF/F0 as a function of stimulus contrast shows little difference with the original analysis ( compare to Figure 2d ) . Paired t-test over test contrasts ( 0 . 5–32% ) showed a significant difference between response and no-response trials ( p<0 . 05 ) . Our original results are very similar to the current analyses and are therefore not dependent on movement-induced modulations . ( h ) As ( g ) but for heterogeneity ( still trials only ) ; the overall paired t-test for hit/miss differences grouping 0 . 5–32% contrasts was highly significant ( p<0 . 001 ) , suggesting that our main results are not due to locomotion-related artifacts . ( i , j ) Population correlates of visual detection are not dependent on motor-related and/or feedback signals . ( I ) As Figure 2d; mean population dF/F0 as a function of contrast , now using only the first ~400 ms ( 394 ms; 10 frames ) after stimulus onset . Mean reaction time over animals and contrasts was ~1 . 2 s ( see Figure 1f ) leaving on average about 0 . 8 s between the last data point included in this analysis and the subsequent licking response . ( j ) As ( i ) but for heterogeneity ( compare with Figure 3d ) . Results were qualitatively similar to our original analysis for heterogeneity , but not for dF/F0 ( paired t-test over intermediate contrasts for dF/F0 , p=0 . 543 , for heterogeneity , p<0 . 005 ) . ( k , l ) Population correlates are not explained by eye blinks or saccades . ( k ) As Figure 2d; mean population dF/F0 as a function of contrast using only trials in which the animal’s eye position remained fixed and no blinks were detected during the entire stimulus period . ( l ) As ( k ) but for heterogeneity ( compare with Figure 3d ) . Our results are qualitatively and quantitatively similar for dF/F0 and heterogeneity ( hit–miss paired t-test over test contrasts; p<0 . 05 and p<0 . 005 , respectively ) . All panels: shaded areas show the standard error of the mean . Asterisks indicate statistical significance: *p<0 . 05; **p<0 . 005; ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 006 Including again also nonpreferred neurons for all further analyses ( unless stated otherwise ) , we calculated the hit modulation ( dF/F0 increase during hits relative to misses ) per neuron per hit trial ( see ‘Materials and methods’ ) ( Figure 2e ( I ) ) and investigated whether this hit modulation could be explained by a subgroup of neurons that consistently enhances its activation during detection trials , random trial-by-trial population fluctuations , or both ( Figure 2e ) . Hit modulation was explained to a small but significant extent by neuronal identity [R2=0 . 059; p<0 . 05; Figure 2e , f] , and to a larger extent by population fluctuations across trials [R2=0 . 248; Figure 2e ( III ) ] or both processes together [R2=0 . 281; Figure 2e ( IV ) ] . The number of consistently hit-modulated neurons ( which could be either up- or downregulated ) was significantly above chance ( Figure 2f; p<0 . 05 ) . This pattern was robust over animals as hit modulations could be explained with above-chance accuracy by neuron identity , population fluctuations , or both in 7/8 , 8/8 , and 8/8 animals , respectively ( Figure 2g ) . The fraction of significantly hit-modulated neurons was above chance ( at p<0 . 05 ) for 7/8 animals . Although significant for most subjects , the variance explained by the consistency of neuronal responses was fairly low ( always R2<0 . 1 ) , and even the combination of trial-by-trial population fluctuations and neuronal identity never exceeded R2=0 . 35 . This could indicate either that detection-related neural correlates in V1 are minor or that a simple enhancement of mean activity is an index ill-suited to describe potentially strong , but more complex changes in neuronal population dynamics . In particular , we hypothesized that correlates of stimulus detection may be unfolding by multineuron interactions at the single trial level and rely on the relative contrast in activation between neurons . Several metrics aim to quantify response heterogeneity within neuronal populations , such as the sparseness ( Field , 1994 ) , or variance ( Seung and Sompolinsky , 1993 ) . However , such metrics are rarely studied in the context of behavioral relevance , and in the few cases where they are , their ability to predict behavior appeared modest ( Froudarakis et al . , 2014 ) . Therefore , we developed an alternative measure of population heterogeneity that aims to capture the spread in normalized population activity ( Figure 3a , b; see also ‘Materials and methods’ ) : by subtracting the z-scored response ( each trial being a single data point per neuron , see Equation 2 ) of each neuron from that of all other neurons in that same trial , we obtained a Δz-score matrix where high values indicate high pairwise dissimilarity in neuronal activation . Taking the mean over all pairwise Δz-scores provides a measure of population heterogeneity that can in theory be computed over an arbitrarily small time interval ( but note that for all analyses , except those shown in Figure 5 , we used a single trial as time unit ) . This way , similarly strongly activated as well as similarly weakly activated pairs of neurons will decrease heterogeneity . By contrast , dissimilarly activated neuronal pairs ( i . e . one strong , one weak ) will increase it . Therefore , population heterogeneity incorporates both trial-by-trial fluctuations and intra-population differences in a neuronal pairwise manner . Its dependence on z-scored activity means that a neuron’s contribution to heterogeneity is scaled to its relative level of activation—and because highly active neurons are often highly variable ( Baddeley et al . , 1997; Montijn et al . , 2014 ) also to its signal-to-noise ratio . 10 . 7554/eLife . 10163 . 007Figure 3 . Neuronal response heterogeneity within populations correlates better with visual detection than mean preferred population ( pref . pop . ) activity . ( a ) Neuronal activity as in Figure 2d , but presented as z-score normalized per contrast to be able to compare relative changes across contrasts . ( b ) Schematic representation of the method to compute heterogeneity on an example trial ( see also ‘Materials and methods’ ) . The dF/F0 response of each neuron is z-scored per contrast and the distance ( absolute difference ) in z-scored activity between all pairs of neurons is calculated for each trial ( color-coded ΔZ-score ) . The population heterogeneity in a given trial is defined as the mean ΔZ-score over all neuronal pairs . ( c ) Population activity heterogeneity in an example animal shows a strong correlation with visual detection . Comparison between detected ( resp . ) and undetected ( no resp . ) trials for test contrasts as a group ( paired t-test , p<0 . 001 ) was highly significant . ( d ) As ( c ) , but showing mean over all animals ( n=8 ) . Stimulus detection correlated with higher heterogeneity; test contrast group hit–miss comparison was highly significant ( p<0 . 001 ) . ( e ) As ( d ) , but for heterogeneity within the preferred ( left panel ) and within the nonpreferred ( right panel ) population only . Hit–miss differences were found in the preferred population ( test contrast group , p<0 . 01 ) and nonpreferred population ( test contrast group , p<0 . 01 ) similar to the whole population ( d ) . ( f ) Using a measure of effect size analysis ( Cohen’s d ) , heterogeneity was found to show a stronger correlation with stimulus detection than mean dF/F0 within the whole population ( Cohen’s d=0 . 114 vs . d=0 . 218 ) ; within the preferred population ( Cohen’s d=0 . 119 vs . d=0 . 213 ) and within the nonpreferred population ( Cohen’s d=0 . 110 vs . d=0 . 206 ) [paired t-tests over animals ( n=8 ) whole population; p<0 . 05 , preferred population; p<0 . 05 , nonpreferred population; p<0 . 01] . ( g ) Example receiver operating characteristic ( ROC ) curve showing the linear separability of single-trial hit and miss trials using population heterogeneity ( see ‘Materials and methods’ ) . The separability can be quantified by the area under the curve ( AUC; blue shaded area ) . True positive rate: fraction of hit trials classified as hit . False positive rate: fraction of miss trials classified as hit . ( h ) Statistical quantification of hit/miss separability using either mean dF/F0 ( black ) or heterogeneity ( red ) across animals ( n=8 ) . Both measures predict the animal’s response above chance ( FDR-corrected paired t-test dF/F0 and heterogeneity AUC vs . 0 . 5 , p<0 . 05 and p<0 . 001 , respectively ) but behavior can be predicted better using heterogeneity ( paired t-test , dF/F0 vs . heterogeneity AUC , p<0 . 01 ) . All panels: shaded areas/error bars show the standard error of the mean . Statistical significance: *p<0 . 05; **p<0 . 01; ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 00710 . 7554/eLife . 10163 . 008Figure 3—figure supplement 1 . Contrast-dependent responses of the preferred , nonpreferred , and whole population show distributed hit-correlated modulations in heterogeneity , but modulations in mean dF/F0 are smaller and only significant for the preferred population . Single animal neural response examples of the whole ( a , b ) preferred ( c , d ) and nonpreferred ( e , f ) population of neurons . Mean dF/F0 shows hit–miss differences only within the preferred population ( p<0 . 05 ) ( c ) but not within the population as a whole ( a ) nor within the nonpreferred population ( e ) ( paired t-tests , n . s . ) . However , heterogeneity shows significant differences across the whole population ( b ) as well as within the preferred ( d and nonpreferred ( f ) population ( paired t-test , p<0 . 001 for all comparisons ) . ( g–l ) As ( a–f ) but for across-animal comparison of hit–miss differences . All panels: shaded areas show the standard error of the mean . Asterisks indicate statistical significance: *p<0 . 05; **p<0 . 01; ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 00810 . 7554/eLife . 10163 . 009Figure 3—figure supplement 2 . Analysis of hit/miss effect size ( Cohen’s d ) shows that simple average perform worst at separating hit from miss responses . Heterogeneity performed significantly better than mean whole-population ( z-scored ) dF/F0 ( black and blue , FDR-corrected paired t-tests , p<0 . 05 ) and mean preferred population ( z-scored ) dF/F0 ( gray and light blue , FDR-corrected paired t-tests , p<0 . 05 ) . Variance ( green ) sparseness ( orange ) mean and SD of instantaneous Pearson-like correlations ( dark blue and dark purple ) and mean and SD of sliding window correlations ( navy blue and light purple ) did not differ significantly from heterogeneity . Error bars show the standard error of the mean . Asterisks indicate statistical significance: *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 009 We applied this metric to the activity of neurons from the entire population during hit and miss trials and found a stronger correlation with behavioral stimulus detection than for mean response strength ( see Video 1 and Figure 3c for a single animal example ) . Test contrasts ( 0 . 5–32% ) showed a highly significant overall increase in heterogeneity for hit trials ( Figure 3c , paired t-test , p<0 . 001 ) , but such modulations were absent for probe trials . This difference was consistent over animals ( Figure 3d ) and showed similar patterns for the within-preferred and within-non-preferred population heterogeneity ( Figure 3—figure supplement 1 ) . Using a measure of effect size over animals ( Cohen’s d ) , we observed that heterogeneity showed a stronger correlation with visual detection than mean dF/F0 ( Figure 3f; three paired t-tests vs . dF/F0 were all p<0 . 05 ) . Linear single-trial prediction of hit or miss responses with a receiver operating characteristic ( ROC ) analysis on either mean dF/F0 or heterogeneity showed that behavioral responses could be predicted above chance at single-trial basis with both metrics , but heterogeneity showed a significantly higher prediction score [area under curve ( AUC ) , t-test across animals , dF/F0 vs . 0 . 5; p<0 . 05 , heterogeneity vs . 0 . 5; p<0 . 001 , dF/F0 vs . heterogeneity; p<0 . 01] ( Figure 3g , h ) . These results show that correlates of visual detection are better captured by the strength of pairwise response dissimilarities within the neuronal population than to overall increases in mean activation ( but note that correlation-based measures also work well for hit–miss differentiation; Figure 3—figure supplement 2 ) . 10 . 7554/eLife . 10163 . 010Video 1 . Typical raw data example showing the first 10 , 000 recorded frames from animal 5 . The left-hand side shows xy-corrected , but otherwise unaltered , raw fluorescence data . The legend above this raw data video shows the time after start of the experiment in seconds and the number of acquired frames . The top two panels on the right-hand side show ( left ) a depiction of the stimulation screen ( gray isoluminant background or oriented drifting grating during stimulus presentation ) , and ( right ) whether the mouse is making a licking response . The two panels below show a live updated summary of mean dF/F0 ( left ) and heterogeneity ( right ) during each trial . Green indicates a licking response , and red indicates no response . The two lower panels show a live trace of mean population dF/F0 and heterogeneity . Licking responses are shown as red dotted lines , and stimulus presentations are shown as a gray shaded area . Note that the recording is very stable , except during periods of heavy licking , such as after hit responses , when reward is delivered . Also note that neural data acquired during licking are not used for any of our analyses and do , therefore , not influence our results . The mouse is licking vigorously during the initial period of the recording , but more typical behavior sets in less than 2 min after start of the recording . Near the end of the video , it can be seen that hits and misses are more easily separable using heterogeneity than dF/F0 ( although this difference is stronger in the example video than in the entire data set as a whole ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 010 Our observations suggest that not a general gain increase in population activity , but rather more complex changes in response strengths within a population determine the behavioral accuracy . Behavioral reaction time is often used as a proxy for salience , attention , and readiness ( Beck et al . , 2008 ) , so we hypothesized that similar dissociations for fast/slow responses may be found as for hit/miss trials . We performed linear regressions per animal for dF/F0 ( Figure 4a ) and heterogeneity ( Figure 4d ) as a function of reaction time . Similarly to hit/miss differences , the preferred population dF/F0 was not significantly associated with behavioral performance ( regression slopes vs . 0 , FDR-corrected one-sample t-test , n . s . ) , nor were preferred population z-scored activity , variance , sparseness , instantaneous Pearson-like correlations ( see ‘Materials and methods’ ) , whole-population ( raw and z-scored ) dF/F0 , and sliding-window based correlations ( Figure 4—figure supplement 1 ) . However , heterogeneity and the spread in instantaneous Pearson-like correlations were inversely correlated with reaction time ( p<0 . 001 , p<0 . 01 , respectively ) and explained significantly more reaction-time-dependent variance in the data than all other measures ( FDR-corrected pairwise t-tests , heterogeneity vs . all , p<0 . 05 ) . This relationship holds when analyzed over animals ( Figure 4c ) as well as per individual animal ( Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 10163 . 011Figure 4 . Heterogeneity is correlated with reaction time . ( a ) Reaction time shows no correlation with mean preferred population dF/F0 activity during stimulus presentation for any individual animal ( left panel; example animal ) nor for the meta-analysis over animals ( right panel; FDR-corrected one sample t-test over individual regression slopes per animal , n=8 , n . s . ) . ( b ) Same as ( a ) but for heterogeneity . Heterogeneity shows a strong correlation with reaction time ( left panel; example animal , p<0 . 001 ) as well as well for the meta-analysis over animals ( p<0 . 001 ) . ( a , b ) Note different y-axis scaling per panel for display purposes . ( c ) Comparison of the explained variance of several neural metrics . Only heterogeneity ( FDR-corrected one sample t-test , p<0 . 001 ) and spread ( SD ) in instantaneous Pearson-like correlation ( see ‘Materials and methods’ ) ( p<0 . 01 ) correlate significantly with reaction time; all other metrics do not [preferred-population ( Pref . P . ) dF/F0 , preferred-population ( P . P . ) z-scored dF/F0 , variance , sparseness ( kurtosis ) mean instantaneous Pearson-like correlation , whole-population ( z-scored ) dF/F0 , mean and SD of sliding window correlation ( width 1 . 0 s ) ; all n . s . ] . Heterogeneity explains more reaction-time-dependent variance than any other metric ( FDR-corrected paired t-tests , all p<0 . 05 ) . ( d ) Decoding of stimulus presence shows similar accuracy as actual behavioral performance by the animals ( Figure 1e ) . When the animal has detected the stimulus ( resp . ; green line ) , the decoder is better able to correctly judge its presence ( a value of 1 indicates perfect performance , paired t-test , p<0 . 001 ) . Shaded areas show the standard error of the mean . ( e ) Behavioral detection performance is more similar ( sim . ) to the optimal decoder’s performance than expected by chance ( paired t-test , n=8 animals , shuffled vs . real similarity , p<0 . 001 ) . Gray: single animal; blue: mean across animals . All panels: error bars/shaded regions show standard error of the mean . Statistical significance: *p<0 . 05; ***p<0 . 01; ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 01110 . 7554/eLife . 10163 . 012Figure 4—figure supplement 1 . Of several tested neural metrics , only heterogeneity and spread ( SD ) in instantaneous Pearson’s correlations show a significant relationship with behavioral reaction time ( RT ) . ( a–k ) RT relationship of several metrics for example single animal . Each gray point represents a single trial . ( a ) Heterogeneity plotted as a function of RT ( Bonferroni–Holmes corrected regression slope vs . 0 , p<0 . 001 ) . ( b ) population dF/F0 ( n . s . ) . ( c ) z-scored dF/F0 ( n . s . ) . ( d ) variance ( n . s . ) . ( e ) population sparseness ( kurtosis ) ( n . s . ) . ( f ) mean instantaneous Pearson’s correlation ( n . s . ) . ( g ) spread ( SD ) in instantaneous Pearson’s correlation ( p<0 . 01 ) . ( h ) Mean dF/F0 of preferred population ( pref . pop . ) ( n . s . ) . ( i ) Mean z-scored dF/F0 of preferred population ( P . P . ) ( n . s . ) . ( j ) mean of sliding window correlations ( n . s . ) . ( k ) Spread ( SD ) of sliding window correlations ( n . s . ) . ( l–v ) As ( a–k ) but for analysis across animals . Each gray line represents the linear regression of a single animal . Nongray lines represent the mean linear regression over all animals . Statistical significance of behavioral reaction time relationship was determined using Bonferroni–Holmes corrected one-sample t-tests of the slopes of all animals versus 0 . Heterogeneity ( p<0 . 001 ) and spread in instantaneous Pearson’s correlation ( p<0 . 01 ) were significant; all other metrics were not ( p>0 . 3 ) . All panels: shaded areas show the standard error of the mean . Asterisks indicate statistical significance: *p<0 . 05; **p<0 . 01; ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 01210 . 7554/eLife . 10163 . 013Figure 4—figure supplement 2 . Fidelity of stimulus feature representation and population heterogeneity are correlated with accurate visual detection and are not influenced by neuropil contamination . ( a ) Analysis with a maximum-likelihood-naive Bayes decoder shows that during hit trials stimulus features ( orientation and contrast ) are more accurately represented in the population response pattern than during miss trials ( paired t-test , n=8 animals , p<0 . 05 ) . ( b ) Orientation decoding as a function of stimulus contrast shows a sigmoid curve . Statistical analysis revealed significantly above-chance orientation decoding for contrasts higher than 2% [post hoc FDR-corrected one-sample t-tests vs . chance level ( 25%; four orientations were used ) ; 0% , p=0 . 724; 0 . 5% , p=0 . 721; 2% , p=0 . 410; 8% , p<0 . 05; 32% , p<0 . 05 , 100% , p<0 . 05] . ( c , d ) Orientation decoding accuracy does not increase when only strong responses to stimuli are taken into account ( c , one-sample t-test , p=0 . 669 ) but does increase for high population heterogeneity ( d ) ( p<0 . 05 ) . ( e ) Mean pupil size 1 s preceding stimulus onset is correlated with neuronal population heterogeneity during stimulus presentation , suggesting pre-stimulus arousal is related to heterogeneity ( regression analysis per animal , one-sample t-test of slopes vs . 0 , p<0 . 05 , n=8 animals ) . ( f ) Comparison of noise correlations ( NCs ) during slow and fast behavioral response trials for an example animal shows a significant reduction in NCs when the animal responds fast ( two-sample t-test , p<0 . 05 , n=2211 pairs ) . ( g ) Analysis across animals shows that this reduction is consistent ( p<0 . 05 , n=8 animals ) . ( h ) Difference in heterogeneity between hit and miss trials is a population-distributed process and does not critically depend on selecting the most , or least , active neurons . On a single-trial basis , we removed a single quintile of neurons within a z-scored activity bracket and recalculated the hit/miss difference after removal of this quintile ( see ‘Materials and methods’ ) . While removal of the least ( 1st quintile ) or most ( 5th quintile ) active neurons per trial led to a decrease in absolute heterogeneity , the differences between hit and miss trials remained intact ( paired t-tests hit vs . miss , n=8 , 1st quintile p<0 . 05 , 2nd p<0 . 005 , 3rd <0 . 005 , 4th p<0 . 005 , 5th p<0 . 001 , no removal , p<0 . 005 ) . Data show mean and standard error over animals ( n=8 ) . ( i ) As Figure 4c , but when computed for neuropil-subtracted data ( see ‘Materials and methods’ ) . Explained variance values were quite similar for most measures and the difference between heterogeneity and the other metrics appeared larger than without neuropil subtraction . All panels: error bars and shaded areas indicate the standard error of the mean . Asterisks indicate statistical significance: *p<0 . 05; **p<0 . 005; ***p<0 . 001 . ( j ) Chi-square analysis of stimulus presence decoding versus behavioral response show a strong correspondence at single-trial level between the decoder’s output and the animal’s response ( p<10–30 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 013 Our definition of heterogeneity is computationally somewhat similar to the width of the distribution of pairwise neuronal correlations in a population ( see ‘Materials and methods’ ) . However , whereas the spread in instantaneous Pearson-like correlations is based on multiplying z-scored pairwise neuronal responses and taking their standard deviation , the heterogeneity metric ( which instead uses the mean absolute distance in z-scored dF/F0 between pairs of neurons ) was an even better predictor of behavioral reaction times ( Figure 4c , p<0 . 05 ) . As such , it is more closely related to the population mean of nondirectional neuron-pairwise Mahalanobis ( i . e . normalized Euclidian ) distances than Pearson’s correlations . Our analysis shows that visual detection correlates well with large mean Mahalanobis distances in neural activity between pairs of neurons; that is , a high heterogeneity in population activity . Nonetheless , it could be argued that changes in population activity might be uncorrelated with the fidelity with which the population code represents visual stimuli . To address this , we used a Bayesian maximum-likelihood decoder to assess the presence of a stimulus from V1 population activity ( see also ‘Materials and methods’; Montijn et al . , 2014 ) . Decoding performance was higher for behaviorally correct detection trials ( Figure 4d; hit vs . miss , p<0 . 05 ) , and the performance was similar to the animals’ actual behavioral performance at a global level across contrasts ( shuffled vs . nonshuffled , p<0 . 001; Figure 4e ) , as well as at a single-trial level ( chi-square similarity analysis of hit/miss trials for behavioral response and stimulus presence decoding , χ2=135 . 36 , p<10−30; Figure 4—figure supplement 2 ) . Moreover , additional analyses revealed that stimulus features ( orientation , contrast ) were better decodable when the animal made a correct detection ( Figure 4—figure supplement 2a ) and when heterogeneity was high ( Figure 4—figure supplement 2b–d ) . Thus , stimulus features , such as orientation , are represented more accurately by neuronal populations in V1 during hit trials , even though the specific orientation was irrelevant for the animal to perform the visual stimulus detection task . During higher levels of arousal , it has been observed that neuronal activation is more desynchronized ( Cohen and Maunsell , 2009; Froudarakis et al . , 2014 ) . Based on our current observations , this led us to hypothesize that a high heterogeneity in V1 populations reflects a brain state conducive to stimulus detection . If correct , heterogeneity immediately prior to stimulus presentation should be predictive of reaction time . To test this , we split all hit trials into the slowest 50% and fastest 50% per contrast ( e . g . see Figure 5a–d ) and calculated a measure of predictability of slow versus fast responses based on the 3 s preceding stimulus presentation ( Figure 5—figure supplement 1 ) . Using pre-stimulus-onset heterogeneity , fast response trials were highly predictable ( FDR-corrected one-sample t-tests , p<0 . 01 ) , while slow versus miss trials were not ( p=0 . 799 ) . Behavioral responses were not predictable based on population dF/F0 ( slow-miss , p=0 . 157; slow-fast , p=0 . 811; fast-miss , p=0 . 924 ) , and the difference in predictability between heterogeneity and dF/F0 was significant for slow-fast ( p<0 . 01 ) and fast-miss ( p<0 . 05 ) , but not for slow-miss ( p=0 . 477 ) trials . 10 . 7554/eLife . 10163 . 014Figure 5 . Heterogeneity preceding stimulus onset predicts behavioral reaction time . ( a , b ) Example traces from one example animal for population dF/F0 ( a ) and heterogeneity ( b ) of fast ( F , green ) , slow ( S , purple ) , and miss ( M , red ) responses ( mean ± standard error over trials ) also showing mean stimulus offsets . ( c , d ) As ( a , b ) but showing mean ± standard error over animals . ( e ) Fast behavioral responses are predictable before stimulus onset using heterogeneity [FDR-corrected one-sample t-tests vs . chance level ( 0 ) ; S–M , p=0 . 799; F–S , p<0 . 001; F–M , p<0 . 01] but not using dF/F0 ( FDR-corrected one-sample t-tests vs . chance level; S–M , p=0 . 157; F–S , p=0 . 811; F–M , p=0 . 924; FDR-corrected paired t-test for heterogeneity vs . dF/F0; S–M , p=0 . 477; S–F , p<0 . 01; F–M , p<0 . 05 ) . ( f ) The population heterogeneity during stimulus presentation does not merely reflect a continuation of pre-stimulus neural state; detected stimuli ( slow and fast responses ) elicit a faster rise to the maximum heterogeneity level than undetected stimuli ( miss trials ) ( paired t-tests , n=8 animals , p<0 . 05 ) . Slow and fast responses do not differ significantly ( p>0 . 05 ) . All panels: error bars/shaded areas indicate standard error of the mean . Statistical significance: *p<0 . 05; **p<0 . 01; ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 01410 . 7554/eLife . 10163 . 015Figure 5—figure supplement 1 . Behavioral response is predictable on single trials when using heterogeneity , but not when using dF/F0 . ( a ) Predictive decoding of miss trials using mean population dF/F0 ( blue ) or heterogeneity ( black ) during 3 s preceding stimulus onset . Each point is the mean prediction for a single animal ( see ‘Materials and methods’ ) . ( b ) same as ( a ) but for slow trials . ( c ) same as ( a ) but for fast trials . ( d ) Quantification of predictability shows chance-level prediction using dF/F0 ( one-sample t-test , p=0 . 643 ) but above-chance prediction of behavioral responses using heterogeneity ( p<0 . 05 ) as well as a significant difference in prediction performance between dF/F0 and heterogeneity ( paired t-test , p<0 . 05 ) . Points with error bars are mean ± standard error ( n=24; three points per animal: miss/slow/fast ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 015 Although heterogeneity before stimulus onset thus predicts behavior , we also found a dissociation between detected ( slow and fast responses ) and undetected stimuli ( miss trials ) in the rise time latency to maximum heterogeneity upon stimulus onset ( p<0 . 05 ) . Detected trials correlated with fast rise times , while neuronal response heterogeneity to undetected stimuli ramped up much more slowly ( Figure 5f ) . This argues against the interpretation that heterogeneity merely reflects a tonic brain state that can be fully gauged before stimulus onset . The formation of nonhomogenous response patterns within neuronal populations is also related to the actual detection of visual stimuli and constitutes a second effect in addition to background heterogeneity . So far , we have mainly addressed static differences in population activity structure correlating with behavioral responses . However , population codes can show complex temporal properties , such as transient formation of assemblies ( Miller et al . , 2014; Harris and Mrsic-Flogel , 2013 ) . After confirming the stability of our recordings to avoid potential confounds ( Figure 1—figure supplement 1 ) , we addressed whether such temporal population structures might offer additional insight in neural mechanisms of visual detection . We again split the data into miss , fast , and slow response trials , and computed the correlations between response patterns from different trials separately for preferred and nonpreferred neuronal populations ( Figure 6a , b ) . Note that this analysis is not sensitive to potential nonstationary effects that might create artificial differences because all stimulus types and behavioral responses are intermingled in time . 10 . 7554/eLife . 10163 . 016Figure 6 . Consistency in population activation patterns across trials is increased during hit trials ( fast and slow ) compared to miss trials . ( a ) Data from example animal showing inter-trial correlations ( Pearson’s r ) between population responses to same-orientation stimuli ( pooled over test contrasts only ) . ( b ) Data from same animal as in panel ( a ) , showing population higher activity pattern consistency ( mean ± standard error over trial pairs ) for fast and slow response trials than miss trials within both the preferred and nonpreferred population . Colored lines show real data , and black lines show shuffled data ( see text and ‘Materials and methods’ ) . ( c , d ) Inter-trial correlations ( mean ± standard error over all animals ) as quantification of population activation pattern consistency are significantly higher during fast trials ( with a trend for slow trials ) than during miss trials within the preferred as well as the nonpreferred neuronal population , suggesting that visual stimulus detection is correlated with the occurrence of more stereotyped population responses ( FDR-corrected paired t-tests , preferred population; miss-slow , p=0 . 081; miss-fast , p<0 . 05; slow-fast , n . s . ; nonpreferred population; miss-slow , p<0 . 05; miss-fast , p<0 . 05; slow-fast , n . s . ) . Comparison with correlations of shuffled data yielded similar results ( both preferred and nonpreferred populations; paired t-test real vs . shuffled: miss; p>0 . 2 , slow and fast; p<0 . 05 ) . Error bars indicate standard error . Statistical significance: * p<0 . 05; ‡ 0 . 05<p<0 . 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 016 Within the preferred population , as well as within the nonpreferred population , we found that neuronal population activity patterns were more similar during fast trials , with a trend for slow trials , than during miss trials ( preferred population , miss-slow; p=0 . 081 , miss-fast; p<0 . 05 , nonpreferred population , miss-slow; p<0 . 05 , miss-fast; p<0 . 05 ) ( Figure 6c , d ) . To rule out that this effect might arise from biases in the analysis , we also compared these population pattern consistencies to those obtained from a shuffling procedure within stimulus types ( see ‘Materials and methods’ ) . Similarly , population pattern correlations were significantly higher than shuffled for fast and slow trials , while miss trial consistency was not statistically different from the shuffled control ( both preferred and nonpreferred populations , paired t-tests shuffled vs . real , miss; p>0 . 3 , slow and fast; p<0 . 05 ) . However , note that these pattern consistencies are relatively low: they cannot fully account for the population activity structure and must therefore be interpreted as happening against a background of dynamic population activity . Most of our results so far have focused on the experimentally observed differences in hit-miss and reaction-time-dependent effect sizes between heterogeneity and mean population activity , but have not addressed the question how this metric might be interpreted within a theoretical framework . Although heterogeneity correlates better with behavioral responses , and especially with reaction time , than many other metrics , this does not exclude that heterogeneity might be an epiphenomenon . We will address this issue next ( see also ‘Materials and methods’ , section ‘Analysis of multidimensional inter-trial distance in neural activity’ ) using an alternative definition of heterogeneity extended to multidimensional space . This alternative definition is required to study multidimensional properties of population responses , but yields a very similar correlation with stimulus detection as our pairwise definition of heterogeneity ( Figure 3—figure supplement 2 ) . Neuronal population activity during any time epoch can be visualized as a single point in multidimensional neural space . For instance , the mean output in spikes per second of a ‘population’ of two neurons will always be somewhere within a two-dimensional space bounded by the minimal and maximal neural activity of these two neurons ( Figure 7a ) . Within a normalized version of this space , a change in mean neural activity will always be parallel to the main diagonal that crosses the origin ( minimal neural activity of both neurons ) and the point of maximum activity for both neurons . This is true for populations consisting of two neurons , but can be readily extended to any number of dimensions ( Figure 7d ) . Heterogeneity , on the other hand , does not change when a point moves across this diagonal ( the difference will be zero regardless of whether all neurons are firing at 0 spikes per second , or their maximum ) , but rather changes as a point moves orthogonally to this diagonal ( Figure 7c ) . 10 . 7554/eLife . 10163 . 017Figure 7 . Conceptual interpretation of heterogeneity as neuronal population coding phenomenon . ( a–d ) The mean population activity during a certain time epoch can be visualized a single point in multidimensional neural space , where every axis represents the activity of a single neuron . ( a ) For an example population of two neurons , the main diagonal ( arrow ) represents the line along which the mean population activity changes . Orthogonal to this line is the gradient along which heterogeneity changes , representing the distance of each point to the main diagonal . The effects of heterogeneity on hit/miss differentiation as reported in this study could be epiphenomenal if the real underlying differentiation depends on localized , segregated clusters of neural activity for hits ( green cloud ) and misses ( red cloud ) . ( b ) This principle can be extended to multidimensional space; segregated clusters activity will show asymmetrical distributions of population activity around the diagonal . ( c , d ) Alternatively , heterogeneity itself could represent a fundamental characteristic of hit/miss differences; in this case , population responses should be distributed symmetrically around the diagonal ( see text for more explanation ) . ( e ) Calculating the pairwise inter-point distance ( each point being the population activity during a single trial ) can reveal information about the underlying multidimensional structure of neuronal population activity . Green: distribution of inter-point distances for hit trials; red: same for miss trials . ( f ) Population responses during hit trials are distributed within a larger volume of neural space , as shown by the on average larger inter-point distance for hits than misses [paired t-test , difference in center of mass ( d ( CoM ) hit vs . miss , p<0 . 05] . Mirroring point across the diagonal to assess symmetry shows a small , but significant asymmetry for hits and miss ( both p<0 . 05 ) and larger asymmetry for hits than misses ( p<0 . 05 ) . This suggests that neuronal populations during hit trials show more structured behavior in a more extended neural space than during miss trials . ( g ) Schematic representation of how the mean , heterogeneity , or both can be removed from population responses to assess the effect they have on hit/miss separability ( see also text and ‘Materials and methods’ ) . ( h ) Removing heterogeneity impairs hit/miss decoding more than removing the mean ( paired t-test , p<0 . 05 ) but in all cases ( including removing both ) the hit/miss separability is still well above chance ( 0 . 5 ) . This suggests that heterogeneity is more important than population mean activity for differentiating stimulus detection from non-detection , but that other more complex neural phenomena account for most of the population response structure . All panels: error bars indicate standard error . Statistical significance: *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 10163 . 017 The observation we present in our current study—viz . that heterogeneity correlates with hit responses—can be explained by two mutually exclusive hypotheses: ( 1 ) the basis for hit and miss-related responses in V1 resides in specific regions in multidimensional neural response space ( i . e . discrete states in the neural circuit ) and therefore heterogeneity is an epiphenomenon ( Figure 7a , b ) , or ( 2 ) neuronal response heterogeneity per se is important for stimulus detection . The latter implies that neuronal population response patterns during hit and miss trials should be distributed symmetrically around the main diagonal ( which is the gradient along which the mean changes , as well as the axis where heterogeneity is zero ) . This is because regardless of the specific location in multidimensional space , heterogeneity only captures the distance to this diagonal; rotation or mirroring around this diagonal , therefore , does not change heterogeneity , but does in fact change the distribution in multidimensional space ( Figure 7c , d ) . Accurately quantifying a distribution’s shape in multidimensional space requires exponentially more data points as the number of dimensions increases . For direct quantification , our current data set is unfortunately insufficiently large . Therefore , in order to estimate multidimensional symmetry around the main diagonal , we decided to study the effect of mirroring points across this diagonal . First , we calculated the distribution of pairwise inter-point distances in neural response space without mirroring ( Figure 7e ) . In this case , each point again represents the population response during a single trial . The data show that the inter-point distance was slightly , but significantly larger for hit than miss trials ( paired t-test , p<0 . 05 ) ( Figure 7f ) . This indicates that population responses during hits encompass a larger volume of neural response space than during misses , which will increase heterogeneity values and allows more information to be encoded with the same number of neurons . To assess symmetry specifically , we mirrored each trial one at a time across the main diagonal and recomputed in each case the distribution of pairwise inter-point distances . If the population responses are distributed asymmetrically around the diagonal , then mirroring will increase the pairwise distance , while if they are distributed symmetrically no change should be observed . There was a small but significant increase in inter-point distance for both hits and misses , and mirroring increased the inter-point distance more for hits than misses ( p<0 . 05; Figure 7f , inset ) . Although the effect sizes are small , they are significant , and we can therefore—at least based on this analysis—not ( yet ) conclude that heterogeneity is more than an epiphenomenon . In fact , the difference between hits and misses suggests that population responses during hit trials are more asymmetrical ( i . e . more clustered in discrete states of neural activity ) than during miss trials ( Figure 7f , inset ) . Considering that inter-trial population pattern consistencies were lower during miss trials ( Figure 6 ) , we can conclude that neuronal populations during miss trials show more random behavior within a limited neural space , while neuronal populations during hit trials show more structured behavior in a more extended neural space . Noting the small effect size of the previous analysis , we asked whether the removal of the mean of the neuronal population response was as detrimental to encoding hit/miss differences as the removal of heterogeneity . Again visualizing population responses as points in neural space , one can remove any differences in mean response between trials by projecting all population responses to a plane orthogonal to the diagonal or remove any differences in heterogeneity by projecting all points onto a manifold at a fixed distance from the diagonal ( see ‘Materials and methods’ and Figure 7g ) . To test the effect of these removals on hit/miss differences , we performed a decoding procedure of hit versus miss trials ( i . e . we decoded the animal’s response ) on the original data , and on data with the mean , the heterogeneity or both aspects removed . The results show no difference between the original data and the mean-removed data , but removing heterogeneity ( or both heterogeneity and the mean ) led to a small but significant decrease in decoding performance ( heterogeneity-removed vs . original data , p<0 . 05; heterogeneity-removed vs . mean-removed , p<0 . 05 ) ( Figure 7h ) . However , even with heterogeneity removed , decoding performance was still well above chance ( 63% correct for original data , 60% correct for both removed; 50% is chance ) . We , therefore , conclude that heterogeneity contributes significantly as a non-epiphenomal population property to the differentiation in neural responses between visual detections and failures to detect a stimulus , but also that most information resides in neuronal population response patterns other than its mean or heterogeneity . Moreover , the mean response of a neuronal population is less important than its heterogeneity . While neural response heterogeneity may be an important factor and useful metric for its strong correlation with especially reaction times , further research is required to discover which other neural properties may be important for visual stimulus detection .
Our analyses show differences between hit and miss trials that we interpret as being related to perception and visual processing . However , in principle , the observed differences could be subject to a number of confounds that might limit their interpretability . First , the relatively mild water restriction led to a behavioral performance at the 100% probe trials that is lower compared to other studies using similar tasks ( Glickfeld et al . , 2013 ) . This could mean that the observed differences in heterogeneity between hit and miss trials are due to changes in motivation rather than visual detection . If this were the case , however , hit–miss differences should be as large during 100% probe trials as during test contrast trials . Figure 3 and Figure 3—figure supplement 1f , 2 shows that this is not the case; during intermediate test contrasts ( 2% and 8% ) , the hit–miss difference in heterogeneity is largest . A similar reasoning applies to 0% contrast probe trials , where a heterogeneity difference between ( false alarm ) responses and correct rejections was lacking , which strongly argues against heterogeneity being due to response emissions per se . Thus , it is highly unlikely that the neural correlates of behavioral responses as reported in this study are due to differences in motivation . Given this caveat on suboptimal performance , one may predict that a better behavioral performance would have only increased the hit–miss effect sizes we report in this study . Other potential confounds include instabilities in z-plane focus , other locomotor-related artifacts and running-induced modulations , as it has been reported that behavioral activity and running can induce instabilities in the plane of focus with awake two-photon calcium recordings , as well as changes in the neuronal responses in mouse V1 ( Dombeck et al . , 2007; Niell and Stryker , 2010; Saleem et al . , 2013 ) . To address potential z-shifts during the acquisition of neural data , we compared each imaging frame to 3D anatomical z-stacks acquired after recording the neural data ( see ‘Materials and methods’ ) . Slight changes in z-location were detected after the onset of hit responses by licking , but these were mostly confined to the reward presentation period ( which was not used for any of our analyses ) and rarely exceeded more than a couple of microns ( Figure 1—figure supplement 1 ) . Moreover , exclusion of trials where mice were running did not qualitatively change hit–miss differences in neuronal activity ( Figure 2—figure supplement 1g , h ) , nor did using only the first 400 ms after stimulus onset to avoid licking-preparation feedback from other brain regions ( Figure 2—figure supplement 1i , j ) . To control for potential confounds related to eye movements and blinking , we analyzed eye-tracking videos to detect blinks and saccades . After removing all trials where the animals were making saccades or were blinking , we again found no qualitative difference with the results we observed previously ( Figure 2—figure supplement 1k , l ) , but did observe a small but significant correlation between pupil size and heterogeneity , suggesting higher heterogeneity with increased arousal states ( Figure 4—figure supplement 1e ) . Overall , we conclude that the neural correlates we report here , and interpret as related to perception , are most likely not due to recording instability , changes in motivation , locomotion , motor-related signals associated with licking , or eye movements and blinking . Importantly , our results only pertain to L2/3 of the primary visual cortex in mouse , which does not exclude the possibility that the mean population response of , for example , deeper layers ( L5 ) in V1 would correlate better with visual stimulus detection . Previous research has shown that extensive differences exist between superficial and deep layers: whereas L2/3 neurons often show relatively low peak firing rates and sparse responses to sensory stimulation , L5 neurons show denser response patterns with on average higher peak firing rates ( De Kock and Sakmann , 2008; Harris and Mrsic-Flogel , 2013 ) . In somatosensory cortex , it has been shown that hits and correct rejections in a go-no-go object localization task can be better separated using mean spiking rates in L5 than in L2/3 ( O'Connor et al . , 2010 ) . Our result that L2/3 populations show only a small differentiation between hit and miss responses in mean activation should therefore not be taken as proof for a canonical principle also applicable to other cortical layers . Future validation of our results in deep layers is necessary for a decisive answer whether our results are indeed applicable to different layers of primary sensory cortex . Our task design included drifting gratings with different orientations , with the qualification that orientation was task-irrelevant for the mice , as they were only required to detect stimuli whenever they appeared . Our observation that stimulus features are represented more accurately ( as quantified by decoding accuracy ) during hit than miss trials may therefore be somewhat surprising ( Figure 4—figure supplement 2a ) . This suggests that mechanisms that increase the likelihood of stimulus detection may be acting through a general enhancement of stimulus processing intensity , corroborating previous research in monkey showing that attention can lead to horizontal shifts in contrast response curves , as if the stimulus were of higher contrast ( Martı́nez-Trujillo and Treue , 2002 ) . It is interesting to ask whether our results on heterogeneity can be cast in terms of dynamic range effects . Neurons are expected to climb in this dynamic range when visual contrast increases , which is confirmed by the rise in dF/F0 ( Figure 2d ) . However , if heterogeneity would be primarily determined by neurons being able to operate along the steep slope of their dynamic range , then the large difference in heterogeneity between hits and misses ( Figure 3c ) along the test contrasts ( 0 . 5–32% ) would not be expected . Of further interest is to compare our results on heterogeneity with studies reporting that sparseness in L2/3 populations of rodent V1 is high during passive viewing ( Barth and Poulet , 2012 ) depends on cortical state and improves neural discriminability during passive processing of natural scenes ( Froudarakis et al . , 2014 ) . Although in our analysis sparseness and variance explained more behavioral variability in reaction time than ( z-scored ) mean population activity ( Figure 4c ) , these measures perform much worse than heterogeneity and the spread of instantaneous Pearson-like correlations . Possibly , a sample of ~60–70 tuned neurons is insufficient to estimate instantaneous sparseness accurately . An alternative explanation for this poor correlation could be that sparseness of L2/3 populations results from anatomical wiring required for efficient stimulus coding and to enable locally selective synaptic plasticity without immediately changing the coding of stimulus features within the population response ( Rao and Ballard , 1999 ) . Correlates of visual detection that depend on accurate stimulus feature representation might then be better captured by a maximization of Mahalanobis distances in neural activity between pairs of neurons within this already sparse network . This latter interpretation is in line with the data we recorded and suggests that sparse stimulus representation by L2/3 neurons reflects a structural optimization of the population code to represent stimulus features , while heterogeneity captures more temporally dynamic modulations related to perception . In addition to our approach based on pairwise relations in neuronal responses , we investigated multidimensional patterns of population activity ( Figure 7 ) . These results indicate that , while heterogeneity is more important for separating stimulus detections from nondetection in neural response space than the population mean , these properties combined still cannot capture the full set of neuronal response characteristics that define the accurate detection of visual stimuli in L2/3 of mouse V1 . This suggests that other patterns of population activity , such as potentially transient assembly formation , may be important for visual stimuli to be correctly detected . From our multidimensional analyses , we can conclude that simple bulk approaches ( i . e . correlating a population’s mean response with behavioral output ) are insufficient when one aims to address how early sensory cortex areas are involved in the processing and detection of visual stimuli . Related to these findings is our observation of behavioral-state-specific consistencies in population activation patterns across trials . This provides some constraints on how population heterogeneity is modulated at a neurophysiological level . Neuromodulators such as acetylcholine ( ACh ) and noradrenaline are correlated with attention and arousal , and may influence cortical population dynamics ( Metherate et al . , 1992; Coull et al . , 2004; Pinto et al . , 2013 ) , such that they facilitate repeated activation of similar subnetworks of neurons within a population responding to the same stimulus . Without such neuromodulators , neurons within the same preferred population would randomly participate in representing the current stimulus . This interpretation is compatible with previous work; for instance , ACh has been observed to influence burst spiking , membrane potential fluctuations , cortical oscillations , and desynchronization . These processes have been implicated in modulating competitive inhibition effects within neuronal populations and may very well influence the consistency of specific neuronal subnetworks being activated ( Borgers et al . , 2008; Fries , 2009; Bosman et al . , 2014 ) . If heterogeneity in a recurrently connected V1 population is in part determined by suppression of the most weakly by the most strongly stimulus-driven neurons , then behaviorally correlated heterogeneity enhancements may be another facet of arousal as well as perception-related modulations of stimulus-evoked population activity . Population coding phenomena have long been hypothesized to be important for sensory processing , but so far few studies have investigated their relevance for perceptual decisions . Here , we show that population heterogeneity is correlated with behavioral stimulus detection and that it predicts correct behavioral performance . Our results imply that neurophysiological measures dependent on population averages ( i . e . multiunit activity , EEG , and fMRI ) may underestimate the correlation between visual detection and V1 L2/3 activity because the assumption of population response homogeneity is violated especially during active processing of visual information . In short , our results support contrast-sensitive changes in mean population activity during visual task performance ( Figure 3c , d ) , but stress the importance of population recordings with single-cell resolution ( Figure 4c–f ) .
All experimental procedures were conducted with approval of the animal ethics committee of the University of Amsterdam ( cf . Goltstein et al . , 2013; Montijn et al . , 2014 ) . Experiments were performed on eight adult , male wild-type C57BL/6 mice ( Harlan ) , 128–164 days old at the day of calcium imaging ( 29 . 1–32 . 7 g ) . Prior to the imaging experiment , all animals were surgically fitted with a head-bar implant and trained head-fixed for up to 3 months to perform a visual go/no-go detection task . At the day of the imaging experiment , we performed intrinsic signal imaging to define the area corresponding to the retinotopic region in V1 responsive to the visual stimulus . We performed a small ( 1 . 5–2 . 0 mm ) craniotomy at that location and used multicell bolus loading with Oregon Green BAPTA-1 AM to record calcium transients and sulforhodamine-101 ( SR101 ) to label astrocytes ( Stosiek et al . , 2003; Nimmerjahn et al . , 2004 ) . Mice were trained 5 days per week , each for approximately 45 min per day , on a head-fixed go/no-go visual detection task over a period of 10–12 weeks , where we aimed to get sufficient hit as well as miss trials for test contrasts . Mice were water-deprived for 6 h preceding training and otherwise had ad libitum access to water . Weight was monitored three times per week and never dropped below 90% of their nonrestricted growth curve . Behavioral training was performed inside four dark , sound attenuated chambers and occurred during the active ( dark ) cycle of the animals; each animal was always trained in the same behavioral setup . We did not observe any deviant learning effects associated with any specific behavioral setup ( data not shown ) . During the first five days of training , we conditioned licking in response to visual stimulation by pairing passive stimulation with reward delivery ( ~9 µl of water with 15% sucrose with 1% vanilla extract ) ( stage 1 ) . After the conditioning phase , visual stimuli ( 100% contrast drifting gratings as described in the previous paragraph ) were presented indefinitely until mice made a licking response that was monitored using a custom-built infrared LED-based lick detector . When animals made a response , the visual stimulus presentation terminated and reward was available for 5 s . This shaping phase ( stage 2 ) lasted for a maximum of 5 days or less if the animals were often making clear lick responses . After this ~2-week initial phase , we started training the animals on a simple version of the final task ( stage 3 ) ; maximum stimulus presentation was reduced to 5 s and subsequent trials would only start if the mice did not make any lick responses for at least a random interval of 1–3 s . During this stage , reward size was gradually reduced to ~3 µl per trial . When animals would consistently perform at least 80 trials within a period of 45 min , they would be moved to the next stage . In stage 4 , we introduced 0% contrast probe trials to monitor the behavioral performance of animals by testing for false-alarm responses and calculating if they showed statistically significant above-chance performance . In this stage , we also lengthened the inter-trial interval to any random duration between 6 and 8 s . Once mice attained a sufficient ratio of hit/miss trials , we moved them to training stage 5 , where we increased the inter-trial interval to 10–12 seconds and presented mild air puffs as a negative reinforcer whenever mice would lick outside the stimulus presentation or reward delivery period . At this stage , animals were required to not lick for a random interval of 1–3 s in order to gain access to the next stimulus presentation . Stage 5 lasted until the mice had been trained for 8–10 weeks in total . Finally , if mice performed consistently and significantly above chance during stage 5 ( n = 12 / 21 animals ) , then in the 2-week period preceding the imaging experiment mice were trained on the microscope setup , and our setup’s resonant mirrors were activated to produce the characteristic 8000 Hz sound that would also be present during calcium imaging . In this final stage , all possible efforts were made to simulate surroundings of the eventual calcium imaging experiment as closely as possible to habituate the mice to the two-photon laser lab’s environment . Mice were always allowed to take up to 3 s after stimulus onset to respond and were thus not explicitly trained to make fast behavioral responses . On the day of the two-photon calcium imaging experiment , buprenorphine ( 0 . 05 mg/kg ) was injected subcutaneously 30–60 min before induction of anesthesia with isoflurane ( 4 . 0% induction , 0 . 8% maintenance during intrinsic signal imaging , 1 . 5–2 . 5% maintenance during invasive surgical procedures ) . After induction , the animal was placed in a custom-built head-bar holder designed for performing surgical procedures . We removed the cover glass , silicon elastomer , and layer of glue covering the skull in the cranial window before performing intrinsic signal imaging to localize the precise location of our stimulus’ receptive field location in the primary visual cortex ( V1 ) . We subsequently performed a small ( 1 . 5–2 mm ) craniotomy above the retinotopic area responding to visual stimulation with drifting gratings . After the craniotomy , the dura was kept wet with an artificial cerebrospinal fluid ( ACSF: NaCl 125 mM , KCl 5 . 0 mM , MgSO4 * 7 H2O 2 . 0 mM , NaH2PO4 2 . 0 mM , CaCl2 * 2 H2O 2 . 5 mM , glucose 10 mM ) buffered with HEPES ( 10 mM , adjusted to pH 7 . 4 ) . After making the craniotomy , multicell bolus loading with Oregon Green BAPTA-1 AM ( OGB ) and SR101 was performed 230-270 µm below the dura as previously described in Montijn et al . , 2014 and Goltstein et al . , 2013 . After injection of the dyes , the exposed dura was covered with agarose ( 1 . 5% in ACSF ) and sealed with a circular cover glass that was fixed to the skull using cyanoacrylate glue . The animal was allowed to recover for a minimum of 90 min before starting the behavioral task and two-photon calcium imaging . Of the 12 mice that learned the task , 2 animals were rejected due to insufficient imaging quality . All visual stimulation was performed on a 15 in . TFT screen with a refresh rate of 60 Hz positioned at 16 cm from the mouse’s eye , which was controlled by MATLAB using the PsychToolbox extension ( Brainard , 1997; Pelli , 1997 ) . Stimuli consisted of sequences of eight different directions of square-wave drifting gratings that were monocularly presented in randomized order . Visual stimulus duration started at infinite during the initial training phase and was gradually reduced to a maximum duration of 3 s for the final task stage . Stimuli were alternated by a blank inter-trial interval of variable duration ( random minimum of 10–12 s ) during which an isoluminant gray screen was presented . Visual drifting gratings ( diameter 60 retinal degrees , spatial frequency 0 . 05 cycles/° , temporal frequency 1 Hz ) were presented within a circular cosine-ramped window to avoid edge effects at the border of the circular window . A field-programmable gate array ( OpalKelly XEM6001 , Opal Kelly Incorporated , Portland , OR ) was connected to the microscope and behavioral setup and interfaced with the visual stimulus presentation computer to synchronize the timing of visual stimulation with the microscope frame acquisition and behavioral setups . Slow z-drifts were quantified by comparing the similarity of 100 frames in the beginning , middle and end of each stimulus repetition set to slices recorded at different cortical depths ( step size ~1–2 µm ) before or after functional calcium imaging was performed for five of eight animals . If z-drifts larger than 10 µm occurred slowly over multiple repetition blocks , or if slow z-drift was detected manually , the entire recording of a single animal was split into multiple analysis periods ( n=2 populations for animals 1 and 7; n=1 population for all other animals ) and analyzed independently ( Figure 1—figure supplement 1 ) . For the two animals for which we split the recordings , we afterwards averaged all measures over the two populations , yielding a single independent data point also for these animals for each measure . To confirm the stability of our recordings , we performed a further analysis quantifying the discriminability of neurons relative to their surroundings over time ( Figure 1—figure supplement 1 ) . Therefore , we calculated during each imaging frame the mean fluorescence of the pixels within the neuron’s soma ( Fsoma ) and the fluorescence of a neuropil annulus surrounding the soma ( Fneuropil ) , which we defined as all pixels within a concentric band from 2–5 µm away from the soma . For each frame , we then calculated the discriminability ratio Dr as Dr = Fsoma / ( Fsoma+ Fneuropil ) , and set a threshold at Dr=0 . 5 ( equal luminance of soma and neuropil ) . Whenever this measure dropped below the threshold , we calculated the duration of this epoch until it would return to above the threshold , and took the maximum duration of all these epochs as a single measure per neuron . Most neurons from all sessions showed maximum below-threshold durations near 0 s , and no neurons showed durations longer than 1 s ( Figure 1—figure supplement 1 ) . To address the potential confound of fast changes in z-plane due to anticipatory fidgeting behavior by the animals , we calculated the depth of each imaging frame and analyzed whether responses to visual stimuli were preceded by shifts in z-plane that could influence our results . As can be seen in figure supplement 1-1L–O , z-shifts were mostly confined to the epoch immediately following hit responses , which are not used in our analyses , and in general z-shifts were very small and rarely exceeded more than 1 µm . We recorded eye movements during the entirety of the calcium imaging experiment to be able to correct for possible contamination of our results by excessive blinking and/or saccades . For this purpose , we placed a near-infrared light sensitive camera ( JAI CV-A50IR-C Monochrome 1/2" IT CCD Camera , JAI A/S , Germany ) with a large-aperture narrow-field lens ( 50 mm EFL , f/2 . 8 ) above the visual stimulation screen directed at the mouse’s visually stimulated eye . Images were acquired at 25 Hz and pupil tracking was performed offline using custom-written MATLAB scripts . Eye position was used to control for possible saccade effects ( Figure 2—figure supplement 1k , l ) , and pupil diameter was used to assess its correlation with heterogeneity ( Figure 4—figure supplement 2e ) . Dual-channel two-photon imaging recordings ( filtered at 500–550 nm for OGB and 565–605 nm for SR101; see Figure 1d ) with a 512 x 512 pixel frame size were performed at a sampling frequency of 25 . 4 Hz . We used an in vivo two-photon laser scanning microscopy setup ( modified Leica SP5 confocal system ) with a Spectra-Physics Mai-Tai HP laser set at a wavelength of 810 nm to simultaneously excite OGB and SR101 molecules , as previously described ( Montijn et al . , 2014 ) in cortical layer 2/3 at depths from the pia mater ranging from 140 to 170 µm ( Figure 1—figure supplement 1 , Video 1 ) . During data acquisition , mice were performing a go/no-go stimulus detection task where the animals had to lick whenever a visual stimulus was presented . Stimulus parameters were equal to those described above . We varied the contrast of the drifting grating ( 0% , 0 . 5% , 2% , 8% , 32% , and 100% ) to elicit a wide range of hit/miss ratios . Responses to 0% contrast probe trials were not rewarded , but responses to all other contrasts were . We did not explicitly aim for very high detection performance ( high hit rates and low miss rates ) to avoid overtraining and associated habitual or automated responding ( Balleine and Dickinson , 1998 ) . A complete set of visual stimuli , therefore , consisted of 48 trials ( 6 contrasts times 8 directions ) . The order of presentation of these 48 trials was randomized independently for each repetition block . After the experiment was completed , we tested for statistically significant stimulus detection performance by calculating the binomial 2 . 5th–97 . 5th percentile intervals ( henceforth 95% CI ) of response proportion to the two probe trial types—100% and 0% contrast stimuli—using the CP method . Of the 10 animals from which we recorded calcium imaging data during task performance , one was rejected because of excessive variability in responses due to brain movement and one was rejected due to insufficient discriminability between the two types of probe trials ( overlapping CIs ) . All data we present in this paper are from the remaining eight animals . The number of repetitions per stimulus type ( unique orientation x contrast ) ranged from 6 to 16 . For most analyses , we took the mean over all orientations ( n=4 ) , so each contrast was presented 24–64 times . For all analyses of single-animal data , each trial was taken as a single data point , where its value was the mean dF/F0 over all recorded frames during stimulus presentation ( which was dependent on the reaction time of the mouse ) . To avoid the confound of having higher signal-to-noise ratios for miss than hit trials due to longer data acquisition , within each contrast group we randomly assigned to all miss trials a duration randomly selected from the reaction time distribution of hit trials . After a recording was completed , small x–y drifts were corrected offline with an image registration algorithm ( Guizar-Sicairos et al . , 2008 ) . To retrieve dF/F0 values from the recordings , regions of interest ( ROIs; neurons , astrocytes , and blood vessels ) were determined semiautomatically using custom-made MATLAB software for each repetition block separately ( see https://github . com/JorritMontijn/Preprocessing_Toolbox ) . For these ROIs , we subsequently calculated dF/F0 values as previously described ( Montijn et al . , 2014 ) : For each image frame i , a single dFi/F0i value was obtained for each neuron by calculating the baseline fluorescence ( F0i ) , taken as the mean of the lowest 50% during a 30 s window surrounding image frame i . dFi is defined as the difference between the fluorescence for that neuron in the given frame and the sliding baseline fluorescence ( dFi = Fi – F0i ) ( Montijn et al . , 2014 ) . The mean number of simultaneously recorded neurons/session was 92 . 6 [range 68 – 130 ( SD: 19 . 0 ) neurons] . After this initial analysis , all neurons were tested on consistency for preferred stimulus orientation and any neurons that showed inconsistencies over different repetition blocks ( i . e . more than one-third showing different preferred orientations ) were rejected from further analysis [mean number of consistently tuned neurons per animal was 66 . 3 ± 18 . 6 ( 70 . 8% ± 7 . 75% of all neurons ) ( mean ± SD ) ] . Unless otherwise specified , all analyses shown in this paper are based on across-animal meta statistics based on a set of eight independent data points ( one data point/animal ) and all multiple comparison t-test p-values were adjusted by the Benjamini and Hochberg FDR correction procedure and were deemed significant if the resultant p-value was <0 . 05 . For quantification and control procedures related to z-drift and recording stability , see Figure 1—figure supplement 1 . For control analyses where we performed neuropil fluorescence subtraction ( Figure 4—figure supplement 2i ) , we used similar procedures as described previously ( Greenberg et al . , 2008; Mittmann et al . , 2011 ) ; we calculated the correlation ( r ) between each neuron’s somatic fluorescence and surrounding neuropil ( annulus between 2 and 5 µm from soma ) and corrected on each frame the neuron’s fluorescence as follows: Fcorr = Fsoma – r * Fneuropil . Estimated neuropil contamination varied widely between neurons , but was generally in the range between 0 . 1 and 0 . 6 , similar to previously reported values ( Greenberg et al . , 2008; Mittmann et al . , 2011 ) . We recomputed the explained variance of several metrics as a function of reaction time ( see Figure 4c , Figure 4—figure supplement 1 ) and found that neuropil correction did not affect our main conclusions ( Figure 4—figure supplement 2i ) . All linear regressions were performed on single-animal data sets , yielding regression coefficients for the intercept and slope through minimizing the error between a linear function and the single animal’s data points . Statistical significance was quantified by performing a one-sample t-test of the coefficients from all animals ( n=8 ) . Significance level was set at an α of 0 . 05 and p-values were adjusted if necessary by a post hoc Bonferroni–Holmes correction . We presented eight directions of visual drifting gratings and calculated the preferred stimulus orientation of all neurons by summing opposite directions as belonging to the same stimulus type because the vast majority of mouse V1 neurons is tuned sharply to an axis of movement , but much less so to a specific direction within that axis ( i . e . most neurons are strongly orientation-tuned , but less direction-tuned; e . g . Andermann et al . , 2011 ) . For these four orientations , we took each neuron’s mean response over all trials and defined its preferred orientation as the stimulus that caused the highest mean dF/F0 value . For most analyses , we used the neuronal responses to all orientations , except for Figure 2c , d , where we used only the response of the preferred orientation , as we hypothesized the preferred population might yield stronger hit/miss differences in neuronal activity . To investigate the source of hit-related increases in population dF/F0 and determine whether there might exist a subgroup of neurons that consistently enhances its activation during detection trials ( as compared to nondetection ) , we defined a dF/F0 hit modulation index Ψ for each hit trial ( t ) for each neuron ( i ) as the neuron’s dF/F0 activity ( R ) relative to the mean ( µ ) and standard deviation ( σ ) of its response during miss trials ( m ) of the same type [identical orientation ( θ ) and contrast ( c ) ]: ( 1 ) Ψi , t = ( Ri , t - μm , c , θ ) / σm , c , θ In other words , Ψi , t of a given trial represents the z-scored dF/F0 activity relative to the neuron’s response to the same stimulus when the stimulus remained undetected ( Figure 2e , left panel ) . The hit-modulation matrix Ψ of all hit trials and all neurons can then be approximated by neuron identity ( mean over trials ) , trial-by-trial fluctuations ( mean over neurons ) , or both ( addition of the matrices yielded by the two previous approximations ) ( Figure 2e ) . We then calculated the explained variance ( R2 ) of the population response pattern by its canonical equation based on the residual ( SSres ) and total sum of squares ( SStot ) . We defined SStot as the sum of all squared values in Ψ , and SSres as the sum of the squared differences between Ψ and the approximation matrix as defined above ( by neurons , trials , or both ) . To assess significance , we performed 1000 shuffle iterations where we randomized neuronal identities per trial ( for approximation by neuron identity ) , randomized trial identities per neuron ( for approximation by trial identity ) , or randomized both ( for approximation by both ) . Per shuffle iteration , we calculated the explained variance , which yielded a shuffled distribution per prediction ( e . g . Figure 2f ) . A prediction was defined as significantly above chance when the real explained variance was at least 2 SDs away from the shuffled distribution mean ( corresponding to p<0 . 05 ) . We calculated heterogeneity of population activity as follows ( see also Figure 3d ) . For each independent data source i ( i . e . a neuron ) that provides a certain measurement R at each time point t ( i . e . dF/F0 activity of a single trial ) , we first z-scored the responses of i over all trials T ( i . e . all contrasts and orientations ) . For all analyses we took t to be a single trial , except those shown in Figure 5 , where t corresponds to a data acquisition point ( i . e . a single calcium imaging frame ) , and calculated heterogeneity as follows . First , we z-scored all trial responses per neuron over all trial types ( therefore high-contrast , preferred orientation stimuli yield higher z-score values than low-contrast , nonpreferred orientations ) : ( 2 ) Zi , t = ( Ri , t - μi ) /σi Z is therefore a matrix containing n ( number of neurons ) by T ( trials ) measurements of standard deviations ( σ ) from the mean over all trials ( μ ) . Next , for each trial t , we calculated the pairwise distance ( in standard deviations ) from each independent source to each other independent source ( pairwise neuronal Δσ ) : we repeated the z-scored population response vector zt over its singular dimension n times , where n is the number of neurons in zt ( yielding a square matrix ) , subtracted this matrix from its own transpose ztT , and took the absolute of the result , giving the heterogeneity matrix Ht: ( 3 ) Ht = | zt - ztT | To get a single measure of population heterogeneity per trial ( ht ) , we next took the mean of all z-scored distances between neuronal pairs ( i , j ) in the heterogeneity matrix; this provides a measure of the mean distance in activation levels within our population at a single trial t: ( 4 ) ht = ∑i=[1 … n-1]∑j=[i+ 1 … n] ( Ht , i , j ) / ( ( n · ( n - 1 ) ) /2 ) We used a measure of effect size using Cohen’s d to quantify which metric ( mean dF/F0 or heterogeneity ) showed a stronger correlation with visual detection . We calculated for both metrics per animal the effect size for all intermediate contrasts ( 0 . 5–32% ) between hit and miss trials and took the mean over these four values , yielding a mean hit/miss effect size for dF/F0 and heterogeneity per animal . This allowed us to perform a paired t-test between the dF/F0 effect sizes and heterogeneity effect sizes to test for statistical significance . Cohen's d is defined as the difference between the two means ( hit; µh , miss; µm ) divided by the pooled standard deviation for the data: ( 5 ) d = ( μh - μm ) / σp , where σp is defined as ( 6 ) σp = √[ ( nh - 1 ) ·varh + ( nm - 1 ) · varm]/ ( nh + nm- 2 ) For a pair of neurons x and y , the Pearson’s correlation ( R ) of their activity can be calculated by z-scoring each neuron’s response vector ( as in Equation 2 ) and taking the mean of the element-wise multiplication of the two vectors: ( 7 ) Rx , y =∑t=[1 … T] ( Zx , t · Zy , t ) /T Here , notations are the same as for Equations 3–5; t is a single trial and T is the total number of trials . Using this equation , it is impossible to obtain an instantaneous correlation value between two neurons for each trial because its calculation requires taking the mean over all trials . This poses a problem if we want to estimate the instantaneous correlation value between a pair of neurons for a given trial . Therefore , we computed a modified measure , the instantaneous Pearson-like correlation ( Ř ) . For each pair of neurons , we calculated the z-scored element-wise product ( each element being a single trial ) , which yields a three-dimensional matrix Ž with size [n by n by T] , where n is the number of neurons: ( 8 ) Žx , y , t = Zx , t · Zy , t Taking the mean over the matrix’s third dimension ( trials ) gives the conventional Pearson’s pairwise correlation matrix over neuronal pairs . However , the matrix also allows us to approximate the mean pairwise correlations within the whole population at any given trial ( Řt ) by taking the mean over all unique neuronal pair values in matrix Ž: ( 9 ) Řt=∑i=[1 … n-1]∑j=[i +1 … n] ( Ži , j , t ) / ( ( n · ( n - 1 ) ) /2 ) Similarly , we can take the standard deviation instead of the mean over all unique pairs per trial to estimate the spread of the instantaneous pairwise correlation distribution . However , note that while the instantaneous Pearson-like correlation is similar to the conventional Pearson correlation , Ř is not bounded within the interval [−1 1] , because the z-scored element-wise product and the mean-operator work over different sets of values ( i . e . matrix dimensions ) . We additionally used for comparison a more conventional measure of correlations across time by using a wavelet-based sliding-window correlation ( Cooper and Cowan , 2008 ) . The time scale of the wavelet used in all sliding-window analyses was set to 1 . 0 s as this was similar to the animals’ median reaction times and should therefore maximize the stimulus-driven change in neuronal pairwise correlations . We quantified the single-trial behavioral response predictability using an ROC approach by calculating the area under the curve ( AUC ) for a false positive rate versus true positive rate plot . All ROC curves were computed separately per contrast and animal for both heterogeneity and mean population dF/F0 ( Figure 3g ) . For comparison across animals , we averaged the AUC of the four test contrasts per animal , yielding a single AUC value per animal for both heterogeneity and dF/F0 ( Figure 3h ) . To ascertain the performance of a decoder on the same task as we required the mouse to perform , we created an algorithm that calculated the probability of a stimulus being present . This decoder was based on a previously published maximum-likelihood-naive Bayes decoding algorithm ( for a more complete description , see Montijn et al . , 2014 ) . For each neuron and stimulus orientation , we computed the mean and standard deviation of mean dF/F0 during presentation of a 100% contrast stimulus as well as the mean and standard deviation during 0% probe trials . For each test trial and neuron with the preferred orientation as the trial’s stimulus orientation , we calculated the probability a stimulus was present by reading out the likelihood density function for 0% and 100% contrast trials . The product over neurons in the preferred population for each trial then yields a population posterior probability value for stimulus absence ( 0% likelihood ) and presence ( 100% likelihood ) . The decoder’s read-out was the posterior with the highest probability . Because the likelihood was only based on 0% and 100% contrast responses , automatic cross-validation was ensured for decoding test contrast stimuli . After decoding stimulus presence for all trials , we split the trials into hits and misses and calculated the percentage for which the decoder indicated a stimulus was present per response type and contrast , averaging over repetitions and orientations . This yielded two curves per animal ( see Figure 4d ) . We tested for statistically significant differences between response and no-response trials by performing a paired t-test over animals on the intermediate contrasts ( 0 . 5–32% ) . Furthermore , we quantified the similarity of our decoder’s performance to the animal’s performance in the visual stimulus detection task by calculating the similarity per animal of its actual behavioral performance to the decoder’s performance ( Pearson’s correlation over contrasts ) . We compared this value to the similarity obtained with a bootstrapped shuffling procedure ( 1000 iterations ) . Here , we shuffled the animal’s behavioral and decoder performance over contrasts , recalculated the similarity index , and took the mean over all iterations as the resultant shuffled similarity . To test for statistical significance , we performed a paired t-test over animals between the shuffled and real similarities ( Figure 4e ) . Moreover , we investigated the similarity between the animal’s and decoder’s output at a single-trial level with a chi-square analysis . Pooling all trials across animals showed significant correspondence between the decoder and animal’s judgment of stimulus presence; hit trials were more often decoded as ‘stimulus present’ and miss trials more often as ‘stimulus absent’ ( Figure 4—figure supplement 2j ) . Note that this decoding procedure is not optimal; the absolute decoding performance therefore should not be interpreted as reflecting the actual amount of information present in the neural responses . The purpose of this decoder is merely to test—in coarse terms—the similarity between the neural signal and the animal’s behavior . We analyzed the predictability of behavioral responses before they occurred based on either the mean population dF/F0 response or population heterogeneity between 3 and 0 s before stimulus onset ( Figure 5e ) . Hit trials were split into the 50% fastest and 50% slowest reaction times per contrast per animal and then averaged over contrasts , yielding 6 data points per animal: the mean pre-stimulus population dF/F0 and mean population heterogeneity preceding fast , slow . and miss trials . We then quantified the consistency of differences over animals by calculating the distance of these points per animal to the mean of their own response group and the other two . We defined the predictability metric per point i ( animal ) for two response types r1 and r2 ( i . e . two types out of fast , slow , or miss ) as ( 10 ) δr1 , r2 , i= ( ( ||d ( ir1 , μr2 ) ||/ ( ||d ( i r1 , μr2 ) ||+ ||d ( i r1 , μr1¬i ) || ) ) - 0 . 5 ) · 2 , where ‖d‖ is the absolute Euclidian distance ( vector magnitude ) , µr is the mean location of lr– where lr is the group of points for response r – and µr¬i indicates the mean location of lr without point i . This analysis yields a vector δr1 , r2; the separability between response type r1 and r2 . Random placement would lead to a separability of δ = 0 , so we quantified statistically significant predictability of responses by performing FDR-corrected one-sampled t-tests ( vs . 0 ) for each separability vector and both neuronal population metrics ( heterogeneity and mean dF/F0 ) . We also tested whether the separability was higher for heterogeneity or dF/F0 by performing FDR-corrected paired t-tests between dF/F0 and heterogeneity separability vectors for the same response type comparisons ( Figure 5e ) . We defined the rise time to maximum stimulus-driven heterogeneity as the time it took the population heterogeneity to rise from 10% to 90% of the difference between pre-stimulus baseline levels and maximum heterogeneity during the stimulus period . This rise time was calculated on the mean curves per animal and contrast as shown in Figure 5d . To create the graph shown in Figure 5f , we took the rise time across test contrasts per animal ( n=8 ) and behavioral response type ( miss , slow , fast ) . We tested for significant differences in average rise times between response types with paired t-tests across animals . Detection of a visual stimulus might be associated with consistencies in population activity . We , therefore , analyzed whether the inter-trial correlation of population activity varies depending on the behavioral performance of the animal . We again separated fast , slow , and miss trials , and for each stimulus orientation calculated the correlation of the dF/F0 response vector between pairs of trials with the same type of behavioral response ( Figure 6a ) . We separated the neuronal responses for that orientation’s preferred and nonpreferred population of neurons , also to address whether consistency across trials might be restricted to the preferred population or would also occur in the nonpreferred population ( Figure 6d–d ) . Note that because we calculated the correlations separately for preferred and nonpreferred populations , the relative contribution of the orientation signal is fairly low , which explains the relatively low correlation values . To assess above-chance similarities , we compared these values to correlations obtained from shuffled data . By shuffling within each stimulus orientation all trial identities randomly for each neuron , the orientation signal is preserved , but other similarities across trials are destroyed . We repeated this shuffling procedure 100 iterations and took the mean of these 100 iterations as shuffled correlation value per animal ( Figure 6b–d ) . To test for statistically significant consistencies in population activation patterns , we performed FDR-corrected paired t-tests between the real and shuffled correlation values over animals for the different response types and the two neuronal population types . We also quantified the differences between response groups in the real data with paired t-tests ( miss vs . slow , miss vs . fast , and fast vs . slow ) . To study the theoretical implications of our results relating to heterogeneity , we proceeded with an analysis of the question whether heterogeneity forms a special case of population codes that do not merely reflect an increased activity of all neurons upon visual detection . For the specific purpose of these analyses ( shown in Figure 7 ) , we use as definition for multidimensional heterogeneity the distance in neural space from the population’s activity to the closest point on the main diagonal ( see text and below for further explanation ) . Although this definition is computationally different from our pairwise definition of heterogeneity , it also captures the overall dissimilarity of responses within a population of neurons . Moreover , applying this procedure to z-scored dF/F0 values yields Pearson's correlations of r > 0 . 9 when compared with our original definition of heterogeneity ( Equations 3 and 4 ) and gives very similar hit/miss Cohen’s d values ( Figure 3—figure supplement 2 ) . The two metrics , therefore , likely capture the same neural phenomenon and show that heterogeneity can be studied by different , but related computational definitions . To assess the distribution of neuronal population activity in multidimensional neural response space ( where each dimension represents the activity of a single neuron; see Figure 7a–d ) , we calculated the inter-point distance ( each point representing the population activity during a single trial ) between all hit trial pairs and between all miss trial pairs . The distance in neuronal activity for a population of n neurons between a pair of trials x and y in multidimensional space can be calculated as the n-dimensional Euclidian: ( 11 ) d ( x , y ) = ( x1 − y1 ) 2+ ( x2 − y2 ) 2+…+ ( xn − yn ) 2 The pairwise inter-point distance is then given in units of neural activity ( dF/F0 , Figure 7e ) . Note that this formula can also be used to calculate the multidimensional heterogeneity , as defined above , by taking the distance between any trial ( x ) and the closest point on the diagonal ( y ) . Next , we investigated the symmetry of population responses around the main diagonal as this symmetry gives an indication of whether heterogeneity is an epiphenomenal observation or a fundamental neural characteristic underlying visual detection ( see text ) . In order to do so , we mirrored each point across the diagonal and recalculated the inter-point distances for the mirrored data . Mirroring across the diagonal was achieved by direct inversion of the signs per neuron relative to the main diagonal . For a population response r = [r1 r2 … ri … rn] , where n is the number of neurons , the mirrored version r’ = [r’1 r’2 … r’i … r’n] was calculated as follows: ( 12 ) r'= ( μr- ( r-μr ) ) where µr is the mean population response over r . For the analyses displayed in Figure 7g , h , we removed the mean and/or heterogeneity from the population responses and assessed the effect on decoding accuracy of hit/miss responses during test contrast stimuli . As mentioned before , for these analyses heterogeneity was defined as the distance to the main diagonal . As such , removal of the mean without influencing heterogeneity is trivial and can be achieved by simply subtracting the mean population response from all neuronal dF/F0 values obtained for each trial . Briefly , heterogeneity was removed from each trial without affecting the mean in two steps; first heterogeneity was removed , and next any influence on the mean was remedied by adding the difference between the new and old mean . First , heterogeneity was removed by dividing each neuron’s response during that trial by the square root of the sum of the squared differences between the neuronal responses and the mean ( i . e . by dividing by the heterogeneity ) : ( 13 ) r'¬H=r ( r1-μr ) 2+ ( r2-μr ) 2+ … + ( rn-μr ) 2 Next , changes in the mean were corrected by removing the new mean of the heterogeneity-removed population activation ( μr'¬H ) and adding the old population mean µr: ( 14 ) r¬H= r'¬H+ μr-μr'¬H This way , the heterogeneity ( i . e . the Euclidian distance of that trial’s population activity to the main diagonal ) is normalized to 1 . 0 for all trials . The multidimensional location relative to the diagonal is preserved , but its distance is always the same; all trials now fall on a cylinder with a radius of 1 . 0 dF/F0 around the main diagonal . In other words , the population activation during a trial is projected as a vector from the closest point on the diagonal to the trial’s position , and the vector’s angle is preserved , but its magnitude is normalized to 1 . 0 . Both properties ( mean and heterogeneity ) can be removed by subtracting the mean from the heterogeneity-removed responses . Removing the mean as well as the heterogeneity collapses this cylinder onto a circle through multidimensional space around the origin . To control for potential locomotor confounds , we split all data sets into trials where the mouse was still ( 90 . 9% ± 3 . 6% of trials ) and where it was moving during stimulus presentation ( 8 . 1% ± 3 . 6% of trials ) , and reanalyzed our data . Our results with exclusion of running trials ( Figure 2—figure supplement 1g , h ) are very similar to our original analysis ( Figure 2a , b ) , showing that the effects we observed cannot have been due to running-induced modulations ( paired t-test , hit vs . miss , 0 . 5–32% , p<0 . 05 ) . Another potential confound for our results could be that response trials induce signals related to motor feedback or motivation to initiate motor actions because the animal initiates licking as a behavioral response . This also seems unlikely; because 0% contrast probe trials did not induce neuronal activity during false alarms ( Figure 2—figure supplement 1a , green line ) . Theoretically , however , such signals could still be present and influence population activity only when occurring concurrently with visual stimulation . To control for this , we re-performed our analyses shown in Figure 2a , b , but now used data only from the first 0 . 4 s after stimulus onset; approximately 0 . 8 s before the mean reaction time . Leaving a window of 0 . 8 s between the latest frame included in the data analysis and the licking response should also eliminate potential modulatory activity from motor cortex related to the preparation of licking . The results from this control analysis were slightly noisier due to the shorter data acquisition duration per trial , but showed no qualitative differences to the original analysis regarding heterogeneity ( Figure 2—figure supplement 1i , j ) . The intermediate contrasts still showed significant enhancements in heterogeneity ( p<0 . 01 ) during hit trials , but we found no significant differences for mean population dF/F0 ( p=0 . 543 ) . We , therefore , conclude that our results regarding heterogeneity are not confounded by motor-related modulations due to running or licking , nor by reward-expectation prior to licking responses , and confirm that the mean population dF/F0 is not or less useful as a measure of neural correlates of perception . To control for possible effects of blinking and saccades , we performed pupil detection on our eye-tracking data and removed all trials in which the animals blinked or made saccades during any time of the stimulus presentation [10 . 2% ± 4 . 6% of trials removed ( mean ± SD ) ] . We re-performed our analyses on only the trials where no contamination by incorrect eye position and/or closing of the eyelids was possible ( Figure 2—figure supplement 1k , l ) and observed that our results regarding heterogeneity were qualitatively and quantitatively similar to our original analyses , but that the dF/F0 results were again more sensitive to a conservative analysis ( hit/miss difference for intermediate contrasts , paired t-test , n=8; dF/F0 , p=0 . 136; heterogeneity , p<0 . 005 ) . We conclude that our main results are not biased by incorrect eye position and blinking . We addressed whether the orientation information contained in the population responses was dependent on the mean dF/F0 and heterogeneity during stimulus presentation . We decoded the presented stimulus orientation for each contrast separately ( i . e . 100% contrast trials based on likelihood from 100% contrast trials , etc . ) by a leave-one-out cross-validation and afterwards split all trials into correctly and incorrectly decoded ones ( Figure 4—figure supplement 2b ) . To quantify the dependence of decoding accuracy on dF/F0 during stimulus presentation , we took for each contrast the trials with highest and lowest 50% of dF/F0 and calculated the mean decoding accuracy for both groups ( high and low activity ) . Next , we took the mean for these groups over contrasts per animal and calculated a percentage decoding accuracy increase for the highest versus lowest 50% dF/F0 trials ( see Figure 4—figure supplement 2c ) . To test for statistical significance , we performed a one-sample t-test of the percentage increase values over animals . For heterogeneity , we performed the same steps and performed a t-test versus 0% increase ( Figure 4—figure supplement 2c ) . To address whether visual stimulus features ( i . e . orientation and contrast ) were more accurately represented by neuronal population activity during correct versus incorrect behavioral performance , we used a Bayesian maximum-likelihood decoder as previously described to extract those features from the population activity ( for a more complete description , see Montijn et al . , 2014 ) . We defined all combinations of orientations and contrasts as different stimulus types , yielding a total of 21 different stimulus types ( four orientations times five contrasts plus probe trials ) . Next , we performed a leave-one-out cross-validated decoding procedure for all trials and calculated the mean percentage correct decoding trials for hits and misses per stimulus type; then we averaged the percentage correct over stimulus types , yielding an accuracy per animal for hit and miss trials . We tested for a statistical difference between hits and misses with a paired t-test over animals ( Figure 4—figure supplement 2a ) . To investigate detection-related increases or decreases in noise correlations ( Figure 4—figure supplement 2f , g ) , we first calculated a response vector for each stimulus orientation θ that was presented during a test contrast trial . Here , each element in the vector is the neuron’s response to a single presentation t ( i . e . a trial ) of that stimulus orientation: ( 15 ) Rθ=[Rθt…Rθn] where n is the number of repetitions per response type per orientation . Because we aim to compare a single noise correlation value per neuronal pair i , j , we took the mean noise correlation over all four stimulus orientations: ( 16 ) ρi , jnoise=∑θ=0135corr ( Ri , θ , Rj , θ ) 4 The noise correlation is , therefore , an index of the mean trial-by-trial variability shared by pairs of neurons over all stimulus orientations . To verify that the behavioral predictability before stimulus onset that we found ( Figure 5e ) was not merely a group-level effect , but was indeed also a single-trial phenomenon , we subsequently performed single-trial decoder-based predictions of fast/slow/miss behavioral responses that occurred during the subsequent stimulus presentation ( see Figure 5—figure supplement 1 ) . We used a similar leave-one-out cross-validated naive Bayes decoder as described above for fast , slow , and miss trials , and calculated per trial the relative likelihood that the subsequent stimulus presentation would lead to a miss , fast , or slow response . We then split the predictive decoding results per actual behavioral response group and averaged the relative prediction likelihood per animal . This yields three relative probability values per actual response type per animal . Assigning an angle to each of these behavioral responses that are separated by 2/3π on the unit circle and taking the relative likelihood as the vector magnitude , it is then possible to calculate a resultant prediction vector per actual response type per animal . To quantify statistical significance , we multiplied an angle-based correctness index ( +1 when the resultant prediction vector angle is perfectly aligned to the actual response angle and –1 when they are separated by 1π ) with the vector magnitude , giving a normalized decoding accuracy index between –1 . 0 and +1 . 0 , where chance level is 0 . Lastly , we performed one-sample t-tests on the normalized decoding accuracy indices over animals and response types for heterogeneity and dF/F0 , and a paired t-test between dF/F0 and heterogeneity ( Figure 5—figure supplement 1 ) . | Seeing is not the same as perceiving , where an object is recognized and information about it is interpreted by the brain . Things might be in your field of view , but not actively perceived; for example , when daydreaming with your eyes open . Many researchers have investigated how the brain responds differently to a perceived object compared with something that is seen but not perceived . However , using relatively coarse techniques , only small differences in brain activity have been found . Many of the techniques used to investigate brain activity only look at the average activity of a group of neurons – the cells in the brain that process information . This raises the possibility that the perception of an object relies on more subtle or complex interactions in brain activity . To investigate this , Montijn et al . trained mice to lick a reward spout that gave out sugar water when they perceived a particular image . A technique called two-photon calcium imaging was then used to simultaneously record the activity of tens to hundreds of neurons in part of the brain called the visual cortex as the mice performed the perception task . This revealed that the average activation of a group of neurons was only weakly related to whether a mouse had perceived the image . However , differences in the strength of the responses of the individual neurons in the group reflected perception more strongly: when a mouse perceived the image and licked in response , a heterogeneous ( non-uniform ) set of neuronal responses occurred . The diversity of the neuronal responses could also be used to predict how quickly a mouse would respond to an image . These activity differences would not be picked up by techniques that detect the average activity of many neurons , explaining why these effects had not previously been seen . These findings shed light on which patterns of activity in the visual region of the brain lead to objects being perceived or not . Whether similar mechanisms operate in different regions of the brain remains to be investigated . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Mouse V1 population correlates of visual detection rely on heterogeneity within neuronal response patterns |
A key step in the de novo formation of the embryonic vasculature is the migration of endothelial precursors , the angioblasts , to the position of the future vessels . To form the first axial vessels , angioblasts migrate towards the midline and coalesce underneath the notochord . Vascular endothelial growth factor has been proposed to serve as a chemoattractant for the angioblasts and to regulate this medial migration . Here we challenge this model and instead demonstrate that angioblasts rely on their intrinsic expression of Apelin receptors ( Aplr , APJ ) for their migration to the midline . We further show that during this angioblast migration Apelin receptor signaling is mainly triggered by the recently discovered ligand Elabela ( Ela ) . As neither of the ligands Ela or Apelin ( Apln ) nor their receptors have previously been implicated in regulating angioblast migration , we hereby provide a novel mechanism for regulating vasculogenesis , with direct relevance to physiological and pathological angiogenesis .
In the vertebrate embryo , the formation of the large axial vessels , namely the dorsal aorta ( DA ) and the cardinal vein ( CV ) , establishes a first circulatory loop and thereby the core of the developing cardiovascular system . Angioblasts are initially specified in the lateral plate mesoderm and migrate between the somites towards the midline , where they coalesce and assemble the DA and the CV underneath the notochord ( NC ) ( Figure 1A , B–F , Video 1 ) . 10 . 7554/eLife . 06726 . 003Figure 1Angioblast migration to the midline is not regulated by Vegf . ( A–F ) Schematic ( A ) and confocal ( B–F ) in vivo time-lapse imaging of angioblasts ( green ) and their migration in a Tg ( fli1a:EGFP ) y1 embryo , injected with H2B-mCherry mRNA ( purple , all nuclei ) , dorsal views at indicated time points . Arrows indicate the initial migration to the midline , and the coalescing into the dorsal aorta ( DA ) . ( G–J ) Loss of Vegf-signaling components by vegfaa/vegfab double morpholino ( MO ) injection ( H ) or in vegf receptor 2 ( kdrl ) mutants ( J ) does not affect angioblast migration . Confocal projections of Tg ( fli1a:EGFP ) y1 embryos in dorsal views at 17 hpf . For each condition , n > 15 . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 00310 . 7554/eLife . 06726 . 004Figure 1—source movie 1 . Time-lapse movie showing angioblast ( green ) migration to the midline analyzed using Tg ( fli1a:EGFP ) y1 embryos injected with H2B-mCherry mRNA ( purple , all nuclei ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 00410 . 7554/eLife . 06726 . 005Figure 1—figure supplement 1 . Inhibition of Vegfa signaling resulted in normal angioblast migration to the midline , but impaired sprouting angiogenesis . ( A–J ) Vegfa signaling was impaired using either ligand MOs ( vegfaa/vegfab MO ) , chemical inhibition of Vegfr2 signaling ( 2 . 5 μM SLKB1002 [Zhang et al . , 2011] ) or mutations in the vegfr2 gene ( kdrl ) . Confocal projections of Tg ( fli1a:EGFP ) y1 embryos in dorsal views at 17 hpf ( A , B , C , G , H , I ) , 24 hpf ( F , J , K , L ) and 28 hpf ( D , E ) . At 17 hpf , angioblasts in control embryos as well as Vegfa signaling deficient embryos arrived at the midline ( A , B , C , G , H , I ) . Control embryos showed sprouting of the intersegmental vessels ( D , F , K ) whereas Vegfa signaling deficient embryos lacked the intersegmental vessels ( E , J , I ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 00510 . 7554/eLife . 06726 . 006Figure 1—figure supplement 2 . Abrogation of Shh signaling abolishes vegfaa expression , but does not influence angioblast migration . ( A , B ) Embryos treated with 100 μM Cyclopamine ( CyA ) from 11 hpf to 18 hpf to block Sonic hedgehog signaling showed no defects in medial angioblast migration . Confocal projections of Tg ( fli1a:EGFP ) y1 embryos in dorsal views at 18 hpf . c-d ) vegfaa expression in the somites is abolished in embryos treated with 100 μM Cyclopamine ( CyA ) from 11 hpf to 18 hpf . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 00610 . 7554/eLife . 06726 . 007Video 1 . Angioblast migration between 12 . 5 hpf and 16 hpf observed by time-lapse imaging using Tg ( fli1a:EGFP ) y1 to visualize angioblasts . Angioblasts migrate to the midline in Wt embryos . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 007 It has been previously proposed that this process is regulated by Vascular endothelial growth factor A ( VEGF-A ) ( Coultas et al . , 2005; Verma et al . , 2010; Gore et al . , 2012 ) , a master regulator of vascular growth in the embryonic and adult organism . In Xenopus embryos , transcripts for the vegfa gene are expressed in the midline and migrating angioblasts were guided towards sites of ectopic Vegfa expression ( Cleaver and Krieg , 1998 ) . Global inactivation of the genes encoding VEGF-A or the corresponding receptor VEGFR2 in mice led to strong vascular defects affecting the whole vascular network including the DA ( Shalaby et al . , 1995; Ferrara et al . , 1996 ) . Based on these data it was concluded that in vertebrates VEGF-A signaling regulates angioblast migration to the midline ( Coultas et al . , 2005; Verma et al . , 2010; Gore et al . , 2012 ) . To get direct , mechanistic and dynamic insight into angioblast migration and formation of the great vessels , we decided to investigate these processes in zebrafish embryos . As previously published ( Nicoli et al . , 2008 ) , expression analysis of vegfaa , the gene for the main VEGF-A ortholog in zebrafish , showed the presence of transcripts in the somites between 12 and 15 hr post fertilization ( hpf ) , that is the stage when angioblasts migrate to the midline ( see Figure 1B–F ) . To analyze whether Vegfa signaling , as previously proposed , is required for angioblast migration to the midline , we analyzed loss-of-function zebrafish embryos with deficiencies in the Vegfa signaling pathway resulting from genomic mutations as well as morpholino ( MO ) -mediated knockdown . To rule out developmental delays , we counted the somites of the embryos and fixed them for confocal analysis , when they had developed 16–17 somites ( equivalent to 17 hpf ) . Surprisingly , neither Vegfa ligand depletion ( vegfaa/vegfab MO ) nor a null mutation in the gene encoding the VEGFR2 ortholog ( kdrl ) interfered with angioblast migration in zebrafish embryos ( Figure 1H , J ) , although Vegfa signaling dependent phenotypes , like failure to form the intersegmental vessels , could be observed after 24 hpf ( Figure 1—figure supplement 1 ) . Additionally , pharmacological inhibition of Vegfr2 signaling from 6 to 17 hpf did not impair angioblast migration to the midline ( Figure 1—figure supplement 1 ) . As activity of Sonic Hedgehog ( Shh ) in the midline has been shown to induce vegfaa expression ( Lawson et al . , 2002 ) , we repeated the published experiments and blocked this process by chemical inhibition of Shh signaling using cyclopamine . In line with our previous experiments , angioblast migration was not affected after Shh inhibition ( Figure 1—figure supplement 2 ) . The sum of these data strongly indicates that Vegfa signaling is dispensable for angioblast specification or migration in zebrafish . In order to identify the endogenous signal ( s ) guiding angioblast migration , we further analyzed the expression pattern of factors involved in cell motility and directed cell migration . As previously published ( Zeng et al . , 2007 ) , we detected the expression of Apelin , a short secreted peptide encoded by the apln gene , by the NC ( Figure 2—figure supplement 1C ) . At the same time , transcripts of its two paralogous receptors apelin receptor a ( aplnra ) and apelin receptor b ( aplnrb ) were present in angioblasts ( Figure 2—figure supplement 1A , B; [Scott et al . , 2007; Zeng et al . , 2007] ) . Using MO-mediated gene knockdown , we observed that both aplnra and aplnrb are important for midline migration of angioblasts . While absence of each individual receptor reduced the efficiency of migration towards the midline , double depletion completely abolished this process ( Figure 2A , Video 2 ) . We used TALEN or CRISPR/Cas9-mediated gene editing ( Hwang et al . , 2013; Jao et al . , 2013 ) to introduce mutations in either the aplnra or the aplnrb gene ( Figure 2—figure supplement 2 ) . As our MO-based analysis already indicated an additive effect of both receptor genes , we analyzed the phenotypes of the offspring of aplnra+/−;aplnrb+/− double heterozygous parents . The embryos were all individually scored for their phenotype regarding angioblast midline migration into four categories ( normal , mild , strong , stronger; Figure 2B ) and then subjected to genotyping . Homozygous mutant embryos phenocopied the MO-induced phenotypes , additional loss of one or two copies of the second receptor increased the severity of the phenotype ( Figure 2C ) . Additionally , we detected aplnrb transcripts together with the angioblast specification marker etv2 indicating that both genes are expressed very early during angioblast specification . However , when we analyzed aplnra/aplnrb double deficient embryos , no difference in etv2 expression was detectable by in situ hybridization indicating that Apln receptors only regulate angioblast migration , but not specification ( Figure 2—figure supplement 1E ) . 10 . 7554/eLife . 06726 . 008Figure 2 . Ela/Apelin receptor - signaling guides angioblast migration to the midline . ( A ) Angioblasts have migrated to the midline at 17 hpf in wild type , aplnmu267 mutant or apln MO injected zebrafish embryos . MO-mediated knockdown of apelin receptor a or b partially inhibited midline migration , while simultaneous loss of both apelin receptor genes completely abolished midline migration of angioblasts . Likewise , embryos with homozygous mutations in the ligand ela display impaired migration of angioblasts . Arrows indicate aberrant positions . ( B , C ) Mutations in apln receptor genes impair angioblast migration . Analysis of the offspring of aplnra+/−;aplnrb+/− double heterozygous parents resulted in four phenotypic categories ( normal , mild , strong , stronger ) . ( B ) Genotyping of individual embryos revealed an additive effect , while mild phenotypes were observed when one receptor gene was homozygously mutant , phenotypic strength increased with additional loss of functional receptor genes ( c; Ra , aplnra; Rb , aplnrb ) . ( D , E ) Ela deficiency can partially be compensated by Apln . Analysis of the offspring of apln+/−;ela+/− double heterozygous parents resulted in four phenotypic categories ( normal , mild , strong , severe ) . ( D ) Genotyping of individual embryos revealed a dose dependency , with increasing phenotypic strength correlating with additional loss of apln alleles in ela mutant embryos . ( E ) ela−/−; apln−/− double mutants phenocopied apln receptor deficiency . Angioblasts ( green ) were labeled by Tg ( fli1a:EGFP ) y1 expression , scale bars represent 30 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 00810 . 7554/eLife . 06726 . 009Figure 2—source data 1 . Excel table showing the phenotype categories and the number of embryos for each genotype in these phenotypic categories ( for the apln/ela double mutant analysis ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 00910 . 7554/eLife . 06726 . 010Figure 2—figure supplement 1 . Expression of aplnra , aplnrb , apln and ela coincides with the formation of the DA and PCV . ( A–D ) In situ hybridizations detecting expression of aplnra ( A ) , aplnrb ( B ) , apln ( C ) and ela ( D ) at the indicated time points ( 12 hpf , 18 hpf shown in dorsal views , 20 hpf in lateral views ) . aplnrs are expressed in the migrating angioblasts and continue to be expressed during formation of the DA and PCV . At 12 hpf , aplnrb is strongly expressed in the angioblasts ( prior to migration , arrowheads ) and within the lateral plate mesoderm and at the notochord somite boundary . aplnra expression is strong in the paraxial mesoderm ( A ) . Both receptors are expressed in the angioblasts at 18 hpf ( A , B , arrowheads ) and in the developing axial vessels at 20 hpf ( Arrowheads point to the DA , arrows to the PCV ) . Prior to angioblast migration apln and ela are both expressed by the notochord , with ela showing a much stronger expression ( C , D ) . After angioblasts reached the midline ( 18 hpf , 20 hpf ) only apln is still expressed by the notochord , whereas notochord derived ela expression is strongly reduced to absent . ( E ) Knockdown of aplnr does not influence the specification of angioblasts . In situ hybridizations detecting expression of the angioblast marker etv2 at 13 . 5 hpf ( dorsal views ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 01010 . 7554/eLife . 06726 . 011Figure 2—figure supplement 2 . Generation of zebrafish aplnra , aplnrb and apln mutants . Targeted indel mutations in the aplnra gene were induced using TALEN mutagenesis , and in the aplnrb and apln genes were induced by engineered sgRNA:Cas9 . The wild type nucleotide sequences are shown at the top with the TALEN ( A ) or CRISPR ( B , C ) target sites highlighted in yellow and the spacer ( A ) or PAM ( B , C ) sequences highlighted in red . ( A ) aplnra coding sequence and TALEN mutagenesis target site . An 8 base pair insertion , shown as lower case letters in blue leads to a premature stop and a shortening of the Aplnra protein to 13 AA . ( B ) aplnrb coding sequence and CRISPR mutagenesis target site . A 4 base pair insertion , shown as lower case letters in blue leads to a premature stop and a shortening of the Aplnrb protein to 65 AA . ( C ) apln coding sequence and CRISPR mutagenesis target site . An 1 baise pair loss / 9 base pair insertion , shown as lower case letters in blue causes a frameshift leading to a premature stop codon and a shortening of the Apln protein to 11 AA . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 01110 . 7554/eLife . 06726 . 012Video 2 . Angioblast migration between 12 . 5 hpf and 16 hpf observed by time-lapse imaging using Tg ( fli1a:EGFP ) y1 to visualize angioblasts . No migration to the midline , but minor movements and strong filopodia formation can be visualized in Aplnra/b double deficient embryos ( aplnra/b MO ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 012 Next , we analyzed the requirement for the ligand Apln by MO-mediated gene knockdown and , unexpectedly , observed no difference to control MO injected embryos ( Figure 2A ) . To obtain undisputable genetic confirmation of this result , we again used gene editing to introduce mutations in the apln gene ( Figure 2—figure supplement 2 , most likely resulting in functional null mutants ) . Consistent with the MO-mediated depletion , homozygous aplnmu267 mutant embryos ( Figure 2A ) showed no vasculogenesis defects , which indicated that endogenous Apln is not sufficient to regulate angioblast migration to the midline . Previous studies have identified a requirement for Apln receptors in myocardial development in zebrafish ( Scott et al . , 2007; Zeng et al . , 2007 ) and mice ( Charo et al . , 2009 ) , which could not completely be phenocopied by Apln deficiency ( Scott et al . , 2007; Zeng et al . , 2007; Charo et al . , 2009 ) . These results led to the proposal that Apln may not be the only ligand for Aplnrs . Recently , a novel peptide hormone named Elabela ( Ela , also known as Apela or Toddler ) was identified in zebrafish and shown to bind and activate Aplnrs ( Chng et al . , 2013; Pauli et al . , 2014 ) . By in situ hybridization we detected ela expression by the NC ( Figure 2—figure supplement 1D , [Pauli et al . , 2014] ) , consistent with a possible role in attracting angioblasts to the midline . We next analyzed angioblast migration in zebrafish embryos carrying a homozygous null mutation in the ela gene ( elabr13 ) . Ela deficiency led to a strong impairment of angioblast migration ( Figure 2A ) , indicating that indeed Ela-Aplnr represent a novel bona fide ligand–receptor pair and therefore a novel signaling pathway regulating angioblast migration . Given that the ela-/- phenotype was not as severe as the depletion of both receptors and that apln expression increases later in the NC ( while ela expression becomes progressively reduced , Figure 2—figure supplement 1C , D ) , we analyzed the phenotypes of the offspring of apln+/−;ela+/− double heterozygous parents . The embryos were all individually scored for their phenotype regarding angioblast midline migration into four categories ( normal , mild , strong , severe ) and then subjected to genotyping . Interestingly a mild phenotype was observed , when either one ela or both apln alleles were mutant , a strong phenotype when ela was homozygously mutant with increasing severity ( severe phenotype ) upon loss of the apln alleles ( Figure 2D , E ) . Homozygous ela−/−; apln−/− double mutants phenocopied apln receptor deficiency ( compare in Figure 2A , B ) . The observed dose dependency is indicative , that in this context Ela and Apln might act as novel chemoattractants for angioblast migration to the midline , with Ela acting as the major endogenous attractant and Apln having a minor additive role . Previous analysis of zebrafish embryos lacking the NC showed , that signals from the midline guide angioblast migration ( Fouquet et al . , 1997; Sumoy et al . , 1997 ) . We could show that indeed angioblast migration as well as ela expression is perturbed in the zebrafish mutants lacking a NC ( notochord homeobox , noto; previous name: floating head ) or exhibiting defective NC precursor differentiation ( T , brachyury homolog a , ta; previous name: no tail ) ( Figure 3 ) ( Halpern et al . , 1993; Odenthal et al . , 1996 ) . While angioblasts in nototk241 mutant embryos failed to migrate to the midline , tab160 angioblasts migrated towards the midline but failed to reach this structure ( Figure 3A ) . In line with these results , ela expression was strongly diminished in the tab160 midline and almost absent in nototk241 mutant embryos ( Figure 3B ) . We next wanted to test whether midline ela expression would be sufficient to restore angioblast migration in tab160 mutants . For this purpose , plasmid DNA with a shh promoter fragment enabling ela expression in the floorplate was injected together with Tol2 transposase mRNA to obtain mosaic Ela overexpression in the midline . Indeed , Ela overexpression did restore angioblast migration to the midline in tab160 embryos ( Figure 3C ) , demonstrating that midline Ela expression is not only necessary , but sufficient to regulate angioblast migration during vasculogenesis . 10 . 7554/eLife . 06726 . 013Figure 3 . Notochord ( NC ) ela expression is sufficient to guide angioblasts to the midline . ( A ) Angioblasts fail to migrate to the midline position in the NC mutants noto and ta . Confocal projections of Tg ( fli1a:EGFP ) y1 embryos in dorsal view at 17 hpf ( scale bars: 20 μm ) . ( B ) In situ hybridization in 17 hpf old zebrafish embryos showing ela expression from the NC and somites ( white arrows ) . NC deficiency abolishes or strongly reduces ela expression in noto or ta mutant embryos . ( C ) Mosaic expression of Ela in the midline , achieved by injection of shh:ela DNA , led to rescue of angioblast migration in tab160 mutant embryos ( n = 61 embryos , ctr injected n = 32 embryos ) . Significance was calculated by chi-square ( χ2 = 19 . 65 ) equivalent to p < 0 . 001 ( p = 9E-06 ) . Scale bars: 30 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 013 Next , we wanted to analyze the function of Ela-Aplnr signaling more mechanistically . Recently , it was published that Ela does not act as a chemoattractant , but instead acts as a motogen in regulating mesoderm migration during gastrulation ( Pauli et al . , 2014 ) . In contrast , the aplnr deficient ( aplra/b MO injected ) angioblasts were still partially motile and did send out filopodia , but failed to migrate to the midline ( Video 2 ) . To test whether Ela does act as a chemoattractant or just enables motility of angioblasts , we overexpressed Ela at ectopic locations in noto or ta mutant embryos , as both of these lack endogenous ela and apln expression in the NC . We used a heatshock promoter to control timing of expression and labeled Ela overexpressing cells by coexpression of RFP through an internal ribosome entry site ( IRES ) element ( Figure 4 ) . We measured the distance of control ( Cherry ) or Ela/RFP expressing cells from angioblasts in a 40 μm radius of the red labeled cell and considered an angioblast as attracted , if it was less than 5 μm away from it . While only 20% of the angioblast were found in the vicinity of control cells , more than 52% of the angioblasts were close to Ela overexpressing cells ( Figure 4D , G ) . Together with the observed dose dependency of Ela-Apln , the motility of Aplnr deficient embryos and the rescue of NC mutants by midline Ela expression , our data support the interpretation , that Ela acts as a chemoattractant for angioblasts during midline migration . 10 . 7554/eLife . 06726 . 014Figure 4 . Ela overexpressing cells attract angioblasts . ( A ) F1 embryos from nototk241+/− or tab160+/− parents were injected with 250 pg Tol2 mRNA as well as 10 pg of DNA constructs , in which the heatshock promotor was used to drive either control ( ctr , Cherry ) or ela expression . Individual Ela overexpressing cells were labeled by IRES mediated RFP expression . Expression was induced by two consecutive heatshocks ( incubation at 39°C ) at 12 hpf and 14 hpf for 1 hr each . ( B–C′ ) Angioblast migration in nototk241−/− mutant embryos injected with the ctr ( B , B′ ) or the ela overexpression ( C , C′ ) construct . ( D ) Quantification of Ela/RFP or ctr ( Cherry ) positive cells showed significantly more Ela overexpressing cells attracting angioblasts than ctr cells ( n = 10 embryos for hs:ela , n = 20 embryos for hs:ctr ; p*** = 0 . 0001 ) . ( E–F′ ) Angioblast migration in tab160 −/− mutant embryos injected with the ctr ( E , E′ ) or the ela overexpression ( F , F′ ) construct ( G ) Quantification of Ela/RFP or ctr ( Cherry ) positive cells showed significantly more Ela overexpressing cells attracting angioblasts than ctr cells ( n = 8 embryos for hs:ela , n = 10 embryos for hs:ctr ; p** = 0 . 0089 ) . Significance was calculated by chi-square test . Maximum intensity projections give an overview of the analyzed embryos . Single planes visualize the closeness of Ela or ctr ( Cherry ) expressing cells to angioblasts . White arrows point to Ela overexpressing cells with less than 5 μm distance to angioblasts . <5 μm distance between an Ela/ctr positive cell and an angioblast was counted as ‘attractant’; 5–40 μm distance was counted as ‘not attractant’ . Scale bars represent 30 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 014 To prove that Ela-Aplnr signaling is directly required in angioblasts and rule out that the observed migration defects were caused by changes in other cell types , we performed transplantation experiments . Mini-ruby injected ( red fluorescence ) transgenic fli1a:EGFP positive ( angioblasts , green ) cells from control MO injected , aplnra/b MO injected or wild type ( WT ) donor embryos were transplanted into either WT or aplnr loss-of-function ( apnlra/b MO injected ) host embryos ( Figure 5A ) . Analysis of angioblast migration to the midline at 17 hpf revealed that in average 93 . 6% of control MO-injected transplanted angioblasts migrated to the midline in a WT host embryo ( Figure 5B , C; n = 39 GFP+ angioblasts , 7 embryos ) . In contrast , only 9 . 4% of aplnra/b MO injected angioblasts arrived at the midline , whereas the majority of these cells remained in more lateral positions ( Figure 5B , D; n = 62 GFP+ angioblasts , 5 embryos ) . Transplanted WT angioblasts were perfectly capable to migrate to the midline in an aplnr deficient environment ( average of 81 . 9% , Figure 5B , E; n = 15 GFP+ angioblasts , 6 embryos ) , which shows that there is indeed a cell autonomous requirement for Aplnr signaling in angioblasts . 10 . 7554/eLife . 06726 . 015Figure 5 . Cell autonomous requirement for Apelin receptor signaling in angioblasts . ( A ) Experimental design: mini-ruby injected cells from ctr MO ( C ) aplnra/b MO ( D ) or WT ( E ) Tg ( fli1a:EGFP ) y1 embryos were transplanted into WT or aplnra/b MO host embryos and scored for their migration to the midline . ( B ) Quantification: 93 . 6% of ctr MO injected ( n = 39 GFP + angioblasts , 7 embryos ) , but only 9 . 4% of aplnra/b deficient ( aplnra/b MO injected; n = 62 GFP+ angioblasts , 5 embryos ) donor angioblasts migrated to the midline in WT host embryos . In contrast , 81 . 9% of WT donor angioblasts ( n = 15 GFP+ angioblasts , 6 embryos ) migrated to the midline in aplnra/b deficient host embryos . Error bars represent SEM , calculated using the standard deviation of percent of angioblasts in the midline per embryo . Statistical analysis showed significance using 2 way ANOVA and t-test , with p *** = 3 . 44577E-08 for aplnra/b MO in WT , and p = 0 . 232995135 ( not significant , n . s . ) for WT in aplnra/b MO . see also Figure 5—source data 1 . ( C–E ) Confocal projections showing representative embryos of the transplantation experiments at 17 hpf . Arrows indicate transplanted angioblasts; asterisks label transplanted cells , which are not angioblasts; dashed lines indicate the midline . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 01510 . 7554/eLife . 06726 . 016Figure 5—source data 1 . Statistical analysis and single values for the number of angioblasts and embryos analyzed . DOI: http://dx . doi . org/10 . 7554/eLife . 06726 . 016 Based on the sum of the results presented above , we propose that Ela expression at the embryonic midline provides a chemoattractive signal for Aplnr-positive angioblasts . While Ela and Apln both signal to the Apln receptors , the initial ela expression at the midline appears stronger than apln , which peaked at later developmental time points ( Figure 2—figure supplement 1C , D ) . Thus , ela and apln may have partially redundant functions , but are required at different stages of vascular morphogenesis . Surprisingly , while Ela-Aplnr signaling was strictly required for the guided migration of angioblasts , this process was not controlled by Vegfa-mediated signaling . Our studies identified Ela-Aplnr signaling as a novel signaling pathway for angioblasts regulating vascular patterning in the developing vertebrate embryo . As therapeutic intervention with vascular growth is clinically highly relevant and , so far , predominantly focuses on the VEGF-A signaling cascade , the Ela-Aplnr signaling pathway may also represent a novel therapeutic target in human disease .
Zebrafish ( Danio rerio ) were maintained in a recirculating aquaculture system under standard laboratory conditions ( Westerfield , 1993 ) . Embryos were staged by hours post fertilization ( hpf ) at 28 . 5°C ( Kimmel et al . , 1995 ) , for 17 hpf embryos were incubated after gastrulation at 21°C and staged on the following morning by counting somites ( 16 somites equal 17 hpf [Kimmel et al . , 1995] ) . Zebrafish strains used were Tg ( fli1a:EGFP ) y1 ( Lawson and Weinstein , 2002 ) , kdrlhu5088 ( Bussmann et al . , 2010 ) , tab160 ( Halpern et al . , 1993 ) , nototk241 ( Odenthal et al . , 1996 ) and elabr13 ( Chng et al . , 2013 ) . For apln and aplnrb Crispr mediated mutagenesis was used . Nls-zCas9-nls mRNA was synthesized as previously described ( Jao et al . , 2013 ) . The target sequences of apln exon 1 ( 5′-GAATGTGAAGATCTTGACGC-3′ ) and aplnrb exon 1 ( 5′-CTACATGCTCATCTTCATCC-3′ ) were cloned into the gRNA expression vector pDR274 as previously described ( Hwang et al . , 2013 ) . The apln sgRNA or aplnrb sgRNA were transcribed using DraI-digested gRNA expression vector as template and the T7 mMessage mMachine kit ( Ambion , Life Technologies , Germany ) . apln sgRNA or apnrb sgRNA and nls-zCas9-nls-encoding mRNA were co-injected into one-cell stage zebrafish embryos . Each embryo was injected with 2 nl of solution containing 12 . 5 ng/μl sgRNA and 300 ng/μl nls-zCas9-nls mRNA . A HgaI ( New England BioLabs , Frankfurt , Germany ) restriction site was used for genotyping of aplnmu267and a FokI ( New England BioLabs ) restriction site was used for genotyping of aplnrbmu270 . aplnra mutants were generated by TALEN mutagenesis . TALENs were assembled using the Golden Gate method ( Cermak et al . , 2011 ) . For targeting the aplnra locus , a 5′ RVD ( NI HD NI HD HD NH NI NH NI HD NI NG NI HD NH NI NG ) and a 3′ RVD ( HD NI HD NI HD HD HD NI NH NI NH NG HD NI NG NG NI NG NI ) were generated . A BsrI ( New England BioLabs ) restriction site was used for genotyping of aplnramu296 . mRNA was generated using the T3 mMessage mMachine Kit ( Ambion ) and injected using 100 pg of the TALEN mix . Microinjections of mRNA or MOs were performed as previously described ( Nasevicius and Ekker , 2000 ) . mRNA was transcribed using SP6 polymerase ( Sp6 mMessage mMachine kit [Ambion] ) . 100 pg H2B-cherry mRNA ( Santoro et al . , 2007 ) , 2 ng aplnra MO ( Scott et al . , 2007 ) , 0 . 5 ng aplnrb MO ( Zeng et al . , 2007 ) , 2 ng apln MO ( Scott et al . , 2007 ) or 2 ng vegfaa/ 2 . 3 ng vegfab MO ( Ober et al . , 2004 ) were injected . Simultaneous knockdown of aplnra and aplnrb was done by coinjection of 2 ng aplnra MO and 0 . 5 ng aplnrb MO ( referred to as aplnra/b MO ) . 4 ng of standard control MO ( Gene tools , Philomath , Oregon ) were injected as control . Plasmid DNA was injected together with Tol2 mRNA ( Kawakami et al . , 2004 ) . A 2 . 2 kb fragment of the shh promoter ( Gordon et al . , 2013 ) was used to drive either rfp or ela expression ( shh:ctr or shh:ela ) . The heatshock promoter ( hsp70l ) was used to drive either Cherry ( hs:ctr [Hesselson et al . , 2009] ) or ela-IRES-RFP ( hs:ela ) . In situ hybridizations were performed as previously described ( Helker et al . , 2013 ) using the following probes: cdh5 ( Larson et al . , 2004 ) , etv2 ( Helker et al . , 2013 ) , vegfc ( Hogan et al . , 2009 ) , vegfaa ( Lawson et al . , 2002 ) . apln , aplnra and aplnrb probe templates were amplified from cDNA of 19 somite stage old embryos using the following primers: apln-forward 5′-GAAAGGCCCAAGTCACAGAG-3′ and apln-reverse 5′-GAGTTCACTATCTGATGTCAAACCA-3′ , aplnra-forward 5′-GAAAGGCCCAAGTCACAGAG-3′ and aplnra-reverse 5′-GAGTTCACTATCTGATGTCAAACCA-3′ , aplnrb-forward 5′-GAAAGGCCCAAGTCACAGAG-3′ and aplnrb-reverse 5′-GAGTTCACTATCTGATGTCAAACCA-3′ . The T7 promotor was added to the 5′- end of the reverse primer in a second round of amplification ( T7-apln-reverse 5′-GTAATACGACTCACTATAGGGAGTTCACTATCTGATGTCAAACCA-3′ , ( T7-aplnra-reverse 5′-GTAATACGACTCACTATAGGGAGTTCACTATCTGATGTCAAACCA-3′ , ( T7-aplnrb-reverse 5′-GTAATACGACTCACTATAGGGAGTTCACTATCTGATGTCAAACCA-3′ ) . Embryos were manually dechorionated and mounted in 0 . 3% agarose ( with subsequent removal of agarose from the head/tail region ) . Medium and agarose were supplemented with 19 . 2 mg/l Tricaine and 30 mg/l Phenylthiourea . For time-lapse imaging , embryos were kept in a 28 . 5°C heated chamber surrounding the microscope stage . All fluorescent images were acquired using an upright Leica Sp5 DM 6000 or a Zeiss LSM 780 Confocal microscope . Confocal stacks and confocal movies were assembled using Imaris Software ( Bitplane , Switzerland ) . Donor embryos were injected with ctr MO or aplnra/b MO and labeled by the injection of 0 . 1 ng mini-Ruby ( Tetramethylrhodamine , Invitrogen/Life Technologies ) . At sphere stage , cells were removed from donor embryos and transferred to wild type hosts using a glass capillary . Transplanted angioblasts were identified by transgenic EGFP expression together with mini-Ruby stain . By counting the number of EGFP positive donor angioblasts in the midline vs all EGFP positive donor angioblasts ( 100% ) , the percentage of donor angioblasts in the midline was determined for each individual embryo . | The circulatory system enables blood to move around the body and deliver substances including nutrients and oxygen to the cells that need them . In the embryos of animals with a backbone , blood flows from the heart through the aorta into branching smaller vessels to the cells . The blood then gets collected by progressively bigger vessels and flows back to the heart via the cardinal vein . The cells that make up these blood vessels develop from cells called angioblasts—but first , during development these angioblasts must move to the place where the vessels will form . A protein called Vascular endothelial growth factor ( VEGF ) had been suggested to help guide and align the angioblasts as the embryo develops . Now , Helker , Schuermann et al . have examined developing zebrafish embryos using new technologies . This revealed that VEGF is in fact not essential for the dorsal aorta and cardinal vein to develop . Instead , the angioblasts only move to the correct part of the embryo if they can produce the Apelin receptor protein , which forms part of a signaling pathway . There are two hormones—called Apelin and Elabela—that can bind to and activate the Apelin receptor . Helker , Schuermann et al . show that Elabela alone is needed to guide the angioblasts to the right part of the embryo during blood vessel development . However , in embryos where there is not enough Elabela , the Apelin hormone can compensate for this deficiency and the first blood vessels will later develop correctly . Future research will address whether this signaling pathway not only guides angioblasts to establish a circulatory system , but also guides blood vessel growth . As blood vessel growth is very relevant to human disease , identifying the mechanisms that regulate it will have an impact on biomedical research . | [
"Abstract",
"Main",
"text",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"short",
"report"
] | 2015 | The hormonal peptide Elabela guides angioblasts to the midline during vasculogenesis |
Actin , spectrin , and associated molecules form a periodic sub-membrane lattice structure in axons . How this membrane skeleton is developed and why it preferentially forms in axons are unknown . Here , we studied the developmental mechanism of this lattice structure . We found that this structure emerged early during axon development and propagated from proximal regions to distal ends of axons . Components of the axon initial segment were recruited to the lattice late during development . Formation of the lattice was regulated by the local concentration of βII spectrin , which is higher in axons than in dendrites . Increasing the dendritic concentration of βII spectrin by overexpression or by knocking out ankyrin B induced the formation of the periodic structure in dendrites , demonstrating that the spectrin concentration is a key determinant in the preferential development of this structure in axons and that ankyrin B is critical for the polarized distribution of βII spectrin in neurites .
Neurons are highly polarized cells with their somatodendritic regions receiving synaptic inputs and axons propagating electrical signals and sending synaptic outputs to target cells . Cytoskeletal proteins are important for maintaining the polarity of neurons . For example , actin and microtubules are essential for the growth and stabilization of axons , the trafficking of cargos to specific neurites , and the stabilization and plasticity of synapses ( Luo , 2002; Dent and Gertler , 2003; Cingolani and Goda , 2008; Barnes and Polleux , 2009; Kapitein and Hoogenraad , 2011; Stiess and Bradke , 2011 ) . Transient destabilization of actin at the tip of a neurite is sufficient to induce a dendrite to become an axon ( Bradke and Dotti , 1999 ) . Increasing evidence also suggests an important role for spectrin in the maintenance of neuronal polarization , as well as the development and stabilization of axons ( Hammarlund et al . , 2007; Galiano et al . , 2012 ) . αII and βII spectrin are enriched in axons ( Riederer et al . , 1986; Galiano et al . , 2012 ) . Spectrin is known to be important for providing the mechanical stability for axons ( Hammarlund et al . , 2007 ) and protecting them from mechanical stress ( Krieg et al . , 2014 ) , for axon path finding ( Hulsmeier et al . , 2007 ) , for the stabilization of pre-synaptic terminals ( Pielage et al . , 2005 ) , and for maintaining specific membrane domains in axons ( Susuki and Rasband , 2008 ) . Mice lacking either αII or βII spectrin die in the embryo , highlighting the crucial function of these proteins ( Tang et al . , 2003; Stankewich et al . , 2011; Galiano et al . , 2012 ) . Spectrin has also been shown to play a role in human neurological diseases ( Ikeda et al . , 2006; Writzl et al . , 2012 ) . Recently , we discovered a periodic sub-membrane lattice structure made of actin , spectrin , and other associated molecules in the axons of mammalian neurons ( Xu et al . , 2013 ) . In this membrane skeleton , actin filaments form a ring-like structure that wraps around the circumference of axons . These actin rings are evenly spaced along the axon shaft with a period of ∼190 nm and show a remarkably long-range order . Actin filaments in the rings are capped by adducin . Adjacent actin rings are connected by spectrin , likely in the form of αII-βII-spectrin heterotetramers , given the observations that the periodic βII spectrin rings alternate with the actin–adducin rings along axons and that the observed 190 nm period matches the length of the spectrin tetramer . The ultrastructural organization of this quasi-one-dimensional , periodic lattice structure is different from the previously observed , two-dimensional polygonal membrane skeletal structure found in red blood cells ( Byers and Branton , 1985; Liu et al . , 1987; Bennett and Lorenzo , 2013 ) , whereas an erythrocyte-like polygonal membrane skeletal was observed in the axon terminals at the Drosophila neuromuscular junction ( Pielage et al . , 2008 ) . Interestingly , this periodic structure preferentially forms in axons , with the actin in dendrites primarily adopting the form of long filaments running along the dendrite shaft ( Xu et al . , 2013 ) . In Caenorhabditis elegans that lack β spectrin or carry a β spectrin mutant , axons break more easily during animal movement ( Hammarlund et al . , 2007 ) and exhibit impaired touch sensation ( Krieg et al . , 2014 ) , suggesting that this structure may be important for the mechanical stability of axons and for sensing mechanical stimuli . This periodic lattice also organizes the axonal membrane by placing important membrane proteins , such as the voltage-gated sodium channels , into a periodic distribution ( Xu et al . , 2013 ) . However , it is unknown how this highly regular membrane skeleton structure develops in axons and how its formation is regulated . For example , it is unclear whether the periodic lattice develops during early or late stages of axon differentiation . Although protein factors previously identified to be important for axon differentiation tend to be enriched and function at the growing tips of axons ( Arimura and Kaibuchi , 2007; Barnes and Polleux , 2009; Stiess and Bradke , 2011; Cheng and Poo , 2012 ) , it is unknown whether the actin–spectrin lattice also initiates at the distal ends of axons or instead forms first in the proximal region near the cell body . Finally , the molecular mechanism that regulates the specific formation of this periodic structure in axons , instead of dendrites , remains a mystery . In this study , we addressed these important questions concerning the development of this newly discovered neuronal structure . We found that the periodic membrane skeleton initiated early during axon differentiation . The lattice structure originated in the axonal region adjacent to the cell body and propagated to the distal ends of axons . The lattice structure further matured by recruiting other components , and the matured membrane skeleton was highly stable . Multiple molecular factors played roles in regulating the formation of this structure . The lattice structure depended on intact microtubules . The high local concentration of βII spectrin in axons was the key determining factor for the specific formation of the lattice structure in axons , and artificially increasing the concentration of βII spectrin in dendrites was sufficient to induce the formation of the periodic lattice structure in dendrites . Remarkably ankyrin B was important for the polarized distribution of βII spectrin in neurites; in ankyrin B knockout mice , βII spectrin was evenly distributed in axons and dendrites , giving rise to a highly regular , periodic membrane skeleton in both dendrites and axons .
Neurons exhibit distinct developmental stages with different morphological characteristics during polarization ( Dotti et al . , 1988; Arimura and Kaibuchi , 2007; Barnes and Polleux , 2009; Cheng and Poo , 2012 ) . In dissociated hippocampal neuronal culture , neurons first display intense lamellipodial protrusive activity in stage 1 , which then leads to the emergence of multiple immature neurites in stage 2 ( ∼1 Day in Vitro [DIV] ) . In stage 3 ( DIV 2–4 ) , one of these neurites breaks the symmetry and extends rapidly to become an axon . The other neurites then gradually acquire dendritic properties in stage 4 ( DIV 4–7 ) . In stage 5 ( >DIV 7 ) , neurons continue to mature and form axon initial segments , dendritic spines , and synapses . In order to determine the developmental course of the periodic membrane skeletal structure , we fixed dissociated neurons at different developmental stages , immunostained for βII spectrin , and imaged using stochastic optical reconstruction microscopy ( STORM ) , a super-resolution imaging method that relies on switching and localizing single molecules to acquire sub-diffraction limit images ( Betzig et al . , 2006; Hess et al . , 2006; Rust et al . , 2006; Huang et al . , 2008 ) . To illustrate how we systematically imaged and quantified this periodic structure in axons , we first imaged a neuron at DIV 10 . Consistent with our previous findings ( Xu et al . , 2013 ) , βII spectrin adopted a highly regular , periodic pattern in all regions of the axonal shaft ( Figure 1—figure supplements 1 , 2 ) . Both Fourier transform and autocorrelation analyses showed that the βII spectrin adopted a periodic distribution with a period of ∼190 nm ( Figure 1—figure supplement 1 ) . Similarly , actin filaments exhibited a highly periodic distribution along axon shafts ( Figure 1—figure supplement 3 ) . Depolymerizing the actin filaments with latrunculin A ( LatA ) disrupted the periodic distribution of βII spectrin ( Figure 1—figure supplement 2 ) , and knocking-down βII spectrin using shRNA led to a loss of the periodic actin distribution ( Figure 1—figure supplement 3 ) . These results indicate that the periodic organizations of actin and spectrin are interdependent , consistent with the model that adjacent actin rings are connected by the spectrin tetramers . Next , we quantified the distribution of βII spectrin at earlier developmental stages in DIV 2 , 4 , and 6 neurons ( Figure 1 ) . Figure 1A shows a typical stage 3 neuron at DIV 2 , with one neurite outgrowing the others and becoming an axon . Interestingly , the periodic pattern of βII spectrin has already formed in the proximal region of this axon near the cell body , as shown by both Fourier transform and autocorrelation analyses of the STORM image ( Figure 1A , C ) . However , the periodic distribution did not extend far—the middle and distal parts of the same axon did not exhibit the periodic pattern ( Figure 1A , C ) . Similar results were observed for other stage 3 neurons that we imaged . As neurons continued to mature , the periodic βII spectrin distribution extended to more distal regions of axons . By DIV 6 , the periodic βII spectrin distribution extended for nearly the entire length of the axon , except for the very distal region ( Figure 1B , D ) . Using the autocorrelation amplitude at the first peak ( ∼190 nm ) to quantify the degree of periodicity for the βII spectrin distributions , we found that the periodicity degraded quickly along axons in DIV 2 neurons but gradually extended to the distal end of the axon in later developmental stages until the structure eventually occupied nearly the entire axon ( Figure 1E and Figure 1—figure supplement 4 ) . Taken together , these results demonstrate that the periodic membrane skeleton forms early during development , originates in proximal axon regions close to the cell body , and propagates toward the distal end of the axon . 10 . 7554/eLife . 04581 . 003Figure 1 . Early development and propagation of the periodic lattice structure in axons . ( A ) A DIV 2 neuron was stained with βII spectrin antibody and imaged by 3D STORM . The single long process from the cell is the axon . A1 , A2 , and A3 are 3D STORM images taken from arrow-indicated regions from A . The Fourier transform analyses of the βII spectrin distribution along the axon shaft are shown on the right . ( B ) Similar to ( A ) , but for a DIV 6 neuron . ( C–D ) Autocorrelation analysis of βII spectrin distributions of DIV 2 ( C ) or DIV 6 ( D ) neurons at the proximal , middle , and distal regions of axons . Shown are the averaged autocorrelation from multiple segments of axons for each condition . ( E ) The average amplitude of autocorrelation analysis for different axonal regions of DIV 2 , 4 , and 6 neurons . The amplitude was measured as the difference between the first peak and the average of the two first valleys of the autocorrelation curve . Error bars are standard deviation from measurements of multiple neurons ( n = 7 neurons for DIV 2; n = 6 neurons for DIV 4; n = 9 neurons for DIV 6; from three independent experiments at each DIV ) . The color bar for 3D STORM image indicates the z-depth of the image and is the same for all of our STORM images . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 00310 . 7554/eLife . 04581 . 004Figure 1—figure supplement 1 . βII spectrin structure in a DIV 10 neuron . ( A ) A DIV 10 neuron was immunostained for βII spectrin ( green ) and Map2 ( red ) and imaged by conventional fluorescence and 3D STORM microscopy . STORM images of the two arrow-indicated regions along the axon are shown below . ( B ) Histograms of the βII spectrin localizations from the boxed regions in STORM images . ( C ) Fourier transform analyses of the βII spectrin localizations from the boxed regions . ( D ) Autocorrelation analyses of the βII spectrin localizations from the boxed regions . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 00410 . 7554/eLife . 04581 . 005Figure 1—figure supplement 2 . The periodic distribution of βII spectrin depends on actin . ( A ) STORM image of βII spectrin in axons of control DIV 10 neurons . ( B ) STORM image of βII spectrin of LatA-treated DIV 10 neurons ( 20 µM LatA , 1 hr ) . Similar results were observed in at least seven independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 00510 . 7554/eLife . 04581 . 006Figure 1—figure supplement 3 . The periodic distribution of actin depends on βII spectrin . ( A ) STORM image of actin in axons of control DIV 10 neurons . In some of the thinnest axons , the presence of long actin filaments running along the axons can disguise the periodic structure due to finite resolution . ( B ) STORM image of actin in axons of βII spectrin-shRNA expressing neurons . Though we picked axon-enriched sample regions for imaging , we cannot preclude the possibility that a small number of the processes may be dendrites . Similar results were observed in 8 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 00610 . 7554/eLife . 04581 . 007Figure 1—figure supplement 4 . The autocorrelation analysis is not sensitive to the number of localizations present in the periodic structure within the range of the localization numbers that we detected in axon segments . Since the localization density of βII spectrin decreased towards the distal end of the axons , we tested whether the autocorrelation amplitude measured in Figure 1E would be sensitive to the localization numbers detected in various axon segments . ( A ) Filled circles: the autocorrelation amplitudes for different axonal regions of DIV 2 neurons reproduced from Figure 1E; open circles: the estimated dependence of the autocorrelation amplitude on the detected localization numbers if the periodic structure remained the same as that in the proximal axon region . To obtain these data , we randomly removed localizations from the proximal axon region to match the localization numbers measured in distal axons at various distances to the cell body , calculated the corresponding autocorrelation amplitudes , and plotted them here as the open circles . ( B–C ) Same as ( A ) except that the analyses were for DIV 4 and DIV 6 neurons . Note that the autocorrelation amplitude only decreases slightly when the localization number of a periodic structure was artificially reduced to match those measured in distal axon ends ( open circles ) , and this slight change does not contribute substantially to the measured dependence of the autocorrelation amplitude on the distance to cell body ( filled circles ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 00710 . 7554/eLife . 04581 . 008Figure 1—figure supplement 5 . Distributions of dendritic βII and βIII spectrin . ( A–B ) DIV 2 and DIV 5 neurons were immunostained with βII spectrin , and both axonal and dendritic processes were imaged by 3D STORM . The reconstructed neuron image , 3D STORM images of axon and dendrite from indicated regions ( arrows ) , and autocorrelation analyses are shown . ( C–E ) DIV 14 neurons were immunostained with βII spectrin and a dendritic marker Map2 , and the βII spectrin was imaged by 3D STORM . The conventional images , 3D STORM images of axons and dendrites from indicated regions ( arrows ) , and autocorrelation analyses are shown . ( F ) DIV 14 neurons were immunostained with βIII spectrin . The conventional image , STORM images of βIII spectrin from two indicated dendritic regions , and autocorrelation analyses are shown . βIII spectrin only stains for dendrites , and we did not observe detectable βIII spectrin in axons . ( G ) Average autocorrelation analyses of βII spectrin in axons and dendrites of neurons at DIV 5 , DIV 10 , and DIV 14 ( n = 14 neurons for DIV 5; n = 11 neurons for DIV 10; n = 16 neurons for DIV 14; at least three independent experiments at each DIV ) . ( H ) Average autocorrelation analysis of βIII spectrin of DIV 14 neurons ( n = 12 neurons , four independent experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 00810 . 7554/eLife . 04581 . 009Figure 1—figure supplement 6 . The periodic actin structure is disrupted if the neurons are subjected to membrane extraction prior to fixation . DIV 10 neurons were extracted with 1% Triton X-100 for 3 min , followed by two quick rinses , and then fixed with 0 . 2% GA for 20 min . See ‘Materials and methods’ , ‘Fluorescence labeling of neurons’ section , for detailed procedure . 3D STORM images of actin and βII spectrin are shown in ( A ) and ( B ) , respectively . Similar observations were found in six independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 009 The periodic structure was only observed in axons , but not in dendrites , during early developmental stages , whereas isolated patches of periodic βII spectrin patterns were observed in dendrites during later developmental stages ( Figure 1—figure supplement 5A–E ) . However , unlike in axons , these patches did not form a cohesive lattice structure with a long-range order . Quantitatively , the average autocorrelation analysis showed much smaller amplitudes in dendrites than those in axons ( Figure 1—figure supplement 5G ) , indicating a much poorer regularity of the structure in dendrites . Similar results were observed for βIII spectrin ( Figure 1—figure supplement 5F , H ) , an isoform of β spectrin that is enriched in dendrites instead of axons ( Sakaguchi et al . , 1998; Stankewich et al . , 1998; Gao et al . , 2011 ) . Similar to the lattice structure in mature axons , the periodic pattern of βII spectrin depended on actin during early development . Treatment of neurons with actin-depolymerizing drugs , cytochalasin D ( CytoD ) , or LatA disrupted the periodicity of βII spectrin in DIV 3 neurons ( Figure 2A–E ) . The effect of actin-depolymerizing drugs set in quickly with the periodic βII spectrin distribution substantially disrupted after several minutes of LatA treatment ( Figure 2F , G ) , consistent with the drug acting directly on the lattice structure . These results indicate that actin is involved in the lattice structure during early neuronal development . The form of actin , however , appeared to be different during the early developmental stages as compared to that in mature axons . We have previously shown that the periodic pattern of actin was not directly observed in the STORM images during DIV 1–4 . In DIV 5 , the periodic actin pattern begins to appear in some neurons and become robustly observed in neurons at DIV 7 ( Xu et al . , 2013 ) . Similar results were observed here ( data not shown ) . One possible interpretation is that actin existed in a less stable form during the early developmental stages and was not preserved by our sample treatment ( fixation and extraction ) prior to imaging . Consistent with the notion that actin filaments in the lattice structure were less stable during early developmental stages , the periodic structure of βII spectrin was more quickly disrupted by LatA treatment during early development than in older neurons ( Figure 2G ) . 10 . 7554/eLife . 04581 . 010Figure 2 . The periodic structure of βII spectrin depends on actin during early development . ( A–C ) DIV 3 neurons were either untreated or treated with latrunculin A ( LatA , 20 µM ) or cytochalasin D ( CytoD , 50 µM ) for 1 hr and subsequently immunostained with βII spectrin antibody for 3D STORM imaging . Shown here are representative images of βII spectrin in proximal axonal regions from control ( A ) , LatA-treated ( B ) and CytoD-treated ( C ) neurons . ( D ) Average autocorrelation analyses of βII spectrin from multiple axon segments of control , LatA-treated , and CytoD-treated DIV 3 neurons ( n = 6 neurons for control; n = 7 neurons for LatA-treated; n = 7 neurons for CytoD-treated conditions; at least three independent experiments for each condition ) . ( E ) The average autocorrelation amplitudes from control , LatA-treated , and CytoD-treated DIV 3 neurons . ( F ) Average autocorrelation analyses of βII spectrin from multiple axon segments of control and LatA-treated DIV 3 neurons at different treatment time ( n > 5 neurons for each condition , three independent experiments ) . The axon segments are taken from the proximal axonal regions near the cell bodies . ( G ) DIV 3 and DIV 10 neurons were treated with 20 µM LatA for indicated amount of time , and the average autocorrelation amplitude of βII spectrin from these neurons are shown . Error bars are standard deviation from measurements of multiple neurons ( n = 6 neurons for DIV 3; n = 8 neurons for DIV 10; four independent experiments at each DIV ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 010 We also observed a relatively slow developmental time course for the periodic pattern of adducin , an actin-capping protein ( Kuhlman et al . , 1996 ) . The periodic pattern of adducin was not observed in axons at DIV 2 or DIV 4 ( Figure 3 ) . A periodic pattern began to appear at ∼DIV 6 and became obvious after DIV 7 ( Figure 3 and Figure 3—figure supplement 1 ) . The lack of adducin capping may have contributed to the lower stability of actin during early development stages , although it is also possible that the lower stability of actin during the early development stages made it difficult to maintain the adducin pattern during cell fixation and extraction . Finally , it is formally possible that actin and adducin are not present in the periodic lattice structure during early developmental stages . However , we consider such a scenario to be less likely as it is difficult to imagine how spectrin tetramers themselves could self-assemble into a periodic lattice structure without the help of actin to crosslink multiple spectrin tetramers . 10 . 7554/eLife . 04581 . 011Figure 3 . Recruitment of adducin into the periodic lattice structure during development . ( A–C ) Neurons were immunostained for adducin and imaged by 3D STORM . Shown here are representative images of adducin at proximal axonal regions from DIV 2 , 4 , and 8 neurons . ( D ) Average autocorrelation analyses of adducin at proximal regions of axons near the cell body . The autocorrelation curves are averaged from multiple neurons ( n = 5 neurons for DIV 2; n = 7 neurons for DIV 4; n = 6 neurons for DIV 8; three independent experiments at each DIV ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 01110 . 7554/eLife . 04581 . 012Figure 3—figure supplement 1 . Distribution of adducin in axons . Neurons were immunostained for adducin and imaged by 3D STORM . The reconstructed neuron image , 3D STORM images of indicated regions ( arrows ) , and corresponding autocorrelation analyses of DIV 4 ( A ) and DIV 8 ( B ) neurons are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 012 Together , the above results indicate that the periodic membrane skeleton continued to mature after formation . Consistent with this notion , the autocorrelation amplitudes of the periodic βII spectrin distribution also continued to increase with time as the neuron matured ( Figure 1E ) . As neurons further mature , axon initial segment ( AIS ) starts to assemble at the axonal region proximal to the cell body . Ankyrin G is the master regulating protein for AIS assembly and recruits other molecular components such as βIV spectrin and sodium channels to the AIS ( Zhou et al . , 1998; Jenkins and Bennett , 2001; Yang et al . , 2007 ) . Next , we examined whether βIV spectrin and ankyrin G were also recruited to the periodic membrane skeleton and , if so , during which developmental stage these components were incorporated . We labeled ankyrin G using an antibody against its spectrin-binding domain near the N-terminus and βIV spectrin using an antibody against its N-terminal domain . Both ankyrin G and βIV spectrin signals were weak during early developmental stages , and the signals became stronger in the proximal region of axons at DIV 8 ( Figure 4—figure supplement 1 ) . Notably , the expression level of βII spectrin remained high throughout the axons during this time ( Figure 4—figure supplement 1 ) . At this time , the distributions of ankyrin G and βIV spectrin were not periodic ( Figure 4A–D ) , in contrast to the highly periodic βII spectrin in the proximal region of axons ( Figure 1 ) . Over time , both ankyrin G and βIV spectrin signals further increased in the proximal region of axons , and by DIV 12 , the N-terminal domains of both ankyrin G and βIV spectrin adopted highly periodic distributions , indicating that these molecules were incorporated into the periodic lattice structure ( Figure 4A–D ) . Interestingly , the periodicity was substantially less pronounced for the C-terminal domain of βIV spectrin and undetectable for the C-terminal domain of ankyrin G ( Figure 4A–D ) . These results suggest that the N-terminal regions of these molecules were tightly incorporated in the periodic lattice structure , but their C-terminal regions were likely hanging off from the lattice structure and moving relatively freely . Notably , as ankyrin G and βIV spectrin became incorporated into the periodic lattice , we observed a decrease in the local concentration of βII spectrin at the AIS ( Figure 4E and Figure 4—figure supplement 1C ) . The decrease of the βII spectrin concentration was associated with a loss of periodicity for βII spectrin at AIS ( Figure 4F ) , suggesting that as βIV spectrin was incorporated into the periodic structure in the AIS region , βII spectrin was displaced . 10 . 7554/eLife . 04581 . 013Figure 4 . Assembly of AIS components into the periodic lattice structure during late developmental stages . ( A and B ) Neurons were immunostained with antibodies against βIV spectrin N-terminus ( βIV spectrin ( N ) ) , βIV spectrin C-terminus ( βIV spectrin ( C ) ) , ankyrin G spectrin-binding domain ( ankyrin G ( N ) ) , or ankyrin G C-terminus ( ankyrin G ( C ) ) , and imaged at various DIVs by 3D STORM . Representative conventional images and STORM images from the boxed region at different developmental stages are shown . ( C ) Left: average autocorrelation analyses of βIV spectrin N-terminus from neurons at different developmental stages ( n = 9 neurons for DIV 4; n = 12 neurons for DIV 8; n = 16 neurons for DIV 12; at least three independent experiments at each DIV ) . Right: average autocorrelation analyses of βIV spectrin N-terminus and C-terminus of DIV 12 neurons ( n = 13 neurons for N-terminus and n = 15 neurons for C-terminus , three independent experiments ) . ( D ) Left: average autocorrelation analyses of ankyrin G spectrin-binding domain ( near N-terminus ) from neurons at different developmental stages ( n = 12 neurons for DIV 4; n = 15 neurons for DIV 8; n = 14 neurons for DIV 12; at least three independent experiments at each DIV ) . Right: average autocorrelation analysis of ankyrin G N-terminus and C-terminus of DIV 12 neurons ( n = 9 neurons for N-terminus and n = 10 neurons for C-terminus , three independent experiments ) . ( E and F ) A DIV 13 neuron was immunostained with βIV and βII spectrin antibodies . βII spectrin was subjected for 3D STORM imaging . ( E ) Left: conventional image of βII and βIV spectrin in axon . Right: fluorescent intensity profile of βII and βIV spectrin along the axon . ( F ) Left: STORM image of βII spectrin in the same region . Right: autocorrelation analyses of βII spectrin from red- and green-boxed regions . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 01310 . 7554/eLife . 04581 . 014Figure 4—figure supplement 1 . The expression profile of βII spectrin and ankyrin G in axons at different developmental stages . Neurons at DIV 4 ( A ) , DIV 8 ( B ) , and DIV 12 ( C ) were immunostained for βII spectrin and ankyrin G . Reconstructed neuron images based on βII spectrin fluorescence , ankyrin G fluorescence , and the fluorescence intensity profiles of both βII spectrin and ankyrin G along axons are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 01410 . 7554/eLife . 04581 . 015Figure 4—figure supplement 2 . The formation of the periodic βIV spectrin structure in the AIS is dependent on βII spectrin . ( A ) Neurons were infected with βII spectrin-shRNA expressing adenovirus at DIV 3 and subsequently stained for βII spectrin and βIV spectrin at DIV 12 . βIV spectrin was subjected for STORM imaging . Infected neurons have a GFP signal . The conventional images of βII spectrin , GFP , and βIV spectrin and the overlay image are shown . Efficient knockdown is demonstrated by lack of βII spectrin in GFP-positive process . ( B ) STORM image of βIV spectrin from boxed region in ( A ) . ( C ) Average autocorrelation analysis of βIV spectrin from βII spectrin knockdown neurons ( n = 11 neurons , three independent experiments ) . ( D–F ) Similar to ( A–C ) except that the virus was added at DIV 7 ( n = 9 neurons , four independent experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 015 To test whether the assembly of βIV spectrin into the periodic structure may rely on βII spectrin , we knocked down βII spectrin at various DIVs using a shRNA-expressing adenovirus , which also expressed GFP , and subsequently imaged βIV spectrin at DIV 12 . The efficiency of knockdown was demonstrated by a lack of βII spectrin signal in virus-infected , GFP-positive neurons ( Figure 4—figure supplement 2 ) . When infected by the virus at DIV 3 , the enrichment of βIV spectrin in the AIS region appeared partially impaired by βII spectrin knockdown , though at least 60% of the neurons still exhibited enrichment of βIV spectrin in AIS . For these neurons , the periodicity of βIV spectrin was also partially disrupted in the βII spectrin-depleted neurons ( Figure 4—figure supplement 2A–C ) , indicating that βII spectrin is important for the periodic assembly of βIV spectrin . On the other hand , when neurons were infected with the virus at DIV 7 , βIV spectrin remained periodic even though βII spectrin was depleted ( Figure 4—figure supplement 2D–F ) . Because it takes several days for pre-existing βII spectrin molecules to degrade ( Susuki et al . , 2011 ) , it is likely that βIV spectrin was already incorporated into the periodic lattice before the eventual depletion of βII spectrin when the virus was added late . We next probed the dynamics of the periodic lattice structure in live neurons . To this end , we genetically fused βII spectrin with mMaple3 , a recently developed photoactivatable fluorescent protein ( Wang et al . , 2014 ) . In neurons moderately expressing βII spectrin-mMaple3 , the periodic pattern of βII spectrin-mMaple3 was readily observable in axons and the spacing of ∼190 nm was identical to that observed for endogenous βII spectrin in fixed neurons ( Figure 5A ) . The periodic pattern was smeared in neurons with high expression levels of βII spectrin-mMaple3 , presumably by the excess , freely diffusing βII spectrin-mMaple3 molecules that were not incorporated into the lattice structure . 10 . 7554/eLife . 04581 . 016Figure 5 . The periodic lattice structure is stable in live neurons . ( A ) 3D STORM image of βII spectrin-mMaple3 in live neurons at DIV 10 . ( B ) The STORM movie was segregated into four different time windows . Fourier transform analysis of each time window is shown . The baseline of Fourier traces is shifted manually for clear visualization . ( C ) Cross-correlation analysis of βII spectrin across different time windows . The black curve is the autocorrelation of the image during 0–100 s . The color curves are the cross-correlation between 0–100 s and later time windows . Similar results were found in six independent experiments . ( D–E ) FRAP analyses of βII spectrin in DIV 10 neurons . Neurons were transfected with βII spectrin-GFP at DIV 8 . ( D ) Representative neurons at a low βII spectrin expression level , where βII spectrin-GFP molecules were incorporated into the periodic structure . The images before photo-bleaching , 0 s post-bleaching , 1600 s post-bleaching , and the fluorescence recovery trace are shown . ( E ) The fluorescence recovery of representative neurons with a high βII spectrin expression level , where most βII spectrin-GFP molecules were not incorporated into the periodic structure . Similar results were found in at least nine independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 01610 . 7554/eLife . 04581 . 017Figure 5—figure supplement 1 . Incorporation of βII spectrin-GFP into the periodic lattice structure in overexpressing neurons . Neurons were transfected with βII spectrin-GFP at DIV 8 and immunostained for GFP at DIV 10 . Representative conventional images of βII spectrin-GFP in low expression and high expression neurons are shown in ( A ) and ( B ) , respectively . The STORM image and autocorrelation analyses of boxed regions are shown in ( C ) and ( D ) . ( E ) Quantification of the expression levels of βII spectrin-GFP in transfected neurons that show ( red ) or do not show ( black ) the periodic lattice structure . The expression levels of the neurons subject to the FRAP analysis in Figure 5D , E are indicated by arrows . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 017 The periodic βII spectrin pattern appeared to be mostly static . Fourier analysis of the patterns showed that the spatial frequency ( i . e . , the period ) of the structure did not change over the imaging time of several minutes ( Figure 5B ) . Cross-correlation analysis of the patterns taken at different time points showed no phase shift of the periodic structure during the imaging time ( Figure 5C ) . As an alternative approach to probe the stability of the structure , we used fluorescence recovery after photo-bleaching ( FRAP ) . For this analysis , we transfected neurons with a βII spectrin-GFP fusion construct , bleached the GFP signal in local regions of axons , and measured the signal recovery rate . In neurons that exhibited moderate expression levels of βII spectrin , where the majority of βII spectrin-GFP molecules were incorporated into the periodic lattice structure ( Figure 5—figure supplement 1 ) , the recovery rate was extremely slow and essentially undetectable after 30 min ( Figure 5D ) . In contrast , the fluorescence recovery was much faster ( 75% recovery in 5 min ) in neurons , where the expression level of βII spectrin-GFP was high and the majority of βII spectrin-GFP molecules were not incorporated into the periodic structure ( Figure 5E and Figure 5—figure supplement 1 ) . These data indicate that periodic lattice structure was highly stable in live neurons . Microtubules are essential for the establishment of neuronal polarity . Local stabilization of microtubules is sufficient to induce axon formation ( Witte et al . , 2008 ) . Moreover , tubulin binds to ankyrin B , a molecule that also interacts with βII spectrin ( Bennett and Davis , 1981 ) . We thus tested whether microtubules play a role in the formation of the periodic membrane skeleton structure . In neurons treated with the microtubule-disrupting drug nocodazole ( 50 µM for 1 hr ) , the periodic pattern of βII spectrin was largely disrupted ( Figure 6A , B ) . On the other hand , when microtubules were stabilized with taxol ( 5 nM for 3 days ) , a treatment that is known to induce multiple axon-like processes in neurons ( Witte et al . , 2008 ) , we found that βII spectrin exhibited a periodic pattern in all of these axon-like processes ( Figure 6C ) . We also treated neurons with SB 216763 , a drug that stabilizes microtubules and promotes axonal growth by inhibiting glycogen synthase kinase-3 beta ( GSK-3β ) ( Jiang et al . , 2005; Yoshimura et al . , 2005 ) . Similarly , in neurons treated with SB-216763 , we observed that the periodic lattice structure was formed in multiple axon-like long processes ( Figure 6—figure supplement 1 ) . 10 . 7554/eLife . 04581 . 018Figure 6 . The periodic structure of βII spectrin relies on intact microtubules . ( A and B ) DIV 4 neurons were either untreated or treated with nocodazole ( 50 µM ) for 1 hr and immunostained with βII spectrin antibody for 3D STORM . A reconstructed neuron image from nocodazole-treated neurons and two STORM images are shown in A . ( B ) Average autocorrelation analyses of βII spectrin from multiple control and nocodazole-treated neurons ( n = 9 neurons for control and n = 8 neurons for nocodazole-treated conditions; three independent experiments for each condition ) . ( C ) Neurons were treated with taxol ( 5 nM ) at DIV 3 for 3 days , which induces the growth of multiple axon-like processes . A representative STORM image of a treated neuron at DIV 6 , the enlarged images of the boxed regions , and the autocorrelation analyses ( n = 10 neurons , four independent experiments ) are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 01810 . 7554/eLife . 04581 . 019Figure 6—figure supplement 1 . The periodic βII spectrin structure can be induced in multiple processes by axon-promoting small molecules . Neurons were treated with SB 216763 at DIV 1 , a drug that inhibits GSK3β and promotes the growth of multiple axon-like processes , and immunostained for βII spectrin at DIV 5 . A representative conventional image of βII spectrin ( A ) , STORM image of boxed region ( B ) , and autocorrelation analyses of the arrow-indicated regions ( B1 and B2 ) are shown . Similar results were found for three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 019 Given the dramatically different actin–spectrin organizations in axons and dendrites , an interesting question arises as to what molecular factors are critical for promoting the formation of the highly regular , periodic lattice structure in axons and/or suppressing it in dendrites . Ankyrin B ( ANK 2 ) is a molecule that binds to βII spectrin ( Bennett and Lorenzo , 2013 ) . It is highly enriched in axons ( Chan et al . , 1993; Kunimoto , 1995; Engelhardt et al . , 2013 ) and recently found to be potentially linked with autism ( De Rubeis et al . , 2014; Iossifov et al . , 2014 ) . We have shown previously that ankyrin B also adopts a partially periodic pattern in axons albeit with a lower regularity ( Xu et al . , 2013 ) , potentially due to an incomplete occupancy of the ankyrin B binding sites on the lattice structure and the presence of ankyrin B on intracellular membranes ( Bennett and Lorenzo , 2013 ) . We thus asked whether ankyrin B is involved in regulating the formation of this periodic membrane skeleton in axons . To address this question , we performed STORM imaging on disassociated hippocampal neurons from ankyrin B knockout mice ( Scotland et al . , 1998 ) at DIV 10 . Similar to wild-type neurons , ankyrin B knockout neurons still showed enrichment of MAP2 in dendrites with similar dendritic morphology , making dendrites easy to identify in these neurons ( Figure 7A and Figure 7—figure supplement 1 ) . The periodic pattern of βII spectrin in axons was not perturbed by ankyrin B deletion and appeared quantitatively similar to that observed in control wild-type neurons ( Figure 7—figure supplement 1 ) . Surprisingly , βII spectrin also adopted a highly regular , periodic distribution in all dendrites , with the periodicity quantitatively similar to that observed in axons ( Figure 7A , B ) . This is in stark contrast to what we observed in wild-type neurons , where the distributions of βII spectrin in dendrites were largely irregular ( Figure 1—figure supplement 5 ) . The actin-capping protein adducin also adopted a periodic distribution in dendrites of ankyrin B knockout neurons , with quantitatively similar periodicity to that of βII spectrin . Knocking-down βII spectrin disrupted the periodic distribution of adducin , indicating that the periodic lattice structure in the dendrites of the ankyrin B knockout neurons also depended on βII spectrin ( Figure 7—figure supplement 2 ) . These data indicate that the formation of the periodic membrane skeleton does not require ankyrin B . Instead , ankyrin B is important for inhibiting the formation of this periodic lattice structure in dendrites . 10 . 7554/eLife . 04581 . 020Figure 7 . Role of ankyrin B in the regulation of the periodic lattice structure . ( A–B ) DIV 10 neurons from ankyrin B knockout ( KO ) mice were immunostained for βII spectrin and a dendritic marker Map2 and imaged . ( A ) Conventional image of βII spectrin and Map2 . ( B ) 3D STORM of βII spectrin . The image is taken from the green-boxed region of the neuron in Figure 7—figure supplement 1 . The enlarged STORM image and autocorrelation analyses of boxed regions are shown in ( B1 ) and ( B2 ) . Similar results were found in four independent experiments . ( C ) Conventional βII spectrin image from wild-type ( control ) and ankyrin B KO DIV 10 neurons . The fluorescence intensity is coded by color , with red indicating higher expression . ( D ) The relative fluorescence intensity of βII spectrin in dendrites and axons of wild-type and ankyrin B KO neurons ( n = 14 neurons for wild-type and n = 13 neurons for ankyrin B KO; three independent experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 02010 . 7554/eLife . 04581 . 021Figure 7—figure supplement 1 . The βII spectrin structure in axons of ankyrin B KO neurons . ( A ) Reconstructed image of an ankyrin B KO DIV 10 neuron . βII spectrin was subjected for 3D STORM image in both dendrites and axons . The STORM image of dendrites in green-boxed region is shown in Figure 7 . The STORM image of axon in white-boxed region is shown in ( B ) . ( C ) Enlarged STORM image of the boxed region in ( B ) . ( D–E ) Fourier transform and autocorrelation analyses of the axon segment shown in C . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 02110 . 7554/eLife . 04581 . 022Figure 7—figure supplement 2 . Formation of the periodic lattice structure in dendrites of ankyrin B knockout neurons depends on βII spectrin . ( A–B ) Ankyrin B knockout neurons were immunostained for adducin ( green ) and MAP2 ( magenta ) , and imaged by conventional ( A ) and 3D STORM ( B ) microscopy . Magnified STORM images of adducin for arrow-indicated regions are shown in ( B-1 ) and ( B-2 ) . ( C–D ) Ankyrin B knockout neurons were infected with βII spectrin-shRNA expressing adenovirus , immunostained for adducin and MAP2 , and imaged by conventional ( C ) and 3D STORM ( D ) microscopy . Infected neurons were marked by GFP signal expressed from the virus ( not shown ) . Magnified STORM images of adducin from arrow-indicated region are shown in ( D-1 ) and ( D-2 ) . ( E ) Average autocorrelation analysis of adducin distribution in dendrites of ankyrin B knockout neurons ( n = 9 neurons , four independent experiments ) . ( F ) Average autocorrelation analysis of adducin distribution in dendrites of ankyrin B knockout neurons treated with βII spectrin-shRNA expressing adenovirus ( n = 15 neurons , three independent experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 022 In addition to the induction of the periodic lattice structure in dendrites , we noticed that ankyrin B knockout also induced a dramatic redistribution of βII spectrin in neurites . In wild-type neurons , the local concentration of βII spectrin , as indicated by immunofluorescence intensity , was ∼twofold higher in axons than that in dendrites ( Figure 7C , D ) , consistent with previous results ( Riederer et al . , 1986; Galiano et al . , 2012 ) . However , the expression level of βII spectrin was substantially increased in dendrites by the ankyrin B knockout to a point that the local concentration of βII spectrin in dendrites became indistinguishable from that in axons , and both were comparable to the βII spectrin concentration observed in wild-type axons ( Figure 7C , D ) . We thus hypothesized that the increased local concentration of βII spectrin caused the formation of this periodic lattice structure in dendrites . To test this hypothesis , we increased the expression level of βII spectrin in all neurites by transiently transfecting neurons with a HA-tagged βII spectrin construct and performed STORM imaging on βII spectrin in transfected neurons at DIV 11 . As expected , the local concentration of βII spectrin in the dendrites of βII spectrin-HA expressing neurons was higher than that observed in control neurons that did not express βII spectrin-HA ( Figure 8A–C ) . Remarkably , whereas βII spectrin appeared mostly irregular in the dendrites of control neurons ( Figure 8D , E ) , in βII spectrin-HA overexpressing neurons , βII spectrin displayed a periodic pattern in nearly all dendritic processes ( Figure 8F , G ) . Autocorrelation analysis showed that the periodicity in the dendrites of overexpressing neurons was substantially enhanced compared to the dendrites of control neurons ( Figure 8H ) . 10 . 7554/eLife . 04581 . 023Figure 8 . Local βII spectrin concentration determines the formation of the periodic lattice structure . DIV 9 neurons were either mock-transfected or transfected with βII spectrin-HA and immunostained for HA and βII spectrin . βII spectrin were subsequently imaged by 3D STORM . ( A–B ) Conventional images of βII spectrin in dendrites of a control neuron and a βII spectrin-HA overexpressing ( OE ) neuron . The HA image is shown in the insets . ( C ) The relative fluorescence intensity for βII spectrin in dendrites and axons of control and βII spectrin-HA overexpressing neurons ( n = 10 neurons for control and n = 15 neurons for βII spectrin-HA overexpressing conditions; three independent experiments for each condition ) . ( D–E ) STORM images of βII spectrin from green-boxed regions in A . ( F–G ) STORM images of βII spectrin from green-boxed regions in B . ( F-1 ) and ( G-1 ) are enlarged images of the white-boxed regions in F and G , respectively . ( H ) Average autocorrelation analyses of βII spectrin in dendrites of control and βII spectrin-HA overexpressing neurons ( n = 8 neurons for control and n = 12 neurons for βII spectrin-HA overexpressing conditions; three independent experiments for each condition ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04581 . 023 Taken together , these data suggest that the local concentration of βII spectrin is a key determining factor for the formation of the periodic membrane skeleton in axons and that ankyrin B was critical for setting the polarized distribution of βII spectrin in axons and dendrites .
Actin , spectrin , and associated molecules form a periodic lattice structure with long-range order underneath the axonal membrane . Many molecular components , including actin , βII spectrin , adducin , ankyrin B , βIV spectrin , ankyrin G , and sodium channels , are present in this structure . We have observed this periodic membrane skeleton with STORM imaging of fixed cultured neurons ( Figure 1 and Figure 1—figure supplements 1–3 ) and fixed brain tissue slices ( Xu et al . , 2013 ) using actin-binding phalloidin and immunolabeling of endogenously expressed proteins , as well as in live cultured neurons using fluorescent fusion proteins ( Figure 5 ) . Recently , this periodic structure has also been observed in live neurons using a cell-permeable actin-binding dye and STED imaging ( Lukinavicius et al . , 2014 ) . Since a single actin filament can interact with multiple spectrin tetramers , and a single spectrin tetramer can bind to two actin filaments , one at each end of the symmetric tetramer ( Bennett and Lorenzo , 2013 ) , we reason that these crosslinking interactions are responsible for the formation of the lattice structure . Indeed , depolymerizing actin filaments disrupted the periodic distribution of βII spectrin , and knocking-down of βII spectrin disrupted the periodic distribution of actin ( Figure 1—figure supplements 2–3 ) . Because this lattice structure is associated with the axonal membrane , proper preservation of the membrane structure is essential for observing this structure . For example , a recent electron microscopy study of the AIS in neurons that has been subjected to detergent extraction of membrane before fixation did not show such periodic membrane skeleton ( Jones et al . , 2014 ) . Indeed when we applied the same ‘fixation after membrane extraction’ protocol to neurons , the structure was destroyed and not observed in STORM images ( Figure 1—figure supplement 6 ) . In this study , we investigated the developmental mechanism of this newly discovered axonal membrane skeleton . We found that this periodic membrane skeleton started to form early during axon development . In stage 3 neurons at DIV 2 , when one neurite just broke the symmetry and became an axon ( typically several times longer than other neurites ) , the periodic pattern of βII spectrin already emerged ( Figure 1 ) . It originated in the proximal axon regions near the cell bodies and gradually propagated to the distal ends of axons ( Figure 1 ) . This spatial distribution is in contrast to most previously identified signaling molecules involved in axon differentiation and development , which are enriched and function at the growing tip of axons ( Arimura and Kaibuchi , 2007; Barnes and Polleux , 2009; Stiess and Bradke , 2011; Cheng and Poo , 2012 ) . After its initial appearance , the lattice structure continued to mature with the actin filaments becoming more stable , potentially because of capping by adducin ( Figures 2 , 3 ) . Once matured , the structure appeared highly stable with little movement and extremely slow turnover of its molecular components was observed in live neurons ( Figure 5 ) . This highly stable membrane skeleton may function to provide a stable mechanical support for axons . Indeed , deletion of β spectrin from C elegans causes axons to break when the animals move ( Hammarlund et al . , 2007 ) . The emergence of this periodic lattice during early axon development and its origination in the proximal axon region near the cell body suggest that the periodic membrane skeleton may function as an independent mechanism for establishing or maintaining neuronal polarization in addition to the previously identified pathways that function at the distal ends of axons . However , the periodic membrane skeleton is not required for the initiation of axon differentiation because the periodic structure only started to form in stage 3 , but not stage 2 , a stage at which most signaling molecules for axon initiation exhibit high activity ( Arimura and Kaibuchi , 2007; Cheng and Poo , 2012 ) . Moreover , neurons depleted of βII spectrin are also capable of forming long axons ( Galiano et al . , 2012 ) , though these axons may not be fully functional . Indeed , removal of αII or βII spectrin is embryonically lethal in mice ( Tang et al . , 2003; Stankewich et al . , 2011 ) and AIS , the structure important for action potential generation , fails to assemble properly in neurons that are depleted of βII spectrin ( Galiano et al . , 2012 ) . Interestingly , molecules important for the specification of AIS , ankyrin G , βIV spectrin , and sodium channels were all incorporated into this periodic membrane skeleton . During the early stages of axon development , the expression levels of ankyrin G and βIV spectrin were low in the proximal axon region , which was instead occupied by the periodic lattice comprising βII spectrin ( Figure 1 ) . Ankyrin G and βIV spectrin began to enrich in the proximal axon region later during axon development , at ∼DIV 8 ( Figure 4—figure supplement 1 ) , consistent with previous observations ( Galiano et al . , 2012 ) . The incorporation of ankyrin G and βIV spectrin into the periodic lattice was observed even later , at around DIV 12 , replacing βII spectrin from the structure in the AIS ( Figure 4 ) . The enrichment and periodic assembly of the AIS components appeared to depend on βII spectrin ( Figure 4—figure supplement 2 ) . Our data suggest the following developmental course for the AIS formation . Before the AIS is formed , actin and βII spectrin form a cohesive periodic lattice structure that covers the entire axonal shaft including the proximal axonal region . The AIS then forms by the enrichment of ankyrin G and βIV spectrin in the proximal axonal region and the replacement of βII spectrin by βIV spectrin in the periodic membrane skeleton . Potentially , ankyrin G , the master regulator of AIS that recruits other AIS components ( Zhou et al . , 1998; Jenkins and Bennett , 2001; Yang et al . , 2007 ) , is first enriched in the proximal axon region . Ankyrin G then recruits βIV spectrin to the same region and causes it to be incorporated in the periodic lattice . βIV spectrin in turn anchors ankyrin G into a periodic pattern as well . As an adaptor , ankyrin G then places sodium channels into a periodic distribution pattern in the AIS . Finally , we addressed the question why the cohesive , periodic lattice structure preferentially formed in axons , with dendrites showed at most isolated patches of periodic structure with much less regularity ( Figure 1—figure supplement 5 ) . We found multiple molecular factors participate in this regulation . The periodic lattice structure depended on intact microtubules . Treatment with a microtubule-depolymerizing drug disrupts the structure in axons , whereas treatment with microtubule-stabilizing drugs induces the formation of the lattice structure in multiple neurites ( Figure 6 ) . Importantly , we found that the local concentration of βII spectrin is a determining factor for the formation of the lattice structure . The local concentration of βII spectrin is ∼2 times higher in axons than in dendrites ( Figure 7 ) . Remarkably , increasing the dendritic concentration of βII spectrin by overexpression induced the formation of the periodic lattice structure in all dendrites ( Figure 8 ) . Interestingly , ankyrin B was critical for maintaining the polarized distribution of βII spectrin . Knocking out ankyrin B led to an even distribution of βII spectrin in dendrites and axons and the formation of a highly regular and cohesive periodic lattice structure in all dendrites ( Figure 7 ) . Consistent with the notion that the increased concentration of βII spectrin in dendrites is responsible for inducing the formation of the periodic lattice structure in dendrites , knocking-down βII spectrin from ankyrin B knockout neurons disrupted the lattice structure ( Figure 7—figure supplement 2 ) . These results indicate that ankyrin B is critical for establishing a polarized distribution of βII spectrin in neurites with a higher concentration of βII spectrin in axons than in dendrites , which in turn promotes the formation of the periodic membrane skeleton in axons . It is interesting to speculate how ankyrin B may establish such a polarized distribution of βII spectrin . The predominant form of ankyrin B during early neuronal development is a 440-kDa splice variant that is preferentially targeted to axons ( Kunimoto et al . , 1991; Chan et al . , 1993; Kunimoto , 1995 ) . Given that ankyrin B specifically binds to βII spectrin , we speculate that the distribution of ankyrin B in axons may help establishing the enrichment of βII spectrin in axons during early neuronal development . Moreover , we recently found that ankyrin B is a major cargo adaptor for dynactin and promotes axonal transport of proteins and organelles and that disruption of ankyrin B–dynactin interaction significantly impairs the axonal transport of many proteins ( Lorenzo et al . , 2014 ) . It is thus possible that ankyrin B may also preferentially transport βII spectrin into axons instead of dendrites . By maintaining a polarized distribution of βII spectrin in neurons , ankyrin B functions as a negative regulator for preventing the formation of the periodic membrane skeleton in dendrites . The exact mechanism by which ankyrin B maintains a polarized distribution of βII spectrin remains an interesting question for future investigation .
Primary hippocampal cultures were prepared from wild-type neonatal ( E18 ) rat embryos ( timed pregnant SD rats from Charles River Laboratories , Wilmington , MA ) or ankyrin B knockout mice as reported previously ( Scotland et al . , 1998 ) . Hippocampi were isolated and digested with 0 . 05% trypsin–EDTA ( 1× ) ( Invitrogen 25300-054 , Grand Island , NY ) at 37°C for 15 min . The hippocampi were transferred to the Hib A solution ( BrainBits HA-Ca , Springfield , IL ) , washed several times with the Hib A solution , and pipetted up and down until the tissues were mostly dissolved . The solution was then passed through a cell strainer ( VWR 21008-949 , Philadelphia , PA ) to remove the residual undissociated tissue and collected in a 50 ml conical tube . Neurons were spun down to the bottom of the tube , resuspended with the culture media made of 96 ml Neurobasal ( Life Technologies 12349-015 ) , 2 ml B-27 Supplement ( Life Technologies 17504-044 , Grand Island , NY ) , 1 ml Penicillin-Streptomycin ( Life technologies 15140-122 ) and 1 ml Glutamax ( Life technologies 35050-061 ) , and then plated onto poly-L-lysine/laminin-coated 12-mm coverslips ( BD bioscience BD354087 , San Jose , CA ) or poly-L-lysine coated 8-well chambers . 5 μM cytosine-D-arabinofuranoside ( Sigma C1768 , St . Louis , MO ) was added to the culture media to inhibit the growth of glial cells 3 days after plating . The neurons were fed twice a week with freshly made culture media until use . The following primary antibodies were used in this study: guinea pig anti-Map2 antibody ( Synaptic Systems , 188002 , Goettingen , Germany ) ; mouse anti-βII spectrin antibody ( BD Biosciences , 612563 ) ; mouse anti-ankyrin G antibody ( Santa Cruz , Sc-12719 , Dallas , Texas; epitope mapping the spectrin-binding domain of ankyrin G near the N-terminal ) ; goat anti-ankyrin G antibody ( Santa Cruz , Sc-31778 , epitope mapping the C-terminal of ankyrin G ) ; rabbit anti-adducin antibody ( Abcam , ab51130 , Cambridge , MA ) ; rabbit anti-HA antibody ( Abcam , ab9110 ) ; rabbit anti-GFP antibody ( Abcam , ab290 ) ; goat anti-βIII spectrin ( Santa Cruz , sc-9660 ) . Rabbit antibodies targeting the C- or N-terminus of βIV spectrin were kind gifts from Dr Matt Rasband at Baylor College of Medicine . The following secondary antibodies were used in this study for conventional imaging: Alexa Fluor 647 donkey anti-mouse ( Invitrogen , A31571 ) , Alexa Fluor 555 donkey anti-mouse ( Invitrogen , A31570 ) , Alexa Fluor 488 donkey anti-mouse ( Invitrogen , A21202 ) , Alexa Fluor 647 donkey anti-rabbit ( Invitrogen , A31573 ) , Alexa Fluor 568 donkey anti-rabbit ( Invitrogen , A10042 ) , Alexa Fluor 488 donkey anti-rabbit ( Invitrogen , A21206 ) , Alexa Fluor 488 goat anti-guinea pig ( Invitrogen , A11073 ) , Alexa Fluor 647 donkey anti-goat ( Invitrogen , A21447 ) , Alexa Fluor 568 donkey anti-goat ( Invitrogen , A11057 ) . For STORM imaging , secondary antibodies were custom-labeled with a photoswitchable reporter dye , Alexa Fluor 647 , and an activator dye Alexa Flour 405 , which facilitates the photoswitching of the reporter dye . Donkey anti-mouse and donkey anti-rabbit secondary antibodies ( Jackson ImmunoResearch , West Grove , PA ) were each labeled with a mixture of amine-reactive activator and reporter dyes in a one-step reaction , as described previously ( He et al . , 2013 ) . In some experiments , commercial Alexa Fluor 647 conjugated donkey anti-mouse , anti-rabbit or anti-goat secondary antibodies were used . βII spectrin-GFP and βII spectrin-HA plasmids ( addgene , 31070 , Cambridge , MA ) used in this study were reported previously , with GFP inserted at the N-terminal of βII spectrin and the HA tag at the C-terminal of βII spectrin , respectively ( Galiano et al . , 2012 ) . To make the βII spectrin-mMaple 3 construct , we replaced the HA tag sequence with the mMaple 3 sequence ( Wang et al . , 2014 ) . Plasmids were transfected into neurons using a calcium phosphate transfection kit from Invitrogen ( K2780-01 ) . The protocol for transfection was modified slightly for our neuronal cultures . Briefly , neurons were plated at a density of 40 , 000 cells/well in 12-well plates and cultured for 6–10 days before the transfection . After changing media from original neuronal culture media to Minimum Essential Media ( MEM , Life Technology , supplemented with 20 mM HEPEs , pH 7 . 15 ) , 100 µl plasmid mixture was added , and neurons were incubated at 37°C for 20 min . The media was subsequently aspirated and replaced with a lower pH MEM ( supplemented with 20 mM HEPEs , pH 6 . 8 ) at 37°C for 4 min . After dissolving all calcium phosphate crystals , we added back the original neuronal culture media . Experiments were performed 2 or 3 days after transfection . The βII spectrin-shRNA adenoviral construct used in this study was a kind gift from Dr Matt Rasband at Baylor College of Medicine and described previously ( Hedstrom et al . , 2008 ) . The two sense sequences of shRNA are: 5′-GCATGTCACGATGTTACAA-3′ and 5′-GGATGAAATGAAGGTGCTA-3′ . For assaying the effect of βII spectrin knockdown on actin and adducin structure , neurons were infected with the virus at DIV 3 and fixed for STORM imaging at around DIV 9 or 10 . For assaying the effect of βII spectrin knockdown on βIV spectrin structure , neurons were infected with the virus at DIV 3 or DIV 7 and fixed at DIV 12 for STORM imaging of βIV spectrin . Infected neurons were marked by a GFP signal expressed from the adenoviral construct . The knockdown efficiency was validated through immunostaining against βII spectrin . The following chemicals were used in this study with their concentration and treatment time stated: latrunculin A ( Sigma , L5163 , 20 µM , 1 hr or indicated time series ) , cytochalasin D ( Sigma , C8273 , 50 µM , 1 hr ) , nocodazole ( Sigma , M1404 , 50 µM , 1 hr ) , taxol ( Sigma , T7402 , 5 nM ) , SB-216763 ( Sigma , S3442 , 5 µM ) . For taxol treatment , neurons were treated with the drug at DIV 3 with the indicated concentration and fixed at DIV 6 for STORM imaging . For SB216763 treatment , neurons were treated with the drug at DIV 1 with the indicated concentration and fixed at DIV 5 for STORM imaging . Cultured neurons were fixed at various days in vitro ( DIV ) . For imaging of actin , we used a similar method to label actin as reported previously ( Koestler et al . , 2008; Xu et al . , 2013 ) . Briefly , the samples were simultaneously fixed and extracted for 1 min using a solution of 0 . 3% ( vol/vol ) glutaraldehyde ( GA ) and 0 . 25% ( vol/vol ) Triton X-100 in cytoskeleton buffer ( CB , 10 mM MES , pH 6 . 1 , 150 mM NaCl , 5 mM EGTA , 5 mM glucose , and 5 mM MgCl2 ) and then post-fixed for 15 min in 2% ( vol/vol ) GA in CB , a previously established protocol for maintaining actin ultrastructure ( Koestler et al . , 2008; Xu et al . , 2013 ) . The GA-fixed samples were treated with freshly prepared 0 . 1% ( wt/vol ) sodium borohydride for 7 min to reduce background fluorescence caused by GA fixation . To label actin filaments , samples were labeled with Alexa Fluor 647 conjugated phalloidin ( Invitrogen A22287 ) overnight at 4°C or ∼1 hr at room temperature . A concentration of ∼0 . 5 µM phalloidin in PBS was used . To minimize the dissociation of phalloidin from actin during washing steps , actin labeling was performed after all other labeling steps ( i . e . , immunofluorescence of other molecular targets ) were completed . The sample was washed 2–3 times with PBS and then immediately mounted for imaging . To test whether strong membrane extraction prior to fixation ( Jones et al . , 2014 ) disrupts the membrane skeleton structure , neurons were extracted with 1% Triton X-100 in PEM buffer ( 100 mM Pipes–KOH , pH 6 . 9 , 1 mM MgCl2 , and 1 mM EGTA ) containing 2% polyethylene glycol , 2 µM phalloidin , and 2 µM taxol for 3 min at room temperature after a quick rinse with the PEM buffer containing 2 µM taxol and subsequently fixed with 0 . 2% GA in PBS for at least 20 min , as described previously ( Jones et al . , 2014 ) . Fixed samples were treated with freshly prepared 0 . 2% ( wt/vol ) sodium borohydride for 5–10 min to reduce background fluorescence caused by GA fixation and washed in PBS . Actin labeling was performed similarly as described above . For imaging of molecular components not including actin ( MAP2 , βII spectrin , βIII spectrin , βIV spectrin , ankyrin G , and adducin ) , the samples were fixed using 4% ( wt/vol ) paraformaldehyde in phosphate buffered saline ( PBS ) for 15 min . Fixed neuron samples were then permeabilized and blocked in blocking buffer ( 3% wt/vol bovine serum albumin or 10% wt/vol donkey serum , 0 . 2% vol/vol Triton X-100 in PBS ) for 1 hr and subsequently stained with primary antibodies in blocking buffer overnight at 4°C . The samples were washed three times and then stained with secondary antibodies ( described above ) in blocking buffer for ∼1 hr at room temperature . The imaging buffer was PBS containing 100 mM cysteamine , 5% glucose , 0 . 8 mg/ml glucose oxidase ( Sigma-Aldrich ) , and 40 μg/ml catalase ( Roche Applied Science , Indianapolis , IN ) for fixed neurons . To image the samples from 12-mm coverslips , approximately 4 μl of imaging buffer was dropped at the center of a freshly-cleaned #1 . 5 rectangular coverslip ( 22 mm by 60 mm ) , and the sample on the 12-mm coverslip was mounted on the rectangular coverslip and sealed with nail polish or Cytoseal . To image samples from 8-well chambers , 400 μl of imaging buffer was added to the imaging chamber . The STORM setup was based on an Olympus IX-71 inverted optical microscope as described previously ( Jones et al . , 2011 ) . 405-nm ( CUBE 405-50C; Coherent , Santa Clara , CA ) , 460-nm ( Sapphire 460-10; Coherent ) , 532-nm ( GCL-200-I; CrystaLaser , Reno , NV ) , and 657-nm ( RCL-300-656; CrystaLaser ) lasers were introduced into the sample through the back focal plane of the microscope . A translation stage allowed the laser beams to be shifted towards the edge of the objective so that the emerging light reached the sample at incidence angles slightly smaller than the critical angle of the glass–water interface , thus illuminating only the fluorophores within a few micrometers of the coverslip surface . A T660LPXR ( Chroma , Bellows Falls , VT ) was used as the dichroic mirror and an ET705/72M band-pass filter ( Chroma ) was used as the emission filter . For 3-dimensional ( 3D ) STORM imaging , a cylindrical lens was inserted into the imaging path so that images of single molecules were elongated in x and y for molecules on the proximal and distal sides of the focal plane ( relative to the objective ) , respectively ( Huang et al . , 2008 ) . During imaging , continuous illumination of 657-nm laser ( ∼2 kW/cm2 ) was used to excite fluorescence from Alexa Flour 647 molecules and switched them into the dark state . Continuous illumination of the 405-nm laser ( when Alexa Flour 405 was used as the activator dye ) or 532-nm laser ( when Cy3 was used as the activator dye ) was used to reactivate the fluorophores to the emitting state . The power of the activation lasers ( typical range 0–1 W/cm2 ) was adjusted during image acquisition so that at any given instant , only a small , optically resolvable fraction of the fluorophores in the sample was in the emitting state . A typical STORM image was generated from a sequence of about 30 , 000–60 , 000 image frames at a frame rate of 60 Hz . The recorded STORM movie was analyzed according to previously described methods ( Rust et al . , 2006; Huang et al . , 2008 ) . The centroid positions and ellipticities of the single molecule images provided lateral and axial positions of each activated fluorescent molecule , respectively ( Huang et al . , 2008 ) . Super-resolution images were reconstructed from the molecular coordinates by depicting each location as a 2D Gaussian peak . Live-cell STORM experiments were performed on the same STORM setup as described earlier ( Jones et al . , 2011; Shim et al . , 2012 ) . Neurons were initially transfected with βII spectrin-mMaple 3 at DIV 8 and imaged in an extracellular solution containing: 128 mM NaCl , 5 mM KCl , 2 mM CaCl2 , 1 mM MgCl2 , 25 mM HEPES , 30 mM glucose , pH 7 . 3 at DIV 10 or DIV 11 . Continuous illumination of the 405-nm laser was used to activate the mMaple 3 fluorescent protein . Continuous illumination of 561-nm laser was used to excite mMaple 3 and switched them to the dark state . Imaging analysis was performed as described above . FRAP experiments were performed on the same STORM setup as described above . Neurons were transfected with βII spectrin-GFP at DIV 8 and then imaged at DIV 10 . Directly before FRAP experiments , neuronal culture media were replaced with an extracellular solution . After recording an image before photo-bleaching , a small region of the sample was bleached by shrinking the size of iris at the excitation light path for 10 s with the maximum laser power . Subsequently , the sample was imaged using the same power as that of pre-bleached image . The image was recorded at a frequency of 1 Hz . We used the unbleached regions in the image to calibrate the photo-bleaching effect during the entire recording time . The fluorescence recovery fraction was measured as the fluorescence intensity at the bleached region at indicated time vs the original intensity before photo-bleaching . | The brain contains hundred types of neurons , but they are all variations on the same basic structure . Each neuron consists of a cell body that is covered in short protrusions called dendrites and a long thin structure called the axon . The dendrites receive incoming signals from neighboring neurons and they transmit these signals via the cell body to the axon , which in turn relays them to the dendrites of the next neuron ( or neurons ) . Like all cells , neurons maintain their structure with the help of an internal cytoskeleton made up of many different proteins . However , it was discovered recently that axons have an additional lattice-like structure underneath their outer membrane . This structure , which consists of rings of actin filaments separated by molecules of a protein called spectrin , is preferentially formed in axons and is found much less frequently in dendrites . Now Zhong , He et al . , who are members of the research group that discovered the axonal skeleton , have used ‘super-resolution imaging’ to figure out how this skeleton forms and why it predominantly forms in axons . In brief , a basic version of the sub-membrane periodic skeleton is laid down early in development , starting next to the cell body before gradually spreading down the axon . The skeleton then continues to mature throughout development with the incorporation of several additional types of proteins . The periodic skeleton only forms in regions which contain enough βII spectrin . Under normal conditions , dendrites contain too little βII spectrin to support the growth of such a periodic skeleton . However , artificially increasing the amount of βII spectrin present by overexpressing the corresponding gene , or by knocking out ankyrin B ( a molecule that is important for establishing the preferential distribution of βII spectrin in axons ) , is sufficient to trigger periodic skeleton formation in dendrites . Given that axons and dendrites have distinct roles in neuronal signaling , this uneven distribution of spectrin is likely to be one way in which these regions maintain the specific structures that support their individual functions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2014 | Developmental mechanism of the periodic membrane skeleton in axons |
Circadian clocks serve as internal pacemakers that influence many basic homeostatic processes; consequently , the expression and function of their components are tightly regulated by intricate networks of feedback loops that fine-tune circadian processes . Our knowledge of these components and pathways is far from exhaustive . In recent decades , the nuclear envelope has emerged as a global gene regulatory machine , although its role in circadian regulation has not been explored . We report that transcription of the core clock component BMAL1 is positively modulated by the inner nuclear membrane protein MAN1 , which directly binds the BMAL1 promoter and enhances its transcription . Our results establish a novel connection between the nuclear periphery and circadian rhythmicity , therefore bridging two global regulatory systems that modulate all aspects of bodily functions .
Most organisms , ranging from cyanobacteria to humans , are governed by their circadian rhythms: endogenous and self-sustained oscillations with a period of roughly 24 hr , which manifest in diverse metabolic , physiological , and behavioral processes ( Ueda et al . , 2005 ) . This internal pacemaker is charged with two important roles: to perpetuate its own rhythms and to regulate the expression of genes that are under circadian control . In mammals , this internal pacemaker consists of a complex network of transcriptional regulations , at the core of which is transcription activators BMAL1 ( also known as ARNTL1 or MOP3 ) and CLOCK , which form heterodimers and regulate gene expression . Up to 15% of the organism's genome is regulated in a circadian manner ( Panda et al . , 2002; Emery and Reppert , 2004; Zhang and Kay , 2010 ) . Well-studied examples include the transcription repressors ( PERIODs and CRYPTOCRHOMEs ) that bind to CLOCK/BMAL1 and suppress their own transcription , thereby forming a feedback loop . Since the identification and cloning of the first mammalian clock gene , CLOCK , two decades ago ( Vitaterna et al . , 1994 ) , the field of chronobiology has uncovered many additional players that regulate circadian rhythms on a transcriptional and/or post-transcriptional level , and many more such candidates are currently being evaluated . Recently , mutations in nuclear envelope ( NE ) proteins have been shown to cause a surprisingly broad range of inherited diseases , thus shedding light on roles played by the NE in global regulations at cellular and organismal levels ( Padiath et al . , 2006; Dauer and Worman , 2009 ) . These diseases ( often referred to as nuclear envelopathies or laminopathies ) can impact muscle , nerve , fat metabolism , bone formation , and others . NE consists of outer and inner nuclear membranes ( connected by nuclear pore complexes ) and nuclear lamina . The inner nuclear membrane proteins ( such as MAN1 , LBR , LAP2 , etc ) include approximately 60 putative transmembrane proteins specifically retained in the inner nuclear membrane and most of them are poorly characterized ( Schirmer et al . , 2005 ) . In metazoan , nuclear lamina is a protein mesh-like structure composed of type V intermediate filament proteins lamins ( including A , C , B1 , B2 , and B3 types ) and sits primarily underneath the inner nuclear membrane ( Zwerger and Medalia , 2013 ) . The idea of nuclear envelope components as transcription regulators in mammals is relatively new , conceived from the observation that gene-rich chromosomes are generally located in more internal nuclear regions , whereas gene-poor chromosomes are relegated to the periphery ( Spector , 2003 ) . Many NE components such as inner nuclear membrane proteins , nuclear lamina , and the nuclear pore complex , harbor DNA-binding domains that are involved in anchoring chromatin to the periphery ( Ulbert et al . , 2006; Mekhail and Moazed , 2010 ) . Functional relevance of these positional distinctions became apparent as studies with yeast and flies revealed that the NE can sequester factors that affect gene transcription in both repressive and , surprisingly , activating manners ( Akhtar and Gasser , 2007 ) . Although recent findings highlight the important functions of the nuclear periphery , its relationship with the circadian clock has not been probed . Given the increasing awareness of the global roles that these two systems play in myriad pathways , we set out to investigate the possibility that these seemingly separate pathways are connected and can work synergistically in regulating diverse functions .
In order to investigate whether NE proteins are involved in circadian regulation , we began by focusing on lamin B1 since it has been shown to play a role in transcriptional regulation ( Hutchison , 2002; Shevelyov et al . , 2009 ) . In vivo oscillation of lamin B1 ( Lmnb1 ) expression patterns ( both RNA and protein levels ) were confirmed using mouse tissues from suprachiasmatic nuclei ( SCN ) , kidney , and liver ( Figure 1A , B ) . To test whether the level of lamin B1 affects the molecular clock , we examined the protein expression patterns for the core clock gene PERIOD2 ( PER2 ) using Lmnb1 heterozygous knock out ( homozygosity is lethal ) and LMNB1 wild-type BAC transgenic mice ( Vergnes et al . , 2004; Heng et al . , 2013 ) . Oscillating PER2 expression patterns were phase delayed in Lmnb1 heterozygous knock out mice and phase advanced in LMNB1 BAC transgenic mice ( overexpression ) when compared to wild-type control mice ( Figure 1C ) , suggesting that the level of lamin B1 can modulate circadian clock . However , neither Lmnb1 heterozygous knock out mice nor LMNB1 BAC transgenic mice demonstrated significant output behavioral change ( Figure 1—figure supplement 1 ) . To expand the investigation , we chose to include two additional NE proteins that are known to associate with lamin B1 , LBR , and MAN1 . We found that LBR and MAN1 expression also oscillate , albeit mildly for MAN1 ( Figure 1D ) . 10 . 7554/eLife . 02981 . 003Figure 1 . Lamin b1 regulates the circadian clock . Expression levels of lamin b1 from SCN , kidney , and liver extracts in C57BL/6J mice ( A and B ) . ( A ) mRNA levels of Lmnb1 and Gapdh were assayed at indicated circadian times ( CT , n = 4 ) . ( B ) Representative immunoblots show the levels of LMNB1 , GAPDH , and β-ACTIN . ( C ) Representative immunoblots show PER2 ( with intensity values indicated at the bottom ) and LMNB1 abundance in Lmnb1+/Δ , wild-type and LMNB1BAC liver extracts . ( D ) Expression patterns of LMNB1 , LBR , and MAN1 in C57BL/6 mouse livers at indicated Zeitgeber times ( ZT ) ( n = 3 ) . Quantifications ( top panels ) of Western blots ( bottom panel ) were obtained by using GAPDH as a loading control . Data represent means ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 00310 . 7554/eLife . 02981 . 004Figure 1—figure supplement 1 . Mice with altered LMNB1 levels do not exhibit altered behavioral rhythms . The period ( A ) , activity level ( B ) , and amplitude ( C ) of wheel-running rhythms under DD condition ( n = 9–27 , Student's t test , *p < 0 . 05 ) . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 004 To determine if these NE genes passively receive cues from the core clock apparatus or if their protein products also actively play a role in maintaining circadian rhythms , we altered their protein levels in human osteosarcoma U2OS cells that express a luciferase reporter gene under the control of mouse Bmal1 promoter ( Bmal1-Luc ) and examined circadian period in cell culture ( Vollmers et al . , 2008 ) . siRNA-induced reduction of LMNB1 , LBR , and MAN1 in this cell-based system resulted in a longer circadian period ( τ ) ( Figure 2A ) , whereas the over-expression of all three led to a shorter τ ( Figure 2B ) . Cells transfected with LBR , LMNB1 , and MAN1 siRNA lengthened τ by 54–69 min ( n ≥ 4 , *p < 0 . 05 ) , when compared with control siRNA ( τ = 27 . 39 ± 0 . 22 hr , n = 8 ) ( Figure 2C ) . On the other hand , overexpression of FLAG-tagged LBR , LMNB1 , or MAN1 shortened τ by 25 . 8–37 . 8 min ( n ≥ 4 , *p < 0 . 05 ) compared to empty vector controls ( τ = 27 . 7 ± 0 . 15 , Figure 2D ) . These changes in τ together with the altered phase of Lmnb1 heterozygous knock out mice and over-expressing mice suggest that these NE proteins participate in modulating circadian clock and therefore could impose significant impacts on downstream biological pathways . 10 . 7554/eLife . 02981 . 005Figure 2 . LBR , LMNB1 , and MAN1 are necessary for normal circadian rhythms . Two representative traces of real-time bioluminescence analyses are shown for each , and Western blot verification of down-regulation or over-expression is demonstrated in the inset images . ( A ) Period was lengthened when LBR , LMNB1 , or MAN1 was knocked down . ( B ) Over-expression of FLAG-tagged LBR ( F-LBR ) , LMNB1 ( F-LMNB1 ) , or MAN1 ( F-MAN1 ) shortened period compared to cells transfected with empty vector ( ctrl ) . ( C and D ) Summary of period in ( A and B ) . CRYPTOCRHOME2 ( CRY2 ) and PER2 siRNA knockdowns served as positive controls . Data represent means ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 00510 . 7554/eLife . 02981 . 006Figure 2—figure supplement 1 . Over-expressing or knocking down nuclear envelope components alters circadian rhythms in flies . ( A ) Normalized locomotor activity profiles of flies over-expressing MAN1 , Lam , or LBR during LD for 1 day followed by 4 days of DD . ( B ) Normalized locomotor activity profiles of flies with MAN1 , Lam , or LBR knocked down by RNAi . White box indicates light period , black box indicates dark period , and gray box indicates subjective light period . Error bars represent SEM ( n = 13–76 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 00610 . 7554/eLife . 02981 . 007Figure 2—figure supplement 2 . The mRNA levels of nuclear envelope genes are reduced in the corresponding knockdown flies . Plots of relative mRNA abundance for MAN1 , Lam , and LBR from whole head extracts of timGAL4/+;UASdicer2/+ , timGAL4/UASMAN1RNAi;UASdicer2/+ , timGAL4/+;UASdicer2/UASLamRNAi and timGAL4/UASLBRRNAi;UASdicer2/+ flies determined by qRT-PCR . Error bars represent SEM ( n = 3–6 ) . Significant differences indicated by asterisks ( Student's t test , *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 ) . The value of timGAL4/+;UASdicer2/+ for one experiment was set to 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 007 The effects of NE proteins on Drosophila circadian clock were also examined . Consistent with the mammalian data , over-expressing dLamin ( dLam ) ( Padiath et al . , 2006 ) in circadian neurons with cryptochrome ( cry ) GAL4-39 and cryGAL4-16 ( Emery et al . , 2000 ) in vivo resulted in substantially shortened periods of behavioral rhythms in constant darkness compared to GAL4 controls ( but not to UASLMNB1/+ , Table 1; Figure 2—figure supplement 1A ) . Knocking down dLam in circadian neurons lengthened the period ( Table 2; Figure 2—figure supplement 1B ) . On the other hand , over-expressing dMAN1 and dLBR lengthened the period ( Table 1; Figure 2—figure supplement 1A ) , while knocking down dMAN1 also lengthened period ( Table 2; Figure 2—figure supplement 1B ) . Besides altering the period , most of these manipulations reduced the amplitude of behavioral rhythms as indicated by the reduced power values . In addition , we have assessed the mRNA levels of dMAN1 , dLam , and dLBR to confirm knockdown ( Figure 2—figure supplement 2 ) . Taken together , these results indicate that NE proteins also participate in the regulation of fly clock . 10 . 7554/eLife . 02981 . 008Table 1 . Over-expressing NE genes alters the behavioral period in fliesDOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 008GenotypePeriod ( hr ) PowerRhythmic%NUASMAN1/+23 . 8 ± 0 . 090 ± 59676cryGAL4-39/UASMAN126 . 4 ± 0 . 3* , †31 ± 6* , †6428UASMAN1/+; cryGAL4-16/+26 . 8 ± 0 . 2* , †40 ± 7* , †6632UASLam/+23 . 9 ± 0 . 087 ± 69355cryGAL4-39/UASLam23 . 8 ± 0 . 2†33 ± 7* , †7322UASLMNB1/+; cryGAL4-16/+24 . 4 ± 0 . 2†5 ± 2* , †2631UASLBR/+23 . 7 ± 0 . 059 ± 57967cryGAL4-39/UASLBR25 . 6 ± 0 . 1* , †44 ± 6†7933UASLBR/+; cryGAL4-16/+N/A1 ± 0* , †032cryGAL4-39/+24 . 8 ± 0 . 186 ± 59357cryGAL4-16/+25 . 6 ± 0 . 186 ± 59464*One-way ANOVA compared to UAS control lines , p < 0 . 001 . †One-way ANOVA compared to GAL4 control lines , p < 0 . 001 . 10 . 7554/eLife . 02981 . 009Table 2 . Knocking-down NE genes lengthens the behavioral period in fliesDOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 009GenotypePeriod ( hr ) Power%RhythmicNUASMAN1RNAi24 . 4 ± 0 . 1110 ± 1010014UASMAN1RNAi;cryGAL4-39/+; UASdcr2/+26 . 3 ± 0 . 3* , †29 ± 6* , †8015UASMAN1RNAi; UASdcr2/+;cryGAL4-16/+27 . 9 ± 0 . 4* , †39 ± 8* , †6916UASLamRNAi/+23 . 6 ± 0 . 1129 ± 1210016cryGAL4-39/+; UASLamRNAi/UASdcr225 . 1 ± 0 . 4*26 ± 17* , †5014UASLamRNAi/UASdcr2; cryGAL4-16/+26 . 8 ± 0 . 1* , †126 ± 1710013UASLBRRNAi/+23 . 5 ± 0 . 0112 ± 1410015cryGAL4-39/UASLBRRNAi; UASdcr2/+24 . 9 ± 0 . 161 ± 12*8015UASLBRRNAi/UASdcr2; cryGAL4-16/+24 . 6 ± 1 . 712 ± 4* , †4416cryGAL4-39/+; UASdcr2/+24 . 9 ± 0 . 169 ± 128816UASdcr2/+;cryGAL4-16/+26 . 1 ± 0 . 187 ± 109416*One-way ANOVA compared to UASRNAi control lines , p < 0 . 05 . †One-way ANOVA compared to control lines with GAL4 and UASdcr2 , p < 0 . 05 . dicer2 ( dcr2 ) is co-expressed to enhance the effects of RNAi . We next explored the relationship of LBR , LMNB1 , and MAN1 by examining mRNA and protein levels while knocking them down one at a time . Both LBR and LMNB1 knockdown significantly decreased the transcript level of MAN1 ( by 15% and 40% , respectively ) ( Figure 3A ) . The effects of LBR or LMNB1 knockdown on MAN1 expression are even more dramatic at the protein level , with 54% and 44% reductions , respectively ( Figure 3B ) . Moreover , knockdown of LBR expression reduces the amount of LMNB1 protein by 32% , which is consistent with the observation that reduction of LBR expression in the fibroblasts of patients harboring a heterozygous LBR mutation results in the abolition of LMNB1 protein ( Gaudy-Marqueste et al . , 2010 ) , whereas a decrease in LMNB1 does not significantly affect LBR expression . MAN1 knockdown also does not change the expression of LBR and LMNB1 , either at the mRNA or protein level ( Figure 3 ) . These results suggest that MAN1 is modulated by LBR and LMNB1 , and thus the effects of LBR and LMNB1 on the clock are at least partially through MAN1 . Therefore , we further investigated the effects of MAN1 on the molecular clock . 10 . 7554/eLife . 02981 . 010Figure 3 . Knocking down LBR/LMNB1 reduces MAN1 mRNA and protein levels but not vice versa . Assessing mRNA ( A ) and protein ( B ) levels of LBR , LMNB1 , and MAN1 while knocking them down one at a time in U2OS cells via RNAi . ( A ) mRNA levels of LBR , LMNB1 , and MAN1 in each of the three knockdown conditions were quantified using qRT-PCR ( n = 14 , *p < 0 . 05 ) . ( B ) MAN1 was significantly down-regulated when LBR or LMNB1 was knocked down ( n = 14 *p < 0 . 001 ) . The error bars represent SEM ( left panel ) . Representative immunoblots show the protein levels of LBR , LMNB1 and MAN1 ( right panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 010 A lengthened period due to decreased MAN1 may arise from altered regulation of core clock proteins and/or altered transcription of core clock genes . Either of these effects would result in disruptions of the stoichiometry and temporal control of the dynamics of the core circadian feedback loops . Given what is known regarding the role of NE proteins in transcriptional regulation , we tested whether reductions in MAN1 expression would affect the transcription of clock genes . We examined the circadian oscillation of core clock genes at the mRNA level and found that only BMAL1 showed a clear difference wherein overall mRNA levels were down-regulated to half the levels of controls ( Figure 4 ) . Western blots also showed lower expression of BMAL1 when MAN1 was knocked down ( Figure 4—figure supplement 1 ) . The non-oscillatory CLOCK showed no significant change of either transcript or protein levels ( Figure 4 , Figure 4—figure supplement 1 ) . The conserved mRNA levels of REV-ERBα and RORα in MAN1 knockdown cells suggest that the reduced BMAL1 expression is not caused by altered transcriptional activation of REV-ERBα , a BMAL1 repressor , or transcriptional repression of RORα , a BMAL1 activator . 10 . 7554/eLife . 02981 . 011Figure 4 . Knocking down MAN1 reduces the levels of BMAL1 mRNA . Each graph shows cells transfected with MAN1 siRNA vs ctrl siRNA . Time 0 indicates the moment that U2OS cells were treated with dexamethasone ( 100 nM ) . Data are presented as means ± SEM , n = 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 01110 . 7554/eLife . 02981 . 012Figure 4—figure supplement 1 . Knocking down MAN1 reduces BMAL1 protein levels . ( A ) Time course evaluations of U2OS cells transfected with MAN1 siRNA showed robust reduction of BMAL1 protein levels compared to ctrl siRNA . CLOCK levels were unaffected by MAN1 abundance ( n = 3 ) . ( B ) The quantification results of ( A ) were graphed ( n = 14 ) . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 01210 . 7554/eLife . 02981 . 013Figure 4—figure supplement 2 . Over-expressing Bmal1 suppresses the period lengthening effect of MAN1 knockdown . Periods of Bmal1-Luc U2OS cells transfected with MAN1 siRNA or control ( ctrl ) siRNA together with either empty vector or varying amounts of Bmal1 . Error bars represent SEM ( n = 2–4 , Student's t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 013 To confirm MAN1 regulates the clock by targeting BMAL1 , we over-expressed Bmal1 in MAN1 knockdown U2OS Bmal1-Luc cells . Knocking down MAN1 lengthened the period in control cells as described above ( Figure 2 ) , whereas cells over-expressing sufficient Bmal1 did not demonstrate period lengthening compared to cells without MAN1 knockdown ( Figure 4—figure supplement 2 ) , suggesting that the lengthened period caused by MAN1 deficiency is due to reduction of BMAL1 . Together , these results indicate that MAN1 functions to promote BMAL1 expression , and thus exerting effects on the clock . Consistent with the cell culture data , over-expressing MAN1 in all clock cells in flies using a timeless ( tim ) GAL4 driver ( Emery et al . , 1998 ) resulted in a significantly increased level of cycle ( cyc ) mRNA , the Drosophila BMAL1 homologue ( Rutila et al . , 1998; Figure 5A ) . The mRNA level of core clock gene tim was also significantly elevated . In addition , we have assessed the levels of MAN1 mRNA to confirm over-expression ( Figure 5B ) . We also examined the effect of knocking down MAN1 in clock cells but did not observe altered cyc mRNA levels ( Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 02981 . 014Figure 5 . MAN1 increases cyc mRNA levels . ( A ) Plots of relative mRNA abundance vs CT for clock genes from whole head extracts of timGAL4/+ and timGAL4/UASMAN1 flies during the first day of DD determined by qRT-PCR ( n = 2 ) . Gray box indicates subjective light period and black box indicates dark period . Significant effect of genotypes ( Two-way ANOVA ) were found for cyc ( p = 0 . 0278 ) and tim ( p = 0 . 0161 ) . For each time series , the value of the lowest time point was set to 1 . ( B ) Plot of relative mRNA abundance for MAN1 from whole head extracts of timGAL4/+ and timGAL4/UASMAN1 flies determined by qRT-PCR ( n = 6 , Student's t test , ***p < 0 . 001 ) . The value of timGAL4/+ in one experiment was set to 1 . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 01410 . 7554/eLife . 02981 . 015Figure 5—figure supplement 1 . cyc transcript level is not altered in MAN1 knock-down flies . Plot of relative mRNA abundance for cyc from whole head extracts of timGAL4/+;UASdicer2/+ and timGAL4/UASMAN1RNAi;UASdicer2/+ flies determined by qRT-PCR . Error bars represent SEM ( n = 6 ) . The value of timGAL4/+;UASdicer2/+ for one experiment was set to 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 015 A luciferase reporter assay using HEK293 cells was used to further investigate the effect of MAN1 on BMAL1 transcription . MAN1 knockdown decreased Bmal1-Luc activity by 72% ( Figure 6A ) , whereas overexpression of FLAG-MAN1 increased the luciferase activity by more than twofold vs cells transfected with empty vector ( Figure 6B ) . Similar results were obtained with a longer , human BMAL1 promoter ( Figure 6C ) . These data indicate that MAN1 may play a role in circadian regulation by activating the promoter of BMAL1 . 10 . 7554/eLife . 02981 . 016Figure 6 . MAN1 promotes BMAL1 transcriptional activity . ( A ) Reduction of MAN1 transcripts ( 13 nM siRNA ) reduced Bmal1 promoter activity ( n = 3 , *p < 0 . 001 ) . ( B ) Over-expression of FLAG-tagged MAN1 ( F-MAN1 ) enhanced Bmal1-Luc activity ( n = 3 , *p < 0 . 001 ) . ( C ) Over-expression of FLAG-tagged MAN1 ( F-MAN1 ) enhanced luciferase activities driven by mBmal1 promoter ( 530 bp ) or hBMAL1 promoter ( 3 . 4 kb ) ( n = 3 , *p < 0 . 05 ) . Error bars represent SD . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 016 Previously , MAN1 has been shown to exert antagonistic regulatory functions on signal transduction through its binding to R-SMADs ( Osada et al . , 2003; Raju et al . , 2003; Hellemans et al . , 2004; Lin et al . , 2005; Pan et al . , 2005; Cohen et al . , 2007 ) and two types of R-SMADs are found in mammals: TGFβ-responsive ( SMAD2 and SMAD3 ) and BMP-responsive ( SMAD1 , SMAD5 , and SMAD8 ) . To determine whether R-SMADs have an effect on BMAL1 transcription , we first expressed R-SMADs individually in HEK293 cells transfected with BMAL1-Luc to determine which R-SMAD/s is/are involved in regulating BMAL1 transcription . Expressing SMAD1 , SMAD5 , SMAD8 , and SMAD3 had no significant effect on BMAL1 transcription but SMAD2 showed significant enhancing effect , suggesting a possible regulatory function by SMAD2 in BMAL1 regulation ( Figure 7A ) . The enhancing action of SMAD2 was then examined together with MAN1 to determine whether there is interplay between MAN1 and SMAD2 on BMAL1 promoter activity . Intriguingly , MAN1 further augmented the enhancing effect of SMAD2 on BMAL1 in an additive manner , indicating that the positive regulatory function of MAN1 and SMAD2 on BMAL1 might be independent of each other ( Figure 7B ) . 10 . 7554/eLife . 02981 . 017Figure 7 . MAN1 and SMAD2 enhance BMAL1 transcription . ( A–D ) Luciferase reporter activities in transfected HEK293 cells . Cells transfected with indicated constructs in the presence of the 3 . 4 kb hBMAL1 promoter for 48 hr and relative luciferase activities were measured in extracts and normalized to Renilla luciferase activities . Relative luciferase activities were shown on the y-axis . Values are means ± SEM , n = 3 , **p < 0 . 01 compared to control . †p < 0 . 05 , ††p < 0 . 01 compared to MAN1 0 , one-way ANOVA with Newman–Keuls test . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 017 Since BMAL1 transcription is regulated by RORα and REV-ERBα , we wondered whether the effect of MAN1 on BMAL1 is influenced by RORα/REV-ERBα . MAN1 increases BMAL1 transcription in the HEK293 luciferase reporter assay but this effect was overshadowed by the presence of either RORα or REV-ERBα , and the impact of RORα and REV-ERBα on BMAL1 was not influenced by the presence of MAN1 ( Figure 7C ) . In addition , the effect of MAN1 was also not significantly altered by mutating the RORE sequence ( +3 ∼ +13 and +39 ∼ +48 ) in the BMAL1 promoter ( Figure 7D ) , which serves as the DNA binding target of RORα and REV-ERBα . Together , these data suggest that the effect of MAN1 on BMAL1 transcription is not through the RORE and does not require or impact RORα and REV-ERBα . Since MAN1 does not execute its function through RORE , we investigated the promoter region of BMAL1 to determine what is necessary for the enhancing effect of MAN1 . A series of deletion constructs of the BMAL1 promoter were generated for luciferase assays and a 900 bp region ( −795 ∼ +106 ) was identified to be the region harboring the necessary DNA sequence for the regulatory effect of MAN1 on BMAL1 ( Figure 8A , Figure 8—figure supplement 1A ) . 10 . 7554/eLife . 02981 . 018Figure 8 . MAN1 binds to the BMAL1 promoter to enhance its transcription . ( A ) Luciferase activities of deleted hBMAL1-promoter constructs in the absence or presence of MAN1 expression vectors . n = 3 , Student's t test , **p < 0 . 01 , ***p < 0 . 001 . ( B and C ) Luciferase activities of the 3 . 4 Kb hBMAL1-Luc in the presence of MAN1 constructs as indicated . n = 3 , **p < 0 . 01 , ***p < 0 . 001 compared to control; †p < 0 . 05; ††p < 0 . 001 compared to WT MAN1 . ( D ) ChIP analysis of MAN1 ( WT or DNA binding truncation ) for 14 segments of hBMAL1 promoter region . Data represent pull-down relative to input . n = 6 , †p < 0 . 05 , compared to WT MAN1 . One-way ANOVA with Newman–Keuls test . All data are presented as ratio of means ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 01810 . 7554/eLife . 02981 . 019Figure 8—figure supplement 1 . Domain-specific interactions between MAN1 and the hBMAL1 promoter . ( A ) Schematic representation of the indicated constructs generated from the 3 . 4 kb hBMAL1 promoter and cloned into the luciferase reporter vector pGL3 . Histogram of luciferase activity in HEK293 cells transfected with the deleted hBMAL1 promoter constructs in the absence or presence of MAN1 expression vectors . Cells transfected for 48 hr and relative luciferase activities measured in extracts and normalized to Renilla luciferase activities . Activities ( relative luciferase activity ) are shown on the y-axis . Values are means ± SEM , n = 3 , Student's t test , **p < 0 . 01 , ***p < 0 . 001 compared to control . ( B ) Schematic representation of the deletion constructs generated from the MAN1 expressing construct . ( C ) Sequences of constructs with point mutations generated from the MAN1 expressing construct . DOI: http://dx . doi . org/10 . 7554/eLife . 02981 . 019 MAN1 has an N-terminal LAP2-Emerin-MAN1 ( LEM ) domain , two transmembrane segments in the middle , a unique DNA binding domain , and a C-terminal RNA recognition motif ( RRM ) that is required for its binding with R-SMADs ( Caputo et al . , 2006 ) . We next examined the domain of MAN1 necessary for its effect on BMAL1 . Two truncation constructs of MAN1 were generated , one without the DNA binding domain ( amino acids 707–725 ) and the other without the RRM domain ( amino acids 760–911 ) ( Pan et al . , 2005; Figure 8—figure supplement 1B ) . In addition , we utilized a substitution mutant of MAN1 ( YV-DD ) ( Pan et al . , 2005 ) , containing two amino acid alterations in the RRM that nullify the ability of MAN1 to antagonize R-SMADs . These constructs were then used to test their transcriptional enhancing effect on BMAL1 promoters ( either full length 3 . 4 kb or 2 . 4 kb [−2300 ∼ +105] ) . Intriguingly , the RRM truncation mutation lost the enhancing effect on the BMAL1 promoter ( Figure 8B ) . The DNA binding domain truncation mutation also lost the activating effect on BMAL1 , and this effect can be produced by simply mutating three positively charged amino acids ( RKK ) at amino acids 709–711 ( Figure 8C , Figure 8—figure supplement 1C ) . These results imply that the effect of MAN1 on BMAL1 transcription requires potential DNA binding ability of MAN1 as well as interaction with a protein partner ( possibly SMAD2 ) through RRM . Since the DNA binding domain is required for the effect of MAN1 on the BMAL1 promoter , we next tested whether MAN1 directly binds to the BMAL1 promoter . Chromatin immunoprecipitation ( ChIP ) analysis revealed that the region −237 bp to +45 bp from transcriptional start site was pulled down by MAN1 indicating a direct interaction ( Figure 8D ) . All together , these data suggest that MAN1 binds to the BMAL1 promoter region ( −237 bp to +45 bp ) to enhance its transcription .
The nuclear envelope plays essential roles in diverse cellular functions including global regulation of gene expression . Interestingly , another global regulatory mechanism is the molecular clock that modulates our body and cellular daily rhythm . We sought to see whether there is cross talk between these two regulatory mechanisms . Our studies revealed that some components of nuclear envelope do show daily oscillations , indicating that nuclear envelope is subject to clock control . On the other hand , we found that one of the major transcription activators of the molecular clock , BMAL1 , is regulated by one of the nuclear envelope proteins , MAN1 . Thus , there is reciprocity between these two global regulatory mechanisms . The nuclear envelope physically separates genomic DNA from the cytoplasm and functions as a signaling control center . An increasing number of human diseases are recognized to be caused by mutations in genes encoding nuclear envelope proteins and hence , termed ‘envelopathies’ ( Dauer and Worman , 2009 ) . Several inner nuclear membrane proteins are known to regulate critical signaling pathways and act as intranuclear regulators of signaling pathways that receive and transduce signals from extracellular cues . The nuclear lamina provides structural support for the nucleus and the nuclear envelope; however , lamins and their associated proteins are also involved in most of the nuclear processes . Lamin B1 is essential for brain development and is required for the integrity of the nuclear lamina ( Padiath et al . , 2006; Coffinier et al . , 2011 ) . Interestingly , no abnormality has been reported for the heterozygous Lmnb1 knock out mice ( Stewart et al . , 2007; Coffinier et al . , 2011 ) , and it is estimated that these heterozygous mice still express 70% of lamin b1 ( SG Young , personal communication ) . Consistent with these reports , we did not find significant phenotype including circadian behavior change in these mice despite the clear phase shift on PER2 oscillation in the liver . It is possible that the oscillation of one or more additional core clock components are also altered by reduced lamin b1 level and these additional alterations can compensate the effect of PER2 phase shift on output behavior . Alternatively , it is also possible that the SCN clock is resilient to perturbations caused by reduced lamin b1 level , hence we can only observe alterations in peripheral clocks . Further investigation is necessary to reveal the mechanism leading to the findings we report here . MAN1 belongs to the inner nuclear membrane LEM protein family ( Worman , 2006; Bengtsson , 2007 ) . LEM domains mediate the interaction with a chromatin-binding protein , barrier-to-autointegration factor ( BAF ) , which has the ability to bind dsDNA , chromatin , histones , lamin binding proteins , and various transcription factors ( Liu et al . , 2003; Worman , 2006; Bengtsson , 2007 ) . Therefore , LEM proteins have roles in gene regulation , chromatin organization , regulation of transcription factor activity at the nuclear periphery , and regulation of specific signal pathways . Both amino and carboxyl termini of MAN1 are nucleoplasmic domains . The amino-terminal nucleoplasmic region of MAN1 ( including LEM ) binds to the nuclear lamins and emerin in addition to BAF . It is also necessary for efficient localization of MAN1 to the inner nuclear membrane ( Mansharamani and Wilson , 2005 ) . The carboxyl-terminal nucleoplasmic region ( residues 655–911 ) exhibits two globular domains ( Pan et al . , 2005; Caputo et al . , 2006 ) . The first globular domain contains a winged helix ( including the sequence RKKMKKVWDR ) which is mainly used for DNA binding and recognition . The second domain ( amino acids 782–911 ) is an RRM-like protein interaction domain where it can interact with R-SMADs . The entire carboxy-terminal region of MAN1 was shown to participate in DNA binding , and this interaction is synergistic to the binding of MAN1 to different transcriptional regulators , including R-SMADs ( Osada et al . , 2003; Raju et al . , 2003; Lin et al . , 2005; Pan et al . , 2005 ) . Consistently , we found that both RRM and DNA binding domains are required for the activation of the BMAL1 promoter by MAN1 . Also congruent with previous findings , we found that mutating three of the highly positively charged and conserved amino acids within the winged helix region is sufficient to dampen the activation efficiency of MAN1 . In contrast to previous reports , we found that MAN1 further augments ( but not antagonizes ) the positive effect of SMAD2 on BMAL1 . Consistent with this result , we also found that the substitution mutation YV-DD of MAN1 does not influence its effect on BMAL1 ( though RRM domain is required ) . Collectively , these results suggest that the effects of MAN1 and SMAD2 on BMAL1 might not be completely independent to each other and that MAN1 might interact with SMAD2 in more than one way ( presumably through other protein partners ) to modulate transcription of target genes . Many of our body functions manifest a daily rhythm which is maintained by the rhythmic regulation of approximately 15% of genes by the core molecular clock ( Vollmers et al . , 2009; Menet et al . , 2012 ) . However , cells must be flexible enough to allow for responses to exogenous and endogenous stimuli . This regulation is likely to be mediated not solely by the molecular clock , but also by many additional global and local mechanisms including at the level of chromatin and genome organization ( Aguilar-Arnal et al . , 2013; Hubner et al . , 2013 ) . Genetic loci associated with the nuclear lamina through large regions of chromatin ( lamin associated domains–LADs ) are associated with changes in transcriptional status ( Peric-Hupkes et al . , 2010 ) . Circadian genes or genes involved in rhythmic processes display robust rhythmic expression patterns at the level of nascent mRNA and mRNA ( Menet et al . , 2012; Rodriguez et al . , 2012 ) , suggesting prominent contribution of transcriptional regulation to clock gene expression . Interestingly , genes harboring this expression pattern are dramatically enriched for specific function in transcriptional regulation and chromatin organization ( Menet et al . , 2012 ) . These unbiased genome-wide transcriptome reports raised the possibility that components of nuclear envelope may modulate oscillation of clock genes at transcriptional levels . A recent study has also demonstrated that the molecular clock drives circadian changes in spatial and temporal chromosomal organization ( Aguilar-Arnal et al . , 2013 ) . Indeed , our study links the nuclear periphery with circadian regulation via the regulatory effects of MAN1 on BMAL1 through transcription . This effect is evident not only in mammalian systems , but also in flies , as over-expressing MAN1 resulted in significantly increased levels of cyc mRNA . MAN1 over-expression also increased tim mRNA levels , which may be at least partially due to increased cyc . Knocking down MAN1 did not alter cyc levels , possibly because the residual MAN1 is still sufficient to maintain normal cyc levels . However , this manipulation lengthened behavioral period , suggesting that MAN1 may target other clock components in addition to cyc . Over-expressing MAN1 and LBR in flies led to a lengthened behavioral period , while in U2OS cells , these manipulations resulted in moderate shortening of the period . On the other hand , over-expressing and knocking down Lam in flies shortened and lengthened the period , respectively , which is consistent with the mammalian data . These discrepancies reflect the differences between mammalian and insect clock , which has been implied in previous work as well ( Lowrey et al . , 2000; Preuss et al . , 2004; Xu et al . , 2005 ) . Nevertheless , our results indicate that there is a conserved role for NE components in setting the clock in organisms ranging from invertebrates to humans . Although only 22–30% of cycling mRNA is driven transcriptionally ( Koike et al . , 2012; Menet et al . , 2012 ) , demonstration of involvement of the nuclear envelope in regulation of the molecular circadian clock suggests one pathway through which the nuclear envelope may globally and temporally regulate large numbers of genes . Interestingly , expression of the genes for many nuclear envelope proteins also oscillates . This finding sheds new light on the interconnectedness of these biological processes and provides further insight into the mechanism whereby cellular , metabolic , physiological , and behavioral processes that oscillate are modulated in a highly coordinated manner .
To generate FLAG-tagged proteins , human LBR , LMNB1 , and MAN1 genes were subcloned into pCMV-Tag2A ( Stratagene; La Jolla , CA ) . HA-tagged SMAD1 , FLAG-MAN1 ( YV-DD ) , and FLAG-MAN1 ( 1–759 ) were gifts from Dr Kunxin Luo ( Pan et al . , 2005 ) . mBmal1-Luc was generated in Dr Satchidananda Panda's lab and kindly provided by Dr John Hogenesch ( Vollmers et al . , 2008 ) , while the human BMAL1-luc construct was a generous gift from Dr Toru Takumi ( Akashi and Takumi , 2005 ) . Syrian hamster Bmal1 in pcDNA3 . 1 vector was provided by Dr David Weaver ( Kume et al . , 1999 ) . pRL-TK was purchased from Promega ( Madison , WI ) . Mutations of expression constructs were introduced by PCR , and all constructs used in this study were verified by sequencing . A stable U2OS-B6 cell line that expresses a destabilized firefly luciferase gene under the control of the mBmal1 promoter was obtained from Dr Satchidananda Panda ( Vollmers et al . , 2008 ) . siRNAs targeted to LBR , LMNB1 , MAN1 , or SMAD1 ( 10 nM , Invitrogen; Carlsbad , CA; see Supplementary file 1A ) were individually transfected into 35-mm culture dishes using Lipofectamine RNAiMAX ( Invitrogen ) . For overexpression of FLAG-tagged constructs , plasmid ( 2 µg ) was distributed into each well along with FuGENE HD ( 4 μl , Roche; Switzerland ) . For co-transfection of MAN1 siRNA and Bmal1 , Lipofectamine 2000 ( Invitrogen ) was used . 24 hr after transfection , cells were synchronized with 100 nM dexamethasone in serum-free DMEM containing 10 mM HEPES ( pH 7 . 5 ) at 37°C for 2 hr . Following synchronization , the media were replaced with phenol red-free DMEM supplemented with 10 mM HEPES and 40 µM Luciferin-EF ( Promega ) . Cells sealed with coverslips were incubated in a 32-channel LumiCycle ( Actimetrics; Evanston , IL ) to monitor real-time bioluminescence for 5 days . Data were analyzed using Lumicycle Analysis ( Actimetrics ) . To over-express NE , cryGAL4-39 and cryGAL4-16 ( Emery et al . , 2000 ) were crossed to MAN1GS2297 ( Kyoto Stock Center; Japan ) , UASLam ( Padiath et al . , 2006 ) and LBRGS2162 ( Kyoto Stock Center ) . To knock down NE , cryGAL4-39;UASdcr2 and UASdcr2;cryGAL4-16 were crossed to UASMAN1RNAi ( 3167R-1 , NIG; Japan ) , UASLamRNAi ( 45 , 635 , VDRC; Vienna ) and UASLBRRNAi ( KK110508 , VDRC ) . For controls , the UAS and GAL4 lines were crossed to w1118 or yw strains . Male progenies were assayed for behavior . Locomotor activity levels of flies were monitored using Trikinetics Activity Monitors ( Waltham , MA ) for 7 days of 12 hr light-12 hr dark ( LD ) conditions followed by 7 days of constant darkness ( DD ) . For DD rhythmicity , chi-squared periodogram analyses were performed using Clocklab ( Actimetrics ) . Rhythmic flies were defined as those in which the chi-squared power was ≥10 above the significance line . Period calculations also considered all flies with rhythmic power ≥10 . RNeasy Mini Kit ( Qiagen ) was used to isolate total RNAs from synchronized U2OS-B6 cells that were collected at interval of 4 hr over the course of 48 hr . Purified RNA ( 2 µg ) was applied in 20-μl reactions for RT primed with Oligo ( dT ) 20 using Super-Script III First-Strand Synthesis System ( Invitrogen ) . All qPCR reactions were carried out on a Rotor-Gene RG-3000 ( Corbett Research; Netherland ) /or 7900HT Fast Real-Time PCR System ( Life technologies; Carlsbad , CA ) using FastStart SYBR Green Master ( Rox ) ( Roche ) . The templates were denatured at 95°C for 10 min , followed by forty cycles with 15 s at 95°C , 10 s at 58°C ( Rotor-Gene ) or 60 s at 60°C ( HT7900 system and data acquisition at the end of this step ) , or 40 s at 72°C , and an additional 2 s for data acquisition ( Rotor-Gene ) . The standard curve and delta–delta CT methods were used for quantification ( Applied Biosystems; Carlsbad , CA ) . Primers used for expression analysis are listed in Supplementary file 1B . Primers used for ChIP assay are labeled as the nucleotide distance from the transcriptional start site ( TSS ) and +1 indicates the starting of TSS . Fly heads were isolated at the indicated time points and total RNA was isolated with TRIzol reagent ( Invitrogen ) . After the removal of contaminating genomic DNA by RQ1 DNase ( Promega ) digestion , total RNA was directly amplified with the QuantiTect SYBR green RT-PCR kit ( QIAGEN ) . The following primers were used: for cyc , cyc_110 . f 5′-GAGGTCTTCGTCGGAAAGG-3′ and cyc_347 . r 5′-AAAGCACATGGGAATCATGG-3′; for tim , tim . f 5′-CTGGGGAGTGACCATGG-3′ and tim . r 5′-GCTGGAATCGCCACTG-3′; for dMAN1 , dMAN1_148 . f 5′-ATTTTGGCCTGTGACACTGC-3′ and dMAN1_303 . r 5′-GAAGCCGCTCTGGATTAGC-3′; for dLam , dLam_446 . f 5′-CGAGGAGCTCAAGAACAAGC-3′ and dLam_675_r . 5′-GCGACAGTGTCTCCTGTTCC-3′; for dLBR , dLBR_645 . f 5′-CATTGACCACCAACACATCC-3′ and dLBR_825 . r 5′-GTTATGCGTTTGCGAATGG-3′; for dActin , dActin . f1 5′-CTAACCTCGCCCTCTCCTCT-3′ and dActin . r1 5′-GCAGCCAAGTGTGAGTGTGT-3′ . All other primers used for fly tissues are previously published ( Lim et al . , 2007; Kilman et al . , 2009 ) . Lmnb1+/Δ mice with a targeted disruption of the Lmnb1 gene ( Vergnes et al . , 2004 ) or LMNB1BAC mouse model overexpressing lamin B1 ( Heng et al . , 2013 ) was generated as previous described . The animals used here were derived from these mice and have been backcrossed to a C57BL/6J background for at least 10 generations . Wild-type littermates were used in pairs for subsequent experiments . Mice housed in light-tight , sound-attenuated cabinets were entrained to LD cycle for 14 days and then released into DD . Wheel-running activity of mice were monitored using Clocklab ( Actimetrics ) . For DD rhythmicity , chi-squared periodogram analyses were performed using Clocklab . Experiments were approved by the Institutional Animal Care and Use Committee at University of California , San Francisco . Brains or livers were collected from mice that were entrained in LD cycle for 14 days and were then released into DD . Total cellular proteins were extracted from mouse brain or liver using RIPA buffer ( 150 mM NaCl , 1% NP-40 , 0 . 1% SDS , 0 . 5% sodium deoxycholate , 20 mM Tris , pH 7 . 5 , and 5 mM EDTA ) . Protein lysates from cells were prepared in SDS-PAGE loading buffer . Equal amounts of protein were resolved on 8% SDS-PAGE gels and then transferred to nitrocellulose membrane . After incubation with primary antibody at 4°C overnight , membranes were incubated with secondary antibodies at room temperature for 1 hr . The primary antibodies were anti-LBR rabbit polyclonal antibody ( 1:500; Abcam ) , anti-LMNB1 rabbit polyclonal antibody ( 1:1000; Abcam; England ) , anti-MAN1 rabbit polyclonal antibody ( 1:3000; from Dr Kunxin Luo ) ( Pan et al . , 2005 ) , anti-mPER2 rabbit polyclonal antibody ( 1:500; Alpha Diagnostic International; San Antonio , TX ) , anti-GAPDH mouse monoclonal antibody ( 1:5000; Chemicon; Billerica , MA ) , anti-BMAL1 goat polyclonal antibody ( 1:500; Santa Cruz; Dallas , TX ) , anti-CLOCK rabbit polyclonal antibody ( 1:1000; Santa Cruz ) , and anti-FLAG M2 antibody ( 1:5000; Sigma; St Louis , MO ) . The MAN1 antibody was generated by immunizing rabbits with C-terminal peptide ( SHLRLRTGLTNSQGSS ) of human MAN1 ( 1:1000; Covance and Agbio , Inc; Princeton , NJ ) . Secondary antibodies were conjugated either with IRDye 680 or IRDye 800 ( LI-COR Biosciences; Lincoln , NE ) and visualized with an Odyssey Infrared Imaging System ( LI-COR Biosciences ) . HEK293 cells were cultured in 24-well plates in DMEM containing 10% fetal bovine serum 24 hr prior to transient transfection with FuGENE HD ( Roche ) for overexpression ( 50–200 ng cDNA constructs ) , or Lipofectamine 2000 ( Invitrogen ) for siRNA knockdown ( 8–13 pmol , Invitrogen ) . All transfection mixtures included a Renilla luciferase plasmid ( pRL-TK; 0 . 7 ng ) , as well as a reporter construct consisting of firefly luciferase driven by mouse Bmal1 or human BMAL1 promoter ( 50 ng ) . We assayed the Bmal1/BMAL1 promoter luciferase activity using the Dual-Luciferase Reporter Assay System ( Promega ) , modifying the protocol to use 30 μl of luciferase substrate and Stop-n-Glo/substrate mix for each reaction . The luciferase activity was quantified with a TECAN GENios Pro Microplate Reader ( TECAN; Switzerland ) 48 hr after the initial transfection . Luciferase reporter vector used is pGL3-basic . We performed ChIP assays using Millipore's EZ-ChIP assay kit ( cat . # 17–371; Millipore; Billerica , MA ) and protein-G sepharose . In brief , HEK293 cells were transfected with hBMAL1-luciferase ( 3 . 4 Kb ) plus vector , FLAG-tagged WT or truncated DNA-binding constructs of MAN1 as indicated . Cell lysates were sonicated on ice using Branson digital sonifier #250 and 1% of cell lysate was taken as input sample . After incubation with FLAG M2 antibody ( Sigma ) , antibody-loaded protein G agarose beads were washed with cold wash buffer six times followed by low-salt buffer , high salt wash buffer , LiCl wash buffer , and then once with TE ( 10 mM Tris–HCl at pH 8 . 0 , 1 mM EDTA at pH 8 . 0 ) . After washing , the beads were re-suspended in 100 μl of ChIP elution buffer supplemented with proteinase K and incubated for 2 hr at 65°C followed by 10 min at 95°C . The beads were spun down and the supernatant was saved . DNA was recovered from the spin column and resuspended in 50 μl of TE , and a 1 μl portion was used for qRT-PCR . The PCR products were analyzed by qRT-PCR and quantitated using 7900HT Fast Real-Time PCR System ( Life technologies ) . Statistical analyses were performed using the unpaired Student's t test , one-way ANOVA with Newman–Keuls test , or two-way ANOVA ( Prism5 , GraphPad; La Jolla , CA ) . Data are presented as Mean ± SEM or SD . Significant differences ( p < 0 . 05 ) are marked with asterisks in figures . | If rodents , or indeed humans , are kept in constant darkness for a number of days , they continue to show patterns of sleep and wakefulness that repeat roughly every 24 hr . This internal ‘circadian rhythm’ controls many aspects of animal physiology , including body temperature , blood pressure , and hormone levels . It does so by regulating the expression of key genes: this means that the activity of the proteins encoded by these genes also varies in accordance with the circadian rhythm . A second mechanism used by the body to coordinate gene expression on a large scale entails making adjustments to the membrane that surrounds the cell nucleus . This ‘nuclear envelope’ consists mainly of lipids , but it also contains proteins that bind DNA . These proteins regulate gene expression by controlling how easy it is for other proteins that activate or repress genes to gain access to specific DNA sequences . Lin et al . now reveal that these mechanisms work together . The first evidence for this was the discovery that the levels of three specific nuclear envelope proteins influence , and are influenced by , circadian rhythms . In particular , two of these proteins control the activity of the third , which is known as MAN1 . This protein in turn triggers the expression of a gene called BMAL1 , which is one of the small number of ‘clock genes’ that are responsible for generating the internal circadian rhythm . As well as adding to our knowledge of circadian biology and the nuclear envelope , this study reveals a mechanism by which cells can orchestrate the expression of large numbers of genes . Such mechanisms allow a wide range of physiological and behavioral processes to be co-ordinated . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | Nuclear envelope protein MAN1 regulates clock through BMAL1 |
Sugars that contain glucose , such as sucrose , are generally preferred to artificial sweeteners owing to their post-ingestive rewarding effect , which elevates striatal dopamine ( DA ) release . While the post-ingestive rewarding effect , which artificial sweeteners do not have , signals the nutrient value of sugar and influences food preference , the neural circuitry that mediates the rewarding effect of glucose is unknown . In this study , we show that optogenetic activation of melanin-concentrating hormone ( MCH ) neurons during intake of the artificial sweetener sucralose increases striatal dopamine levels and inverts the normal preference for sucrose vs sucralose . Conversely , animals with ablation of MCH neurons no longer prefer sucrose to sucralose and show reduced striatal DA release upon sucrose ingestion . We further show that MCH neurons project to reward areas and are required for the post-ingestive rewarding effect of sucrose in sweet-blind Trpm5−/− mice . These studies identify an essential component of the neural pathways linking nutrient sensing and food reward .
Animals and humans generally prefer sugars containing glucose , such as sucrose , compared to non-nutritive sweeteners such as sucralose ( Smiciklas-Wright et al . , 2002; Jacobson , 2005; Domingos et al . , 2011; Sicher , 2011 ) as a result of the post-ingestive rewarding effect of sucrose ( Domingos et al . , 2011 ) . This post-ingestive rewarding effect of sucrose was first described by showing that non-nutritive liquids that are paired to glucose administration either in the intra-gastric tract or in plasma , are greatly preferred over liquids that are not paired with nutrients ( de Araujo et al . , 2008; Ren et al . , 2010; de Araujo et al . , 2010; Oliveira-Maia et al . , 2011; Sclafani et al . , 2011; Fernstrom et al . , 2012; de Araujo et al . , 2013 ) . In addition , sweet-blind Trpm5 knockout mice can still sense the nutrient value of sucrose ( de Araujo et al . , 2008 ) . These studies have indicated that the nutrient value of sucrose is sensed and in turn establishes a preference for nutritive sugars ( Ren et al . , 2010; de Araujo et al . , 2010; Sclafani et al . , 2011; Fernstrom et al . , 2012 ) . These data further indicate that the post-ingestive rewarding effect plays an important role in driving nutrient choice ( in addition to sweet taste ) . However , despite the substantial evidence that they play a major , perhaps dominant , role in driving food intake and sweetener preference , the neural pathways that sense glucose and mediate the post-ingestive rewarding effect of sucrose have not been identified . Rodent studies have further shown that sucrose but not artificial sweeteners such as sucralose can drive dopamine ( DA ) release in the midbrain even in the absence of taste ( de Araujo et al . , 2008; Ren et al . , 2010; de Araujo et al . , 2010; Oliveira-Maia et al . , 2011; Sclafani et al . , 2011; Fernstrom et al . , 2012; de Araujo et al . , 2013 ) . The combination of sweet taste plus an increase of dopamine accounts for the preference for natural vs artificial sweeteners ( Domingos et al . , 2011 ) . We previously reported that the artificial sweetener sucralose is preferred to sucrose only if supplemented by a proxy for this post-ingestive reward in the form of optogenetic activation of DA neurons ( Domingos et al . , 2011 ) . However , the elements of the neural circuit that convey the post-ingestive rewarding effect of sucrose and activate DA neurons are unknown . Melanin-concentrating hormone-expressing neurons ( Pmch or , MCH neurons; in accordance with previous literature [Shimada et al . , 1998; Alon and Friedman , 2006; Kong et al . , 2010] , we adopt the later nomenclature throughout this report ) in the lateral hypothalamus ( LH ) are glucose sensitive , and show increased activity when extracellular glucose levels increase ( Burdakov et al . , 2005; Kong et al . , 2010 ) . Pmch knockout and MCH neuronal ablation lead to reduced body weight , indicating a critical role of these neurons in the regulation of energy balance ( Shimada et al . , 1998; Whiddon and Palmiter , 2013 ) . In addition , MCH neurons send dense projections to reward centers in the striatum and midbrain where dopaminergic neurons are located ( current report ) . This strong anatomical connection between MCH neurons and reward nuclei , as well as the fact that MCH neurons sense glucose levels , led us to hypothesize that these hypothalamic neurons could play a role in conveying the reward value of sucrose .
We first tested whether optogenetic activation of MCH neurons could alter an animal’s preference for sucrose vs sucralose using a BAC transgenic Pmch-CRE mouse line that we generated ( see ‘Materials and methods’ ) . Pmch-CRE mice were crossed to the channelrhodopsin-2 ( ChR2 ) reporter mouse line B6;129S-Gt ( ROSA ) 26Sortm32 ( CAG-COP4*H134R/EYFP ) Hze/J ( Madisen et al . , 2012 ) , herein abbreviated Rosa26-LSL-ChR2-YFP , to generate Pmch-ChR2 mice . We characterized Pmch-CRE mice and thus confirmed tissue- and cell-specific expression of ChR2-YFP in MCH neurons as shown ( Figure 1A ) . The YFP signal was seen in the LH with the characteristic appearance of MCH neurons ( Figure 1A ) , and there was a 92 ± 8% overlap of YFP and MCH . In addition , 97 ± 3% of MCH neurons expressed ChR2-YFP ( Figure 1B ) . Whole-cell patch-clamp recordings in slice preparations confirmed light-evoked spiking at 5 , 10 , and 20 Hz ( Figure 1C ) , as well as during continuous light pulses of one second ( Figure 1D ) . The spike rate of MCH neurons was higher with light pulses of 20 Hz vs 5 Hz . Note , glucose has been shown to evoke similar high-frequency bursting of MCH neurons ( see inset , Figure 1D ) ( Burdakov et al . , 2005 ) . We also recorded voltage responses to consecutive pulses of continuous light in order to test the capacity of these cells to resist repeated trains of light stimulation ( Figure 1D ) . Spike attenuation ( Figure 1D , inset ) during optogenetic stimulation was similar to what has been previously reported for glucose-triggered responses ( Burdakov et al . , 2005; Kong et al . , 2010 ) in MCH neurons , and changes in membrane potential were resilient to optical stimulation ( Figure 1—figure supplement 1B ) . 10 . 7554/eLife . 01462 . 003Figure 1 . Optogenetic control of MCH neurons . ( A ) Pmch-CRE mice were mated to Rosa26-LSL-ChR2-YFP , and expression of ChR2-YFP ( green—right panel ) in MCH neurons ( red—middle panel ) in the LH are shown individually as well as in a merged panel ( left panel ) ; scale bar ( Scale bar: 15 µm ) . ( B ) Quantification of co-expression of MCH and YFP shows that 97 ± 3% of MCH positive neurons expressed YFP and that 92 ± 8% of ChR2-YFP neurons expressed MCH ( n = 1200 cells in four mice ) . ( C ) The effect of light stimulation on spike activity , evoked by light stimulation at 5 Hz , 10 Hz , and 20 Hz . ( D ) The response to 1 s continuous light stimulation , repeated 10 times . Inset , spike train in response to continuous light stimulation similar to what has been described for glucose-induced responses ( see text for references and Figure 1—figure supplement 1 for quantification ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 00310 . 7554/eLife . 01462 . 004Figure 1—figure supplement 1 . Optogenetic activation of MCH neurons . ( A ) In voltage clamp mode , typical lightinduced ChR2 currents ( n = 10 , 1 ms pulse ) and average amplitude of ChR2-induced inward currents . ( B ) Quantification of Figure 1D: left plot—continuous light stimulation generates a spiking rate of 20 ± 1 . 37 spikes/s ( n = 20 light pulses ) . Right plot–Vm before , during and after continuous light stimulation shows that continuous light stimulation for 1 s depolarizes MCH neurons , which promptly recover to native membrane potential ( Vm ) upon termination of light pulses . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 004 We implanted optical fibers into the LH of Pmch-ChR2 and Pmch-CRE control mice ( respectively , ChR2[+] and ChR2[–] ) and assayed their preferences in a series of two-bottle choice tests ( Figure 2 , Figure 2—figure supplement 1 and ‘Materials and methods’ ) . After five licks at a designated sipper , laser pulses of 5 Hz , 20 Hz , were delivered for 1 s , followed by a refractory period of another second ( ‘Materials and methods’ and Figure 2—figure supplement 1 ) . ChR2 ( + ) and ChR2 ( – ) mice had equal preference for water+laser vs water alone at all stimulation frequencies tested ( Figure 2A ) . We next compared an animal’s preference for sucrose vs sucralose plus optogenetic activation of MCH neurons ( see ‘Materials and methods’ for the rationale of concentrations chosen ) . In the absence of ChR2 ( gray bars in Figure 2 ) or at a low stimulation frequency of 5 Hz ( Figure 2A ) , animals still preferred sucrose to sucralose . However , consistent with the greater effect of 20 Hz on spiking of MCH neurons in slice preparation , a light frequency of 20 Hz , as well as continuous light , inverted an animal’s preference with a strong preference for sucralose plus MCH activation relative to sucrose ( Figure 2B , Figure 2—figure supplement 2 ) . At 20 Hz , ChR2 ( – ) mice ( Figure 2B , middle panel , gray bars ) displayed a preference ratio for sucrose of 82 . 2 ± 3% , whereas ChR2 ( + ) mice had a sucrose preference ratio of 20 . 0 ± 4% ( Figure 2B , middle panel , blue bars ) . This preference ratio for sucrose is significantly lower than isopreference ( p<0 . 0001 one sample T-test against 50% ) . Under continuous light , ChR2 ( – ) mice displayed a preference ratio for sucrose of 76 . 7 ± 3% ( Figure 2B , right panel , gray bars ) , whereas ChR2 ( + ) mice displayed a preference ratio for sucrose of 26 . 8 ± 5% ( Figure 2B , right panel , blue bars ) . This preference ratio for sucrose is significantly lower than isopreference ( p<0 . 0007 one sample T-test against 50% ) . 10 . 7554/eLife . 01462 . 005Figure 2 . Optogenetic activation of MCH neurons inverts preference from sucrose to sucralose . ( A ) Pmch-ChR2 and Pmch-CRE control mice ( respectively , ChR2[+] and ChR2[−] ) were implanted with optical fibers ( Figure 2—figure supplement 1 ) and were given the choice between water paired to laser and water alone . Laser preference is defined as the ratio of the number of licks of the water bottle that was paired to laser ‘ON’ and the total number of licks of both bottles ( ×100 ) . Light stimulation during ingestive behavior was set to 5 Hz , 20 Hz , and continuous ( Figure 2—figure supplement 1 for lick/laser contingency ) . Optogenetic stimulation of MCH neurons during water intake did not influence preference behavior at any of the light stimulation frequencies . ( B ) ChR2 ( – ) and Chr2 ( + ) mice were given the choice between sucralose coupled to laser and sucrose . Sucrose preference is defined as the ratio of the number of licks of the bottle containing sucrose and the total number of licks of both bottles ( ×100 ) . Light stimulation during ingestion of sucralose was set to 5 Hz , 20 Hz , and continuous . 20 Hz and continuous light stimulation , but not 5 Hz , inverts preference from sucrose to sucralose ( see ‘Materials and methods’ and Figure 2—figure supplement 2 for total licks per bottle ) . All data are mean ± SEM and n = 4 mice . ***p<0 . 0001 , ns: p>0 . 28 , t test . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 00510 . 7554/eLife . 01462 . 006Figure 2—figure supplement 1 . Lick/laser contingency . ( A ) Schematic of optical fiber implant in LH of ChR2 ( − ) and ChR2 ( + ) mice . ( B ) The laser was turned ON for one second at every five consecutive licks on the same sipper , and OFF for the following 1 s . The laser was turned ON at three frequency regimens: either at 5 Hz , 20 Hz or continuously on . Lick/laser contingencies were the same for sucralose and water . Each vertical blue stroke in the blue raster represents the laser onset . Bottom schematic of magnified time lines ( zoom in inset ) of licks and laser are not draw to scale . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 00610 . 7554/eLife . 01462 . 007Figure 2—figure supplement 2 . Total licks in Figure 2 . Licks on each bottle , averaged across mice . ChR2 ( − ) and ChR2 ( + ) mice gave a total number of licks , respectively . ( A ) 347 ± 61 and 432 ± 72 on the water+laser and 385 ± 78 and 382 ± 70 on the water . ( B ) 354 ± 71 and 382 ± 70 on the water+laser and 317 ± 69 and 333 ± 52 on the water . ( C ) 395 ± 72 and 334 ± 73 on the water+laser and 320 ± 53 and 326 ± 64 on the water . ( D ) 102 ± 31 and 182 ± 45 on sucralose+laser and 464 ± 92 and 588 ± 81 on sucrose ( E ) 107 ± 13 and 555 ± 87 on sucralose+laser and 494 ± 74 and 139 ± 33 on sucrose . ( F ) 141 ± 56 and 386 ± 68 on sucralose+laser and 463 ± 83 and 141 ± 56 on sucrose . All data are mean ± SEM , ***p<0 . 0002 , **p<0 . 005 , *p<0 . 02 , t tests . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 007 The post-ingestive rewarding effect of glucose is associated with an increase of DA release in the striatum ( de Araujo et al . , 2013; de Araujo et al . , 2008; de Araujo et al . , 2010; Ren et al . , 2010 ) , and we confirmed that MCH neurons densely innervate the striatum and the ventral midbrain , making synapses onto DA neurons ( Figure 3—figure supplement 1 , ‘Materials and methods’ ) . We thus tested whether the inversion of preference to sucralose by activating MCH neurons was correlated with increases in striatal DA release as measured by microdialysis in ChR2 ( + ) mice ( de Araujo et al . , 2008; de Araujo et al . , 2013 ) ( Figure 3A ) . After signal stabilization , samples were collected during a 30-min period when the animals had access to sucralose ( Figure 3A , B ) . As previously reported , animals drinking sucralose ( without light stimulation ) displayed negligible changes in striatal DA levels ( Figure 3B ) . Animals drinking sucralose when the laser was set to OFF displayed an overall 8 . 2 ± 2 . 6% change from baseline DA ( average across all S1–S5 samples ) ; this change was not significantly different from baseline ( gray bars in Figure 3E p>0 . 05 , one sample T-test against 100% baseline ) . However , when sucralose ingestion was coupled to laser stimulation of MCH neurons , DA release significantly increased in the striatum ( **p<0 . 008; Figure 3C–G , blue bars ) . When laser was set to ON during ingestion of sucralose , DA levels increased 68 . 7 ± 9% vs baseline DA ( average across all S1–S5 samples , blue bars in Figure 3E ) . This increase is not only significantly different from baseline and the OFF condition , but is also significantly different from DA release after experimenter-controlled ( i . e . , direct ) delivery of laser pulses , in the absence of sucralose ingestion ( light blue bar in Figure 2E , ****p<5 . 7e10−7 , t-test , with Bonferroni correction for multiple comparisons ) . Each animal received the same number of pulses as the number of laser pulses during ingestion of sucralose in the ON condition ( Figure 3E ) . Mice used in condition laser-ON were the same as in condition laser-OFF . Optogenetic activation of MCH neurons markedly increased sucralose ingestion during the laser ON condition ( *p<0 . 05; Figure 3F ) while activation of MCH neurons did not change intake of water ( Figure 2—figure supplement 2 ) . As mentioned above , optogenetic stimulation of MCH neurons paired to water was not preferred to water alone , showing that activation of MCH neurons is not rewarding in the absence of sucralose . Thus , both MCH activation and the presence of sucralose were required to establish a change of preference . 10 . 7554/eLife . 01462 . 008Figure 3 . Optogenetic activation of MCH neurons increases DA release during sucralose ingestion . ( A ) Schematics of microdialysis sampling of striatal DA release in behaving mice ( left panel ) after intracranial implants of optical fibers in the LH and microdialysis probe in the striatum ( S ) . ( B ) A timeline of licking behavior and DA collection with corresponding HPLC-ECD chromatograms of DA release when a ChR2 ( + ) mouse drank 1 . 5 mM sucralose with the laser OFF . ( C ) A timeline with the laser ON at 20 Hz . ( D ) A timeline of average of DA increases from baseline across mice . ( E ) Overall change from baseline DA averaging across all S1–S5 samples in ( D ) and in the absence of drinking behavior with the laser ON ( lighter blue ) . Each animal received the same number of pulses as the number of laser pulses delivered during ingestion of sucralose in the ON condition . On average , 201 ± 40 pulses were delivered . ( F ) The cumulative licks during microdialysis in both conditions are shown ( see Figure 3—figure supplement 1 for MCH projections to reward centers and Figure 3—figure supplement 2 for requirement of DA transmission in sucrose/sucralose preference ) . All data are mean ± SEM and n = 4 mice , *p<0 . 05 , **p<0 . 008 , ****p<5 . 7e10−7 . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 00810 . 7554/eLife . 01462 . 009Figure 3—figure supplement 1 . MCH axonal projections onto DA neurons and reward areas . ( A ) GFP labeled axons that form synaptic-like boutons ( white arrowheads ) onto dopamine neurons . ( B ) Electron microscopy of gold enhanced detection of GFP in MCH neurons ( black arrowheads ) and silver-gold enhanced detection of TH-expressing neurons ( grey arrow heads ) of MCH-GFP transgenic mice , showing the existence of asymmetrical synapses between MCH and DA neurons , note the presence of dense core vesicles in MCHsilver-gold labeled boutons . ( C ) axonal projections of MCH-GFP neurons onto the dorsal striatum ( DS ) . ( D ) axonal projections of MCH-GFP neurons onto the nucleus accumbens ( Nac ) . ( E ) axonal projections of MCH-GFP neurons onto the substancia nigra ( SN ) . ( F ) axonal projections of MCH-GFP neurons onto the ventral tegmental area ( VTA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 00910 . 7554/eLife . 01462 . 010Figure 3—figure supplement 2 . Preference for sucrose vs sucralose requires DA transmission . ( A ) Blocking dopamine transmission with haloperidol ( hal ) suppressed preference for sucrose versus sucralose . In 10 min , wild-type animals injected with hal ( ip , 1 mg/kg ) showed a preference ratio for sucrose of 51 ± 7% , whereas vehicle treated animals showed a preference ratio for sucrose of 79 ± 4% . ( B ) Licks on each bottle in ( A ) averaged across mice . Vehicle and haloperidol treated mice gave a total number of licks , respectively 272 ± 28 , 104 ± 16 on sucrose and 72 ± 10 , 100 ± 11 on sucralose . All data are mean ± SEM , n= 12 mice . ***p<0 . 0022 , ****p<0 . 0001 , NS = not significant , t test . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 010 We next tested whether a loss of MCH neurons decreased DA release during sucrose intake . We crossed Pmch-CRE mice to C57BL/6-Gt ( ROSA ) 26Sortm1 ( HBEGF ) Awai/J , ( Buch et al . , 2005 ) herein abbreviated Rosa26-LSL-DTR , to generate Pmch-CRE;LSL-DTR mice that specifically express the diphtheria toxin receptor ( DTR ) in MCH neurons ( Figure 4 ) . We injected Pmch-CRE;LSL-DTR mice with diphtheria toxin ( +DT ) or vehicle ( +veh ) intracranially ( 1 ng/g of body weight ) . This dose led to a complete loss of MCH neurons ( Figure 4A , see Figure 4—figure supplement 1 for other doses ) . MCH-ablated ( Pmch-CRE-LSL-DTR+DT ) and control mice ( Pmch-CRE-LSL-DTR+veh , and LSL-DTR+DT ) were subjected to microdialysis sampling of striatal DA during sucrose ingestion ( Figure 4B ) . Control mice drinking sucrose displayed significantly higher dopamine levels compared to MCH-ablated mice ( Figure 4C–F; **p<0 . 008 ANOVA in Figure 4E ) . Control mice showed an overall 118 . 3 ± 0 . 3% increase in striatal DA levels while drinking sucrose ( average across all S1–5 samples , Figure 4F , ***p<1 . 98e10−9 T-test ) . The increase of DA levels was significantly above those at baseline ( p<0 . 0018 , one sample T-test compared to 100% baseline ) . In contrast to animals without DT injection , MCH-ablated mice showed a negligible DA efflux during sucrose intake , and the levels after sucrose exposure did not differ from baseline levels ( p>0 . 4 , one sample T-test compared to 100% baseline ) . Baseline DA levels were similar in both groups ( Figure 4 , Figure 4—figure supplement 2A ) . Consistent with a lower reward value of sucrose , when given free access to sucrose or water , MCH-ablated mice consumed significantly less sucrose than control mice ( Figure 4G , *p<0 . 05 ANOVA ) , while water intake was equivalent between the groups ( Figure 4 , Figure 4—figure supplement 2B ) . 10 . 7554/eLife . 01462 . 011Figure 4 . MCH neurons are required for DA release during sucrose ingestion . ( A ) Pmch-CRE;LSL-DTR mice were treated with 1 ng/g of DT . Complete ablation of MCH neurons by intracranial injection of diphtheria toxin is shown . ( see Figure 4—figure supplement 1 for other doses ) . ( B ) Schematics of microdialysis in behaving mice after intracranial implant of microdialysis probe in the striatum ( S ) . ( C ) A timeline of licking behavior and DA collection , with corresponding HPLC-ECD chromatograms when a control mouse drank 0 . 4 M sucrose is shown . ( D ) A timeline similar to ( C ) when an MCH-ablated mouse drank sucrose ( E ) timeline of average DA increases from baseline across mice are shown . ( F ) For both genotypes , overall change from baseline DA averaging across all S1–S5 samples in ( E ) is shown . ( G ) Cumulative licks during microdialysis in both groups ( see figure supplements for additional controls ) . All data are mean ± SEM and n = 4 mice . *p<0 . 05 , **p<0 . 008 , ***p<1 . 98e10−9 . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 01110 . 7554/eLife . 01462 . 012Figure 4—figure supplement 1 . Titration of intracranial dose of DT . Medium/high systemic doses ( ∼25 ng/g body weight or higher ) of DT are thought to be toxic to wild type mice . We thus assessed whether low doses of DT administered intra-cranially were efficient at ablating MCH neurons , keeping mice viable , normal in size , and free of any gross physical or behavioral abnormalities . Animals were injected intra-crannially at coordinates AP = −1 . 5 , ML = +/−1 . 6 , DV = 5 . 5 ( ‘Materials and methods’ ) . PBS injected mice had 31 ± 2 neurons/section marked with MCH . At a dose of 0 . 5 ng per gram of animal , 3 . 6 ± 0 . 5/section neurons marked with MCH could be visualized under fluorescence microscopy . At a dose of 1 ng per gram of animal , 0 . 3 ± 0 . 18 neuron/section marked with MCH could be visualized under fluorescence microscopy . Statistically , this amount is not significantly different from zero ( p>0 . 05 , One sample t test against zero ) . All data are mean ± SEM , n = 5 mice . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 01210 . 7554/eLife . 01462 . 013Figure 4—figure supplement 2 . MCH-ablated mice have normal baseline DA levels and are not adipsic . ( A ) Baseline DA concentration of ablated MCH-Cre; LSL-DTR+DT mice ( 0 . 34 ± 0 . 09pg/μl ) were not significantly different from control MCH-Cre; LSLDTR+veh mice ( 0 . 25 ± 0 . 05pg/μl ) ( p>0 . 22 , t test ) , thus excluding chronic effects of MCH neuronal ablation on DA levels . ( B ) after 20–23 hr of water deprivation , ablated MCH-Cre; LSL-DTR+DT mice drank , for 1 hr , approximately as much water ( 782 ± 100 licks ) as control MCH-Cre;LSL-DTR+veh mice ( 642 ± 101 licks ) ( difference is not statistically significant p>0 . 29 , t test , n = 4 mice ) . All data are mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 013 MCH-ablated and control mice were also given a series of choices sequentially with studies of ( A ) sucrose vs sucralose , ( B ) sucrose vs water , ( C ) sucralose vs water ( Figure 5A–C ) , and preference ratios for each comparison were computed . Preference ratios for sucrose in Pmch-CRE;LSL-DTR ( +veh ) , LSL-DTR ( +DT ) and Pmch-CRE;LSL-DTR ( +DT ) mice were , respectively , 77 . 1 ± 7% , 82 . 0 ± 4% , and 39 . 9 ± 5% . While control mice preferred sucrose to sucralose , MCH-ablated mice no longer had a preference for sucrose . This reduction in sucrose preference is statistically significant ( Figure 5A , *p<0 . 0012 , ¥p<0 . 009 , T-test with Bonferroni correction for multiple comparisons , same-symbol pairs indicate statistically significant differences; see also Figure 5—figure supplement 1 for total licks in each ) . However , MCH neuronal ablation did not alter an animal’s preference for either sucrose or sucralose vs water indicating that , in contrast to their requirement for establishing the post-ingestive effect of sucrose , MCH neurons are not required for establishing a preference for sweet taste ( Figure 5B , C and Figure 5—figure supplement 1 ) . The postprandial increase in blood glucose after sucrose ingestion was normal in MCH-ablated animals ( Figure 5 and Figure 5—figure supplement 2 ) , demonstrating that the changes in sucrose preference were not a result of any differences in blood glucose levels ( Kong et al . , 2010 ) . Altogether , these data indicate that activation of MCH neurons are necessary and sufficient for establishing the preference of animals for sucrose compared to artificial sweeteners . 10 . 7554/eLife . 01462 . 014Figure 5 . MCH neurons are required for sucrose vs sucralose preference , even in the absence of taste . ( A–C ) Mice with ablated MCH neurons ( red filled bars ) and their respective controls ( blue filled bars ) were given the choice of ( A ) 0 . 4 M sucrose vs 1 . 5 mM sucralose ( *p<0 . 03 , ¥p<0 . 011 , see ‘Materials and methods’ for the rationale of concentrations ) . ( B ) Sucrose vs water . ( C ) Sucralose vs water . All mice preferred either sweetener—sucrose or sucralose—over water . ( D ) Sweet-blind Trpm5−/− mice , with and without ablation of MCH neurons , were subject to a 4-day bottle-conditioning protocol , in which sucrose and sucralose were presented in opposing bottles , on alternate days . Bottle preference was tested on the fifth day with two bottles filled with water . Sweet-blind control mice showed a significant side bias towards the bottle where sucrose was placed during the conditioning sessions whereas MCH-ablated mice did not ( *p<0 . 045 , ¥p<0 . 099 , see Figure 5—figure supplement 1 for total licks in each bottle , Figure 5—figure supplement 2 for blood glucose controls , and ’Materials and Methods’ for details ) . All data are mean ± SEM and n = 8 mice , t test with Bonferroni correction for multiple comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 01410 . 7554/eLife . 01462 . 015Figure 5—figure supplement 1 . Total licks per bottle in MCH-ablated and control mice . Licks on each bottle , averaged across mice , panels follow the same order as in main figure . LSL-DTR+DT and MCH-Cre; LSL-DTR+veh control mice and MCH-Cre; LSL-DTR+DT ablated mice give a total number of licks , respectively . ( A ) 287 ± 31 , 336 ± 11 , 99 ± 14 on sucrose and 63 ± 7 , 100 ± 11 , 149 ± 19 on sucralose . ( B ) 308 ± 43 , 301 ± 33 , 310 ± 34 on sucrose and 150 ± 17 , 80 ± 9 , 73 ± 8 on water . ( C ) 316 ± 34 , 288 ± 31 , 320 ± 35 , on sucralose and 69 ± 8 , 59 ± 6 , 106 ± 12 on water ( d ) Trpm5−/− LSL-DTR+DT and Trpm5−/− MCH-Cre; LSL-DTR+veh control mice and Trpm5−/− MCH-Cre; LSL-DTR+DT ablated mice give a total number of licks , respectively 267 ± 29 , 248 ± 27 , 99 ± 11 on the sucrose-conditioned side and 114 ± 12 , 65 ± 7 , 97 ± 10 on the sucralose control side . All data is mean ± SEM , ****p<0 . 0001 , ***p<0 . 005 , NS = not significant—p>0 . 05 . t test . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 01510 . 7554/eLife . 01462 . 016Figure 5—figure supplement 2 . Ablation of MCH neurons affects neither peak blood glucose after IP challenge , nor acute postprandial blood glucose . ( A ) Blood glucose ( BG ) was measured 10 min before and after an IP bolus injection of glucose ( 10% , ‘Materials and methods’ ) . The BG of MCH-Cre; LSL-DTR+DT MCH-ablated mice and control MCH-Cre;-LSL-DTR+veh before injection was , respectively , 122 ± 11 mg/dl and 151 ± 16 mg/dl . After injection , BG raises , respectively to 325 ± 28 mg/dl and 355 ± 34 mg/dl . The raise in BG was statistically significant before/after injection ( *p<0 . 05 , student t test , n = 4 ) , but not across control and ablated groups of mice ( *p>0 . 05 , student t test , n = 4 ) . ( B ) BG levels were measured ten minutes before and at the end of a behavioral trial in which mice drank 0 . 4 M sucrose for 10 min . The BG of MCH-Cre;-LSL-DTR+DT mice and control MCH-Cre;-LSL-DTR+veh before drinking was , respectively , 121 . 82 ± 11 mg/dl and 138 ± 12 mg/dl . After drinking , BG raised , respectively to 254 ± 20 mg/dl and 295 ± 55 mg/dl . The raise in BG was statistically significant before/after drinking ( *p<0 . 05 , student t test , n = 4 ) , but not across control and ablated groups of mice ( p>0 . 05 , student t test , n = 4 ) . ( C ) MCH-Cre;-LSL-DTR+DT MCH-ablated mice and control MCH-Cre;-LSL-DTR+veh drank comparable amounts of sucrose solution during the behavioral trial in panel ( B ) , respectively , 336 ± 132 and 351 ± 121 licks ( n = 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 01610 . 7554/eLife . 01462 . 017Figure 5—figure supplement 3 . Ablation of MCH neurons affects postingestive DA neuron activation . ( B–C ) Confocal imaging of cFos staining of in Th positive neurons in the VTA of one control Trpm5−/−;MCHCRE; Rosa26-Flox-DTR+veh mouse after drinking sucrose ( 350 licks ) . ( D–F ) Confocal imaging of cFos staining in Th positive neurons in the VTA of one ablated Trpm5−/−;MCHCRE; Rosa26-Flox-DTR+dt mouse after drinking 0 . 4 M sucrose ( 364 licks ) . ( G ) Quantification across groups ( n = 4 ) revealed that the level of cFos expression in DA neurons after exposure to sucrose is significantly lower in sweet blind Trpm5−/− animals in which MCH neurons have been ablated vs controls . Trpm5−/−;MCHCRE; Rosa26-Flox-DTR+dt , show cFos expression in 20 . 3 ± 3 . 7% of DA neurons after sucrose intake vs . 36 . 0 ± 2 . 5% in control Trpm5−/−;MCH-CRE;Rosa26-Flox-DTR+veh ( p*< 0 . 004 , t test ) . Ablated and control group drank , respectively , 356 ± 102 and 341 ± 111 licks of 0 . 4 M sucrose ( n = 4 ) . ( H ) Ablation of MCH neurons does not impact activation in THnegative cells . Quantification of c-Fos in nondopaminergic cells of the same slices as in ( g ) . Trpm5−/−;MCH-CRE;Rosa26-Flox-DTR+veh and Trpm5−/−;MCH-CRE;Rosa26-Flox-DTR+dt have respectively , 16 . 0 ± 4% and 14 . 1 ± 3 . 5% of cFos positive neurons that are TH negative . DOI: http://dx . doi . org/10 . 7554/eLife . 01462 . 017 Finally , to confirm that MCH neurons are required for the post-ingestive rewarding effect of sugar even in the absence of sweet taste , we tested whether sucrose could condition preference in sweet-blind Trpm5–/– mice lacking MCH neurons ( Figure 5D , de Araujo et al . , 2008 ) . Sweet-blind control mice showed a significant side bias towards the side where sucrose was placed during the conditioning sessions , in which sucrose or sucralose are delivered at opposite sides on alternate days , ( Figure 5D ) . In contrast , sweet-blind MCH-ablated mice did not show a preference in this conditioning protocol , indicating that the post-ingestive rewarding effect of sucrose had been lost . Trpm5–/– Pmch-CRE;LSL-DTR+veh and Trpm5–/–LSL-DTR+DT control mice showed conditioned side preferences of 70 ± 3% and 79 ± 5% respectively , towards the side where sucrose was placed . Trpm5–/– Pmch-CRE;-LSL-DTR ( +DT ) mice had 50 ± 7% conditioned side preference towards the side where sucrose was placed . This reduction in sucrose conditioning was statistically significant from that of controls ( Figure 5D , *p<0 . 045 , ¥p<0 . 099 , T-test with Bonferroni correction for multiple comparisons , same-symbol pairs indicate statistically significant differences ) . Finally , the behavioral conditioning by the post-ingestive rewarding effect of sucrose in sweet-blind Trpm5–/– mice positively correlates with the extent of DA neuron activation , as assayed by staining for cFos in DA neurons in the ventral tegmental area ( VTA ) ( Figure 5 , Figure 5—figure supplement 2 ) . These data confirm that MCH neurons are required for sensing the nutrient value of sucrose in the absence of taste .
In this manuscript , we report that MCH neurons are necessary and sufficient for establishing a preference for sucrose vs sucralose , an artificial sweetener . MCH neurons serve as an essential link between glucose sensing and sugar reward , and these data thus identify a key component of the neural circuit that establishes the preference for natural vs artificial sweeteners ( Smiciklas-Wright et al . , 2002; Jacobson , 2005; Domingos et al . , 2011; Sicher , 2011 ) . MCH neurons have previously been shown to be excited by glucose ( Burdakov et al . , 2005; Kong et al . , 2010 ) , suggesting that direct glucose sensing by these neurons regulates reward . However , it is also possible that glucose is sensed elsewhere , such as by putative gastric/intestinal sensors or other nutrient sensors , which inform the brain and MCH neurons about the nutrient content of ingested food ( Sclafani et al . , 2011 ) . Glucose sensing in MCH neurons or elsewhere would explain why , even in the absence of taste ( such as in Trpm5–/– mice ) , sucrose is able to drive significant increase of striatal DA . This increase , in turn , conveys reward and conditions behavior ( de Araujo et al . , 2008 ) . We note , however , that optogenetic stimulation of MCH neurons alone is not sufficient to alter behavior in the absence of taste . The data thus suggest that MCH neurons are components of a reward-encoding network that integrates information from multiple sources , including the nutrients themselves , lingual taste buds and , possibly , other sites of glucose sensing like the gut . Consistent with this possibility , viral tracing from lingual taste buds shows that MCH neurons are part of a circuit processing gustatory information ( Pérez et al . , 2011 ) . This finding is also consistent with the optogenetic data reported here , showing a requirement for both sweet taste and activation of MCH neurons to drive reward . The role of MCH neurons uncovered here contrasts with that of DA neurons , which upon optogenetic stimulation have been shown to be rewarding when paired to water ( Domingos et al . , 2011 ) . A synergy between taste and the post-ingestive rewarding effect would explain why sucrose and other fructose/glucose disaccharides , which are more potent stimulators of sweet-taste receptors than glucose alone , are generally preferred to glucose alone ( Sclafani and Mann , 1987; Nelson et al . , 2001 ) . This synergy would also explain why our optogenetic gain of function experiments lead to an inversion of preference , rather than an isopreference , which could perhaps have been achieved by decreasing the concentration of sucralose . Likewise , the loss of function of MCH neurons leads to isopreference , but this could perhaps be biased toward sucralose by increasing its concentration . Further studies will be necessary to establish the relevant sites of glucose sensing , identify additional elements of the neural circuit integrating gustatory perception with nutrient sensing and reward , as well as elucidating the neural mechanisms by which MCH neurons regulate striatal DA release . Several methods including viral tracing can be used to identify monosynaptic or polysynaptic inputs onto MCH and DA neurons . As mentioned , the activity of MCH neurons is increased by glucose ( Burdakov et al . , 2005 ) . Glucose-activated MCH neurons and pancreatic β cells share signal transduction components necessary for glucose sensing , and both Kir6 . 2 and UCP2 regulate glucose excitability of MCH neurons ( Kong et al . , 2010 ) . Moreover , a loss of function of Kir6 . 1 in MCH neurons leads to alterations in results of a glucose tolerance test with increased plasma glucose at later times , establishing a role for these neurons in glucose homeostasis ( Kong et al . , 2010 ) . Further studies are required to establish whether these or other components of glucose sensing pathways are also required for the ability of MCH neurons to influence sucrose preference . It is possible , however , that additional neural populations are components of this nutrient sensing circuit . For example , orexin/hypocretin-containing neurons can also sense glucose and it is thus possible that these or other neural populations in the mesolimbic system , or higher order centers can also influence the reward value of sugar ( Burdakov et al . , 2005; Karnani and Burdakov , 2011 ) . Previous reports have also explored the relationship between MCH neurons and reward: both MCH knockout and MCH ablated mice show augmented locomotor responses to psychostimulant drugs ( Pissios et al . , 2008; Whiddon and Palmiter , 2013 ) . These augmented locomotor phenotypes contrast with the behavioral effect we see with a loss of sucrose preference in MCH ablated mice . It is possible that the locomotor phenotypes in response to stimulants result from actions in the ventral striatum , where the MCH receptor ( MCHR-1 ) is expressed , and that sucrose preference relies on other brain areas . Further experiments will be required to ascertain which brain areas and MCH projections are relevant for sucrose preference . Loss of function of MCHR-1 recreates the locomotor phenotypes seen in Pmch−/− mice: Mchr1−/− mice are super-sensitive to the locomotor activating effects of d-amphetamine ( Smith et al . , 2005 ) . These studies do not establish whether it is MCH or another neurotransmitter expressed in these neurons that is responsible for the observed phenotypes , and further experiments will also be required to ascertain whether the MCH neuropeptide itself is relevant for sucrose preference . We assayed dopamine release using microdialysis to show that MCH neural activation increases dopamine release in the striatum and that the increase of dopamine in response to sucrose is lost after ablation of MCH neurons . Consistent with this , the levels of cFos in dopaminergic neurons of the VTA are reduced in MCH-ablated , sweet-blind Trpm5 KO mice given sucrose . These assays of taste-blind mice confirm that MCH neurons regulate the activity of dopaminergic neurons , though the data do not establish whether this effect of dopaminergic neural activity and dopamine release is direct and/or indirect . Further studies will be necessary to establish how MCH neurons regulate dopaminergic signaling . The delineation of this neural circuit may also provide a basis for understanding how leptin modulates reward ( Domingos et al . , 2011 ) . The effects of leptin on reward are unlikely to be a result of a direct effect on MCH neurons as they do not appear to express the leptin receptor ( Leinninger et al . , 2011 ) . However , a distinct neural population in the LH expressing neurotensin respond to leptin , and further studies may reveal whether or not leptin reduces MCH activity indirectly by activating these cells ( Leinninger et al . , 2011 ) . Ablation of MCH neurons attenuates the obese phenotype of leptin deficient ob/ob mice indicating that MCH neurons are downstream of leptin action ( Alon and Friedman , 2006 ) . Ablation of MCH neurons causes hypophagia and leanness , and it is possible that the reduced food intake is a result of a loss of the reward value of nutrient in these animals ( Alon and Friedman , 2006 ) . The reward value of sugar is also regulated by leptin , which has been recently reported to have a presynaptic action to suppress excitatory synaptic input onto VTA DA neurons ( Domingos et al . , 2011; Thompson and Borgland , 2013 ) . Further studies will reveal whether leptin modulates excitatory output to the VTA via MCH neurons or influences nutrient preference by a direct effect on DA neurons in the VTA . The importance of brain nutrient sensing for behavior has also been studied in Drosophila ( Dus et al . , 2011 , 2013; Miyamoto et al . , 2012 , 2013 ) . Future studies will likely elucidate the extent to which the cellular mechanisms and neural pathways that regulate nutrient preference are shared between invertebrates and mammals . Brain nutrient sensing may represent an evolutionary adaptation to avoid starvation , by expediting decisions about which foods to consume . In summary , these results confirm that MCH neurons are both necessary and sufficient for sensing the nutrient value of sucrose and suggest that these neurons play a critical role in establishing nutrient preference . The market share of sugared soda is nearly triple that of diet soda ( Smiciklas-Wright et al . , 2002; Jacobson , 2005; Sicher , 2011 ) , and our data suggest a biological approach to potentially regulate sugar consumption . This could be achieved via the development of means for suppressing the activity MCH neurons , or via the development of new artificial sweeteners with neuroexcitatory activity specific to MCH neurons .
In order to restrict Cre expression to MCH neurons we used a BAC clone containing the full-length pro-melanin-concentrating hormone gene ( RP23–129A21 ) with upstream and downstream flanking sequences of 108 kb and 89 kb , respectively . Prior to further manipulation , BAC DNA was prepared and electroporated into E . coli strain SW102 as required for BAC recombineering . An NLS-Cre PolyA construct ( pML78 , Mark Lewandowski , National Cancer Institute ) was targeted to replace the ATG translational start codon of MCH exon 1 and correct insertion was verified by PCR and sequencing . 5′ recombineering homology: TGAAAGTTTTCATCCAATGCACTCTTGTTTGGCTTTATGCAAGCATCAAA 3′ recombineering homology: CTGCAGAAAGATCCGTTGTCGCCCCTTCTCTGGAACAATACAAAAACGAC . All DNA fragments used for recombineering were generated with the FastStart High Fidelity PCR System ( Roche , Indianapolis , IN ) . The modified BAC insert was released by NotI digestion , gel purified and used for pronuclear injection . Rosa26-LSL-ChR2-YFP and Rosa26-LSL-DTR were obtained from Jackson Laboratories . All animal procedures were carried out in accordance with the National Institutes of Health Guidelines on the Care and Use of Animals and approved by the Rockefeller University Institutional Animal Care and Use Committee ( Protocols #13608 , #10005 and #09012 ) , JB Pierce Institutional Animal Care and Use Committee , ( Protocol #101 ) and Yale University Institutional Animal Care and Use Committee ( Protocol #2011-07942 ) . Immunohistochemistry was performed as published elsewhere ( Domingos et al . , 2011 , Pérez et al . , 2011 ) , using chicken anti-GFP ( 1:1000; Abcam , Cambridge , MA ) , rabbit anti-MCH ( 1:1000; Abcam ) , c-Fos ( 1:100; Abcam ) . 30 to 50-day old MCH-ChR2-YFP mice were used for recordings . Mice were euthanized at the beginning of the light cycle ( 9:00 AM ) , and brain slices containing the LH were cut at 300 μm ( 2/mouse ) . Slices were transferred to a chamber at room temperature to stabilize in artificial cerebrospinal fluid ( aCSF ) . Slices were then transferred to a recording chamber after ≥1 hr recovery and constantly perfused at 34°C with bath solution at a speed of 1 . 5 ml/min . Whole cell patch-clamp recording was performed on identified MCH-YFP neurons with a Multiclamp 700B amplifier ( Axon Instruments , New York , NY ) . The patch pipettes were made of borosilicate glass ( Sutter Instruments , Novato , CA ) with a Sutter pipette puller ( P-97 ) . The tip resistance of the recording pipettes was 2–3 MΩ after being filled with a pipette solution containing ( in mM ) : K-gluconate 125 , MgCl2 2 , HEPES 10 , EGTA 0 . 2 , Mg-ATP 4 , Na2- phosphocreatin 10 , and Na2-GTP 0 . 5 , pH 7 . 3 with KOH . The composition of the bath solution was as follows ( in mM ) : NaCl 124 , KCl 3 , CaCl2 2 , MgCl2 2 , NaH2PO4 1 . 23 , glucose 2 . 5 , sucrose 7 . 5 , and NaHCO3 26 . After a gigaohm seal and whole-cell access were achieved , membrane potential and action potentials were recorded under current clamp at 0 pA . ChR2 currents were recorded under voltage clamp mode . Light stimulation ( 470 nm , LED ) ( CoolLED pE-100 , UK ) was performed in the following configurations: 5 , 10 , 20 Hz ( 1 ms pulses , 20 pulses total ) ; 10 x 1 s ( 1 s pulse light ON , 1 s light OFF ) . All data were sampled at 3–10 kHz and filtered at 1–3 kHz with an Apple Macintosh computer using Axograph X ( Axograph X , Berkeley , CA ) . MedAssociates chambers ( MedAssociates , St . Albans , VT ) were equipped with two contact lickometers and a laser source ( solid state Crystal laser , 473 nm wavelength ) controlled by MedPC via a TT impulse to be triggered upon lick detection ( Domingos et al . , 2011 ) . The laser turns on every five consecutive licks on the same bottle ( Figure 1—figure supplement 1 ) . Animals were acclimated to the chambers until side preference for either bottle was even . During the acclimation and exposure periods mice were water deprived for 16–23 hr and were given water through the bottles inside the chamber for half an hour . In addition , stimuli in 10 min two-bottle tests were side balanced across the same genoptype group , being received either through the left bottle or through the right bottle . Two-bottle preference was calculated as the ratio: preference for 1 = number of licks on bottle 1/ ( number of licks on bottle 1+number of licks on bottle 2 ) and expressed as percentage values , with 50% representing the indifference ratio ( referred to as isopreference in the ‘Results’ section ) . Behavioral data was analyzed with Excel and Prism , and expressed as mean ± SEM . Significance tests comparing groups were ANOVAs or t tests and , when appropriate , followed by Bonferroni corrections for multiple comparisons . Two-bottle tests without laser stimulation were carried out in the same setup , with the laser turned off . The size of each animal group is represented by ‘n’ , and each animal was tested three times . The investigator was blind to the genotype . In all cases , concentration of sucrose was 0 . 4M and concentration of sucralose was 1 . 5 mM . These concentrations were based on previous literature ( consult supplementary figure-4 in Domingos et al . , 2011 ) . Briefly , the differences in molarity of sucrose and sucralose reflect differences in ligand-binding affinity of either sweetener to taste receptors , and were chosen among the plateau values of behavioral dose-response curves ( preference for either sweetener versus water in Domingos et al . , 2011 ) . Volume dispensed by the lickometers averages 2 μl/lick ( Domingos et al . , 2011 ) . Locations of optical probes were confirmed histologically ( data not shown ) . For each light stimulation regimen in Figure 2 , mice in top and bottom panels are the same . After animals were corrected for any spontaneous side bias , and prior to the 10-min testing data in Figure 2 , animals had a 10-min pre-exposure to either one of the two stimuli in two consecutive days . The pre-exposure procedure is intended to avoid novelty-related artifacts . On day one animals had exposure to water , followed by water+laser the day after . On the third day animals were tested for water vs water+laser for 10 min . On the fourth day , animals had exposure to sucrose , followed by sucralose+laser the day after . On the sixth day animals were tested for sucrose vs sucralose+laser for 10 min . During the experimental sessions microdialysate samples from the freely-moving mice were collected , separated and quantified by high-pressure liquid chromatography coupled to electro-chemical detection methods ( ‘HPLC-ECD’ ) . Briefly , after recovery from surgery and behavioral habituation , a microdialysis probe ( 2 mm CMA-7 , cut off 6 kDa , CMA Microdialysis , Stockholm , Sweden ) was inserted into the striatum through the guide cannula ( the corresponding CMA-7 model ) . After insertion , probes were connected to a syringe pump and perfused at 1 . 2 μl/min with artificial CSF ( Harvard Apparatus ) . After a 90 min washout period , dialysate samples were collected every 6 min and immediately manually injected into a HTEC-500 HPLC unit ( Eicom , Japan ) . Analytes were then separated via an affinity column ( PP-ODS , Eicom ) , and compounds subjected to redox reactions within an electro-chemical detection unit ( amperometric DC mode , applied potential range from 0 to ∼2000 mV , 1 mV steps ) . Resulting chromatograms were analyzed using the software EPC-300 ( Eicom , Japan ) , and actual sample concentrations were computed based on peak areas obtained from 0 . 5 pg/μl dopamine standards ( Sigma ) and expressed as % changes with respect to the mean dopamine concentration associated with baseline ( i . e . , behavioral task ) samples . Animals were water deprived for 16–23 hr , and rested in their home cages for baseline sample collection until values were stable . Chromatograms shown in Figures 3 and 4 are time-gated to the DA peak at 1 . 7 min . S0 denotes the pre-ingestion sample , and refers to the sample in which the animal was placed inside the behavioral box . Locations of microdialysis probes were confirmed histologically . The pre-embedding dual-labeling protocol of anti-GFP and anti-TH used in this study was adapted from Lane et al . ( 2010 ) . Briefly , vibratome sections were placed in 0 . 1% sodium borohydride and 0 . 1% glycine in 0 . 1 M phosphate buffer to remove excess aldehydes . Sections were incubated in a cryoprotectant solution ( 25% sucrose and 2 . 5% glycol in 0 . 05 M phosphate buffer ) , then immersed successively in liquid Freon and liquid nitrogen to freeze , and thawed at room temperature in 0 . 1 M phosphate buffer to enhance penetration of immunoreagents . Sections were incubated in 2% bovine serum albumin ( BSA ) in PBS to block non-specific labeling and then incubated for 42 hr at 4°C in a primary antibody solution containing both rabbit anti-TH ( P40101; 1:1000; Pel-Freez ) and mouse anti-GFP ( 1:1000; Millipore mab3580 ) antibody in 0 . 1% BSA in PBS . Detection of GFP was done first . Sections were incubated for 2 hr in biotinylated horse anti-mouse IgG ( 1:1000 ) and the immunoperoxidase–DAB procedure was applied using avidin-biotin complex ( Vectastain Elite ABC kit from Vector Laboratories ) , followed by diaminobenzidine and urea tablets ( Sigma ) for 10 min . The DAB reaction product was then silver-gold enhanced for 15 min using the Teclemariam method ( Teclemariam-Mesbah et al . , 1997 ) . After fixation in 0 . 5% glutaraldehyde , the detection for TH began: the sections were incubated for 2 hr in biotinylated horse anti-rabbit IgG ( 1:1000 ) and followed by the same steps used for GFP except , no silver enhancement was used . After post-fixation in 1% osmium tetroxide/1% Potassium ferrocyanide in 0 . 1 M cacodylate buffer ( pH 7 . 4 ) for 1 hr at 4°C , the sections were dehydrated in a graded ethanol series , propylene oxide and embedded in Eponate ( Ted Pella , INC ) . Blocks were cut with a diamond knife on a Leica UltracutE . Ultra-thin ( ∼70 nm ) sections were collected on uncoated 200 mesh grids . Unstained sections were viewed with a TecnaiSpiritBT Transmission Electron Microscope ( FEI ) at 80 KV and images were taken with Gatan 895 ULTRASCAN Digital Camera . Blood glucose tests were performed on mice that had been fasted for 24 hr beginning at the onset of the dark cycle . The following day mice were given an intraperitoneal injection of an aqueous solution of 10% glucose ( 10 ml/Kg body weight ) and blood glucose was measured from the tail vein at 0 and 10 min using an Ascensia Elite XL glucometer ( Bayer Health-Care , Tarrytown , NY ) . Once acclimated to the behavioral chamber , sweet blind Trpm5−/− mice were resented with one bottle containing sucrose or sucralose . The drinking behavior is quantified by monitoring licks with contact lickometers ( MedAssociates ) and the number of licks for each bottle was used to calculate the preference ratio for sucrose , as in Domingos et al . ( 2011 ) . To verify whether Trpm5−/− mice with dysfunctional MCH neurons could detect the post-ingestive effects of sucrose , we adapted a conditioning protocol as in de Araujo et al . ( 2008 ) , that allows the animal to manifest taste-independent preferences ( scheme in Figure 5 ) . All experiments were conducted with naive animals under a 16–23 hr water deprivation schedule . Animals were conditioned for 4 days with daily 30 min sessions of free access to either 1 . 5 mM sucralose or 0 . 4 M sucrose in one-bottle forced-choice training sessions . Either solution was presented on the opposite sides of the chamber on intercalated days . After training , on the 5th day , side bias was tested in 10-min two-bottle water versus water tests . This procedure was executed in Trpm5−/− mice with ablated MCH neurons and control Trpm5−/− mice with normal MCH neurons . | Sales of full-sugar fizzy drinks are almost triple those of diet versions , providing real-world confirmation of the laboratory finding that humans , as well as animals , prefer sugar to artificial sweeteners . However , it is not simply that sugary things taste better . Mice with a mutation that prevents them from perceiving sweet tastes still prefer the natural sugar sucrose over the artificial sweetener sucralose . This is because sugar , unlike artificial sweeteners , has nutritional value , and both humans and animals find it rewarding to consume foods with a high caloric content . Consuming sugar has been known to cause certain parts of the brain to release more of the chemical transmitter dopamine , which is used to signal reward , but exactly how this process produces a preference for sugar has been unclear . Now , Domingos et al . have revealed that a brain region called the lateral hypothalamus is responsible for this effect . This region of the brain—which helps to control appetite and which is also connected to the brain’s reward system—normally contains cells called MCH neurons . Domingos et al . show that the natural preference for sucrose over sucralose can be reversed by stimulating the MCH neurons with light , which in turn stimulates dopamine release in reward centers in the brain . Moreover , mutant mice that do not have any MCH neurons in the lateral hypothalamus show a reduced preference for sucrose over sucralose , compared to normal mice , and they release less dopamine than normal mice when they consume sucrose . By demonstrating that MCH neurons are both necessary and sufficient for sensing the nutritional value of sugar , these results provide new insights into the biological basis of sugar cravings . However , given the health implications of excessive sugar consumption , they may ultimately be used to find ways to make sugar less desirable , or to make artificial sweeteners more closely mimic the real thing . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2013 | Hypothalamic melanin concentrating hormone neurons communicate the nutrient value of sugar |
Apicomplexan actin is important during the parasite's life cycle . Its polymerization kinetics are unusual , permitting only short , unstable F-actin filaments . It has not been possible to study actin in vivo and so its physiological roles have remained obscure , leading to models distinct from conventional actin behaviour . Here a modified version of the commercially available actin-chromobody was tested as a novel tool for visualising F-actin dynamics in Toxoplasma gondii . Cb labels filamentous actin structures within the parasite cytosol and labels an extensive F-actin network that connects parasites within the parasitophorous vacuole and allows vesicles to be exchanged between parasites . In the absence of actin , parasites lack a residual body and inter-parasite connections and grow in an asynchronous and disorganized manner . Collectively , these data identify new roles for actin in the intracellular phase of the parasites lytic cycle and provide a robust new tool for imaging parasitic F-actin dynamics .
Toxoplasma gondii is a wide-spread obligate intracellular parasite that is thought to infect over two billion people worldwide . T . gondii infection of healthy individuals causes no major complications , infection can cause severe disease in immunocompromised individuals and foetuses infected in utero . The pathological manifestation is caused in a large part by repeated rounds of the parasite’s lytic cycle , beginning with active invasion of the host cell by the parasite , replication within a specialized vacuole termed the parasitophorous vacuole ( PV ) , followed by egress and lysis of the host cell . Replication occurs via a unique process of endodyogeny , where two daughter parasites are constructed within the mother , before elongating and budding , leading to the breakdown of the maternal parasite ( Francia and Striepen , 2014 ) . Remnants of the mother cell remain at the posterior end of the daughter parasites in a structure known as the residual body ( RB ) , which has a role in organizing the parasites into their characteristic rosette pattern within the PV ( Muñiz-Hernández et al . , 2011; Hu et al . , 2002 ) . Membrane connections persist between the RB and the parasites until host cell lysis and parasite egress , presumably to allow inter-parasite communication ( Muñiz-Hernández et al . , 2011 ) . To date little is known about the molecular mechanisms underlying these processes . Previous studies focused on the role of microtubules during parasite replication , since treatment with microtubule inhibitors leads to structural collapse of the parasite , while actin-modulating drugs are thought to cause only slight defects in replication ( Shaw et al . , 2000 ) . In contrast , research on parasite actin focused on its crucial role during host cell invasion and egress ( Dobrowolski and Sibley , 1996 ) . However , recent studies have also implicated actin and myosins in intracellular processes including apicoplast division ( Egarter et al . , 2014; Andenmatten et al . , 2013; Jacot et al . , 2013 ) , secretory organelle ( dense granule ) transport ( Heaslip et al . , 2016 ) and parasite replication ( Haase et al . , 2015 ) . Actin is a highly conserved protein , which forms dynamic filaments in eukaryotic cells . Through an association with actin-binding proteins , these filaments are themselves organized into higher order structures that play important roles in a wide variety of cellular functions including muscle contraction , vesicle transport and cytokinesis . T . gondii actin is encoded by a single gene , act1 and has only ~80% sequence identity with mammalian actin isoforms but shares 93% similarity with Plasmodium ACT1 ( Dobrowolski et al . , 1997 ) . Apicomplexan ACT1 is clearly essential , and compared to its counterparts in higher eukaryotes is believed to be intrinsically unstable , resulting in the formation of only short filaments ( Skillman et al . , 2011 ) . Biochemical assays indicate that 97% of the parasites actin is present in the globular form ( Dobrowolski et al . , 1997; Skillman et al . , 2011; Wetzel et al . , 2003 ) . It has been proposed that apicomplexan actin is unique amongst actins as it polymerizes in a highly unusual , isodesmic manner ( Skillman et al . , 2013 ) . According to the isodesmic polymerisation model , monomer addition is governed by a single equilibrium constant , meaning that no ( unfavourable ) activation step is required to initiate the formation of the first dimer leading to polymerisation . In this instance , nucleation and elongation are equally favourable . This contrasts to cooperative polymerisation , where the activation step is the formation of the first dimer/trimer , which has a higher equilibrium constant than polymer elongation ( Smulders et al . , 2010 ) . Therefore polymer formation can only occur above a critical concentration ( Cc ) of monomers ( Pantaloni et al . , 1985 ) . It is this activation step that is regulated by actin nucleators , such as the Arp2/3 complex or formins ( Carlier et al . , 2015 ) . Puzzlingly , formins and other actin nucleating proteins have been shown to have essential roles in Toxoplasma and Plasmodium , begging the question of their function if they are not required to initiate actin polymerisation or accelerate filament elongation ( Baum et al . , 2008; Jacot et al . , 2013 ) . A recent study suggested that the polymerization process of apicomplexan actin needs to be reinvestigated , as heterologously expressed apicomplexan actin , the basis for many of the previous studies , is incorrectly folded ( Olshina et al . , 2016 ) . Furthermore , it was demonstrated that conditional deletion of act1 in T . gondii results in complete abrogation of known actin functions , long before G-actin levels are fully depleted , suggesting that in vivo the formation of F-actin depends on a critical monomer concentration ( Whitelaw et al . , 2017 ) . This raises concerns about previous studies of actin polymerization kinetics based on mis-folded actin . Furthermore , imaging studies on different life cycle stages of the apicomplexan parasite , Plasmodium falciparum , demonstrated the formation of an extensive F-actin cytoskeleton in both gametocytes and ookinetes ( Hliscs et al . , 2015; Siden-Kiamos et al . , 2012 ) . Using 3D-SIM it was demonstrated in gametocytes that these filaments appear to be organized in the cytoplasm , below the Inner Membrane Complex ( IMC; a specialised structure found in apicomplexans that consists of membranous vesicles and structural components located just beneath the plasma membrane ) of the parasite rather than between the IMC and the plasma membrane ( Hliscs et al . , 2015 ) , where it is thought to power parasite gliding motility ( Meissner et al . , 2013 ) . To further address the role of ACT1 in parasite growth , we characterized a recently generated conditional actin knockout parasite line ( act1 cKO ) during the intracellular portion of the parasites lytic cycle ( Andenmatten et al . , 2013 ) . We demonstrate that act1 cKO parasites lack a residual body and grow in an asynchronous and disorganized manner within the PV . Further investigation of actin functions required imaging the filamentous actin cytoskeleton . Previous attempts to visualise F-actin within the parasites have largely been unsuccessful , since conventional actin labelling techniques such as Life-Act , Phalloidin and SiR-Act do not allow detection of F-actin within the parasites . GFP-actin shows a high signal to noise ratio in parasites causing the inability to differentiate actin filaments from the monomeric form Angrisano et al . ( 2012a ) . Thus we sought a new approach to image F-actin in intracellular parasites by expressing actin-chromobody , a single chain anti-actin camel antibody that has been successfully employed in diverse eukaryotes , including plants ( Rocchetti et al . , 2014 ) and animal cells ( Plessner et al . , 2015; Panza et al . , 2015 ) . Chromobodies ( Cb ) were found to have several advantages compared to other actin probes , such as lower toxicity , less influence on F-actin dynamics and a high signal to noise ratio ( Panza et al . , 2015 ) . Using this probe , we were able to visualise filamentous actin in live parasites . While some filamentous structures were seen within the parasite cytosol as expected , we were surprised to observe extensive F-actin networks in the RB and F-actin-containing membranous tubules linking parasites within a vacuole . These tubules appear to be involved in the transport of material between parasites and recycling of the mother IMC at the end of the replication cycle . Collectively , these data identify new roles for actin in the intracellular phase of the parasites lytic cycle and provide a robust new tool for imaging parasitic F-actin dynamics .
We previously characterised a conditional knockout for Toxoplasma actin ( act1 cKO ) and found that , in addition to its important role during gliding motility and host cell invasion , parasite actin is essential for maintenance of the apicoplast , dense granule motility and host cell egress ( Egarter et al . , 2014; Heaslip et al . , 2016 ) . Additionally , a recent study demonstrated a role for actin in parasite replication ( Haase et al . , 2015 ) and thus we used the act1 cKO to further examine the role of actin in parasite growth . Although initially intracellular act1 cKO parasites replicate at comparable rate to wild-type parasites ( Egarter et al . , 2014 ) , they grew asynchronously and appeared disorganised , without the characteristic rosette organization of parasites in the parasitophorous vacuole ( Figure 1A and B ) . While parasites replicate normally up to the 4 cell stage , later divisions are not tightly synchronised ( Figure 1A ) , meaning that within the same PV parasites can be identified that are at different stages of endodyogeny ( Figure 1B , arrow ) . In addition , the intravacuolar network , as visualised using Gra2 antibodies ( Mercier et al . , 1993 ) , appeared disorganised and malformed ( Figure 1A ) . Using scanning electron microscopy ( SEM ) of intracellular parasites , we observed that in wild-type ( RH ) T . gondii a membranous network connects parasites at their posterior pole whereas no obvious connections between individual parasites were observed in the act1 cKO ( Figure 1B ) . In addition , imaging of extracellular parasites by SEM and immunofluorescence showed aberrant morphology at the posterior pole in the act1 cKO parasites , which appears generally flattened ( Figure 1C , arrow ) . Together , these data suggest a role of actin in the formation of the RB , the organization of parasites within the PV and the intravacuolar network . Furthermore , asynchronous replication of parasites within the PV may indicate a loss of signalling between individual parasites . 10 . 7554/eLife . 24119 . 003Figure 1 . Analysis of intracellular development of act1 cKO . ( A ) Conditional null mutants for act1 were induced as previously described in order to excise act1 ( Egarter et al . , 2014 ) . 96 hr post induction , parasites were fixed and stained with anti-Gra2 to visualise the intravacuolar network . Scale bars: 10 µm . Replication assay of loxPact1 and act1 cKO . Number of nuclei per parasitophorous vacuole were counted 24 hr after inoculation on HFF cells . Mean values of three experiments in triplicate are shown . Asynchronous division is indicated by PVs with unusual amounts of parasites ( ≠2n ) . Two-way ANOVA followed by Sidak’s test were used to compare means between groups . ****p<0 . 0001 , ***p<0 . 001 , non-significance ( ns ) p>0 . 05 . For source data see supplemental information ( Figure 1A—source data 1 ) . ( B ) Top image: The developing IMC of parasites is shown using IFA . Note that in case of act1 cKO individual parasites are in different stages of replication ( i . e . white arrow indicates a parasite at the end of replication ) . Scale bars: 10 µm . Bottom image: Scanning EM analysis in intracellular parasites . Scale bars: 2 µm . Note the disordered appearance of parasites within the PV in absence of ACT1 and flattened bottom of the act1 KO parasites ( Red Arrowhead ) . ( C ) Analysis of the IMC of the act1 cKO using IFA and scanning EM analysis in extracellular parasites . act1 cKO parasites have a flattened bottom , torpedo shape ( Red arrowhead ) . The posterior pole of the parasite appears to be misformed , indicating a role of ACT1 during the final stages of replication ( see also Figure 9 ) . Scale bars: fluorescence images: 5 µm , SEM: 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 00310 . 7554/eLife . 24119 . 004Figure 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 004 Based on electron microscopy evidence , the RB has been predicted to be involved in the transfer of material between parasites within a vacuole ( Muñiz-Hernández et al . , 2011 ) although this has not been previously demonstrated . To investigate this hypothesis and the role of actin in this process , we infected host cells with wt or act1 cKO parasites expressing GFP and bleached individual parasites within the PV before measuring the time of fluorescence recovery ( Videos 1 and 2; Figure 2A ) . While the fluorescence signal recovered rapidly in wt parasites ( ~30 s ) , no fluorescence recovery could be observed in parasites depleted of actin , indicating a defect in material transfer ( Figure 2A ) . From this experiment , we concluded that individual parasites remain connected via the intravacuolar network and can exchange cytoplasmic material . Furthermore , these connections appear to be actin dependent . 10 . 7554/eLife . 24119 . 005Figure 2 . Material transport in between individual parasites within a PV . ( A ) FRAP treatment in a conditional act1 cKO 72 hr after excision of act1 ( Left panel ) . FRAP treatment shows no recovery in fluorescence intensity in the bleached cell ( arrow ) over the duration of the experiment ( 90 s ) , indicating that the absence of actin abrogates transport of constitutively expressed YFP between neighbouring cells . FRAP treatment in a wt strain constitutively expressing GFP shows recovery in a bleached cell as soon as 20 s after bleaching ( right panel , control experiment ) . Intensity in the FRAP area was expressed as intensity percentage of the same area in a cell unbleached . Scale bar; 5 µm . See also Videos 1 and 2 . FRAP experiments shown are representative for several biological replicates ( n > 3 ) . ( B ) Extracellular vesicles are actively transported in wt parasites . Vesicles positive for GAPM1a-YFP were seen to be transported . Individual vesicles were tracked and their path color coded by frame according to indicated scale . Time is indicated in minutes . Scale bar: 5 µm , detail 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 00510 . 7554/eLife . 24119 . 006Video 1 . FRAP on act1 cKO . FRAP treatment in a conditional act1 cKO 72 hr after excision of act1 in loxPAct1 . After FRAP treatment in a cell in the PV , no recovery in fluorescence intensity was observed . Scale bar 5 µm . Imaging speed 5 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 00610 . 7554/eLife . 24119 . 007Video 2 . FRAP on RH-YFP . FRAP treatment in a RH strain constitutively expressing GFP shows recovery in a bleached cell after a period of 20 s . Scale bar 5 µm . Imaging speed 5 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 007 To determine if vesicular as well as cytoplasmic material could be transferred between parasites , we performed time-lapse analysis of the integral membrane protein GAPM1a-YFP ( Figure 2B ) and identified vesicles moving in a directed manner , demonstrating that vesicular transport occurs outside of the parasite ( Figure 2B , Video 3 ) . Of note , digital tracking of vesicles suggested a directional movement along a tubular or filamentous structure . 10 . 7554/eLife . 24119 . 008Video 3 . Vesicular motility of GAPM1A-YFP in parasites endogenous expressing GAPM1a-YFP vesicles positive for GAPM1a-YFP were seen to be transported extracellularly . A vesicle was tracked and its path indicated . Time is indicated in minutes , scale bar 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 008 In summary , these data indicate that individual parasites within a vacuole remain connected via a network , possibly through the RB that is required to transfer both cytoplasmic and membrane bound material between individual parasites . We hypothesize that parasite actin is required for the formation and/or maintenance of this interconnecting network . We hypothesised that actin may be involved in residual body formation and inter-parasite communication . However , F-actin has not previously been visualised in live parasites , complicating a functional characterisation of F-actin and F-actin dynamics in apicomplexan parasites . In order to visualize F-actin in T . gondii , we modified a commercially available actin-chromobody ( Cb ) that has been established as a novel tool to study F-actin dynamics in living cells . This single chain camel antibody specifically recognizes F-actin and has been successfully employed to study actin dynamics in diverse eukaryotes . We generated two expression vectors for Cb , where Cb is fused to either Emerald GFP ( Cb-EmGFP; [Shaner et al . , 2007] ) or Halo ( Cb-Halo; [Los et al . , 2008] ) . Upon transient expression of these proteins , we obtained identical staining of filamentous structures within the parasites . Cb was also localised somewhat diffusely in the parasite cytosol , probably corresponding to unbound Cb in the cytosol as the protein was expressed at a high level ( Figure 3A ) . Strikingly , individual parasites within the PV remain connected by a filamentous actin network which stretches throughout the RB and parasitophorous vacuole and can be seen to enter the posterior pole of individual parasites ( Figure 3A , arrow ) . With increasing size of the parasitophorous vacuole , an impressive intravacuolar network consisting of F-actin became apparent ( Figure 3B ) that appeared to not only connect individual parasites , but also reached from the centre to the edge of the PV ( Figure 3B , Figure 7 ) . 10 . 7554/eLife . 24119 . 009Figure 3 . Filamentous actin can be visualized by expression of Cb-Halo and Cb-Emerald . ( A ) Sections of parasite vacuoles in two , four , and eight-cell stage stably transfected with Cb-Halo or Cb-Emerald ( red ) . A filamentous network connecting individual cells ( arrows ) can be visualised using Cb-Halo and Cb-Emerald . Parasites were co-stained with IMC marker , GAP45 ( green ) , DAPI staining ( blue ) Scale bar 5 µm . ( B ) Images of larger parasite vacuoles containing 16 or 32 parasites transfected with Cb-Emerald ( red ) . Parasites were co-stained with IMC marker GAP45 ( green ) . Note the formation of the extensive intravacuolar network with long filamentous tubes ( see also Figure 7 ) . Scale bar 10 µm ( C ) Top . Image of parasites expressing SAG1ΔGPI-GFP to label the dense granules and low and high levels of Cb-Halo . Scale bar 10 µm . Bottom . Directed granule run frequency in control ( non-expressing ) parasites and parasites expressing Cb-Halo at high and low levels . ***p<0 . 0001; students t-test . Total number of directed runs counted in control , low expression and high expression samples were 183 , 150 and 1 respectively . Total number of vacuoles analysed from control , low expression and high expression were 19 , 17 and 14 , respectively , from two independent transfections . Error bars are S . E . M . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 009 During transient expressions of the fusion proteins , the majority of intracellular parasites appeared healthy with no apparent morphological changes , indicating that expression of Cb-proteins is well tolerated by the parasite . To analyse the effects of Cb expression on actin dynamics , we used dense granule motility as a surrogate marker , since it has been shown to critically depend on actin dynamics ( Heaslip et al . , 2016 ) . Parasites stably expressing SAG1-ΔGPI-GFP ( to label the dense granules ) were transiently transfected with Cb-Halo and dense granule motility was analysed and correlated to mean fluorescent intensity of Cb-Halo within the parasite ( Video 4; Figure 3C ) . Strong expression of Cb-Halo leads to an almost complete block of dense granule motility , however , weaker expression levels are well tolerated with no significant change in either dense granule run frequency ( Figure 3C ) or run length compared to controls ( 935 ± 39 nm vs 1013 ± 53 nm for control and Cb-Halo , respectively ) . 10 . 7554/eLife . 24119 . 010Video 4 . Dense Granule dynamics in intracellular T . gondii parasites expressing SAG1△GPIGFP and low or high levels of CB-Halo . Imaging speed 10 fps , playback 6x real time . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 010 In summary , expression of Cb is well tolerated by the parasite indicating that it does not significantly adversely affect normal actin functions within the cell , in agreement to data obtained in other cellular systems ( Belin et al . , 2014; Panza et al . , 2015 ) . However , it cannot be ruled out that actin dynamics are locally affected within the cell due to expression of Cb . As transient transfection resulted in a heterologous population of parasites expressing Cb at varying levels , which may be deleterious for the parasites , we generated stable parasite lines expressing either Cb-Halo or Cb-Emerald and confirmed that parasite growth , invasion , replication and egress rates are indistinguishable from wt parasites ( Figure 4A–E ) . To ensure minimal influence of Cb expression on actin dynamics in our stable line , we also analysed gliding motility , an important process depending on F-actin dynamics ( Dobrowolski and Sibley , 1996; Drewry and Sibley , 2015; Jacot et al . , 2013; Opitz and Soldati , 2002; Skillman et al . , 2011; Wetzel et al . , 2003 ) . We found that the rate of overall motility is slightly increased ( p<0 . 05 ) upon expression of Cb ( Figure 4F ) , indicating an influence of Cb expression on actin dynamics . Intriguingly , average trajectory lengths are identical in wt and Cb expressing parasites ( Figure 4G ) , while average gliding speed is reduced ( Figure 4H ) . We also analysed average invasion speeds and confirmed that parasites penetrate the host cell within 30 s irrespective of Cb expression ( Figure 4D ) . Furthermore , time lapse imaging of Cb-Emerald demonstrated a highly dynamic behaviour of F-actin within the cytosol ( Video 5 ) . Together these data demonstrate that , similar to the situation in other eukaryotes , the expression of this actin binding protein has no or modest effects on established actin-dependent processes in T . gondii . Moreover , as we identified little phenotypic consequence due to stable Cb expression in actin function , this reagent will be a useful tool for detecting parasite F-actin in live cells , and characterizing actin dynamics and organization . 10 . 7554/eLife . 24119 . 011Figure 4 . Phenotypic analysis of parasite expressing Cb-Halo . ( A ) Left . Growth assay of indicated parasites . After 5 days growth no significant difference in growth rates can be observed . Right . Representative image of plaque area size created . Scale bar , 100 μm . ( B ) Replication rates between Cb-Halo and RH parasites are comparable . Indicated parasites were inoculated on HFF cells and number of parasites/vacuole was determined . ( C ) Invasion rates are not significantly different between RH and Cb-Halo expressing parasites . Parasites were allowed to invade for 1 hr , before non-invaded parasites were removed . ( D ) Penetration time of parasites was determined using time-lapse analysis . Both RH and Cb-Halo expressing parasites are capable to invade the host cell within 30 s . In some cases slower parasites can be detected , although on average the difference is insignificant . n = 22 independent events . ( E ) No differences in egress could be detected between RH and Cb-Halo parasites . Egress was triggered using Calcium Ionopore A23187 and the number of egress events was determined after 10 min . ( F ) Trail deposition assay comparing gliding rates between RH and Cb-Halo . Cb-Halo parasites form slightly more trails ( *p<0 . 05 ) . ( G , H ) Comparison of gliding motility between RH and Cb-Halo . Whereas average run length ( G ) is identical , parasites expressing Cb-Halo demonstrate slower gliding speeds for helical and circular motility . Parasites were tracked with Fiji wrMTrck software . N = 20 individual events for each condition . All assays were conducted in triplicates . Datasets were compared using two-tailed Student’s t-test . Error bars for A , B , C , E , F represent S . E . M from three independent , biological replicates . Error bars for D , G , H represent 95 % CI . *p<0 . 05 , ****p<0 . 0001 . For source data see supplementary information ( Figure 4—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 01110 . 7554/eLife . 24119 . 012Figure 4—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 01210 . 7554/eLife . 24119 . 013Video 5 . Parasites expressing Cb-Emerald demonstrate highly dynamic F-actin dynamics . After FRAP treatment in a cell in the PV , recovery of fluorescence intensity is observed due to polymerization of new actin filaments formed from the cell periphery to the bleached cytoplasm region . Scale bar 5 µm . Imaging speed 5 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 013 To further validate Cb for its use in apicomplexan parasites , we next analysed binding characteristics of Cb to F-actin . Since apicomplexan actin cannot be functionally heterologously expressed ( Olshina et al . , 2016 ) , we used skeletal chicken actin to estimate the binding characteristics of Cb to F-actin . Recombinant Cb with a C-terminal 6-His tag ( rCb ) was expressed and purified from bacteria ( Figure 5A , B ) . rCb bound to chicken skeletal actin with Kd of 5 ± 1 µM ( Figure 5B ) . To confirm that Cb-Halo binds specifically to Toxoplasma F-actin when expressed in the parasite , we co-immunoprecipitated actin from extracellular parasites using an anti-Halo antibody ( Figure 5C ) . As expected , only a small proportion of actin was precipitated , which corresponds well with published data suggesting that only ~2% of total parasite actin can be found in its filamentous form in extracellular parasites ( Dobrowolski and Sibley , 1997 ) . Importantly , mass-spectrometric analysis of the immunoprecipitation confirmed that Cb-Halo specifically precipitated parasite actin . Although additional , potential F-actin binding proteins were detected , no host cell actin , actin related or actin-like proteins could be identified ( not shown ) . To assess if expression of Cb stabilises F-actin in parasites , we compared the amount of pelletable actin in parasites expressing Cb and wt parasites ( Figure 5D ) . We confirmed that in both cases F-actin is barely detectable in extracellular parasites , indicating that it is primarily found in globular form . Treatment of parasites with Jas allowed detection of significant amount of F-actin in the pellet . Of note , no significant difference could be observed between control parasites and parasites expressing Cb . Together , these data confirm that expression of Cb in parasites allows detection of F-actin and that Cb expression has only minor effects on F-actin dynamics , as previously shown in other eukaryotes ( Panza et al . , 2015 ) . 10 . 7554/eLife . 24119 . 014Figure 5 . Chromobodies are specific for parasite F-actin and do not influence total amount of F-actin in the parasite . ( A ) Coomassie stained gel showing recombinant Cb purified from bacteria . ( B ) Left . rCb affinity assays . Coomassie stained gels showing in vitro binding of purified Cb ( 4 μM ) to variable range of skeletal chicken F-Actin concentration ( 80 μM down to 0 μM ) . Supernatant ( S ) and pellet ( P ) were separated by ultracentrifugation . Right . Quantitative analysis . Ratio between purified Cb in the supernatant and pellet was determined . Solid line is a fit of the binding equation to the data ( Kd = 5 ± 1 . 2 mM ) . Results obtained from two independent experiments . ( C ) Interaction between Cb-Halo and actin in T . gondii Cb-Halo expressing strain . Western blot comparison of input lysate ( I ) and elution ( E ) obtained from co-immunoprecipitation using beads against the halo-tag with the Cb-Halo strain and RH . Actin pull-down was only detected in the Cb-Halo expressing strain . ( D ) Sedimentation assays . Actin sedimentation , with and without Jasplakinolide ( 1 μM ) was evaluated for Cb-Halo strain and RH . GRA7 was used as loading control and signal intensity normalisation between conditions . Increased amount of F-actin was found in the pellet of parasites incubated in the presence of Jasplakinolide . However , no difference between RH and parasite expressing Cb-Halo could be detected in both control and Jas treated condition ( n = 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 014 Next , we wished to determine if Cb binds specifically to F-actin within the parasites . Using actin-modulating drugs , the actin cytoskeleton was either depolymerized with Cytochalasin D ( Cyt-D ) or stabilized with Jasplakinolide ( Jas ) . Treatment of intracellular parasites with Cyt-D led to the disintegration of the inter-parasite connections with only punctate spots remaining , which were also identified with anti-actin antibody ( Figure 6A ) . In contrast , treatment with Jas led to the stabilisation of F-actin and intra-parasite connections were easily detectable with either Cb or an anti-actin antibody ( Figure 6A ) . Using default settings of the Ridge Detector plugin ( Steger , 1998 ) we determined the apparent number , maximum length and average filament size of PVs treated with Jas , Cyt-D or untreated PVs ( control ) ( Figure 6A right panel ) . In Jas-treated PVs , and in contrast to Cyt-D treated PVs , there is an increase in average and maximum filament size , together with a decrease in total number of short filaments supporting the stabilisation and depolymerisation of Cb-Halo detected filaments in the presence of Jas and Cyt-D respectively . An intermediate situation occurs in untreated PVs with long filaments and high number of short filaments supporting a dynamic regulation in actin polymerisation in control samples . Overall , these results support that filaments detected with Cb-Halo behave as typical actin filaments in the presence of actin regulatory drugs . This supports the evidence that Cb-Halo is specifically labelling F-actin filaments and that Cb binding to actin does not prevent depolymerisation caused by Cyt-D . 10 . 7554/eLife . 24119 . 015Figure 6 . Chromobodies specifically stain parasite F-Actin . ( A ) 3D-SIM imaging of parasites transiently expressing Cb-Halo ( red ) and stained using actin antibody ( green ) ( Angrisano et al . , 2012b ) . Treatment with 100 nM Jas for 1 hr results in formation of an elaborate network , whereas treatment with 2 μM Cyt-D , the filamentous network collapses . Note that in the bottom images a field of view was selected were a transiently transfected PV expressing Cb-Halo is next to a non-transfected PV . Both show identical staining . Scale bars 2 μm . Right: Quantification of the apparent filament size in five representative PVs growth for 24 hr , as measured using default settings in the Ridge Detection Plug In ( ImageJ , see Material and Methods ) , n is the number of filaments , max size is the longest filament found in each condition expressed in µm . Number below indicates average pixels +/- SEM ( 10 pixels correspond to 1 µm ) . ( B ) Expression of Cb-Halo in a conditional act1 cKO . A filamentous network is observed prior to excision of act1 . As soon as 24 hr after excision of act1 the network diminishes . Right: The percentage of vacuoles with >8 parasites containing actin filaments in RH Δhxpgrt and act1 cKO were quantified at 24 , 48 , 72 and 96 hr after induction . Mean values of three experiments in triplicate are shown . One way ANOVA followed by Dunnett’s multiple comparisons test was used to compare means between groups . ***p<0 . 0001 . ( C ) Host cell actin is not involved in formation of the filamentous network . Parasites were allowed to replicate for 24 hr , before being treated for 3 hr with indicated concentration of Latrunculin A . Host cell F-actin was visualised with Phalloidin488 ( green ) . Scale bars; 10 μm . ( D ) Analysis of actin dynamics in a conditional mutant for adf . Left: Immunoblot using indicated antibodies . ADF-HA is depleted upon treatment of parasites with 1 μM ATc . Parasites were grown for 96 hr in HFF cells in the presence and absence of inducer , before being artificially released . Equal amounts of parasites were loaded . Right: Representative still image of Video 6 and 7 . Parasites were grown for 96 hr in the presence or absence of ATc . Note that upon depletion of ADF no actin filaments can be detected in the cytosol of the parasites ( arrows ) and F-actin accumulates at the posterior and ( to a lesser extent ) apical pole of the parasite . Scale bar: 5 μm . For source data ( A , B ) see supplementary information ( Figure 6—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 01510 . 7554/eLife . 24119 . 016Figure 6—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 016 To confirm that parasite F-actin corresponds to Cb-positive structures , we expressed Cb-Halo in act1 cKO cells . Using these parasites , we found that as early as 24 hr after excision of the act1 gene , the filamentous network collapsed , leaving punctate actin spots visible only in the RBs ( similar to vacuoles treated with Cyt-D ) . By 48 hr onwards , no F-actin structures could be observed and Cb-Halo was completely cytosolic , as expected due to cytosolic expression of this reagent ( Figure 6B ) . The loss of filaments within the act1 cKO over-time follows the down-regulation of ACT1 in act1 cKO , indicating a polymerisation mechanism similar to eukaryotic actins ( Whitelaw et al . , 2017 ) . To exclude a role of host cell actin in the formation of the inter-parasite connections , we treated infected host cells with the actin-disrupting drug latrunculin A , which specifically inhibits polymerization of host cell , but not parasite , actin ( Hegge et al . , 2010; Whitelaw et al . , 2017 ) . We found that latrunculin A treatment for 3 hr led to complete disruption of host cell F-actin ( visualised using Phalloidin488 ) , while the parasite F-actin network remained unaffected . This confirms that Cb-positive filamentous structures were not derived from host cell actin ( Figure 6C ) . Next we wished to compare F-actin dynamics in a conditional mutant for the actin depolymerisation factor ( ADF; [Mehta and Sibley , 2011] ) . Previous studies demonstrated that the knockdown of ADF leads to stabilisation of actin filaments ( Mehta and Sibley , 2011 ) and apicoplast loss ( Haase et al . , 2015 ) . When we expressed Cb-Emerald in adf cKD to compare actin dynamics , we found that in absence of anhydrotetracycline ( ATc ) - when ADF is expressed - parasite F-actin shows a highly dynamic behaviour similar to wt parasites ( Figure 6D , compare Videos 6 and 7 ) . In stark contrast , incubation of parasites with ATc for 96 hr leads to depletion of ADF and consequently F-actin dynamics is significantly diminished . Under these conditions F-actin filaments accumulate at the posterior and to a lesser extent at the apical pole of the parasites . However , no dynamic behaviour can be detected and no filaments can be detected within the cytosol of the parasites ( Figure 6D ( red arrows ) , Videos 6 and 7 ) . 10 . 7554/eLife . 24119 . 017Video 6 . Analysis of actin dynamics in adf cKO expressing Cb-Emerald in absence of ATc . Note the highly dynamic F-actin within the cytosol of the parasites . Scale bar 5 µm . Imaging speed 5 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 01710 . 7554/eLife . 24119 . 018Video 7 . Analysis of actin dynamics in adf cKO expressing Cb-Emerald in the presence of ATc ( when ADF is depleted ) . Actin dynamics is almost completely abolished and F-actin can be found concentrated at the posterior end and much less at the apical tip of the parasite . Scale bar 5 µm . Imaging speed 5 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 018 As we had now established that endogenously expressed Cb bound specifically to parasite F-actin , we sought to investigate the dynamics of the inter-parasite F-actin network during the parasite’s life cycle . Using time-lapse analysis , we found that in non-replicating parasites F-actin forms an extended , continuous filamentous network through the RB , connecting the parasites . During parasite replication , this network collapses and F-actin retreats to the RB ( asterisks ) before the network reform again , extending throughout the PV ( Figure 7A , Video 8 ) . Note that these filaments appear relatively static in resting parasites and do not show much reorientation/movement within 8 hr . However , once parasites start to replicate within the PV , the connections disintegrate and F-actin appears to be restricted to the RB ( Figures 7A and 8A , Videos 8 and 12 ) . 10 . 7554/eLife . 24119 . 019Figure 7 . The F-actin network is stable in resting parasites , but highly dynamic during replication and egress . ( A ) Analysis of Cb-Halo during two rounds of replication . Images were taken every 30 min for 20 hr ( Video 8 ) . The network appears dynamic across the intracellular lifecycle , collapsing into rings during daughter cell emergence ( asterisks ) . Time indicated in hours . Scale bar 5 μm . ( B ) Collapse of the F-actin network can be triggered by calcium signaling . Parasites were induced for egress with Calcium Ionophore A23187 and imaged at 1 frame per second ( Video 9 ) . The network collapses before parasites begin to egress . While filaments quickly collapse , the residual body remains intact during egress and is left behind . Box in lower left image shows freshly egressed parasites ( enlarged below ) , where F-actin appears to accumulate at the posterior pole of the parasite . Time indicated in minutes:seconds . Scale bar , 10 μm . ( C ) Correlative light electron microscopy ( CLEM ) . A vacuolar network was imaged with 3D-SIM super-resolution microscopy and the same areas were imaged with TEM . Filaments of 5 nm in thickness were present within the network tubules , extending over 100 nm in length . Scale bars: 200 nm ( 3D-SIM ) ; 50 nm ( TEM ) . ( D ) FRAP treatment in cells stably expressing Cb-Emerald . The F-actin inside the cells ( left panel ) and the nanotubular network connecting the parasites ( right panel ) show different fluorescence recovery times ( 20 and 60 s respectively ) , suggesting the presence of different actin dynamics inside the parasite and the filamentous network of the PVs respectively . Intensity in the FRAP area was expressed as intensity percentage of the same unbleached area ( filament or nanotubular network ) . Time is expressed in seconds . Scale bar 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 01910 . 7554/eLife . 24119 . 020Figure 8 . Parasite-derived extracellular vesicles are transported in an F-actin dependent manner . ( A ) Parasites co-expressing GAPM3-YFP and Cb-Halo were imaged every 6 min for 5 hr , F-actin can be seen initially connecting the basal end of the parasites before accumulating beneath the forming daughter cells where it appears to concentrate towards the rear of the new daughters during emergence and recycling of the maternal IMC . Note the sudden collapse of the mother IMC into vesicles that appear to move towards the IMC of the nascent daughter cells ( arrow ) . Scale bar 10 μm . ( B ) In parasites endogenously expressing GAPM1a-YFP , extracellular vesicles could be observed in close apposition to Cb-Halo labelled filaments ( arrow ) . Parasites expressing Cb-Halo were imaged every second for up to 5 min . Extracellular vesicles positive for GAPM1a-YFP were observed to move along F-actin filaments ( arrows ) . Scale bar , 5 μm . ( C ) The number of vesicles per vacuole that moved within 5 min of imaging were quantified in the presence and absence of 500 nM Cyt-D . At least 60 vacuoles were counted over three independent experiments . ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 02010 . 7554/eLife . 24119 . 021Video 8 . Analysis of Cb-Halo during two rounds of replication . Images were taken every 30 min for 20 hr . The network appears dynamic across the intracellular lifecycle , collapsing into rings during daughter cell emergence before reforming . Scale bar 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 021 As the parasites appeared to be connected by F-actin filaments within a vacuole , we wondered how the organization of this network changes as the parasite exits from the host , especially as act1 cKO parasites are unable to egress ( Egarter et al . , 2014 ) . When we triggered a calcium signalling cascade using Ca2+-Ionophore ( Black et al . , 2000 ) , the F-actin network collapsed rapidly , between 10 and 60 s after Ca2+-Ionophore addition ( Figure 7B , Video 9 ) . Collapse of the network preceded the initiation of motility and egress from the host cell ( Figure 7B , arrow head; Video 9 ) . As the parasites begin to move away from the lysed host cell , F-actin can be detected at the rear of the parasites ( Figure 7B , inset ) and residual filamentous actin is still seen within the RB . 10 . 7554/eLife . 24119 . 022Video 9 . Imaging of F-actin dynamics after addition of Ca2+-Ionophore , images taken every second . Filaments break up in a calcium dependent manner before parasites start to egress . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 022 To investigate the presence of F-actin within this network , we performed correlative light-electron microscopy ( CLEM ) . The network within a vacuole was imaged using 3D structural illumination super-resolution microscopy ( 3D-SIM ) ( Figure 7C ) . Thin sections of the same network were then imaged with transmission electron microscopy ( squares ) . This demonstrated that extracellular F-actin filaments reside within membranous tubules of 50–60 nm in diameter . Within these tubules , several ~5 nm thick filaments ( arrows ) extending over 100 nm in length were observed ( highlighted in magenta ) . Taken together these results confirm the presence of bundles of actin filaments bound within a membranous network , which connects individual parasites within the PV . This situation appears very similar to the formation of tunnelling nanotubes , long filopodia like structures , which consists of thin F-actin-based membranous structures with a small diameter ( 20–500 nm ) that facilitate long range communications between cells ( Abounit and Zurzolo , 2012 ) . Given that the inter-parasite tubules are reorganized during both replication and egress , we investigated the dynamic behaviour of F-actin within individual tubules using fluorescence recovery after photobleaching ( FRAP ) . While F-actin dynamics within the parasite cytosol is very fast ( Video 10 ) and recovery rates very rapid , within 20 s ( Figure 7D ) , we found that F-actin within the tubular network is much more stable and fluorescence labelling of these structures took more than 60 s to fully recover ( Video 11 ) . This demonstrates the presence of highly dynamic , intracellular F-actin and a stable F-actin containing filamentous network ( Figure 7D ) . 10 . 7554/eLife . 24119 . 023Video 10 . FRAP of cytosolic filaments . FRAP treatment in cells stably expressing Cb-Emerald . After FRAP treatment , the F-actin inside the cell shows a fast fluorescence intensity recovery time of 20 s . Scale bar 5 µm . Imaging speed 5 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 02310 . 7554/eLife . 24119 . 024Video 11 . FRAP of filamentous structure . FRAP treatment in cells stably expressing Cb-Emerald . After FRAP treatment in the nanotubular network , F-actin shows a fluorescence intensity recovery time of 60 s . Scale bar 5 µm . Imaging speed 5 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 024 Finally , given that act1 cKO parasites divided asynchronously we wanted to readdress the role of actin in parasite replication . We generated parasites co-expressing GAPM1a-YFP and Cb-Halo and performed live imaging ( Figure 8A , Video 12 ) . During early stages of daughter cell formation , F-actin was found in the RB linking the two parasites at their posterior end ( Figure 8A; 1 . 24–2 . 36 hr ) . At the earliest stages of daughter cell construction , F-actin accumulated at the elongating IMC ( Figure 8A: 1 . 36–2 . 36 hr ) and further concentrated towards the posterior end of the mother cells , where it colocalized with the mother IMC ( Figure 8A , 3 . 00–3 . 24 hr ) . As the daughters begin to bud from the maternal cell , the IMC of the mother disintegrates and appears to be transported towards the daughter cells ( Figure 8A , Arrows , 3 . 24–3 . 48 hr ) . At the end of replication , the now mature parasites remain connected through F-actin structures ( Figure 8A , 4 . 00–5 . 12 hr ) . 10 . 7554/eLife . 24119 . 025Video 12 . GAPM3-YFP expressing parasites expressing Cb-Halo were imaged every 6 min for 5 hr , F-actin can be seen initially connecting the basal end of the parasites before accumulating beneath the forming daughter cells where it appears to concentrate towards the rear of the new daughters during emergence and recycling of the maternal IMC . Scale bar 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 025 Given that GAPM1a-YFP positive vesicles appear to be transported between individual parasites within a PV or between parasites and the residual body ( Figures 2B and 8A ) we wished to further define the role the filamentous network plays in this process . We imaged Cb-Halo and GAPM1a-YFP co-expressing parasites and tracked the motion of vesicles within the residual body . GAPM1a-YFP vesicles moved outside the parasite along filamentous tubules ( Figure 8B ) , demonstrating vesicular transport within the residual body , as also shown for wt parasites ( Figure 2B ) . Importantly , transport of vesicles was dependent on F-actin , as incubation of parasites with Cyt-D significantly abrogated vesicular transport ( Figure 8C ) . In summary , these data demonstrate that individual parasites remain connected via an F-actin containing network that is required to transfer material in an active , F-actin dependent process ( Video 13 ) . 10 . 7554/eLife . 24119 . 026Video 13 . Vesicle tracking on F-actin tubules . Parasites stably expressing GAPM1a-YFP and transiently expressing Cb-Halo were imaged every second and vesicles containing GAPM1a-YFP could be observed to move along Cb-Halo filaments . Time indicated in minutes , scale bar 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 026
Studies on Toxoplasma actin performed in the 90’s suggested that apicomplexan actin is highly divergent and incapable of forming stable filaments ( Dobrowolski and Sibley , 1997; Dobrowolski et al . , 1997; Dobrowolski and Sibley , 1996 ) . Since then , polymerization kinetics of heterologously expressed apicomplexan actin showed that it polymerized in an unusual isodesmic process , in contrast to conventional eukaryotic actin ( Skillman et al . , 2013 ) . These findings have been recently questioned , as heterologously expressed apicomplexan actin was shown to be misfolded ( Olshina et al . , 2016 ) . In parallel , a number of recent genetic studies have demonstrated important functions of parasite actin during intracellular development , including maintenance of the apicoplast ( Andenmatten et al . , 2013; Egarter et al . , 2014 ) , daughter cell replication ( Haase et al . , 2015 ) and motility of secretory organelles ( Heaslip et al . , 2016 ) . These findings cannot be easily reconciled with the current view of parasite actin being incapable of efficient polymerization , and would instead predict the presence of F-actin filaments within the parasite cytosol . Furthermore , there are 11 putative myosin motors within the parasite ( Foth et al . , 2006 ) , driving diverse cellular processes from cell division to motility , which require actin filaments to function . A major impediment to resolving this controversy has been the lack of appropriate reagents for specifically labelling F-actin . Parasite actin does not bind phalloidin , the gold standard reagent in other eukaryotic systems ( Skillman et al . , 2011 ) and , attempts to use genetically encoded actin sensors , such a LifeAct ( Riedl et al . , 2008 ) or Utrophin-CH ( Burkel et al . , 2007 ) have failed thus far ( our and others unpublished data ) . Using a novel tool based on camelid nanobodies , Chromobody , we have visualized F-actin in T . gondii for the first time in live parasites and demonstrated it has important and previously unforeseen roles during the intracellular development of the parasite . While individual actin filament kinetics could not be easily defined , due to their highly dynamic nature within the parasite cytosol ( Videos 5–7 ) , a relatively stable F-actin network was clearly visible that linked individual parasites within the vacuole . Chromobodies have been used successfully in diverse eukaryotes to analyse actin dynamics ( Panza et al . , 2015; Plessner et al . , 2015; Rocchetti et al . , 2014 ) and appear to have less toxicity than other F-actin sensors . However , we were concerned that Cb expression in T . gondii would perturb actin polymerisation , especially given the proposed dynamic and unstable nature of T . gondii actin . Indeed , high levels of transiently expressed Cb inhibited dense granule motility , likely by stabilizing F-actin ( Figure 3C ) . However , upon stable expression of Cb-Halo or Cb-Emerald , no effect on the actin-dependent processes of invasion , replication and egress were observed , while only small alterations in dense granule movement and parasite motility were detectable ( Figure 4 ) . The reason for these discrepancies is currently unknown , but may be related to differing sensitivities of actin dynamics for these processes . Additionally , we demonstrated that the total amount of F-actin in the parasite does not change upon expression of Cb ( Figure 5 ) , indicating that while actin dynamics may be modulated , the overall polymerization state of actin does not change . We show that parasite actin remains susceptible to depolymerising and stabilizing agents Cyt-D and Jas . Moreover , the localization of Cb to filamentous structures in the residual body and inter-parasite network critically depends on the presence of actin ( Figure 6 ) . These data confirm that genetically encoded Cb-Halo and Cb-Emerald does not significantly alter actin dynamics within the parasite and so is an appropriate and robust tool for examining F-actin in this system . The presence of actin in the RB and inter-parasite tubules has not previously been observed or predicted , and so we sought to independently verify that this localization was not an artefact of Cb expression . Labelling parasites with α-actin that preferentially recognises F-actin ( Angrisano et al . , 2012b ) allowed the detection of actin positive structures connecting individual parasites ( Figure 6 ) , though the stain appeared more diffuse . When actin filaments were stabilised by the presence of Jas ( Figure 6 ) this stain becomes more prominent and this was not dependent on the expression of Cb . This suggests that F-actin is less accessible to exogenously added antibodies after fixation , perhaps due to epitope alterations during fixation or the association of actin binding proteins to filaments . Interestingly , while actin has never been previously associated with this membranous network in T . gondii , it appears very similar to structures recently described for the related parasite Theileria annulata , which was shown to contain F-actin in a similar configuration ( Kühni-Boghenbor et al . , 2012 ) . Using a combination of imaging and reverse genetics , we demonstrate that F-actin is required for the formation and/or maintenance of the residual body . Parasites lacking actin do not contain a residual body and are disorganized within the PV . In addition , we show that these extracellular actin-containing structures are required for the transport of material between parasites within a vacuole . The residual body and inter-parasite network allow both the free diffusion of cytoplasmic material between parasites , and transport of membrane bound vesicles between parasites . Further work will be required to determine if these membrane bound vesicles are then taken up by other parasites or are simply transported to the RB to remove them and if a myosin motor powers this for its transportation , as is the case of dense granule transport within the parasite cytosol ( Heaslip et al . , 2016 ) . The organization of F-actin within the parasites and vacuole is highly dynamic throughout the cell cycle ( Figures 8 and 9 ) . In interphase parasites , long inter-parasite contentions extend throughout the PV . At the beginning of the replication cycle this network collapses and is found concentrated in the residual body ( or bodies ) . Within the parasites , F-actin is found concentrated at the IMC of growing daughter cells during elongation ( Figures 8 and 9 ) . As the daughters bud from the mother cell , the inter-parasite connections again extend throughout the PV . As would be expected , the network also collapses in response to calcium ionophore , freeing the parasites and allowing egress . As motile parasites leave the PV , actin appears concentrated at the basal end of parasites ( Figures 8 and 9 ) . 10 . 7554/eLife . 24119 . 027Figure 9 . ( 1 ) After successful invasion , tachyzoites establish a parasitophorous vacuole and initiate replication . ( 2 ) During daughter cell formation , actin labelling is observed at the IMC of the daughter cells and at the posterior pole of the mother . ( 3 ) Once the daughter cells are fully formed , the actin signal strongly localises at the posterior end of the parasites and with the remains of the mothers IMC , as it is recycled . The first actin filamentous network and ring-like structures are visualized . ( 4–5 ) Replication continues and the filamentous network is established between the tachyzoites . The actin ring continues to localize at the residual body . ( 6 ) The filaments between the parasites and the ring break in a calcium dependent manner prior to egress . The network collapses and dots of actin are detected at the posterior end of tachyzoites . ( 7 ) Tachyzoites egressing from the vacuole leave behind an accumulation of actin in the residual body . DOI: http://dx . doi . org/10 . 7554/eLife . 24119 . 027 While Cb expression allowed us to demonstrate F-actin-containing membranous tubules for the first time , we were unable to assess the length of the individual F-actin filaments using these techniques . We speculate that this network consists of either short F-actin bundles that are cross-linked via unknown actin-binding proteins ( based on the formation of short actin filaments in vitro ) ( Skillman et al . , 2011 ) or that the presence of actin binding proteins such as formin , profilin , and coronin may coordinate their activities in vivo to produce longer actin filaments than those formed in vitro ( Skillman et al . , 2011; Olshina et al . , 2016;Salamun et al . , 2014 ) . While this study focuses on the characterization of the membranous network within the PV , it is worth nothing that highly dynamic actin filaments have been also detected within the cytosol of the parasite , and we show that these dynamics are almost completely abolished upon depletion of ADF in good agreement with previous findings ( Haase et al . , 2015; Mehta and Sibley , 2011 ) ( Figure 6; Videos 5–7 ) . Defining the organization of cytosolic filaments will be critical to further elucidating the role of F-actin both during intracellular processes such as vesicle transport and extracellular processes such as motility and invasion , where the role of the parasites acto-myosin system is currently readdressed , since parasites devoid of detectable actin remain invasive ( Whitelaw et al . , 2017 ) . Interestingly , it appears that host cell membrane dynamics , driven by host cell actin , appears to be modulated by the parasite , enabling invasion in the absence of a functional parasite acto-myosin system ( Bichet et al . , 2016 ) . It is possible that F-actin is organized into distinct , higher order structures , perhaps forming large stable bundles in the tubules while existing as single filaments or small bundles in the cytosol . Future work will be needed to further define the organization of the actin cytoskeleton and to identify the actin binding proteins which contribute to the formation of these structures .
The Cb-Halo plasmid consists of a sequence encoding actin chromobody ( Cb ) from Chromotek followed downstream by an in frame sequence encoding Halo ( Promega ) . The vector backbone contains a Toxoplasma tubulin promoter for protein expression and hxgprt resistance cassette . Actin-Cb was amplified with primers FW pG1Cb attaGAATTCCCTTTTTCGACAAAATGGCTCAGGTGCAGCTGGT and Rv pG1Halo TATGTTAATTAATTAACCGGAAATCTCCAGAGTAG using as a template pHTC Halo Tag ( Promega ) containing in frame actin Cb . Actin chromobody ( actin-Cb ) was cloned in frame into pHTC Halo Tag ( Promega ) using a PCR product generated with primers FWvhH GAATTCATGGCTCAGGTGCAGCTGGTGGA , RVvhH CTCGAGGCTTCTTGAGGAGACGGTGACCT using a pAC-TagRFP ( Chromotek ) as a template . To endogenously tag gapm1a , the 3’flank of the gene was amplified using 5’ TACTTCCAATCCAATTTAATgccgccctgttcgtgtagttttatctg 3’ and 5’ TCCTCCACTTCCAATTTTAGCGGATCTGCAGGACAGGCAAGCC 3’ and inserted into LIC-YFP by ligation independent cloning ( Huynh and Carruthers , 2009 ) . To create Cb-EmFP , the emeraldFP coding sequence was amplified using primers ( emeraldFP-F: atgcaccggtatgggactcgtgagcaaggg and EmeraldFP-R: atgccttaagttacttgtacagctcgtcca ) . The emeraldFP PCR product and Cb-Halo plasmid were subcloned using traditional restriction digestion/ligation protocols . Human foreskin fibroblasts ( HFFs ) ( RRID: CVCL_3285 ) , ATCC ) were grown on tissue culture-treated plastics and maintained in Dulbecco’s modified Eagle’s medium ( DMEM ) supplemented with 10% foetal bovine serum , 2 mM Lglutamine and 25 mg/mL gentamycin . Parasites were cultured on HFFs and maintained at 37°C and 5% CO2 . Cultured cells and parasites were regularly screened against mycoplasma contamination using the LookOut Mycoplasma detection kit ( Sigma ) and cured with Mycoplasma Removal Agent ( Bio-Rad ) if necessary . To generate stable Cb-Halo expressing parasites , 1 × 107 of freshly released RH ∆hxgprt parasites were transfected with 20 µg DNA by AMAXA electroporation . Selection was based on mycophenolic acid and xanthine ( Donald et al . , 1996 ) . Gapm1a-YFP parasites were transfected as above and were selected using pyrimethamine . In order to express adf cKO stably expressing Cb-Emerald , parasites were transfected with Cb-Emerald and selection was performed using flow cytometry with a S3 Cell Sorter ( Bio-Rad , Hercules , CA , USA ) . The inducible act1 cKO was obtained by the addition of 50 nM rapamycin to the parental LoxPAct1 strain for 4 hr at 37°C , 5% CO2 and cultured as described in Egarter et al . ( 2014 ) . Cb-Halo plasmid was transiently transfected into LoxPAct1 . Transfected parasites were induced with 50 nM rapamycin for 4 hr ( Andenmatten et al . , 2013 ) , washed and plated on HFFs . Parasites were fixed at the desired time and stained with Halo-TMR ( 1:10 , 000 ) for 15 min . Extracellular parasites were pelleted and then resuspended in RIPA buffer ( 50 mM Tris–HCl pH 8; 150 mM NaCl; 1% Triton X–100; 0 . 5% sodium deoxycholate; 0 . 1% SDS; 1 mM EDTA ) , incubation for 5 min on ice was used to lyse the cells . Afterwards , samples were centrifuged for 60 min at 14 , 000 rpm at 4°C and laemmli buffer was added to the supernatant . 5 × 106 parasites were loaded onto an SDS acrylamide gel . Western blotting was performed as described previously ( Herm-Götz et al . , 2007 ) using IRDye680RD or IRDye800RD ( Li-Cor ) secondary antibodies . Extracellular Wt and Cb-Halo parasites were harvested , filtered and washed before being resuspended in actin stabilization lysis buffer ( 60 mM PIPES , 25 mM HEPES , 10 mM EDTA , 2 mM MgCl2 , 125 mM KCl completed with Pierce Protease inhibitor mini tablets , EDTA Free ( Thermo Scientific ) and Triton X-100 0 . 2% ) . Lysates were incubated on ice for 1 hr , then incubated with equilibrated Magne HaloTag Beads ( Promega ) for 2 hr at 4°C . Beads were washed 5 times with 1 ml of buffer and elution was made using the TEV protease ( Promega ) as instructed in the protocol . Western blot analysis was performed as above . Freshly egressed parasites were harvested and resuspended in buffer A ( 60 mM PIPES , 25 mM HEPES pH7 . 5 , 10 mM EDTA , 2 mM MgCl2 , 125 mM KCl ) containing 1 µM Jasplakinolide or DMSO . Parasites were incubated for 1 hr at 37°C in a water bath . After centrifugation , parasite pellets were resuspended in buffer B ( buffer A complemented , 10% glycerol and 1% Triton X-100 ) . The suspensions were left on ice for 1 hr and centrifuged at 13 , 000 rpm for 30 min at 4°C . Pellets were washed once with buffer B , resuspended in SDS protein loading buffer , and boiled . Western blot analysis and semi quantification was performed as described above using Li-Cor Odysseys Clx with antibodies against ACT1 ( Angrisano et al . , 2012b ) and GRA7 as a loading control . The Cb coding sequence with a C-terminal 6-His tag was cloned into a pET22b bacterial expression vector and transfected into chemically competent Rosetta ( DE3 ) bacteria ( EMD Millipore ) . A 100 ml culture of LB-ampicillin was grown for 24 hr from a single bacterial colony . 25 mls of bacteria was used to inoculate 500 mls of LB-ampicillin and grown at 37°C until OD between 0 . 6 and 0 . 8 . Expression was induced with 0 . 5 µM IPTG at 37°C for 4 hr . Bacterial pellets were frozen overnight at −80°C . Bacteria pellets were resuspended in 40 mls of xTractor buffer ( Clontech ) and then sonicated for 4 min on ice . Extracts were clarified at 9500x g for 20 min at 4°C . Supernatant were added to 1 ml of equilibrated Talon resin and agitated at 4°C for 60 min . Supernatant/resin mix was added to affinity column ( Biorad ) and supernatant allowed to flow through by gravity . Resin was washed with 20mls of Talon equilibration buffer ( Clontech ) followed 10 mls wash buffer ( Equilibration buffer with 1/10th volume of elution buffer ) . Proteins were eluted with 10mls of elution buffer in 1 ml aliquots . Elutions containing rCb were pooled and dialyzed overnight in 125KMEI buffer ( 125 mM KCl , 1 mM MgCl2 , 1 mM EGTA , 10 mM Imidazole pH 7 . 0 , 10 mM DTT ) . All steps were performed at 4°C and proteins were stored at 4°C . Protein concentration was determined using Bradford Assay . rCb was diluted to 20 µM in 125KMEI and clarified at 100 , 000xg for 20 min at 4°C . Protein concentration in supernatant was determined using Bradford Assay and diluted to 8 µM in 125KMEI . Actin was diluted to various concentrations in Actin Buffer ( 25 mM KCL , 1 mM EGTA , 25 mM imidazole pH 7 . 4 , 4 mM MgCl2 , 10 mM DTT ) . Actin and rCb were added together in equal volumes and incubated at room temperature for 30 min before centrifugation at 100 , 000xg for 20 min at 4°C . Equal volumes of supernatant and pellet were run on NuPAGE Bis-Tris 4–12% Gradient gels ( ThermoScientific ) with 1XMES buffer . Gels were stained using Simply Blue Coomassie Stain as per manufacturer’s instructions . The ratio of rCb in supernatants and pellets was determined by densitometry using ImageJ ( RRID: SCR_003070 ) . The data from two independent experiments were used to determine the apparent Kd of rCb for F-actin . Parasites were incubated with either 100 nM jasplakinolide or 2 µM cytochalasin D for 1 hr at 37°C . Parasites were fixed with 4% PFA and counterstained with the respective antibodies . Widefield images were acquired in z-stacks of 2 μm increments and were collected using an Olympus UPLSAPO 100× oil ( 1 . 40NA ) objective on a Deltavision Core microscope ( Applied Precision , GE ) attached to a CoolSNAP HQ2 CCD camera . Deconvolution was performed using SoftWoRx Suite 2 . 0 ( Applied Precision , GE ) . Video microscopy was conducted with the DeltaVision Core microscope as above . Normal growth conditions were maintained throughout the experiment ( 37°C; 5% CO2 ) . Further image processing was performed using ImageJ64 software . FRAP data were recorded using the same microscope as above . The region of interest was photobleached with a 405 laser for an optimised number of events for the cell strain and area investigated . Three pre-bleach and a number of post-bleach images ranging between 20 and 180 s ( one image per second ) were recorded with Exc and Em filter for FITC with an exposure time of 100–200 ms , ND filter 32% . Data were displayed and analysed using ImageJ software ( Sultana et al . , 2007 ) . Fluorescence intensity was expressed as intensity percentage of the same unbleached area ( filament or nanotubular network ) to account for photobleaching and defocussing in the sample . Super-resolution microscopy ( SR-SIM ) was carried out using an ELYRA PS . 1 microscope ( Zeiss ) as described in Harding et al . ( 2016 ) . For filament size analysis default settings of the Ridge Detection Plug-In ( Steger , 1998 ) in ImageJ was used . | Toxoplasma gondii is a parasite that commonly infects most warm-blooded animals and is thought to affect over two billion people worldwide . In most cases , the infection does not cause any symptoms , although it can lead to serious complications in pregnant women or people with a weakened immune system . T . gondii has a complex life cycle that involves different stages . During infection , the parasite invades the host cells and replicates inside a specialized cell structure called a ‘parasitophorous vacuole’ until the host cell bursts . The parasite then spreads and infects more host cells . The replication is synchronised , meaning all parasites in a host cell replicate at the same time . It was unclear how the parasites coordinated this process , but some researchers suggested that the parasites remained connected to each other to communicate by exchanging material and information . A good candidate to form such connections is the protein actin , which in many organisms forms filaments that guide the transport of cargo molecules in the cell . However , previous research indicated that actin in T . gondii is incapable of forming these stable filaments . Periz et al . developed a new tool of fluorescence markers that specifically bind to actin in T . gondii and found extensive actin networks that connected parasites with each other and also to the membrane of the parasitophorous vacuole . Actin was needed to transport molecules between the parasites within a vacuole and was also found to enter the cells of the parasite . When the protein was depleted in the parasite , the network collapsed; the parasites started to replicate at different times and could no longer leave the host cell . A next step will be to further investigate the role of actin in T . gondii and other parasites using the tools developed by Periz et al . A better understanding of replication of T . gondii could provide clues to new treatments for parasitic diseases that cause substantial economic losses worldwide . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2017 | Toxoplasma gondii F-actin forms an extensive filamentous network required for material exchange and parasite maturation |
All motor behaviors require precise temporal coordination of different muscle groups . Breathing , for example , involves the sequential activation of numerous muscles hypothesized to be driven by a primary respiratory oscillator , the preBötzinger Complex , and at least one other as-yet unidentified rhythmogenic population . We tested the roles of Atoh1- , Phox2b- , and Dbx1-derived neurons ( three groups that have known roles in respiration ) in the generation and coordination of respiratory output . We found that Dbx1-derived neurons are necessary for all respiratory behaviors , whereas independent but coupled respiratory rhythms persist from at least three different motor pools after eliminating or silencing Phox2b- or Atoh1-expressing hindbrain neurons . Without Atoh1 neurons , however , the motor pools become temporally disorganized and coupling between independent respiratory oscillators decreases . We propose Atoh1 neurons tune the sequential activation of independent oscillators essential for the fine control of different muscles during breathing .
One of the most essential and seemingly simple behaviors in mammals; breathing , requires state-dependent temporal coordination of multiple muscle groups , including the diaphragm and upper airway , to ensure unobstructed airflow ( Feldman et al . , 2013 ) . All motor behaviors , in fact , require the coordinated activity of various muscles acting with temporal precision ( Grillner , 2006 , 2011 ) , but the neural mechanisms underlying such temporal coordination are only starting to become clear . Identifying how neural networks generate coordinated respiratory behaviors should provide insight into how the brain organizes other complex motor behaviors . Breathing in mammals encompasses several distinct behaviors , including inspiration , expiration , and sighing ( Ramirez and Viemari , 2005 ) . Deflection of the diaphragm ( inspiration ) is the predominant breathing behavior in mammals and requires glutamatergic neurons within the preBötzinger Complex ( preBötC ) ( Bouvier et al . , 2010; Gray et al . , 2010; Gray , 2013 ) . Expiration is a state-dependent behavior in adults and is normally passive but becomes active under some conditions , including exercise or sleep ( Pagliardini et al . , 2012; Feldman et al . , 2013 ) . The appearance of rhythmic abdominal activity , in vivo , or rhythmic facial cranial ( VII ) nerve output , in vitro , after opioid-induced inhibition of inspiration , together with data from transection experiments , have provided evidence for an independent second respiratory oscillator adjacent to the VII motor nucleus ( Mellen et al . , 2003; Janczewski and Feldman , 2006; Onimaru et al . , 2006 , 2009 ) . Whereas the vast majority of excitatory neurons in the preBötC are derived from neural progenitors expressing the transcription factor ( TF ) Developing Brain Homeobox 1 ( Dbx1 ) ( Bouvier et al . , 2010; Gray et al . , 2010; Gray , 2013 ) , the VII region contains three developmentally distinct excitatory neuronal populations that are prime candidates for generating abdominal and VII activity ( Figure 1A ) . Although none has yet been conclusively shown to play a role in abdominal or VII rhythm generation ( Feldman et al . , 2013 ) , all three populations express neurokinin 1 receptors ( NK1R ) and respond to Substance P ( SP ) , and are theorized to couple with preBötC Dbx1-derived glutamatergic neurons to generate coordinated respiratory behaviors ( Figure 1B; Bouvier et al . , 2010; Gray et al . , 2010; Feldman et al . , 2013 ) . 10 . 7554/eLife . 02265 . 003Figure 1 . Perinatal mouse hindbrain produces temporally coordinated respiratory output from cervical and lumbar motor pools , as well as internal intercostal respiratory muscles . ( A ) Cartoon of sagittal view through caudal hindbrain indicating the locations of distinct developmentally defined glutamatergic populations important for breathing . Colored filled circles indicate the relative location of rhombic lip ( RL ) Atoh1 ( orange ) , RTN Phox2b ( purple ) , and Dbx1 ( blue ) derived neurons . Note: RTN Phox2b neurons transiently express Atoh1 , but we have left this out for clarity . Atoh1 RL neurons do not express Phox2b and have not been shown to be chemosensitive . The green circle identifies the location of preBötzinger Complex ( preBötC ) Dbx1 neurons hypothesized to generate the cervical ( inspiratory ) rhythm . The magenta rectangle indicates the location of Retrotrapezoid/Parafacial respiratory group ( RTN/pFRG ) region neurons hypothesized to generate independent lumbar ( expiratory ) rhythm . LRN–lateral reticular nucleus , NA–nucleus ambiguus , rVRG–rostral ventral respiratory group , VIIn ( VII motor nucleus ) . Scale bar = 200 µm . ( B ) Schematic describing the hypothesized glutamatergic coupling between Dbx1 preBötC neurons generating inspiration ( green circle–right ) and the three candidate glutamatergic RTN/pFRG populations generating expiration ( magenta rectangle–left ) . Colored circles indicate genetic lineage as in ( A ) . ( C ) Electrophysiological traces ( upper–integrated , lower–raw ) of spontaneous respiratory-related output from cervical ( C4 , green ) and lumbar ( L1 , magenta ) motor roots as well as EMG recording from the XIth internal intercostal muscle ( IC ) . Cervical roots innervate the diaphragm active during inspiration . Lumbar roots innervate abdominal respiratory muscles active during expiration in adults . Arrows indicate respiratory-related bursts where some respiratory motor pools lack activity . ( D ) In some fictive breaths , cervical ( green ) , lumbar ( magenta ) , and IC ( black ) motor outputs are active nearly simultaneously . ( E ) In other fictive breaths only the cervical root is active . ( F–I ) Single integrated C-L-IC fictive breaths showing breath-by-breath variations in temporal co-activation between respiratory motor outputs . Note that in some respiratory bursts ( G–I ) both lumbar ( magenta ) and IC ( black ) burst peaks occur before the cervical burst . Also IC activity can occur during both cervical and lumbar bursts ( H–I ) . ( J–K ) Cartoons indicating how different patterns of respiratory motor output from cervical and lumbar motor pools could be produced by changes in the synaptic strengths of excitatory ( dots ) or inhibitory ( arrowheads ) synaptic connections between the putative preBötC Dbx1 inspiratory oscillator , the unknown RTN/pFRG expiratory oscillator , and intervening inhibitory interneurons . Solid lines indicate strong connections , dotted lines indicate weaker connection . Scale bar = 10 s ( C ) , 500 ms ( D–I ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 003 First is a population of glutamatergic retrotrapezoid nucleus ( RTN ) neurons expressing the TF Paired-like Homeobox 2b ( Phox2b , hereafter called RTN neurons , Figures 1A and 2A , B ) . In adult rodents , these neurons play an important role in chemosensitivity to CO2 ( Stornetta et al . , 2006; Guyenet and Mulkey , 2010; Feldman et al . , 2013 ) . In late embryonic and perinatal rodent reduced preparations , however , RTN neurons are active prior to and independent of those in the preBötC , are rhythmically active out of phase with phrenic motor output , and have been proposed to constitute an independent respiratory oscillator ( Onimaru et al . , 2009; Thoby-Brisson et al . , 2009 ) . Many of these neurons overlap the functionally defined parafacial respiratory group ( pFRG ) ( Onimaru and Homma , 2003; Onimaru et al . , 2009 ) . RTN neurons transiently express and require the TF Atonal Homolog 1 ( Atoh1 ) post-mitotically for appropriate migration , maturation , and function ( Dubreuil et al . , 2009; Rose et al . , 2009b ) . The second RTN/pFRG population consists of Dbx1-derived glutamatergic neurons developmentally related to those of the preBötC but whose role is unknown ( Figures 1A and 2C , D ) . Loss of Dbx1 leads to perinatal lethality due , it is assumed , to loss of preBötC and other hindbrain glutamatergic neurons ( Figure 1A , B; Gray et al . , 2010 ) . 10 . 7554/eLife . 02265 . 004Figure 2 . The RTN/pFRG contains three candidate expiratory rhythm generating populations . ( A and B ) Phox2b-expressing RTN neurons are located directly ventral to VII motor nucleus ( VII ) ( A–arrow , B–green ) ; many co-express NK1R ( B , red ) . ( C and D ) Dbx1-dependent LacZ-labeled neurons are located ventral and dorsomedial to VII ( C–arrow ) . RTN/pFRG Dbx1-derived LacZ-expressing neurons ( B–blue , D–magenta ) co-express NK1R ( B–red ) and VGlut2 mRNA ( D–green ) in a P0 Dbx1LacZ/+ mouse ( B–D ) . ( E and F ) Atoh1-dependent RL neurons are located medial and ventral to VII , and many express Lhx9 mRNA ( E ) or NK1R ( F–magenta ) in P0 WT ( E ) or Atoh1-CreTg;R26 YFP ( F–green ) mice . Arrows in B and D indicate co-expression , arrowheads indicate lack of co-expression . Squares ( B , D , F ) are enlarged to right . Scale bars = 200 µm . D-dorsal , M-medial . See Figure 2—figure supplement 1 for the effects of loss of Atoh1 on RTN neurons as well as the loss of Dbx1 on RTN/pFRG VGlut2 expressing neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 00410 . 7554/eLife . 02265 . 005Figure 2—figure supplement 1 . Loss of Atoh1 or Dbx1 eliminates specific RTN/pFRG neuronal populations . ( A ) Three color confocal image showing co-expression of Lbx1 ( red ) in Atoh1-dependent , LacZ ( green ) expressing neurons in the RTN , ventral to Islet1 ( Isl1 , blue ) expressing VII motoneurons in a P0 Atoh1LacZ/+ mouse . Schematic ( upper right ) indicates maintenance of all three putative RTN/pFRG developmental lineages . ( B ) RTN Lbx1 ( red ) and LacZ ( green ) co-expressing neurons fail to property migrate ventral to Isl1 expressing VII neurons in E18 . 5 Phox2b-Cre;Atoh1LacZ/F ( RL+/RTN− ) mutant mice . Schematic ( upper right ) indicates selective loss of RTN/pFRG Phox2b neurons ( cross through purple RTN population ) . ( C ) VGlut2 ( green ) expressing neurons fail to form dorsomedial to VII in E18 . 5 Dbx1LacZ/LacZ mutant mice . Note the maintenance of LacZ+ VGlut2− cells in the RTN/pFRG region ( magenta ) . Schematic ( upper right ) indicates selective loss of RTN/pFRG and preBötC Dbx1 neurons ( crosses through blue Dbx1 populations ) . Single color images ( A and B ) are shown to right . Boxes ( A–C ) are expanded to far right . Scale bar = 200 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 005 The third RTN/pFRG population consists of glutamatergic neurons derived from Atoh1-expressing progenitors within the rhombic lip ( hereafter called RL neurons ) , some of which express LIM Homeobox 9 ( Lhx9 ) post-mitotically ( Figure 1A , B , 2E , F ) . These neurons require Atoh1 for their initial neural specification ( Wang et al . , 2005 ) , express the glutamate transporter , solute carrier family 17 ( sodium-dependent inorganic phosphate cotransporter ) , member 6 ( SLC17A6 ) , also known as the vesicular glutamate transporter 2 ( VGlut2 ) , and do not project to the cerebellum ( Wang et al . , 2005; Gray , 2013 ) . Importantly , these neurons do not express Phox2b and have not been shown to be chemosensitive . The germline elimination of Atoh1 in mice leads to the complete loss of RL neurons and abnormal migration and functional elimination of RTN neurons ( Rose et al . , 2009b; Huang et al . , 2012 ) . This produces a completely penetrant perinatal lethality due to the inability to establish respiratory rhythm in vivo , although reduced preparations containing the preBötC can still produce rhythmic inspiratory output ( Rose et al . , 2009b ) . In contrast , the selective elimination of functional RTN neurons by the conditional genetic targeting of Atoh1 in subsets of hindbrain neurons expressing Phox2b does not eliminate breathing in vivo but produces partial neonatal lethality and blunting of chemosensitivity ( Dubreuil et al . , 2009; Ramanantsoa et al . , 2011; Huang et al . , 2012 ) . These observations suggest that both RL and RTN neurons play important roles in breathing . We used multiple transgenic mouse lines to test the role of distinct brainstem populations in generating and coordinating respiratory behaviors in mice . We found that Dbx1-dependent , but not Atoh1-dependent , neurons are necessary for rhythmic motor output from respiratory motoneurons . RL neurons , however , are required for temporal delay and coupling between multiple respiratory oscillators . This suggests that genetically distinct populations independently control two fundamental aspects of motor behavior: rhythm generation and the relative timing between motor pools .
Post-natal rodent hindbrain-spinal cord preparations generate robust respiratory-related rhythmic bursts ( fictive breaths ) from cervical ( diaphragm ) and lumbar ( abdominal ) motor roots , with abdominal output requiring the RTN/pFRG region ( Smith et al . , 1990; Iizuka , 1999 , 2004; Janczewski and Feldman , 2006; Taccola et al . , 2007; Abdala et al . , 2009 ) . These outputs are hypothesized to represent in vitro correlates of inspiratory ( cervical ) and expiratory ( lumbar ) motor output ( Abdala et al . , 2009; Feldman et al . , 2013 ) . Because mouse mutants affecting Atoh1 , Phox2b , or Dbx1 neurons either do not breathe , or have respiratory depression in vivo , we first tested whether the entire hindbrain and spinal cord at E18 . 5 from timed pregnant wild-type mice produced coordinated cervical and lumbar output . Previous work has shown rhythmic cervical ( inspiratory ) output beginning around E15 in mice , but whether lumbar ( expiratory ) activity is also present is unknown ( Thoby-Brisson et al . , 2005 ) . We simultaneously recorded endogenous respiratory-related cervical and lumbar root output as well as electromyographic ( EMG ) output from the XIth internal intercostal muscle ( IC ) . The IC muscle is innervated by motoneurons in the caudal thoracic spinal cord . Like the lumbar innervated abdomen , the IC is active during expiration in vivo and in more mature in vitro preparations ( Miller et al . , 1985 , 1987; Iscoe , 1998; Legrand and Troyer , 1999; Iizuka , 2003 , 2004 ) . As in preparations from older animals , we found rhythmic motor output from both cervical and lumbar roots that corresponded to rhythmic respiratory muscle activity ( Figure 1C ) . These outputs were silenced by transection above the first cervical root ( not shown ) , and thus require the hindbrain . In late fetal ( E18 . 5 ) preparations , however , both lumbar motor roots and intercostal EMG were often co-active with cervical output ( Figure 1C , D , F ) . Importantly , however , cervical , lumbar , and IC activities were not identical arguing against a single source of respiratory drive for all respiratory motor pools . Cervical bursts were present without lumbar motor root or intercostal ( IC ) muscle activity , as is most common in adult rodents at rest ( see arrows in Figure 1C , Figure 1E ) , whereas both lumbar and IC activity were often active prior to the onset of cervical output , as in older preparations ( Figure 1G–I ) . These data are consistent with the presence of functional and independent rhythmogenic sources of respiratory drive to cervical , thoracic , and lumbar motor pools in E18 . 5 mice . It further suggests these different patterns of temporal co-activation from different respiratory motor pools might be the consequence of variations in the coupling between independent respiratory oscillators ( Figure 1J–L ) . While these data are consistent with the coupling of a functional RTN/pFRG with the preBötC to produce coordinated cervical–lumbar motor output , the genetic identity of the essential neurons for lumbar output was unknown ( Figures 1B and 3A ) . We set out to directly test the role of RTN and RL neurons in the generation and coordination of respiratory motor activity by genetically eliminating specific hindbrain populations in transgenic mice . RTN neurons are functionally eliminated ( RTN− ) by targeted deletion of Atoh1 in Phox2b-expressing neurons in Phox2b-Cre; Atoh1LacZ/F mice ( RL+/RTN− ) ( Figure 2—figure supplement 1A , B; Huang et al . , 2012 ) . Because the loss of Atoh1 does not completely ablate RTN neurons , we also examined Phox2b-Cre; VGlut2F/F mice in which Phox2b excitatory neurons are selectively disconnected from hindbrain networks ( RL+/RTNsilent ) ( Rossi et al . , 2011 ) . In both of these lines , preBötC and other hindbrain Dbx1-dependent neurons remain unaffected ( Rose et al . , 2009b; Gray et al . , 2010; Huang et al . , 2012; Feldman et al . , 2013 ) . 10 . 7554/eLife . 02265 . 006Figure 3 . RL and RTN neurons are neither necessary nor sufficient for lumbar respiratory output . ( A ) Schematic cartoon describing populations targeted for genetic elimination to determine necessity for generating cervical ( green ) and/or lumbar ( magenta ) respiratory output . Colored filled circles indicate the developmental origin of RL Atoh1 ( orange ) , RTN Phox2b ( purple ) , or Dbx1 ( blue ) derived neurons with schematic as in Figure 1B . ( B ) 50 s integrated recordings of spontaneous respiratory output from cervical ( C4 , green ) and lumbar ( L1 , magenta ) motor roots in an E18 . 5 WT isolated hindbrain-spinal cord preparation showing normal lumbar-cervical coordinated respiratory output . Upward deflections indicate respiratory-related events . Arrows indicate fictive breaths where cervical output is not matched by lumbar output . Targeted ablation ( C , Phox2b-Cre;Atoh1LacZ/F [RL+/RTN−] ) or silencing ( D , Phox2b-Cre;VGlut2F/F [RL+/RTNsilent] ) of RTN Phox2b neurons does not eliminate lumbar respiratory output . Schematics ( C–D , upper right ) indicate targeted loss of Phox2b RTN neurons ( purple ) . Note the marked increase in respiratory period ( i . e . , slowing of frequency ) . ( E ) Loss of both RTN Phox2b and RL Atoh1 neurons in Atoh1LacZ/LacZ ( RL−/RTN− ) mice similarly does not eliminate lumbar respiratory output . Schematic ( upper right ) indicates targeted loss of both Phox2b RTN ( purple ) and RL Atoh1 neurons ( orange ) . ( F ) Loss of Dbx1-dependent neurons in Dbx1LacZ/LacZ mice eliminates both lumbar and cervical respiratory outputs . Schematic ( upper right ) indicated targeted loss of Dbx1 neurons ( blue ) . Scale bar = 15 s . ( G ) Loss of RTN Phox2b neurons slows the respiratory rhythm , thus increasing the cervical period . Bar graphs show least-squares mean ( LSMEAN ) of respiratory periods in seconds ( ±SEM ) during baseline ( cervical–[C] , dark green , lumbar–[L] , magenta ) or in the presence of 1 µm SP ( cervical–[C] , light green , lumbar– ( L ) , pink ) in wild-type ( WT ) ( n = 6 ) , Phox2b-Cre;Atoh1LacZ/F ( RL+/RTN− ) ( n = 4 ) , and Atoh1LacZ/LacZ ( RL−/RTN− ) ( n = 5 ) mice . Top brackets indicate statistical test groups ( *p<0 . 05 , #p<0 . 001; mixed random effects ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 006 The loss of RTN neurons leads to respiratory depression in vivo and in vitro , and RTN Phox2b neurons have been a leading candidate for an independent respiratory oscillator driving abdominal and , possibly , VII motor output ( Onimaru et al . , 2008; Thoby-Brisson et al . , 2009; Feldman et al . , 2013 ) . WT hindbrain-spinal cord preparations produce highly reliable rhythmic cervical and lumbar motor outputs ( Figures 1C and 3B ) . Surprisingly , both cervical and lumbar respiratory motor outputs were still present after either elimination or silencing of RTN neurons in RL+/RTN− and RL+/RTNsilent mice , although the cycle time between successive fictive breaths ( the respiratory period ) dramatically increased ( p<0 . 01 in RL+/RTN− mice , Figure 3C , D , G ) ( Dubreuil et al . , 2009; Ramanantsoa et al . , 2011; Huang et al . , 2012 ) . Substance P ( SP ) application in RL+/RTN− mice recovered inspiratory periods to WT levels ( p<0 . 0001 , Figure 3G ) , likely due to direct effects upon preBötC neurons and other RTN/pFRG neurons ( Figure 2B , F; Gray et al . , 1999 ) . Given that lumbar motor output persisted after RTN loss , we explored whether the combined loss of both RTN and RL neurons in Atoh1LacZ/LacZ mice ( RL−/RTN− ) would affect respiratory behaviors ( Figure 3E ) . We again found that rhythmic respiratory output from both cervical and lumbar rhythms persisted , but the combined loss of both RL and RTN recovered respiratory periods to near WT levels under both baseline conditions and with SP application ( Figure 3G ) . Lastly , we tested the effect on lumbar output of the complete loss of Dbx1-dependent neurons in Dbx1LacZ/LacZ mice ( Dbx1− , Figure 3F ) , where both RL and RTN populations are unaffected ( Bouvier et al . , 2010; Gray et al . , 2010 ) . Dbx1-derived neurons are essential for the formation of the preBötC and rhythmic cervical output , but their role in lumbar output was unknown . In contrast to mice with loss of RL and/or RTN neurons , Dbx1− mice produced no rhythmic respiratory output from either cervical or lumbar motor roots , even when stimulated with SP or elevated potassium in the hindbrain perfusion solution ( Figure 3F ) . This failure was not due to an inability to generate spinal output of any kind , because global disinhibition of the hindbrain by application of bicuculline and strychnine produced robust but non-respiratory spinal outputs ( not shown ) . These results are consistent with an important role for RTN neurons in modulating breathing , but undermine current hypotheses by indicating that RTN and RL neurons are neither necessary nor sufficient for lumbar respiratory motor output ( Onimaru et al . , 2008; Thoby-Brisson et al . , 2009; Mellen and Thoby-Brisson , 2012 ) . Dbx1-derived neurons , however , are essential for the expression of both cervical and lumbar respiratory behaviors ( Bouvier et al . , 2010; Gray et al . , 2010 ) . The maintenance of lumbar output in RL−/RTN− mice raises the question of whether the networks underlying respiratory outputs to different motor pools in WT mice are generated by multiple oscillators ( Mellen et al . , 2003; Mellen and Thoby-Brisson , 2012; Hägglund et al . , 2013; Moore et al . , 2013 ) or by interactions within a single oscillatory population ( Smith et al . , 2007 ) . In WT mice , most fictive breaths showed cervical and lumbar motor co-activation . Often , however , individual cervical respiratory bursts occurred without a corresponding lumbar burst , and vice versa ( Figure 1E , arrows in Figures 1C and 3B ) . Respiratory bursts in which one or more motor outputs are absent ( termed deletions ) occur both in vivo and in vitro ( Iizuka , 2010; Pagliardini et al . , 2011 ) . Analysis of deletions is used in spinal motor circuits as a test of independent oscillators controlling antagonist flexor and extensor muscles ( McCrea and Rybak , 2008; Stein , 2008; Grillner , 2011 ) . In general , a single rhythmogenic source drives multiple motor outputs nearly identically . We reasoned that decreases in the number of oscillators driving motor output in mutant mice should be revealed by large decreases in , if not elimination of , respiratory deletions . This should be most true of lumbar motor outputs hypothesized to be generated by RTN/pFRG neural populations . We therefore determined the percentage of respiratory bursts in which respiratory motor pools were co-active in WT and Atoh1 mutant mice . We recorded simultaneously from three respiratory motor roots , lumbar , cervical and cranial facial ( VII ) . VII motoneurons are directly adjacent to the location of the hypothesized secondary oscillator and can show rhythmic motor output in the absence of the preBötC ( Onimaru et al . , 2006 ) . WT cervical-lumbar-VII bursts also showed highly associated output ( Figure 4A ) . The outputs from these three roots were not identical , however , as one might assume if they were driven by a single rhythmic source . Cervical bursts , for example , were co-active with both lumbar and VII ( C+L+VII ) only 29 . 2 ± 9% of the time ( n = 4 , Figure 4C , D , G , Figure 4—figure supplement 1B ) . Lumbar and VII bursts were co-active with both other roots ( C+L+VII ) 34 . 9 ± 11 . 7% or 46 . 7 ± 11 . 3% of the time respectively ( Figure 4C , D , G , Figure 4—figure supplement 1A–D; Table 1 ) . All three motor pools could be active either alone or with one or the other motor pool , suggesting a high degree of independence ( Figure 4C , D , G , Figure 4—figure supplement 1A–D ) . Cervical , lumbar , or VII roots were active independently , that is only a single motor pool was active , 23 ± 5 . 8% , 14 ± 6 . 3% , or 18 . 6 ± 10 . 4% of the time , respectively ( Figure 4G; Table 1 ) . This partial independence of motor output was also seen in cervical-lumbar-IC triple recordings , where 88 . 0 ± 5 . 1% of IC muscle contractions were co-active with both cervical and lumbar output ( Table 2 ) . 10 . 7554/eLife . 02265 . 007Figure 4 . RL and RTN neurons are not essential for the expression or independence of cervical , lumbar , or VII respiratory motor outputs . ( A ) E18 . 5 WT mice show respiratory co-activation of cervical ( green ) , lumbar ( magenta ) , and/or VII ( dark blue ) motor roots . Schematic ( top ) indicates maintenance of all RTN/pFRG and preBötC glutamatergic lineages . ( B ) Targeted loss of RL and RTN neurons in Atoh1LacZ/LacZ ( RL−/RTN− ) mice does not eliminate respiratory co-activation of cervical , lumbar , and/or VII respiratory outputs . Schematic ( top ) indicates targeted loss of Phox2b RTN ( purple ) and Atoh1 RL ( orange ) neurons . Arrows ( A and B ) indicate respiratory cervical outputs lacking lumbar or VII output . During single respiratory bursts , each motor pool can be co-active nearly simultaneously , with a temporal delay , or can be silent in relation to each other motor pool in WT ( C and D ) or in the absence of both RTN and RL neurons ( E and F ) . ( G ) Stacked histograms showing the percentage of fictive breaths from each individual motor pool ( cervical–C+ , lumbar–L+ , VII–VII+ ) where that motor pool fires either alone ( bottom–black ) , fires with one of the two other motor roots ( middle two bars , +cervical–green , +lumbar–magenta , +VII–dark blue ) , or where all three roots are co-active ( top , C+L+VII–gray ) from WT ( left ) or in Atoh1LacZ/LacZ ( right ) mice . Note the loss of RL and RTN neurons only affects the relative percentage of coupling . Scale bars = 10 s ( A and B ) , 500 ms ( C–F ) , 1 s ( H ) . See Figure 4—figure supplements 1A–L for additional examples of the independence of respiratory motor pool co-activation in WT and Atoh1LacZ/LacZ mice . See Figure 4—figure supplements 2A–C for examples of independent lumbar activation and quantification of cervical and lumbar deletions in Phox2b-Cre;Atoh1LacZ/F ( RL+/RTN− ) mice . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 00710 . 7554/eLife . 02265 . 008Figure 4—figure supplement 1 . Maintenance and independence of multiple respiratory motor outputs in RL−/RTN− mice . Overlaid integrated cervical ( green ) , lumbar ( magenta ) , and VII ( dark blue ) fictive breaths from WT ( A–D ) or Atoh1LacZ/LacZ ( E–L ) mice showing the range of independence and co-activation of each motor root during single respiratory bursts . Schematics indicate maintenance ( top ) or targeted loss ( middle ) of Phox2b RTN ( purple ) , Atoh1 RL ( orange ) , or Dbx1 ( blue ) populations in WT ( top ) or Atoh1LacZ/LacZ ( RL−/RTN− ) mice . Arrow in L indicates maintenance of active inhibition of lumbar output during cervical and VII co-activation in the presence of 1 µM SP . Scale bar = 500 ms . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 00810 . 7554/eLife . 02265 . 009Figure 4—figure supplement 2 . Independent lumbar oscillator is transiently released in RL+/RTN− mice . ( A ) Loss of RTN neurons transiently releases independent lumbar output ( arrow ) in Phox2b-Cre;Atoh1LacZ/F mice , which is eliminated after application of SP to the hindbrain ( B ) . Schematic ( top ) indicates the targeted loss of Phox2b RTN neurons ( purple ) . Scale bar = 10 s . ( C ) Bar graph showing LSMEAN percentage of cervical bursts without lumbar output ( lumbar deletion ) ( left ) and lumbar bursts without cervical output ( cervical deletions , right ) for E18 . 5 RL+/RTN− mice ( n = 5 ) . Top bracket indicates statistically significant differences between treatments ( *p<0 . 05; mixed random effects ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 00910 . 7554/eLife . 02265 . 010Table 1 . Baseline percentage of simultaneous co-activation for zero , one , or two other motor pools between cervical , lumbar , and VII respiratory outputs in E18 . 5 WT miceDOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 010Percentage of Co-activation+Cervical+Lumbar+VIIC+L+VIICervical23 ± 5 . 825 . 1 ± 11 . 622 . 8 ± 9 . 729 . 2 ± 9Lumbar34 . 7 ± 17 . 514 ± 6 . 316 . 4 ± 8 . 434 . 9 ± 11 . 7VII24 . 9 ± 912 . 3 ± 4 . 518 . 6 ± 10 . 446 . 7 ± 11 . 310 . 7554/eLife . 02265 . 011Table 2 . Baseline percentage of simultaneous co-activation with zero , one , or two other motor pools between cervical , lumbar , and intercostal XI ( ICX1 ) respiratory outputs in E18 . 5 WT miceDOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 011Percentage of co-activation+Cervical+Lumbar+ICC+L+ICCervical27 . 4 ± 9 . 67 . 1 ± 3 . 77 . 1 ± 2 . 858 . 5 ± 11 . 6Lumbar12 . 5 ± 7 . 63 . 2 ± 1 . 5*0 . 084 . 3 ± 7 . 2IC10 . 4 ± 4 . 50 . 01 . 6 ± 0 . 8†88 . 0 ± 5 . 1*p<0 . 026 . †p<0 . 018 . One-way ANOVA vs cervical . Because the WT results indicated that multiple respiratory motor pools received similar but not identical respiratory drives which varied on a breath-by-breath basis ( Carroll and Ramirez , 2013 ) , we went on to test the effect of eliminating RTN and RL populations on respiratory deletions . Similar to WT , in RL−/RTN− mice 30 . 9 ± 7 . 9% of cervical bursts showed lumbar and VII co-activation ( C+L+VII , i . e . , no deletion ) , as compared to 33 . 2 ± 10 . 6% of lumbar and 74 ± 2 . 5% of VII bursts showing triple co-activation ( Figure 4B , E–H , Figure 4—figure supplement 1E–L; Table 3 ) . In some cases , we observed bouts of multiple non-overlapping respiratory bursts over short timescales ( Figure 4H ) . We also observed independent , higher frequency lumbar-only bursts as well as the maintenance of cervical and lumbar deletions in RL+/RTN− mice under baseline conditions ( Figure 4—figure supplement 2A–C ) . Overall , there was no consistent decrease in deletions amongst all three motor roots between control and RL−/RTN− mice . Lumbar only bursts showed a 24 . 3% increase while cervical only bursts decreased by 11 . 7% and independent VII bursts decreased by 51 . 1% although independent VII outputs were still present ( Table 4 ) . Similar variations in two or three root co-activation were also seen ( Table 4 ) . 10 . 7554/eLife . 02265 . 012Table 3 . Baseline percentage of simultaneous co-activation for zero , one , or two other motor pools between cervical , lumbar , and VII respiratory outputs in E18 . 5 Atoh1LacZ/LacZ ( RL−/RTN− ) miceDOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 012Percentage of co-activation+Cervical+Lumbar+VIIC+L+VIICervical20 . 3 ± 1 . 845 . 6 ± 8 . 53 . 3 ± 0 . 8*30 . 9 ± 7 . 9Lumbar46 . 3 ± 7 . 817 . 4 ± 7 . 33 . 2 ± 1†33 . 2 ± 10 . 6VII8 . 9 ± 3 . 18 . 1 ± 3 . 39 . 1 ± 5 . 274 ± 2 . 5‡*p<0 . 04 , vs cervical . †p<0 . 031 vs lumbar . ‡p<0 . 023 vs lumbar . One-way ANOVA . 10 . 7554/eLife . 02265 . 013Table 4 . Percentage change between WT and E18 . 5 Atoh1LacZ/LacZ ( RL−/RTN− ) mice of baseline simultaneous co-activation for zero , one , or two other motor pools between cervical , lumbar , and VII respiratory outputsDOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 013Percentage change of co-activation between WT and E18 . 5 Atoh1LacZ/LacZ ( RL−/RTN− ) mice . +Cervical+Lumbar+VIIC+L+VIICervical−11 . 781 . 7−85 . 55 . 8Lumbar33 . 424 . 3−80 . 5−4 . 9VII−64 . 3−34 . 1−51 . 1164 . 7 The maintenance of independent output patterns from all three motor roots in RL−/RTN− and the presence of independent motor output with different frequencies are evidence for the maintenance of separate extant oscillators . Our present results also match invertebrate preparations where individual oscillators increase their frequency after uncoupling from a network ( Mulloney , 1997 ) . These data suggest that , in WT animals , each respiratory motor pool receives drive from independent but interconnected oscillators and that these oscillators persist in RL−/RTN− mice . These data further suggest that the loss of both RTN and RL neurons accentuates free-running oscillator activity induced by selective neuromodulation ( Doi and Ramirez , 2008 , 2010 ) . The primary neurotransmitter of both RL and RTN neurons is glutamate ( Stornetta et al . , 2006; Rose et al . , 2009a ) . Consistent with their stimulatory role , the selective loss of RTN neurons leads to a significant prolongation of the inspiratory period . After the loss of both RL and RTN populations , however , the respiratory period recovered almost to WT levels ( Figure 3C–E , G ) . This raised the possibility that RL neurons mediate a net inhibitory effect on respiratory networks , possibly via activation of inhibitory neurons . Inhibition is essential for normal respiratory output to control the timing and pattern of the different respiratory muscles ( Marder and Bucher , 2001; Ramirez and Viemari , 2005; Feldman et al . , 2013; Janczewski et al . , 2013 ) . The possibility of a net inhibitory role of RL neurons led us to examine the temporal structure of respiratory motor output in WT and RL−/RTN− mice . Respiratory networks produce several motor behaviors besides the basic cervical ( eupneic ) respiratory pattern . The best characterized of these are sighs ( Ramirez and Viemari , 2005 ) , which are generated within the medulla , persist in reduced preparations , are first seen during late embryonic stages , and are increased after SP application ( Lieske et al . , 2000; Gray et al . , 2001; Ballanyi and Ruangkittisakul , 2009; Chapuis et al . , 2014 ) . Sighs consist of a biphasic inspiratory ( cervical ) burst with an initial peak of typical amplitude that , before decaying , gives rise to a larger amplitude sigh burst . This biphasic burst pattern is present under baseline conditions , both in vivo and in vitro . The sigh component of these biphasic motor outputs , however , can be decoupled from normal ‘eupneic’ bursts either pharmacologically or by blockade of inhibition ( Lieske et al . , 2000; Ramirez and Viemari , 2005; Tryba et al . , 2008; Chapuis et al . , 2014 ) . Our WT E18 . 5 preparations showed biphasic cervical bursts similar to sighs ( Figure 5A , B , D , E , Figure 5—figure supplement 1A , B , F , G , I ) . Under baseline conditions , 6 of 11 preparations produced biphasic cervical bursts ( 54 . 5% ) . In the presence of 1 µM SP , biphasic cervical bursts increased and were present in 3 of 4 preparations ( 75% ) . Alternately , biphasic bursts decreased in the presence of 10−12 M SST and were present in only 3 of 7 ( 42 . 9% ) preparations . This is lower than what was recently found in E18 . 5 mouse in vitro slice preparations but not unreasonable given the elevated potassium used to stimulate respiratory-related output in slices ( Chapuis et al . , 2014 ) . Under baseline conditions in preparations showing biphasic cervical bursts , these augmented bursts occurred in 8 . 87 ± 3 . 16% of total cervical bursts . In addition , biphasic cervical bursts also showed a consistent biphasic lumbar pattern in which activity occurred primarily during the initial cervical peak ( Figure 5A , D , Figure 5—figure supplement 1A , B , F , G , I ) . SP increased the frequency of biphasic cervical bursts , which were often accompanied by an active inhibition of lumbar output during the second cervical peak ( Figure 5B ) . This complex respiratory pattern was also found in cervical-lumbar-intercostal and cervical-lumbar-VII recordings ( Figure 5D , E ) . In contrast to lumbar output , intercostal EMG activity was most active during the augmented burst , whereas VII output could be active during either burst ( van Lunteren et al . , 1988; Lieske et al . , 2000; Ramirez and Viemari , 2005; Figure 5D–E ) . RL+/RTN− mice showed occasional biphasic cervical bursts , although lumbar activity in these events occurred during the larger peak ( Figure 5C ) . Note that some cervical bursts that lack a biphasic component can show either smaller or larger amplitudes than putative sigh-bursts ( Figure 5—figure supplement 1C–E , H ) . 10 . 7554/eLife . 02265 . 014Figure 5 . Atoh1 neurons are necessary for biphasic cervical sigh-like fictive breaths . E18 . 5 WT mice produce biphasic respiratory-related cervical bursts ( C4 , green ) with lumbar ( L1 , magenta ) activation only during the initial normal amplitude cervical burst under baseline ( A ) or in the presence of 1 µM SP ( B ) . Arrow in B indicates likely active inhibition of lumbar root during the larger amplitude cervical burst peak . ( C ) Biphasic cervical bursts persist after loss of RTN neurons in Phox2b-Cre;Atoh1LacZ/F ( RL+/RTN− ) mice with the lumbar burst occurring during the larger amplitude burst . Schematics indicate maintenance ( top left ) or targeted loss ( right ) of Phox2b RTN ( purple ) populations in WT or Phox2b-Cre;Atoh1LacZ/F ( RL+/RTN− ) mice . ( D–E ) In contrast to lumbar respiratory output ( magenta ) , XIth internal intercostal ( D , IC , black ) , or VIIn ( E , VII , dark blue ) can occur during both the initial and larger amplitude cervical burst ( green ) in E18 . 5 WT ( D ) or control Atoh1LacZ/+ heterozygote ( E ) . Note the co-activation of cervical and VII roots during lumbar inhibition ( E ) . ( F–I ) Atoh1LacZ/LacZ ( RL−/RTN− ) mice do not exhibit biphasic respiratory cervical bursts . Schematic indicates targeted loss of Phox2b RTN ( purple ) and Atoh1 RL ( orange ) neurons . ( F–G ) Single integrated traces showing temporal separation between cervical ( green ) and lumbar ( magenta ) peaks due to increased noise ( arrows ) . Numbers under traces indicate time of the lumbar peak in relation to the cervical peak ( in ms ) . In some fictive breaths , Atoh1LacZ/LacZ mice show respiratory doublets with two distinct cervical outputs with likely lumbar inhibition ( arrow in I ) during the initial burst in the presence of 1 µM SP ( H ) or 10−12 M SST ( I ) . Scale bar = 1 s . Figure 5—figure supplement 1 shows the variability in amplitude and pattern of biphasic cervical bursts other respiratory bursts during baseline rhythmic activity as well as average and overlapping standard and biphasic bursts in a E18 . 5 WT mouse preparation . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 01410 . 7554/eLife . 02265 . 015Figure 5—figure supplement 1 . Normal and biphasic cervical respiratory bursts show different cervical and lumbar patterns and vary in amplitude during fictive breathing in an E185 mouse brainstem preparation . ( A ) 100 s recording of integrated endogenous respiratory-related output from cervical ( C4 , green ) and lumbar ( L1 , magenta ) motor roots in an E18 . 5 WT mouse brainstem-spinal cord preparation . Letters above peaks ( B–G ) are expanded below for clarity . Note the variations in amplitude of both cervical and lumbar bursts . ( B–G ) 2 s overlapping traces of integrated cervical and lumbar motor output expanded from A . Peaks have been scaled to the same height for ease of comparison of the patterns of activity . Biphasic cervical bursts ( B , F–G ) show lumbar activity primarily during the initial cervical peak . Note that some fictive breaths that do not show biphasic cervical activity can have either smaller ( C–D ) or larger ( E ) amplitude motor output than biphasics bursts . Scale bar 10 s ( A ) , 500 ms ( B–G ) . ( H ) Integrated average top and overlapped cervical ( middle , green ) and lumbar ( bottom , magenta ) traces from 10 consecutive fictive breaths lacking biphasic cervical output . Note the variation in amplitude in cervical traces . Integrated average traces are scaled for clarity . ( I ) Integrated average top and overlapped cervical ( middle , green ) and lumbar ( bottom , magenta ) traces from 10 sequential , but not consecutive , fictive breaths showing biphasic cervical output . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 015 In contrast , RL−/RTN− mice did not generate fictive breaths with biphasic cervical output ( Figure 5F , G ) . They did show occasional double bursts , with two cervical outputs of similar amplitude separated by a short interval ( Figure 5H , I ) . In these cases , lumbar activity was limited to the second cervical burst . In many cases , cervical-only bursts accompanied by active inhibition of baseline lumbar motor root activity were still present in both RL−/RTN− and RL+/RTN− mice ( Figure 4—figure supplement 1L ) . This suggests that the absence of lumbar bursts can represent an active process similar to observations in vivo ( Pagliardini et al . , 2011 ) and is consistent with the maintenance of the neurons necessary for producing biphasic cervical outputs ( not shown ) . In adult mammals , abdominal muscles are usually silent during quiet breathing . Active abdominal activity during respiration is strongly state-dependent and appears under conditions of hypercapnia , hypoxia , anesthesia , and even sleep . When abdominal respiratory output is present , it usually precedes inspiratory activity , although in some cases it can also be present after inspiration ( Iizuka , 1999; Iizuka and Fregosi , 2007; Iizuka , 2009; de Almeida et al . , 2010; Pagliardini et al . , 2011 ) . In our E18 . 5 WT in vitro preparations , we noticed that lumbar activity either was co-active with , or preceded cervical activity ( Figure 6A–C ) . Similar co-activation patterns were previously reported but not quantified in older in vitro rat preparations ( Taccola et al . , 2007 ) . This suggested the temporal pattern of activation between motor pools might be controlled independently from the generation of each motor burst by unknown neurons . For each fictive breath , we calculated the time interval ( τ ) between the peaks of the integrated lumbar and cervical bursts ( Figure 6A ) . 10 . 7554/eLife . 02265 . 016Figure 6 . Rhombic lip , Atoh1-dependent neurons are necessary for normal cervical-lumbar temporal lag . ( A ) Cartoons showing hypothesized network interactions ( bottom ) between putative independent lumbar ( magenta square ) and cervical ( green circle ) oscillators and unknown inhibitory interneurons , together generating two distinct temporal patterns of cervical ( green ) and lumbar ( magenta ) respiratory co-activation ( top ) . These different networks produce motor outputs that differ in the relative time between their peaks ( lumbar time-cervical time = τ , in ms ) . ( B–D ) Integrated and overlapped cervical and lumbar traces from 10 consecutive ( B , WT–τ <150 ms ) , sequential but not consecutive ( C , WT–τ ≥150 ms ) , or consecutive ( D , Atoh1LacZ/LacZ [RL−/RTN−] ) respiratory-related bursts aligned from the peak of cervical output ( top–single burst , middle–average of 10 breaths , bottom–overlap of 10 bursts , numbers under traces = τ ) . Note the low level pre-inspiratory activity of the cervical trace during the lumbar burst as well as likely inhibition during the cervical burst in wild-type mice ( C ) , but the strong simultaneous co-activation of lumbar and cervical outputs in Atoh1LacZ/LacZ mice ( D ) . Partial peak delays are still present in E18 . 5 Phox2b-Cre; Atoh1LacZ/F ( RL+/RTN− , E ) , and Phox2b-Cre;VGlut2F/F mice ( RL+/RTNsilent , F ) indicating Phox2b RTN neurons are not essential for temporal delay . ( G–H ) Histograms ( 12 ms bins ) showing distributions of τ for WT ( G–H ) and Atoh1LacZ/LacZ ( G and I ) mice to determine whether loss of RL and RTN neurons affects τ . Distributions are scaled to same height for better comparison . Gray dashed line indicates τ = −150 ms cutoff . Note the clear differences in the peak and spread between baseline WT and Atoh1LacZ/LacZ distributions ( G ) . WT but not Atoh1LacZ/LacZ distributions show increased temporal delay after application of 1 µM SP or 10−12M SST ( H–I ) . Histogram color: WT ( baseline–black , SP–red , SST–green ) , Atoh1LacZ/LacZ ( baseline–cyan , SP–purple dashed , SST–orange dashed ) . See Figure 6—figure supplement 1 for traces showing the absence of effect of SP or SST on temporal delay ( τ ) in Atoh1LacZ/LacZ mice . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 01610 . 7554/eLife . 02265 . 017Figure 6—figure supplement 1 . Peptides do not induce cervical-lumbar temporal delay in Atoh1LacZ/LacZ ( RL−/RTN− ) mice . Schematic ( top ) indicates targeted loss of both Phox2b RTN neurons ( purple ) and RL Atoh1 neurons ( orange ) . Overlapped integrated cervical ( green ) and lumbar ( magenta ) bursts showing lack of temporal delay in E18 . 5 Atoh1LacZ/LacZ preparations in the presence of 1 µM SP ( A ) or 10−12M SST ( B ) . Integrated traces from single fictive breaths ( left ) , average trace of 10 consecutive bursts ( middle ) , and overlap of 10 bursts ( right ) aligned from the peak of cervical output . Numbers under traces indicate time of integrated lumbar peak relative to cervical peak ( τ , in ms ) . Scale bar = 500 ms . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 017 In E18 . 5 mice , lumbar bursts preceded cervical bursts with two distinct peaks ( −220 or −60 ms , Figure 6B , C; evidence for both peaks is shown in the delay histogram ( gray line in Figure 6G ) ) . This is similar to intervals recorded in post-natal rat in vitro preparations ( Taccola et al . , 2007 ) and compares to delays of from less than 300 to over 500 ms between abdominal and diaphragm muscle activation in anesthestized adult rats breathing elevated CO2 or decreased O2 at 37° ( Iizuka and Fregosi , 2007 ) . Separately analyzing cycles based on long or short intervals , we found that the peak of lumbar output in our preparations preceded peak cervical output by ≥150 ms in 25 . 9 ± 6 . 6% of respiratory bursts ( arithmetic mean ) ( n = 21 , Figure 6B , C , G , H ) . In many cases , peak lumbar activity overlapped low-level ‘pre-inspiratory’ cervical activity ( Figure 6C ) . Peptidergic modulation of the hindbrain with SP ( 1 µM ) , which stimulates preBötC and both RL and RTN populations , or SST ( 10−12M ) , which inhibits preBötC but not RTN or rostral RL populations ( Gray et al . , 2010; Gray , 2013 ) , did not statistically increase the percentage of respiratory bursts with temporal delay ( 36 . 9 ± 8 . 7% for SP modulation , p<0 . 6 n = 15; 39 . 6 ± 10 . 2% for SST modulation , p<0 . 6 [One-way ANOVA] , n = 7 , Figure 6H ) . These data indicate these reduced preparations can produce the adult-like respiratory pattern of lumbar activity preceding cervical activity ( Iizuka , 2011 ) . In RL−/RTN− mice ( n = 19 ) , in contrast , the percentage of fictive breaths where the lumbar peak preceded the cervical peak by ≥150 ms was significantly decreased ( 4 . 1 ± 2 . 2% , p<0 . 004 , Independent samples t test; Figure 6D , G , I ) . The interval between peaks in this 4 . 1% of respiratory bursts can be attributed to the inherent noise and variability in the output signals and are unlikely to reflect actual temporal disparity ( Figure 5F , G ) . The temporal distribution for RL−/RTN− mice showed a single peak temporal difference that was significantly less than in WT ( median peak lag in WT = −39 ms and RL−/RTN− mice = −3 ms , p<0 . 0016; Figure 6G , I ) . This loss of normal temporal interval was not a simple consequence of low excitability in the network , because neither excitation with SP nor preBötC inhibition with SST ( Onimaru et al . , 2006; Gray et al . , 2010 ) generated any respiratory cycles with clear lumbar-cervical delay ( Figure 5F , G , Figure 6I , Figure 6—figure supplement 1A , B ) . Unlike RL−/RTN− mice , however , RL+/RTN− and RL+/RTNsilent mice maintained the interval between the cervical and lumbar peaks in some fictive breaths ( Figure 6E , F ) . The overall temporal distributions in RL+/RTN− affected preparations were not statistically different from those of either WT or RL−/RTN− mice because their rhythms were extremely slow and they had far fewer respiratory bursts to analyze . These data support previous findings that RTN neurons provide a baseline excitatory drive to maintain respiratory output ( Takakura et al . , 2008 ) but suggest that RL neurons have a different function , possibly related to relative timing between motor pools . The change of the delay interval distribution between lumbar and cervical peaks in RL−/RTN− mice is consistent with a change in connectivity between independent oscillators . The appearance of uncoupled fictive breaths , especially evident in cervical-lumbar-VII recordings , led us to wonder whether what might appear to be deletions at longer time scales ( >1 s ) were , in fact , only changes in the relative timing of respiratory output such that lumbar or VII activity occurs at a different phase of the respiratory cycle ( Figure 7A ) . In vivo , the peak of activity for many respiratory neurons is delayed to late in the expiratory period ( Richter , 1982 ) . This type of phase switching is seen when moving from the simultaneous activation of legs during hopping to alternating activation during walking . To address this , we determined the phase of the peaks of lumbar and VII motor output during the cervical respiratory period ( defined as Φ [0–360°] for each individual respiratory burst , Figure 7A ) ( Strohl , 1985; Hwang et al . , 1988; Plowman et al . , 1990; Huangfu et al . , 1993; Iscoe , 1998 ) . Consistent with our interval analysis in WT mice , we found a single peak for both lumbar and VII root activity just prior to the peak of cervical output ( Figure 7B ) . In RL−/RTN− mice , the peak of both lumbar and VII activity also occurred just prior to cervical output ( Figure 7C ) . Interestingly , VII output showed a broadening of the phase of activity relative to cervical output ( Figure 7C , D ) . In WT mice we found no secondary peaks that would indicate consistent phase differences . In RL−/RTN− mice , however , the lumbar trace showed a small second peak around 50° ( arrows in Figure 7C , E ) . This peak represented a respiratory pattern unique to RL−/RTN− mice that was characterized by an isolated cervical burst followed , after a pause , by lumbar and/or VII output ( Figure 7F ) . Together , however , these data suggest an absence of a longer timescale reorganization of respiratory activity . They also show that both lumbar and VII outputs are still largely coupled to cervical output , with changes in the details of their relative timing . 10 . 7554/eLife . 02265 . 018Figure 7 . Loss of RL and RTN neurons does not change the phase relationship between putative respiratory oscillators . ( A ) Cartoons indicating how relative phase ( Φ , in degrees ) between cervical burst peaks and either lumbar or VII respiratory peaks for each fictive breath are calculated ( left ) and possible change in phase relationships after loss of RL and RTN neurons ( right ) . Schematics ( top right ) indicate targeted loss of Phox2b RTN ( purple ) and Atoh1 RL ( orange ) neurons . ( B ) Histogram ( 5° bins ) showing the distributions of phase relationship ( Φ ) between cervical and either lumbar ( magenta ) or VII ( dark blue ) peaks indicating consistent temporal delays between outputs in WT mice . ( C ) Phase relationships ( Φ ) between cervical and either lumbar ( cyan ) or VII ( tan ) outputs indicate a maintenance in phase relationships between respiratory outputs in Atoh1LacZ/LacZ ( RL−/RTN− ) mice . ( D ) Loss of RL/RTN neurons broadens the phase relationship ( Φ ) of cervical-VII root activation compared to WT . ( E ) . Loss of RL/RTN neurons does not effect phase ( Φ ) of most lumbar bursts but introduces a occasional Atoh1LacZ/LacZ specific burst pattern ( arrow in C and E , F ) with the cervical burst ( green ) followed by a pause then overlapping lumbar ( magenta ) and VII ( dark blue ) bursts . Scale bar = 500 ms . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 018 It is now possible to construct a model for the respiratory network that would explain the role of RL neurons ( Figure 8A ) . We propose that RL neurons inhibit the preBötC via activation of inhibitory interneurons but excite the as-yet unidentified abdominal oscillator . The strength of RL-induced inhibition of preBötC neurons influences the temporal interval between cervical and lumbar output . Similarly , preBötC neurons activate inhibitory interneurons to attenuate lumbar output ( Pagliardini et al . , 2011 ) . This type of asymmetrical inhibition has been suggested to underlie intersegmental coupling of independent oscillators in crayfish and lamprey rhythmogenic networks ( Mulloney , 1997; Hill et al . , 2003; Mulloney and Hall , 2007 ) . This RL neuron-mediated net inhibition would also explain the remarkable slowing of respiratory-related rhythmic output seen with selective RTN ablation or silencing ( Figure 3C , D ) . 10 . 7554/eLife . 02265 . 019Figure 8 . Proposed respiratory network underlying temporal relationship between inspiratory and expiratory respiratory oscillators . ( A ) Putative cervical ( preBötC , green circle ) and lumbar ( RTN/pFRG , magenta square ) oscillators contain Dbx1-derived neurons ( blue filled circles ) and drive expiratory ( magenta , abdominals ) and inspiratory ( green , diaphragm ) muscles . The preBötC is asymmetrically coupled to RTN/pFRG neurons via unknown inhibitory interneurons ( gray filled circles ) . Phox2b RTN neurons ( purple ) provide excitatory drive to both oscillators . Atoh1 RL neurons ( orange ) excite the RTN/pFRG oscillator . Lines with circles indicate glutamatergic connections . Lines with arrowheads indicate inhibitory GABA/glycine connections . Black lines indicate strong synaptic connections . Dashed lines indicate weak synaptic connections . ( B ) EYFP-labeled neurons ( green ) within the BötC area of an Atoh1-Cre;R26 EYFP slice preparation . The recorded neuron was filled with Alexa 568 ( magenta ) through whole-cell patch dialysis . The merged pseudo-color fluorescence and transmitted light IR-DIC images of the recorded neuron are also shown . ( C–F ) Current-clamp recordings of BötC Atoh1 neurons ( black , top ) . The corresponding integrated inspiratory activity is shown as ∫XII ( bottom , green ) . The majority of Atoh1 neurons show tonic non-respiratory activity ( C–D ) . ( E ) Recording of an expiratory Atoh1 neuron ( from B ) with rhythmic inhibition during inspiration . ( F ) Inhibitory inspiratory drive reverses with current bias to achieve a baseline membrane potential of −80 mV . Voltage scale bar = 20 mV . ( G ) Inhibitory inspiratory drive is eliminated with bath application of picrotoxin/strychnine . Scale bar = 1 s . ( H–J ) Cartoons showing hypothesized network reorganizations ( bottom ) underlying cervical ( green ) and lumbar ( magenta ) bursts ( top ) . ( H ) Outputs without temporal lag are mediated by weak Atoh1 RL and inhibitory inputs and strong RTN/pFRG and preBötC excitatory coupling . ( I ) Fictive breaths with temporal delay between cervical and lumbar motor outputs have strong Atoh1 excitation of RTN/pFRG Dbx1 neurons and produce phasic inhibition of preBötC neurons . During the cervical burst , strong preBötC-mediated excitation of inhibitory interneurons leads to silencing of lumbar outputs . ( J ) Fictive breaths with lumbar inhibition during cervical bursts are due to weak Atoh1 RL neuron output and strong preBötC-mediated excitation of inhibitory interneurons leads to silencing of lumbar outputs . Lines define connectivity as in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02265 . 019 To address the role of RL neurons , we recorded from fluorescently labeled RL , Atoh1-derived neurons in the ventrolateral medulla rostral to the preBötC but caudal to the RTN ( Figure 8B ) . Consistent with our model , most RL neurons were tonically active ( 18 of 19 , Figure 8C , D ) . One Atoh1-derived neuron was rhythmically inhibited during inspiration ( Figure 8E ) . This inhibition reversed at potentials more negative than chloride reversal potential; it was eliminated by blockade of fast inhibitory amino acid receptors and consistent rhythmic synaptic input from inspiratory glycinergic neurons ( Figure 8F , G; Winter et al . , 2009; Morgado-Valle et al . , 2010 ) . Phasic inhibition suggests that at least some Atoh1 neurons within the medulla are synchronized to respiratory output . These data suggest a mechanism by which the respiratory network can produce the variety of coordinated motor patterns seen in vivo ( Figure 8H–J ) . Distinct but coupled oscillatory populations generate independent rhythms , but reconfiguration of the relative strengths of Atoh1- and Dbx1-dependent populations produces changes in the strength of both excitatory and inhibitory connections , leading to changes in respiratory pattern .
Breathing is a complex behavior that responds to the homeostatic needs of an organism and involves the coordinated activation of numerous respiratory muscles , which in mammals includes the diaphragm . The activities of other respiratory muscles are partially state-dependent ( Iscoe , 1998; Iizuka and Fregosi , 2007; de Almeida et al . , 2010; Pagliardini et al . , 2012 ) . Respiratory muscles , including the diaphragm , are also active during non-respiratory behaviors such as cough , emesis , valsalva , and hiccup ( Tomori and Widdicombe , 1969; Miller et al . , 1987; Iscoe , 1998; Straus et al . , 2003 ) . In this study , we begin to unravel how the respiratory network enables the temporal coordination of these different muscles . In vitro preparations produce endogenous rhythmic output , which , while not exact replicates of breathing movements in vivo , have shed light on how respiratory , especially inspiratory , behaviors are generated ( Feldman et al . , 2013; Funk and Greer , 2013 ) . Reduced in vitro slice and en bloc preparations , as well as in situ perfused brainstem preparations have been found to express a larger range of fictive respiratory behaviors including sighs , gasps , and active expiration ( Smith et al . , 1990; Iizuka , 1999; Lieske et al . , 2000; Shvarev et al . , 2003; Iizuka , 2004; Taccola et al . , 2007; Abdala et al . , 2009; Funk and Greer , 2013; Chapuis et al . , 2014 ) . These preparations have also allowed us to extend our analysis of respiratory networks to include transgenic mouse models that do not survive birth or early life ( Champagnat et al . , 2011 ) . At late developmental stages , E18 . 5 WT mice delivered by C-section will breathe in vivo and produce robust inspiratory-related output in vitro ( Greer et al . , 2006 ) . We found that at this age , in vitro preparations also produce robust rhythmic output from lumbar motor pools that innervate abdominal muscles active during expiration , thoracic motor pools that innervate intercostal muscles during expiration , and VII motor pools , all of which is consistent with previous work ( Huangfu et al . , 1993; Iscoe , 1998; Onimaru et al . , 2006; Iizuka , 2010 ) . In contrast to experiments in older animals , however , we found this lumbar output did not require altered pH or pharmacological stimulation . Moreover , we found the lumbar output overlapped cervical output for the majority of respiratory-related bursts , although the more mature ‘pre-inspiratory’ pattern was also present as in older rat in vitro preparations ( Taccola et al . , 2007 ) . Two developmentally distinct populations within the RTN/pFRG ( Phox2b- and Atoh1-dependent RTN neurons ) as well as Atoh1-dependent RL neurons have been hypothesized to generate the abdominal and/or VII rhythms ( Onimaru et al . , 2008; Thoby-Brisson et al . , 2009; Gray , 2013 ) . We find that both RL and RTN neurons are important for appropriately coordinated respiratory rhythm in cervical and lumbar motor pools . Respiratory periods are much longer in the absence of RTN glutamatergic neurons , consistent with previous work ( Dubreuil et al . , 2009; Ramanantsoa et al . , 2011; Huang et al . , 2012 ) . The elimination of both RL and RTN neurons , however , recovered baseline respiratory period . Not only does this reveal a previously unknown net inhibitory role for RL neurons in the control of breathing , but it also demonstrates that neither of these populations is necessary or sufficient for expiratory motor output , since both cervical and lumbar respiratory-related outputs persist after their genetic elimination . In the absence of Dbx1-derived neurons , however , all rhythmic respiratory outputs were eliminated . The absence of both cervical and lumbar rhythms in Dbx1 mutant mice suggests that in addition to its inspiratory rhythmogenic role in the preBötC , Dbx1 is essential for the development of neurons necessary for the expression of rhythmic respiratory output from other motor pools , although the exact location of these neurons remains unknown . Dbx1-dependent neurons have recently been shown to play an important role in locomotor pattern generation ( Talpalar et al . , 2013 ) . The temporal overlap of lumbar and cervical bursts raised the question whether the lumbar output seen in E18 . 5 wild-type preparations is generated by the same networks that produce active expiration in more mature in vitro preparations or in vivo . We propose that this is the case for three reasons . First , we found that lumbar and cervical activities were not identical . Either output could occur either in isolation or with a distinct temporal separation . This same pattern of independent motor activity was also seen in respiratory output from internal intercostal muscles or VII motor roots , suggesting it is a general property of respiratory motor pools . Second , output from lumbar and thoracic roots ( which innervate expiratory musculature ) has previously been shown to require structures outside of the preBötC and the rostral ventral respiratory group , which are the well-established sources of rhythmic inspiratory drive and premotor innervation to the phrenic motor pool ( Miller et al . , 1985; Merrill and Lipski , 1987; Sasaki et al . , 1991; Iscoe , 1998; Janczewski et al . , 2002; de Almeida et al . , 2010; Ford and Kirkwood , 2013 ) . These anatomical differences are consistent with independent sources of respiratory drive to different motor pools serving inspiration and expiration , respectively , although we cannot rule out changes in connectivity during maturation ( Iscoe , 1998 ) . Third , simultaneous co-activation of cervical and lumbar roots is also present in older in vitro rat preparations , suggesting it is only the relative percentage of respiratory bursts showing one or the other pattern that changes , not the ability to generate each pattern ( Taccola et al . , 2007 ) . We further provide evidence that each respiratory motor pool receives input from independent but coupled oscillators that interact within the timescale of an individual respiratory event ( Carroll and Ramirez , 2013; Moore et al . , 2013 ) . Respiratory networks are capable of reconfiguring to produce different patterns of respiratory outputs , such as sighs , depending upon the behavioral or modulatory state of the organism ( Lieske et al . , 2000; Doi and Ramirez , 2008 , 2010; Chapuis et al . , 2014 ) . We extend those findings to include activation of lumbar , thoracic , and VII motor pools , which are normally , but not obligatorily , active during different phases of the respiratory cycle . We observed that sigh-like cervical patterns corresponded with lumbar outputs active primarily only during the initial eupneic phase , whereas intercostal and VII outputs were active during both biphasic cervical bursts . In contrast to these motor outputs , 95% of preBötC inspiratory neurons are active during both inspiratory bursts and during sighs , which suggests that the differential gating of motor output occurs outside the putative inspiratory oscillator or by transient reconfigurations within respiratory populations ( Tryba et al . , 2008; Chapuis et al . , 2014 ) . We speculate that the differences in the timing and presence of respiratory outputs from lumbar and thoracic motor pools between these late fetal and more mature preparations are related to differences in the balance of synaptic interactions between hindbrain neurons and not the absence of connectivity ( Thoby-Brisson et al . , 2009; Bouvier et al . , 2010; Perreault and Glover , 2013 ) . The loss of both RTN and RL neurons does not preclude respiratory deletions but does lead to periods of transient decoupling of respiratory motor rhythms ( e . g . , Figure 4H ) . Between control and RL−/RTN− mice , individual motor pools differed in their co-activation . Lumbar outputs increased but cervical decreased their percentage of solitary activity . These data are inconsistent with a role for RL and RTN populations in generating lumbar respiratory output . It is important to note that the strong coupling of the cervical , lumbar , and VII motor roots in RL−/RTN− mice suggests that the excitatory connections likely arising from the preBötC are maintained ( Tan et al . , 2010; Feldman et al . , 2013; Moore et al . , 2013 ) . In RL−/RTN− mice , the temporal overlap between cervical and lumbar output was nearly complete , again raising the question of the independence of the drives to each motor pool . Because there is no evidence in the literature that loss of Atoh1 would affect Atoh1-independent projection patterns , we suggest the temporal gap is lost because of differences in connectivity between hindbrain respiratory populations ( Miesegaes et al . , 2009; Rose et al . , 2009a; Kohl et al . , 2012; Perreault and Glover , 2013 ) . We find that the loss of both RL and RTN neurons eliminates respiratory bursts with intervals longer than 150 ms between respiratory oscillators , that is , each population can be co-active or independent but lacks the capacity for the short temporal gaps necessary for normal breathing patterns . We interpret these results to indicate that the loss of RL neurons eliminates a net inhibitory effect on preBötC Dbx1 neurons , likely via reciprocally connected inhibitory interneurons . Recent work has found that biphasic inspiratory bursts ( sighs ) begin during late embryogenesis and that the temporal delay between eupneic and sigh components is dependent upon the strength of synaptic inhibition ( Chapuis et al . , 2014 ) . Moreover , the density of the co-transporter necessary for setting the chloride equilibrium potential in the medulla increases significantly during the post-natal period ( Liu and Wong-Riley , 2012 ) . This suggests the difference in the relative percentage of respiratory cervical-lumbar bursts showing temporal delays above 150 ms between our E18 . 5 preparations and older animals may be simply a consequence of generally weaker synaptic inhibition ( Greer et al . , 2006 ) . We propose RL neurons play an essential role in the coordination of mammalian breathing behavior by modulating the inhibitory interneurons that produce the temporal lag between independent oscillators . This modulation also helps stabilize coupled oscillators preventing independent or unwanted activation . This has the advantage that temporal delays can be controlled without strongly affecting the rhythmogenic properties of the underlying oscillators ( Hill et al . , 2003; Grillner , 2006 ) . It is also possible , however , that the loss of Atoh1 neurons indirectly affects respiratory networks by modulating the development of inhibitory populations or the overall maturation of hindbrain circuits . It is important to point out that we were unable to selectively eliminate lumbar motor output in any of our genetic manipulations . While our data are consistent with the presence of independent respiratory oscillators , we cannot rule out that the variability we see between respiratory motor pools is the consequence of higher level network interactions between a single distributed respiratory oscillator and small pools of respiratory premotor neurons ( Smith et al . , 2007 ) . Similarly , whether there are discrete anatomical boundaries between these putative independent oscillators is also unknown . Atoh1 is essential for the development of a number of important neural populations in vertebrates . One general characteristic of many of these populations is their importance for the temporal coordination of the activity of different populations , be it fast motor coordination for neurons of the cerebellum or auditory processing in the hindbrain ( Wang et al . , 2005; Maricich et al . , 2009; Miesegaes et al . , 2009; Rose et al . , 2009a ) . This temporal functionality is also consistent with previous data indicating a role for the Atoh1-derived RL neurons of the parabrachial nucleus in the phase-dependent inhibition of inspiratory output ( Richter et al . , 1992; Gray , 2008; Rose et al . , 2009a ) . These data emphasize that phasic inhibition influences respiratory patterns but is not rhythmogenic per se , as has been suggested ( Smith et al . , 2007; Abdala et al . , 2009; Richter and Smith , 2014 ) . An influential model proposes that simple behaviors are generated by glutamatergic interneurons controlling the activation of segmentally organized unit burst oscillators ( central pattern generators ) ( Grillner , 2011 ) . These excitatory unit-burst-oscillator networks are putatively coupled to inhibitory interneurons such that independent modules generate the activity underlying muscle groups that must be properly coordinated , for example , flexors and extensors across hinge joints in locomotion or specific left-right axial muscles in body segments during swimming . Recent work provides evidence for unit burst oscillator modules coordinated by inhibition in the mammalian spinal cord ( Dougherty et al . , 2013; Hägglund et al . , 2013; Talpalar et al . , 2013 ) . We propose that the mammalian respiratory network is organized in a similar fashion , with separate oscillators driving specific cervical and lumbar motor pools , in which the preBötC plays the central organizing role ( Feldman et al . , 2013; Moore et al . , 2013 ) . The fact that RL glutamatergic neurons are not necessary for rhythmogenesis implies that the generation of rhythm and the modulation of phase timing are controlled by separate populations . This shared control is similar to proposed networks of coupled independent oscillators in invertebrate motor neurons but represents a novel viewpoint in vertebrate systems ( Hill et al . , 2003; Jezzini et al . , 2004; Mulloney and Hall , 2007 ) . Whether Atoh1-dependent neurons control temporal intervals between independent oscillators only in breathing or also in other rhythmic behaviors is an interesting question for future research .
Experiments were done in accordance with the Institute for Laboratory Animal Research Guide for the Care and Use of Laboratory Animals ( Care et al . , 1985 ) . All experiments were approved by the Animal Studies Committee at Washington University School of Medicine , the Institutional Animal Care and Use Committee at The College of William & Mary , and the Center for Comparative Medicine , Baylor College of Medicine . We utilized Atoh1-Cre transgenic ( Atoh1-CreTg ) ( Matei et al . , 2005 ) , Atoh1Cre/+ ( Yang et al . , 2010 ) , Atoh1LacZ/+ ( Ben-Arie et al . , 1997 ) , Atoh1Flox/Flox ( Atoh1F/F ) ( Huang et al . , 2012 ) , Dbx1LacZ/+ ( Pierani et al . , 2001 ) , Phox2b-Cre ( Rossi et al . , 2011 ) ; Rosa26-stop-eYFP ( R26;YFP ) ( Srinivas et al . , 2001 ) , Rosa26-stop-TD Tomato ( Madisen et al . , 2010 ) , and VGlut2Flox/Flox ( VGlut2F/F ) ( Tong et al . , 2007 ) mice . Mice were crossed and bred on a C57BL6 or mixed CD1/C57BL6 background . Brainstem-spinal cord ( en bloc ) preparations with an anterior transection near the diencephalon–midbrain junction were made using E18 . 5 embryos delivered under anesthesia ( ketamine/xylazine mixture ) by cesarean section from timed-pregnant female mice . The dissections were done while keeping the embryos submerged in cold ( 4°C ) artificial cerebral spinal fluid ( regular/enhanced = WT [n = 17/5]; RL−/RTN−[n = 13/6]; RL+/RTN−[n = 5/1]; RL+/RTNsilent[n = 3/0]; aCSF ( in mM ) : 124 NaCl , 3/5 KCl , 1 . 5/2 . 4 CaCl2 , 1 . 0/1 . 3 MgSO4 , 25 . 0/26 . 0 NaHCO3 , 0 . 5 NaH2PO4/1 . 2 KH2PO4 , 30 D-Glucose ( Sigma , St . Louis , MO ) equilibrated with 95% O2 and 5% CO2 to pH = 7 . 4 ) and transferred into a partitioned 6 ml recording chamber , which were separately gravity fed by reservoirs of heated ( 25°C–26°C ) and aerated ( 95% O2 and 5% CO2 ) aCSF at a rate of 3–4 ml/min . After transferring the preparation into the recording chamber , the brainstem and spinal cord compartments were rendered mutually impervious with petroleum jelly . Allowing ∼30 min for stabilization , the extracellular electrophysiological recordings were made ( acquisition rate 4 kHz ) simultaneously from a cervical ( C2–C6 ) and a lumbar ( L1 ) ventral spinal motor root using suction electrodes , differentially amplified ( low noise Grass Instruments , band pass filtering [0 . 3–3 kHz] ) , digitized using an analog to digital converter ( AD instruments , Colorado Springs , CO ) and then integrated over time ( absolute value with a 100 ms decay time constant ) using LabChart 7 Pro software ( version 7 . 2 . 4 , AD Instruments ) . Similarly , triple recordings were also made using an additional suction electrode to record from facial nerve or a sharp tungsten microelectrode ( 10 mΩ impedance and 1 µm tip diameter; FHC , inc . ) to record electromyograph from intercostal muscle XI ( IC ) in a non-partitioned bath as previously described in neonatal rats ( Iizuka , 1999 ) . After recording baseline activity , 1 μM substance P ( SP ) or 10−12 M somatostatin ( SST ) in aCSF was added selectively to the brainstem compartment in the partitioned bath preparations . In some early experiments , we used an enhanced aCSF suggested to increase the percentage of respiratory bursts with lumbar activity ( Iizuka , 2004; Ruangkittisakul et al . , 2008 ) . We found no changes in the percent of lumbar bursts nor any differences in the phase distributions of fictive breaths , so we combined both solutions when comparing deletion and time lag . The peak time and amplitude , phase difference and inter-burst intervals ( period ) were determined for cervical , lumbar , and facial nerve bursts as well as for IC . The comparisons of cervical and lumbar periods indicative of respiratory frequency were made between the different genotypes by direct comparison , whereas the amplitudes of the integrated respiratory bursts were first normalized against the baseline arithmetic mean determined individually . The percentage of inspiratory and expiratory deletions was also calculated from each recording and these values were then compared between treatments and mouse lines . In cases of triple recording involving facial nerve , the relative degree of phase separation for the lumbar and facial nerve bursts from the preceding cervical burst ( Φ ) was determined by equating the instantaneous cervical inter-burst interval period to 360° . Moreover , coupling efficiencies of rhythms whose integrated peaks occurred within 1 s period of a cervical inspiratory burst were quantified and compared in facial nerve and IC triple recordings . Transverse slices ( 550 µm thick ) from neonatal ( P0-P2 ) Atoh1-CreTG X Rosa26-stop-YFP or tdTomato mice were dissected and prepared for recordings as described previously ( Hayes and Del Negro , 2007; Picardo et al . , 2013 ) . On-cell and whole-cell patch recordings were obtained using infrared-enhanced differential interference contrast videomicroscopy ( IR-DIC ) after fluorescent identification of Atoh1-derived neurons in both YFP and tdTomato reporter mice . ACSF contained ( in mM ) : 124 NaCl , 9 KCl , 0 . 5 NaH2PO4 , 25 NaHC03 , 30 D-glucose , 1 . 5 CaCl2*2H20 , and 1 MgSO4 . Slices were placed into a 0 . 5-ml chamber within an upright fixed-stage microscope ( Zeiss Microimaging , Thornwood , NY ) and ACSF was perfused at ∼5 ml/min at 27–28°C . Patch recordings employed a Dagan IX2-700 amplifier ( Minneapolis , MN ) . Respiratory-related motor output was monitored from XII nerves with extracellular suction electrodes and a high-gain differential amplifier with band-pass filtering ( 0 . 3–1 kHz ) ( Dagan EX4-400 ) , full-wave rectified and smoothed for display . Data were acquired using Chart software and a Powerlab 4/30 ( ADInstruments ) . A liquid junction potential , which measured 1 mV , was not corrected in current-clamp experiments . We used the following patch solution containing ( in mM ) : 140 K-gluconate , 5 NaCl , 0 . 1 EGTA , 10 HEPES , 2 Mg-ATP , and 0 . 3 Na ( 3 ) -GTP ( pH = 7 . 2 using KOH ) . We added 2–4 µl/ml of Alexa Fluor 568 hydrazide ( Na+ salt , Invitrogen , Carlsbad , CA ) to the patch solution for fluorescent visualization of morphology . Neurons were visually identified using both IR-DIC videomicroscopy and epifluorescence illumination ( X-Cite 120 , EXFO , Montreal , Canada ) and a fluorescent filter to identify YFP or tdTomato labeled cells . Images were acquired of recorded Atoh1 neurons . Tissue sections were washed in PBS with 0 . 2% triton X-100 , blocked in 10% heat inactivated normal horse sera , incubated in antibody overnight at 4°C , incubated in secondary antibody and coverslipped in Vectashield or Prolong Gold . Chicken anti-beta galactosidase ( LacZ ) 1:4000 ( Abcam , Cambridge , MA ) , Chicken anti-green fluorescent protein ( GFP ) 1:1000 ( Aves Labs , Tilgard , OR ) , Lbx1 ( 1:10 , 000 , gift from C Birchmeier ) , Rabbit anti-neurokinin 1 receptor ( NK1R ) 1:2000 ( Millipore , Billerica , MA ) , Goat anti-Phox2b 1:500 ( Santa Cruz Biotechnology ( SCBT ) , Santa Cruz , CA ) , Goat anti-Islet-1 1:500 , ( Neuromics ) . All antibodies used have been previously characterized and no signals were present in genetic or antibody controls . As previously described ( Gray , 2013 ) , slides were immersed in 4% PFA , permeabilized with proteinase K or RIPA Buffer , washed in 0 . 1 M triethanolamine-HCl with 0 . 25% acetic anhydride , blocked in hybridization buffer at 65°C , then placed into slide mailers containing hybridization buffer with DIG-labeled antisense RNA at 1 µg/ml overnight at 65°C . Slides were washed in SSC buffers at 62°C , then washed and incubated in alkaline phosphatase conjugated anti-DIG antibody in 10% NHS and incubated in NBT-BCIP until cellular labeling is clear . For combined immunohistochemistry and in situ hybridization , slides are stained for mRNA expression prior to immunohistochemical labeling . Mice were genotyped by PCR using primers specific for Atoh1F , Atoh1LacZ , Cre recombinase , Dbx1LacZ , GFP/YFP , TD-Tomato , and VGlut2F , as previously described ( Ben-Arie et al . , 1997; Pierani et al . , 2001; Srinivas et al . , 2001; Matei et al . , 2005; Tong et al . , 2007; Madisen et al . , 2010; Yang et al . , 2010; Rossi et al . , 2011; Huang et al . , 2012 ) . Neonatal pups ( P0–P4 ) or embryos from timed pregnant females ( morning of plug = E0 . 5 , E18 . 5 ) were anesthetized and perfused with 4% paraformaldhyde in 0 . 1 M PBS , pH 7 . 4 . Embryos or isolated brainstems were postfixed in PFA overnight at 4°C , cryoprotected in 25% sucrose in PBS , blocked , frozen in OCT , and stored at −75°C . Brainstems were sectioned in sets of six on a Hacker ( Winnsboro , SC ) cryostat at 20 µm and sections are thaw mounted onto Superfrost Plus slides and stored at −20°C until use . Fluorescent and brightfield images were acquired using a Nikon 90i microscope ( Nikon Instruments , Melville , NY ) , Roper H2 cooled CCD camera ( Photometrics , Tucson , AZ ) , and Optigrid Structured Illumination Confocal with a Prior ( Rockland , MA ) motorized translation stage . Pseudo-colored images were acquired in Velocity ( Perkin Elmer , Waltham , MA ) , and modified in Photoshop ( Adobe , San Jose , CA ) and exported as 8 bit JPEG images . Images were filtered and levels were modified for clarity . Temporal interval values ( τ , in ms ) and respiratory phase ( Φ , in ° ) were binned ( delay–12 ms , phase–5° ) , and graphed in Igor Pro ( Wavemetrics , Lake Osego OR ) then exported to Adobe Illustrator and scaled to a standard height for comparison . Delay histograms were generated from 1088 ( WT Baseline ) , 789 ( WT SP ) , 373 ( WT SST ) , 1522 ( Atoh1LacZ/LacZ baseline ) , 318 ( Atoh1LacZ/LacZ SP ) , and 1047 ( Atoh1LacZ/LacZ SST ) respiratory-related bursts . Phase histograms were generated from WT cervical-lumbar or cervical-VII breaths ( 506 , 500 ) and Atoh1LacZ/LacZ cervical-lumbar or cervical-VII bursts ( 2125 , 857 ) . Respiratory periods , the distribution of time lag between inspiratory and expiratory burst peak times , the effect of time lag on the amplitude of respiratory bursts and percentage of respiratory deletions were compared using mixed random effects ANOVA with mouse as the random effect and least-squares mean values were determined . The percentage of fictive breaths with respiratory phase lag were compared between WTs and RL−/RTN− mice using independent samples t test and One-Way ANOVA followed by Tukey's honestly significant difference test for the effect of peptides with-in these genotypes ( *p<0 . 05 , **p<0 . 01 , #p<0 . 001 ) . | A healthy adult at rest will breathe in and out around 20 times per minute . Each breath requires a complex series of coordinated muscle activity . Inhalation begins with the opening of the airway followed by the contraction of the diaphragm and the intercostal muscles between the ribs , causing the chest cavity to expand . As the lungs increase in volume , the pressure inside them drops and air is drawn in . Relaxation of the diaphragm and intercostal muscles compresses the lungs , causing us to exhale . Breathing is driven by the brainstem and it cannot be suppressed indefinitely: holding your breath eventually triggers a reflex that forces breathing to resume . The region of the brainstem that controls breathing is called the preBötzinger Complex . However , there is increasing evidence that a second region in the brainstem is also involved . This region , which is called the retrotrapezoid nucleus/parafacial respiratory group , consists of three types of excitatory neurons—Dbx1 neurons , Phox2b neurons , and Atoh1 neurons—but their roles had not been clear . Now , using multiple lines of genetically modified mice , Tupal et al . have teased apart the roles of these three cell types . These experiments showed that the Dbx1 neurons—which are also found in the preBötzinger Complex—have an essential role in sending the signals from the brain that drive the different muscle activities needed to breathe . The Phox2b neurons modulate breathing based on the level of carbon dioxide in the blood . Atoh1 neurons help control the sequence of respiratory muscle activity during a breath , probably by selectively inhibiting different populations of Dbx1 neurons . The work of Tupal et al . indicates that distinct populations of neurons within the brainstem independently control two different aspects of breathing: the generation of breathing rhythms , and the coordination of these rhythms . Given that many other physiological processes involve rhythmic activity patterns , this model may help us to understand how the brain generates and controls complex behaviors more generally . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | Atoh1-dependent rhombic lip neurons are required for temporal delay between independent respiratory oscillators in embryonic mice |
For coordinated circulation , vertebrate and invertebrate hearts require stereotyped arrangements of diverse cell populations . This study explores the process of cardiac cell diversification in the Drosophila heart , focusing on the two major cardioblast subpopulations: generic working myocardial cells and inflow valve-forming ostial cardioblasts . By screening a large collection of randomly induced mutants , we identified several genes involved in cardiac patterning . Further analysis revealed an unexpected , specific requirement of EGF signaling for the specification of generic cardioblasts and a subset of pericardial cells . We demonstrate that the Tbx20 ortholog Midline acts as a direct target of the EGFR effector Pointed to repress ostial fates . Furthermore , we identified Edl/Mae , an antagonist of the ETS factor Pointed , as a novel cardiac regulator crucial for ostial cardioblast specification . Combining these findings , we propose a regulatory model in which the balance between activation of Pointed and its inhibition by Edl controls cardioblast subtype-specific gene expression .
The heart consists of a variety of cells with distinct molecular and physiological properties in both vertebrates and invertebrates . A complex regulatory network of transcription factors and signaling pathways orchestrates the specification of these different cell populations and their proper arrangement within a regionalized working myocardium or other functional structures such as valves , inflow and outflow tracts ( reviewed in Greulich et al . , 2011; Miquerol and Kelly , 2013; Rana et al . , 2013; for the invertebrate Drosophila heart see for example Bodmer and Frasch , 2010; Lehmacher et al . , 2012; Lovato and Cripps , 2016; Reim and Frasch , 2010 ) . For example , the vertebrate T-box gene Tbx20 promotes working myocardial fate by restricting Tbx2 expression and enabling the expression of chamber myocardium-specific genes ( Cai et al . , 2005; Singh et al . , 2005; Stennard et al . , 2005 ) . By contrast , Tbx2 and Tbx3 repress working myocardium-specific gene expression and chamber differentiation in the non-chamber myocardium and thus contribute to the formation of endocardial cushions and structures of the conduction system ( Christoffels et al . , 2004; Hoogaars et al . , 2007; Singh et al . , 2012 ) . Normal endocardial cushion formation also requires COUP-TFII , an orphan nuclear receptor transcription factor that regulates cell fate decisions in several tissues ( Lin et al . , 2012; Wu et al . , 2016 ) . In the embryonic mouse myocardium , COUP-TFII is restricted to atrial cardiomyocytes , a pattern consistent with a fate determination function that confers atrial over ventricular fate ( Lin et al . , 2012; Wu et al . , 2013 ) . This function appears to involve the up-regulation of Tbx5 ( Wu et al . , 2013 ) , another T-box gene with non-uniform cardiac expression and a fundamental role in heart development and human cardiac disease ( Basson et al . , 1997; Bruneau et al . , 1999; Bruneau et al . , 2001; Ghosh et al . , 2017; Steimle and Moskowitz , 2017 ) . Furthermore , FGF-mediated receptor tyrosine kinase ( RTK ) signaling upstream of the cardiogenic transcription factor Nkx2-5 was recently shown to be required for the maintenance of ventricular chamber identity of cardiomyocytes in zebrafish ( Pradhan et al . , 2017 ) . As emphasized below , spatial restriction of cardiac transcription factors as well as precisely controlled RTK signaling activities are not only important in vertebrate but also invertebrate hearts ( Gajewski et al . , 2000; Lo and Frasch , 2001; Zaffran et al . , 2006; this work ) . The Drosophila heart ( dorsal vessel ) comprises several types of cardiomyocytes ( in the embryo called cardioblasts , CBs ) and non-contractile pericardial cells ( PCs ) ( Bodmer and Frasch , 2010; Lovato and Cripps , 2016 ) . The progenitors of these cells are specified in segmentally repeated heart fields located at the intersection of BMP/Dpp and Wg/Wnt signaling activities ( Frasch , 1995; Reim and Frasch , 2005; Wu et al . , 1995 ) . Subsequent specification of the definitive cardiogenic mesoderm depends on a conserved group of transcription factors , most importantly those encoded by the Nkx2-5 ortholog tinman ( tin ) , the Gata4 ortholog pannier ( pnr ) and the Dorsocross1-3 T-box genes ( three Tbx6-related paralogs that also share features with Tbx2/3/5; in the following collectively called Doc ) ( Alvarez et al . , 2003; Azpiazu and Frasch , 1993; Bodmer , 1993; Gajewski et al . , 1999; Junion et al . , 2012; Reim and Frasch , 2005; Reim et al . , 2003; reviewed in Reim and Frasch , 2010; Reim et al . , 2017 ) . While the identification of cardiogenic factors has greatly improved our understanding of early specification events , much less is known about the mechanisms that lead to the diversification of cardiac cell subpopulations . In this study , we mainly focus on the development of the two major cardioblast subpopulations: generic cardioblasts ( gCBs ) , which build the main portion of the contractile tube ( ‘working myocardium’ ) , and ostial cardioblasts ( oCBs ) , which form bi-cellular valves ( ostia ) for hemolymph inflow . Due to Hox gene inputs , ostial progenitor specification is limited to the abdominal region ( Lo et al . , 2002; Lovato et al . , 2002; Ponzielli et al . , 2002; Ryan et al . , 2005; reviewed in Monier et al . , 2007 ) . Current research suggests that each abdominal hemisegment generates at least seven distinct progenitors that give rise to six CBs ( 4 gCBs + 2 oCBs ) and several types of PCs ( Tin+/Even-skipped[Eve]+ EPCs , Tin+ TPCs , and Odd-skipped[Odd]+ OPCs; Bodmer and Frasch , 2010 and references therein ) . Whereas gCBs ( a . k . a . Tin-CBs ) maintain expression of tin , oCBs ( a . k . a . Svp-CBs ) specifically express the COUP-TFII ortholog seven-up ( svp ) and Doc ( Gajewski et al . , 2000; Lo and Frasch , 2001; Ward and Skeath , 2000; Zaffran et al . , 2006 ) . Previous work has shown that Doc is repressed in gCBs in a tin-dependent manner ( Zaffran et al . , 2006 ) . Robust tin expression in turn depends on the Tbx20 ortholog midline ( mid/nmr2 ) . The mid gene is first activated in gCB progenitors , but later , like its paralog H15/nmr1 , becomes expressed in all cardioblasts ( Miskolczi-McCallum et al . , 2005; Qian et al . , 2005; Reim et al . , 2005 ) . In oCBs , svp represses tin expression thereby permitting continued Doc expression in these cells ( Gajewski et al . , 2000; Lo and Frasch , 2001; Zaffran et al . , 2006 ) . In the abdomen , gCBs and most PCs are preceded by a precursor that undergoes symmetric division , whereas oCBs and half of the OPCs are derived from common , asymmetrically dividing CB/PC progenitors ( Alvarez et al . , 2003; Han and Bodmer , 2003; Ward and Skeath , 2000 ) . The process of progenitor specification in the somatic and cardiogenic mesoderm involves the antagonistic actions of RTK/Ras/MAPK and Delta/Notch signaling ( Carmena et al . , 2002; Grigorian et al . , 2011; Hartenstein et al . , 1992 ) . Two types of RTKs , the fibroblast growth factor ( FGF ) receptor Heartless ( Htl ) and the epidermal growth factor ( EGF ) receptor EGFR , act positively on progenitor selection via MAPK signaling , although they are used by different progenitors to different extents ( Buff et al . , 1998; Carmena et al . , 2002; Michelson et al . , 1998 ) . Htl and its FGF8-like ligands Pyramus ( Pyr ) and Thisbe ( Ths ) have a dual function as regulators of mesodermal cell migration and cell specification , with progenitors of the Eve+ lineage as the most prominent example for the latter ( reviewed in Bae et al . , 2012; Muha and Müller , 2013 ) . EGFR signaling appears to be dispensable for early mesoderm migration events ( Wilson et al . , 2005 ) but has been reported to contribute to the specification of particular cell types within the mesoderm , including subsets of adult muscle precursors ( AMPs; Figeac et al . , 2010 ) and the Eve+ DA1 muscles ( derived from the so-called P15 progenitors in the dorsal mesoderm; Buff et al . , 1998; Carmena et al . , 1998 ) . By contrast , Eve+pericardial cells derived from the P2 progenitor were shown to form independent of EGFR activity . The exact contribution of EGFR signaling to Drosophila heart development has been less clear until now , but it was shown that EGFR loss-of-function results in a severe reduction of the numbers of cardioblasts , pericardial nephrocytes , and blood progenitors ( Grigorian et al . , 2011 ) . Molecularly , the predominant EGFR ligand in the embryo , Spitz ( Spi ) , relies on the protease Rhomboid ( encoded by rho ) and the chaperon Star ( S ) for its conversion from a membrane-bound into its active form ( reviewed in Shilo , 2014 ) . In contrast to spi , rho expression is restricted to a limited number of cells in a complex and dynamic pattern , including cells of the cardiogenic area ( Bidet et al . , 2003; Liu et al . , 2006 ) , which points to rho expression being the most decisive factor for Spi-mediated EGFR activation . Among the most important downstream effectors of RTK/Ras/MAPK pathways are the ETS transcription factors PntP2 ( encoded by pointed/pnt ) and Yan/Aop ( encoded by anterior open/aop ) . While PntP2 becomes an active transcriptional activator upon phosphorylation by MAPK , the transcriptional repressor Yan is negatively regulated by MAPK ( Gabay et al . , 1996; O'Neill et al . , 1994 ) . Unlike PntP2 , a shorter isoform encoded by pnt , PntP1 , is constitutively active but was shown to require activated MAPK for its transcriptional activation at least in some cell types ( Brunner et al . , 1994; Gabay et al . , 1996; Klämbt , 1993; O'Neill et al . , 1994 ) . Notably , chordate Pnt orthologs ( ETS1/2 ) were shown to contribute to cardiac progenitor formation in the tunicate Ciona and during transdifferentiation of human dermal fibroblasts into cardiac progenitors ( Davidson et al . , 2006; Islas et al . , 2012 ) . During early Drosophila cardiogenesis , Pnt favors expression of eve over that of another homeobox gene , ladybird ( lbe , expressed in mesodermal cells immediately anterior of the Eve+ cluster and later in TPCs and two of the four gCBs per hemisegment; Jagla et al . , 1997 ) ( Liu et al . , 2006 ) . In addition , Pnt promotes pericardial cell development and antagonizes CB fate , especially that of oCBs ( Alvarez et al . , 2003 ) . Despite the progress in the understanding of cardiac progenitor specification , the mechanisms that diversify progenitors of the oCB and gCB lineages have remained elusive . We have performed an unbiased large-scale mutagenesis screen to identify genes that regulate cardiac development in Drosophila embryos and found several mutants that show CB subtype-specific defects . On this basis , we discovered a novel and rather unexpected function of the EGF pathway in specifying the gCBs of the working myocardium , thus revealing an intimate link between cardioblast specification and diversification . Furthermore , we identified ETS domain lacking ( Edl a . k . a . Modulator of the activity of ETS , Mae ) as a crucial regulator of the specification of inflow valve-forming oCBs . Edl possesses a SAM domain , which mediates binding to the SAM domain-containing ETS factors PntP2 and Yan , thereby inhibiting their activity as a transcriptional activator or repressor , respectively ( Baker et al . , 2001; Qiao et al . , 2006; Qiao et al . , 2004; Tootle et al . , 2003; Vivekanand et al . , 2004; Yamada et al . , 2003 ) . Our data imply that Edl enables svp expression and thus oCB fate by limiting the activity of PntP2 , thereby blocking subsequent activation of important downstream targets such as pntP1 and mid . Collectively , our data provide the basis for an elaborated model of cardiac cell fate diversification that links MAPK signaling , Pnt activity and the cell-type-specific expression patterns of key cardiac transcription factors .
In order to identify genes involved in heart and muscle development in an unbiased manner , we have performed an EMS mutagenesis screen for chromosome two in Drosophila melanogaster embryos ( Hollfelder et al . , 2014 ) . Several of the isolated mutants display a partial loss or irregular alignment of cardioblasts ( CBs ) . Such defects may potentially result from mutations in genes that regulate the specification or differentiation of all CBs or only a particular CB subtype . In the latter case , disturbances in the characteristic ‘2 + 4’ CB pattern of two ostial cardioblast ( oCBs; Doc+/Tin- ) and four generic CBs ( gCBs; Doc-/Tin+ ) per hemisegment are to be expected . To analyze the cardiac pattern of mutants in more detail , we performed immunofluorescent double stainings for Doc and H15 ( or alternatively Mef2 ) to label oCBs and all CBs , respectively . We then genetically and in part also molecularly mapped the mutations responsible for CB pattern anomalies ( for details see the Materials and methods section and Supplementary file 1-Table S1 ) . The class of mutants characterized by a loss of CBs contained several novel alleles of genes involved in RTK/Ras/MAPK signaling , which is consistent with the assumed role of this pathway in cardiac progenitor selection or maintenance ( Carmena et al . , 2002; Grigorian et al . , 2011 ) . However , no specific role for the specification of a particular cardioblast subtype or diversification of gCB versus oCB progenitors had been previously attributed to RTK/Ras/MAPK signaling . Our phenotypic analysis now shows that diminished EGF/EGFR but not FGF/Htl signaling leads to a preferential reduction of gCB numbers . Embryos with partially reduced FGF/Htl signaling , that is mutants lacking both copies of the FGF-encoding gene pyr and one copy of its paralog ths , as well as hypomorphic htl mutants , show an about equal reduction of gCB and oCB numbers ( Figure 1B , for quantification see Figure 1M; additional examples in Figure 1—figure supplement 1B , C ) . This CB reduction can be explained by uneven spreading of the early mesoderm to Dpp-receiving areas . By contrast , several mutations mapped to EGF signaling components feature a preferential loss of gCBs . In strong Egfr mutants very few CBs can be found ( Figure 1C , Figure 1—figure supplement 1E ) . Remarkably , the overwhelming majority of the residual CBs express Doc . The few remaining Doc-negative CBs are usually located toward the anterior and thus are possibly remnants of the oCB-free anterior aorta . In spitz , rhomboid and Star loss-of-function mutants , the number of Doc-/Tin+ CBs is strongly reduced while that of ostial Doc+/Tin- CBs is nearly normal or in some cases even increased by a few cells ( Figure 1D–G , M , Figure 1—figure supplement 1F , Figure 1—figure supplement 2A–C ) . In the wild type , the two pairs of sibling gCBs within each hemisegment can be further categorized as Lbe+ ( anterior pair ) or Lbe- ( posterior pair ) subtypes . Since the above-mentioned spitz group mutants often feature a single pair of gCBs in each abdominal hemisegment , we tested whether these cells are preferentially Lbe+ or Lbe- , which would indicate that one of the two gCB progenitor types may be more sensitive to impaired EGF signaling . However , our finding that both types are about equally represented in rho mutants ( Figure 1—figure supplement 3 ) argues against this assumption . Moreover , segment-by-segment analysis in homozygous rhoL68 mutants reveals that residual gCBs most frequently occur either as Lbe+ or Lbe- pairs , whereas none of the analyzed residual gCB duplets consisted of a combination of both gCB types . This suggests that EGF function is required for the formation of gCB progenitors prior to their final division . Notably , progenitors of the oCB lineage apparently do not require activity of the ostial marker gene svp to develop and survive independently of EGF , since total CB numbers are similar in Star single and Star svp double mutants ( compare Figure 1H–1G; quantification in Figure 1M ) . Previous studies in EGF pathway mutants suggested that incorrectly specified mesodermal progenitors undergo apoptosis ( Buff et al . , 1998; Grigorian et al . , 2011 ) . Using TUNEL and anti-activated caspase stainings , we could not reliably detect signs of apoptosis in the Tin- or Doc-labeled cardiogenic mesoderm of Star mutants , while numerous signals were observed in other tissues ( Figure 1—figure supplement 4 and data not shown ) . Nevertheless , we obtained indirect evidence for the occurrence of at least some apoptosis by using the baculoviral apoptosis inhibitor p35 ( Zhou et al . , 1997 ) . If p35 is artificially expressed in the mesoderm of S mutants the number of CBs slightly increases in comparison to S mutants without p35 ( Figure 1I , M ) . Although this is consistent with a pro-survival function of EGF signaling , it does not fully account for the gCBs missing in S mutants . Of note , we detect a small , but statistically significant increase in the average number of Doc+ CBs in comparison to the wild type in spi mutants , in p35-expressing S mutants as well as in embryos overexpressing dominant-negative EGFR ( Figure 1M ) , which suggests that at least some presumptive gCB progenitors adopt oCB-like fates at reduced EGFR activity . However , the observed effects are small and additional explanations such as persistence in an uncommitted dorsal mesoderm cell pool must be considered to fully explain the fate of all lost gCB progenitors ( see discussion ) . Collectively , the cardiac patterning phenotypes imply that EGF signaling plays a major role in the correct specification of gCB progenitors , although we cannot exclude an additional function in cardiac cell survival that might be difficult to detect by the applied methods . Because EGF signaling is involved in multiple processes during embryogenesis we next asked whether its impact on gCB specification is directly linked to signaling activity within mesoderm cells . Indeed , mesoderm-specific attenuation of the pathway by expression of a dominant-negative EGFR variant resulted in essentially the same phenotype as with the spitz group mutants ( Figure 1J , M ) . Activation of the EGF pathway in mesoderm cells appears to be largely controlled by the spatially restricted expression of rho ( Bidet et al . , 2003; Grigorian et al . , 2011; Halfon et al . , 2000 ) . Overexpression of rho with the pan-mesodermal how24B-GAL4 driver has been previously reported to affect the number of tin-expressing pericardial cells ( Bidet et al . , 2003 ) , but CBs and their subtypes were not unambiguously labeled in these experiments . We extended these experiments using also other drivers . Consistent with a mesoderm-autonomous function , overexpression of rho in the dorsal ectoderm ( via pnrMD237-GAL4 ) has no significant effect on CB number or pattern ( Figure 1M and data not shown ) . By contrast , all mesodermal rho overexpression setups increase the gCBs:oCBs ratio in comparison to the wild type ( Figure 1K–M and data not shown ) . The impact on the absolute CB numbers depends on the timing and strength of transgene expression . The later rho is activated in mesodermal cells ( with following drivers according to their temporal order and progressive spatial restriction: twist-GAL4 , how24B-GAL4 and tinD +tinCΔ4-GAL4 ) the larger the total number of CBs ( Figure 1K–M and data not shown ) . This implies that rho activity needs to be tightly regulated , spatially as well as temporally . In the wild-type mesoderm , rho expression is first seen in the Eve+ progenitor P2 ( Buff et al . , 1998; Carmena et al . , 1995; Halfon et al . , 2000 ) followed by expression in the adjacent CB progenitor-containing clusters C14 and C16 ( Bidet et al . , 2003; Grigorian et al . , 2011; see also Figure 2A–D ) . Of note , stage 11 rho expression is still robustly observed in all C14/C16 clusters in S mutants ( Figure 2E cf . 2A ) , showing that earlier patterning events are not disrupted in this situation . Later during stage 12 , when rho RNA is normally found in developing CBs along the dorsal mesoderm margin , a reduction of rho expressing cells is apparent in S mutants ( Figure 2F cf . 2C ) , which is consistent with defects in CB progenitor formation . Importantly , detection of active diphospho-MAPK is severely reduced in cardiac cells of S mutants already in the cardiogenic clusters at stage 11 as well as during 12 in which dpMAPK is normally detected in both ostial and generic CB progenitors ( Figure 2H , J cf . 2G , I; later activity in cardiac cells appears to be less affected; Figure 2L cf . 2K ) . Similar observations were made for embryos with pan-mesodermal overexpression of the dominant-negative EGFR ( data not shown ) . Altogether , this demonstrates that EGF signaling serves as the major positive input for MAPK activation during early gCB progenitor formation , whereas input from FGFs may gain importance in developing CBs at later stages for CB fate maintenance as was proposed previously ( Grigorian et al . , 2011 ) . Since half of the odd-expressing pericardial cells ( OPCs ) are siblings of oCBs , we also analyzed PCs in EGF-related mutants by Odd/Eve as well as Odd/Zfh1 double-stainings ( Figure 3A–C , E; Figure 3—figure supplement 1A–D and data not shown ) . Consistent with the results of previous studies on Eve+ progenitor derivatives ( Buff et al . , 1998; Carmena et al . , 2002; Su et al . , 1999 ) , we detected EPCs in almost normal numbers in spi group mutants and in embryos with pan-mesodermal dominant-negative EGFR , whereas spi-dependent Eve+ DA1 muscles were largely absent ( Figure 3B , C , E ) . OPCs are strongly reduced in these loss-of-function backgrounds . Our quantification revealed that about half of the OPCs were lost in rho7M43/L68 and other EGF pathway mutants ( Figure 3B , C , E ) . A converse phenotype with many extra OPCs as well as Tin+ PCs ( TPCs , excluding the unaffected EPCs ) is generated by rho overexpression with tinD +tinCΔ4-GAL4 ( Figure 3D , E; Figure 3—figure supplement 1F ) . Notably , the number of oCB-sibling OPCs ( as identified by svp-lacZ reporter analysis ) is not significantly reduced in Star mutants if compared to the wild type ( Figure 3F , G ) , thus implying that the EGF signaling-dependent OPCs are those derived from symmetrically dividing OPC progenitors . In sum , these data demonstrate that EGF pathway activity is required in the mesoderm specifically for the specification of the symmetrically dividing gCB and OPCs progenitors ( and probably also for those of the TPCs , which we did not quantify in detail ) but is largely dispensable or even detrimental for the specification of the svp-expressing oCB/OPC progenitors . Our EMS screen also yielded mutants in which the number of ostial cardioblasts was specifically reduced . One such complementation group consisting of three alleles was mapped to the numb gene ( alleles listed in Supplementary file 1-Table S1 ) , which is consistent with its well-known function as a Notch suppressor during asymmetric cell division in the oCB lineage ( Gajewski et al . , 2000; Ward and Skeath , 2000 ) . Preferential reduction of oCBs was also observed in the mutant line S0520 . We found that its cardiac phenotype was caused by loss of the gene ETS domain lacking ( edl ) as part of a multi-gene deletion and named this mutant Df ( 2R ) edl-S0520 ( Figure 4A , Supplementary file 2-Table S2 ) . We identified edl as the gene responsible for the oCB losses by obtaining phenocopies with other edl mutants ( Figure 4A–D and data not shown ) . The lacZ enhancer trap insertion allele edlk06602 was used in most edl loss-of-function experiments since its cardiac phenotype is indistinguishable from that of Df ( 2R ) edl-S0520 and Df ( 2R ) edl-L19 ( Figure 4C , D and data not shown ) , and we detected in this strain a small deletion that specifically destroys the edl gene ( Figure 4A , Supplementary file 2-Table S2 ) . Furthermore , we were able to rescue the cardiac phenotype of edl by introducing a genomic edl transgene ( Yamada et al . , 2003; Figure 4E ) . Phenotypic rescue was also achieved , albeit with lesser efficiency , by artificially expressing edl in the dorsal mesoderm cells or in cardioblasts using the drivers tinD-GAL4 and tinCΔ4-GAL4 , respectively ( Figure 4F , G ) , demonstrating that Edl is required directly within these cell types . In accordance , edl mRNA is found within the cardiogenic region during stages 10 to 12 ( Figure 4—figure supplement 1A–C; Figure 4—figure supplement 2A–D ) , including prominent expression in early svp-expressing oCB progenitors ( Figure 4—figure supplement 2E ) . Thereafter edl expression shifts to the pericardial region , where it persists until stage 15 ( Figure 4—figure supplement 1D and data not shown ) . A distinctive feature of edl mutants is that the normal ‘2 + 4’ pattern of 2 Doc+ CBs + 4 Doc- CBs is often transformed into a ‘1 + 5’ pattern ( e . g . bracket in Figure 4D ) , indicating a fate switch from ostial to generic CBs . However , Edl is not a direct activator of Doc expression because Doc is found in CBs of edl double mutants with CB-specific ablation of tin ( Figure 4I ) , a phenotype reminiscent of that of CB-specific tin single mutants ( Figure 4H; Zaffran et al . , 2006 ) . This suggests that edl normally contributes to the activation of Doc in oCBs via suppression of tin . This role of edl in CB patterning is further supported by the observation of some CBs with low levels of both Tin and Doc in edl mutants ( Figure 4K; compare to the strictly complementary distribution of Doc and Tin in the wild type , Figure 4J ) . Next , we analyzed Edl function by ectopic expression . Consistent with a mesoderm-autonomous function , overexpressing edl in the dorsal ectoderm via pnrMD237-GAL4 has no significant effect on cardiogenesis ( data not shown ) . By contrast , overexpression of edl in the entire mesoderm via twist-GAL4 results in an increase of CB numbers ( Figure 5A ) and a decrease of OPCs ( described in the next subsection ) . The increase in Doc+ CBs is disproportionately high . The extra Doc+ CBs in the heart proper also activate ostial cell differentiation markers such as wg ( data not shown ) . In agreement with the proposed function of Edl as a negative regulator of PntP2 ( Yamada et al . , 2003 ) , our overexpression phenotypes of edl are very reminiscent to that of pntP2-specific mutants ( pntRR112 reported in Alvarez et al . ( 2003 ) ; and pntMI03880 shown in Figure 5B ) and amorphic pnt mutants ( pntΔ88 , pnt2; see Figure 5E , I and Alvarez et al . , 2003 ) . Accordingly , overexpression of constitutively active PntP2VP16 ( Figure 5C ) or PntP1 ( not shown ) via tinD +tinCΔ4-GAL4 causes a phenotype similar to that of edl loss-of-function mutants ( Figure 4C , D ) . By contrast , analogous overexpression of the potential Edl target Yan/Aop leads to a loss of heart cells irrespective of their subtype ( Figure 5D ) . These losses may result from a more general block in cell specification and differentiation since Yan has been related to such functions in several other types of MAPK-dependent progenitors ( Bidet et al . , 2003; Caviglia and Luschnig , 2013; Halfon et al . , 2000; Rebay and Rubin , 1995 ) . If the predominant function of Edl during CB specification is the inhibition of Pnt , edl pnt double mutants should mimic pnt mutants . In principle , this is what we observed ( Figure 5E , F; quantifications in Figure 5I ) . By contrast , edl aop double mutants show an additive combination of aop and edl single mutant phenotypes ( compare Figure 5H with 5G and 4D; see also quantifications in Figure 5I ) . Amorphic aop mutants display a reduction in CB number irrespective of CB subtype , which we ascribe to a permissive function during CB development that is probably linked to its well-documented role in restricting eve expression in the early dorsal mesoderm ( Bidet et al . , 2003; Halfon et al . , 2000; Liu et al . , 2006; Webber et al . , 2013 ) . Importantly , and in contrast to edl and pnt activity changes , manipulating aop activities does not lead to significant shifts in the oCBs:gCBs ratio ( Figure 5I ) . Thus , we suggest that Edl acts mainly via negative modulation of PntP2 activity during cardioblast diversification . An additional function of Pnt ( and thereby Edl ) regarding to the total number of CBs is also apparent in Figure 5 . The increase in the total CB number detected in pnt mutants is reminiscent of Notch pathway mutants . Figure 5—figure supplement 1 shows examples of such mutants isolated from our EMS screen . There is an important difference between pnt and Notch pathway mutants regarding the oCBs:gCBs ratio . Whereas oCBs account for about 40–50% of the CBs in pnt mutants ( as compared to 27% in the wild type ) , all Notch pathway mutants for which CB patterning data are available feature a significantly smaller fraction of oCBs than pnt mutants ( Figure 5—figure supplement 1D ) . The maximum fraction of oCBs observed was 33% of the total CB number , found in mamS0669 . In kuz mutants ( data not shown; Albrecht et al . , 2006 ) , oCBs even increase by smaller factors than gCBs resulting in oCB fractions below 27% . ( Some differences in the oCBs:gCBs ratio between various Notch pathway mutants are likely to arise from variable impact on lateral inhibition and specific functions of Notch in asymmetrically dividing lineages ) . On a side note , edl expression , which was found to be positively regulated by Notch signaling in a Drosophila cell culture system ( Krejcí and Bray , 2007 ) , is not negatively affected in the cardiogenic mesoderm of two mam alleles and in bibS1538 mutants ( Figure 5—figure supplement 2 and data not shown ) . The population of oCBs is characterized by expression of svp . In svp mutants all oCBs are converted into Tin+/Doc- CBs due to de-repression of tin ( Gajewski et al . , 2000; Lo and Frasch , 2001; Zaffran et al . , 2006; Figure 6—figure supplement 1A ) . Therefore , we tested the possibility that Edl promotes oCB fate by regulating svp . In the wild type , expression of svp is recapitulated by the enhancer trap svpAE127-lacZ ( Figure 6A; Lo and Frasch , 2001 ) . In edl mutants , svp-LacZ expression is strongly reduced in cardiac cells ( Figure 6B , D ) . The reduction in numbers of both svp-LacZ+ oCBs and OPCs at late stages ( Figure 6D cf . 6C ) suggests that edl already affects the fates of their common progenitors . Consistent with a function in promoting svp expression and oCBs fates , mesodermal overexpression of edl leads to larger numbers of svp-LacZ+ cardiac cells , particularly of CBs , where svp expression correlates with expanded Doc expression ( Figure 6E , F ) . As shown for Doc expression , svp expression can be suppressed by PntP2 hyperactivity ( green asterisks in Figure 6H ) . These observations and further evaluation of the epistatic relations between svp and edl ( Figure 6—figure supplement 1 ) demonstrate that edl affects CB patterning by blocking Pnt activity upstream of svp . Proposing a gCB-specific function of Pnt , we next analyzed its cardiac expression . Boisclair Lachance et al . previously reported that the expression of a fully functional genomic pnt-GFP transgene mirrors the combined expression of all Pnt isoforms ( Boisclair Lachance et al . , 2014 ) . The authors detected Pnt-GFP fusion protein in nearly all cells of the cardiac region , but highest levels were observed in two Yan-negative clusters per hemisegment flanking Eve+ cells . We confirmed and refined these observations showing that high levels of Pnt-GFP are present in the nuclei of gCB progenitors as identified by their position , characteristically enlarged size , presence of only low levels of Doc , and absence of svp-LacZ expression ( Figure 7A ) . We attribute these high total Pnt levels largely to a gCB-specific expression of the PntP1 isoform since PntP1-specific antibodies ( Alvarez et al . , 2003 ) specifically label gCB progenitors ( Figure 7B ) , whereas pntP2 transcripts are present in a rather uniform pattern in the mesoderm including the cardiogenic area ( Klämbt , 1993; and data not shown ) . We further speculated that PntP2 could activate pntP1 transcription in gCB progenitors for a sustained signaling response as found in other tissues ( Shwartz et al . , 2013 ) . This assumption is indeed supported by our genetic data . First , we detect PntP1 in an expanded pattern in the cardiogenic mesoderm of edl mutants in which PntP2 activity is assumed to increase ( Figure 7C ) . Second , overexpression of edl ( i . e . repression of PntP2 function ) as well as genetic disruption of pntP2 resulted in a near-complete loss of cardiac PntP1 ( Figure 7D , E; note persistent expression of PntP1 in other cells located more laterally ) . We conclude that the combined activities of Edl and PntP2 lead to the confined pntP1 expression in gCBs . The EGF Spitz appears to be a major , although not necessarily the sole factor for the MAPK-mediated activation of PntP2 in this context , because PntP1 levels are reduced but not eradicated in cardiac cells of amorphic spi mutants ( Figure 7F ) . According to the common view , we expect Pnt to act as a transcriptional activator also during CB diversification , particularly since overexpression of PntP2 fused to the VP16 activator domain has essentially the same effect on cardiac patterning as PntP1 overexpression ( Figure 6H and data not shown ) . Therefore , its negative impact on svp expression is likely to involve Pnt-dependent activation of a transcriptional repressor . Interestingly , the T-box factor Midline ( Mid ) , like PntP1 , shows expression in early gCB progenitors ( Figure 2I , K; Figure 7G ) . We previously reported that mid functions to maintain tin expression in gCBs , thereby restricting Doc expression to oCBs ( Reim et al . , 2005 ) . Consistent with this function our EMS screen also generated novel mid alleles showing the same CB patterning defects as previously described alleles ( Supplementary file 1-Table S1 , Figure 7H and data not shown ) . While a direct regulation of tin by Mid was previously proposed to be responsible for these changes ( supported by the gain- and loss-of-function phenotypes of mid; Qian et al . , 2005; Reim et al . , 2005 ) , another non-exclusive scenario could involve repression of svp ( encoding a repressor of tin ) by Mid . Consistent with the latter , we observe a Doc-like expansion of svp expression in mid loss-of-function mutants ( Figure 7I ) and a reduction of svp expression upon ectopic overexpression of mid via tinD +tinCΔ4-GAL4 ( Figure 7J ) . Moreover , persistent tin expression in all CBs of mid svp double mutants ( Figure 7—figure supplement 1D , compare to control in A and single mutants in B and C ) demonstrates that mid is not directly required for tin expression in CBs . Furthermore , the wild type-like expression of svp-lacZ ( with nearly no LacZ in gCBs ) observed in the same genetic background argues for the involvement of a Svp-dependent positive feedback loop in ectopic cardiac svp activation in gCBs , as has been predicted previously based on svp overexpression studies ( Zaffran et al . , 2006 ) . The cardiac pattern phenotype of edl mid double mutants is a composite of the single mutant phenotypes . The number of oCBs ( average oCBs: 24 . 4 ± 3 . 6; n = 6 ) is strongly increased as compared to edl mutants , but reduced in comparison with mid mutants , with total CB numbers being similar to those of edl mutants . In some cases , a near wild-type pattern is observed ( Figure 7K ) , although many embryos display an asymmetric arrangement of CBs . While the prevalence of many Doc-negative CBs in this background implies that mid is not the only factor that limits oCB fate , it also indicates that edl is normally required in the oCB lineage to restrict mid activity , possibly by blocking a Pnt-dependent activation of mid transcription . This hypothesis is indeed supported by the reversion of ectopic Doc and svp expression in pnt mutants upon forced mid expression ( Figure 7L , Figure 7—figure supplement 2C ) . By contrast , overexpression of the previously assumed Mid target tin in this background only represses Doc , but not svp ( Figure 7—figure supplement 2D ) . To further test the idea that Mid is a repressor of oCB fate downstream of pnt , we analyzed whether it is a direct target of Pnt . Notably , an enhancer identified as a Tin target and named midE19 ( mid180 for a shorter minimal version ) was recently shown to drive mid expression specifically in gCBs ( Jin et al . , 2013; Ryu et al . , 2011; Figure 8A–C; Figure 9A , C ) . Since this enhancer does not drive reporter expression in oCBs after germ band retraction as detected for mid in the genomic context , additional cis-regulatory regions must be at work to reproduce all aspects of cardiac mid expression . The characteristic activity pattern of the enhancer suggests that this regulatory region may be specifically ( or exclusively ) devoted to the reception of early gCB-specific inputs . Consistent with our assumption that this enhancer is also a target of Pnt , very little midE19-GFP activity is detectable in pnt mutants ( Figure 8D ) , reduced activity is observed in embryos with mesodermal edl overexpression ( Figure 8E ) , and expanded activity is seen upon overexpression of PntP1 ( Figure 8F; note occasional expansion into CBs with no detectable Tin ) or PntP2VP16 ( not shown ) . An observed reduction of midE19-driven GFP levels in many of the retained Tin+ gCBs of rho mutants ( Figure 8G ) corroborates that EGF signaling feeds into mid activation . The idea that mid is a target of Pnt is further supported by the almost complete elimination of reporter activity upon mutating a single ETS binding motif within the mid180 minimal cardiac enhancer ( Figure 9A–D ) as well as the strong reduction of endogenous mid transcription in emerging CBs during germ band retraction stages in pnt mutants ( Figure 9E–H ) . After germ band retraction , endogenous mid is activated independently of pnt in all CBs ( Figure 9J ) as observed in the wild type ( Figure 9I ) indicating that distinct mechanisms regulate mid transcription in early gCB progenitors and maturing CBs . In sum , our data lead to the conclusion that EGF signaling contributes to gCB specification by at least two distinct mechanisms , Pnt-independent specification of a subset of cardiac progenitors as well as Pnt-dependent inhibition of ostial cardioblast fate . Modulation by Edl is needed to inhibit Pnt-dependent gene activation and thus enable formation of ostial cardioblasts .
According to our data , EGF signals are the major source for MAPK activation and progenitor specification in the symmetrically dividing progenitors of gCBs and OPCs ( and likely also TCPs ) . By contrast , EGF signals are dispensable ( in high doses even unfavorable ) for the development of progenitors of oCBs and their sibling OPCs . Thus , EGF signaling clearly has a lineage-specific function , which is most easily explained by a requirement for progenitor selection and cell fate specification . This interpretation does not preclude contributions to cell survival ( which might depend on differentiation ) or lineage-specific divisions ( i . e . correct progenitor specification is a prerequisite of the subsequent final division ) . Notably , in most hemisegments of the analyzed EGF pathway mutants , the number of gCBs is reduced by even numbers and remaining gCB pairs are usually of the same subtype regarding Lbe expression , arguing for a requirement prior to completion of the final mitotic division at the progenitor stage . Since we have only minor evidence for apoptosis and fate conversions into other cell types in EGF-related mutants ( minor increase in oCBs , overall reduction of PCs ) we propose that many of the missing gCBs are not selected as highly Delta-expressing CB progenitors upon reduced MAPK signaling activity ( Carmena et al . , 2002; Grigorian et al . , 2011; Hartenstein et al . , 1992 ) . Instead , they are likely retained by default within a pool of undifferentiated dorsal mesoderm cells . Our overexpression studies demonstrate that the timing of EGF signals is crucial for their function in differential progenitor specification . In previous studies , earlier functions of MAPK signaling might have obscured its specific impact on gCBs and OPC subtypes . While early pan-mesodermal activation of MAPK signaling or expression of constitutive active Pnt forms via the twi-GAL4 driver reduces the numbers of all cardiac cells except the Eve+ progenitors ( Alvarez et al . , 2003; Bidet et al . , 2003; Liu et al . , 2006; and our own data ) , later MAPK activation favors formation of the symmetrically dividing OPC , TPC and gCB progenitor subpopulations ( e . g . as seen in our experiments with tinD-GAL4-driven rho ) . We propose that the specification of these progenitors requires the context of the definitive cardiogenic mesoderm , whereas premature MAPK activation in all mesoderm cells negates any pro-cardiogenic effects due to the massive expansion of Eve+ clusters ( which are normally the first cells in the heart field to display MAPK and rho activity ) at the expense of the cardiac progenitors in the neighboring C14/C16 clusters ( Buff et al . , 1998; Jagla et al . , 2002; Liu et al . , 2006; Qian et al . , 2005; and our own data not shown ) . As discussed above , cardioblast formation as such is independent of pnt . How could this be achieved ? Growth factor-activated MAPK can also phosphorylate the repressor Yan thereby diminishing its activity as an antagonist of progenitor selection ( Halfon et al . , 2000; O'Neill et al . , 1994; Rebay and Rubin , 1995 ) . Therefore , it is conceivable that MAPK activity in the context of CB progenitor selection might be primarily required to eliminate the repressive activity of Yan . This would be consistent with the observed reduction of cardiac cells upon aop/yan hyperactivation ( Halfon et al . , 2000; this study ) . In this context , a minor function of Edl could contribute to the robustness of cardiac progenitor selection and thus total cardioblast and pericardial cell numbers by reducing the repressive Yan activity . Combining previous findings with our new data we have conceived the regulatory model of cardioblast diversification illustrated in Figure 10 . The central element of this model is the differential modulation of Pnt activity in the gCB and oCB progenitors leading to lineage-specific outcomes . Our work clearly identifies Pnt and Edl as crucial transducers of spatio-temporal inputs during cardiac cell diversification , but open questions remain regarding the initial source for the differential activities . Our model proposes that factors which tilt the balance between PntP2 activity and Edl will have a major impact on CB subtype choice ( see Figure 10 ) . Thus , any input that modestly increases MAPK/PntP2 activity within the appropriate window of time would favor gCB fate , whereas factors that have the opposite effect should promote oCB specification . This points to activities that impinge on the highly complex and dynamic expression of rho and/or edl . The Rhomboid protease is a key determinant in the decision of which cells will activate the more broadly expressed EGF Spitz and thus emanate signaling activity . A prime candidate for an instructive cue to anterior-posterior positioning within each segment could be Hh ( indicated in the extended model in Figure 10—figure supplement 1 ) , because it was proposed to be an oCB-promoting and rho/MAPK pathway-modulating signal towards the cardiogenic mesoderm in previous studies ( Liu et al . , 2006; Ponzielli et al . , 2002 ) . In these studies , decreased svp expression and reduced numbers of Tin-negative CBs observed in hh mutants and upon overexpression of constitutive repressor forms of the Hh effector Ci were interpreted as signs of Hh-dependent oCB specification , although no converse effects have been reported using constitutive active Ci forms . However , the role of the Hh pathway in CB diversification is not fully understood , mainly due to complications arising from ectodermal Hh functions , primarily in maintaining pro-cardiogenic wg expression ( Bejsovec and Martinez Arias , 1991; Park et al . , 1996 ) . Furthermore , the effect of Hh on MAPK and rho activities in the dorsal mesoderm was suggested to be positive rather than negative based on an expansion of stage 12 mesodermal rho expression and expanded numbers of cells with activated MAPK upon pan-mesodermal overexpression of hh ( Liu et al . , 2006 ) . This would refute a function favoring oCB fate , but it is an interesting finding in light of our work , which couples rho activity with gCB specification . A modulation of rho expression via Hh signaling , whether direct or indirect , would also be consistent with the phenotype of mutants lacking the function of patched ( encoding a negative regulator of Hh signaling activity ) , in which we observe a strong increase in the gCBs:oCBs ratio ( although absolute CB numbers are highly variable between embryos and alleles; E . Heyland , F . Karama , B . Schwarz and I . Reim , unpublished observations ) . On the other hand , mutants with diminished Hh pathway activity , including some that were recovered by our EMS screen because of their partial CB losses ( i . e . smoothened mutants ) , do not display a biased reduction of either oCBs or gCBs ( E . Heyland , F . Karama , B . Schwarz and I . Reim; unpublished observations ) . Hence , the regulation of rho and the role of hh during CB diversification await more detailed analysis . Factors that regulate edl expression levels might also determine the outcome of the competition between Edl and Pnt . The edl gene was found to be positively regulated by EGF signaling , and to be a target of Pnt and Yan , and thus was proposed to provide a negative feedback system for EGF inputs ( Baker et al . , 2001; Leatherbarrow and Halfon , 2009; Vivekanand et al . , 2004; Yamada et al . , 2003 ) . Our extended model therefore includes regulation by Pnt as a possibility ( dashed arrows in Figure 10—figure supplement 1 ) . Nevertheless , additional or alternative inputs need to be considered to explain the strong edl expression in presumptive oCB progenitors with low Pnt activity . Notably , ChIP-on-chip experiments suggest that edl is also targeted by cardiogenic factors ( Junion et al . , 2012 ) . Furthermore , edl was identified as a positively regulated target of Notch signaling in a Drosophila cell culture system ( Krejcí et al . , 2009 ) . However , observed persistent edl expression in Notch pathway mutants argues against positive inputs from Notch during edl regulation in oCB progenitors . The spatio-temporal dynamics and detailed mechanisms that regulate MAPK and edl activities within the cardiogenic mesoderm remain to be investigated in future studies . Such studies may also help to understand lineage decisions in other tissues and species . Edl/Mae-relatives are also present in non-Dipteran insects ( e . g . Tribolium; Bucher and Klingler , 2005 ) , echinoderms , and the chordate Ciona . Although no clear ortholog of Edl appears to be present in vertebrates , a SAM domain-only isoform of the human Yan-relative TEL2 as well as Drosophila Edl were shown to inhibit transcriptional stimulation by the mammalian Pnt orthologs ETS1/ETS2 in cell culture ( Gu et al . , 2001; Vivekanand and Rebay , 2012 ) . Hence , the restriction of ETS protein activities by protein-protein interactions offers an intriguing mechanism to fine-tune MAPK signaling output in developing tissues of both invertebrates and vertebrates .
The mutants bibS1538 , Df ( 2R ) edl-S0520 , EgfrS0167 , EgfrS2145 , EgfrS2307 , EgfrS2561 , kuzS3330 , kuzS3832 , mamS0669 , mamS4648 , midS0021 , midS2961 , numbS1342 , numbS3992 , numbS4439 , pyrS3547 ( Reim et al . , 2012 ) , spiS3384 , StarS4550 were recovered from our EMS screen . The lines mid1 , UAS-mid-B2 , how24B-GAL4 , pnrMD237-GAL4 , svpAE127-lacZ ( a svp mutant in homozygous condition ) , UAS-svp . I , 2xPE-twi-GAL4 , twi-SG24-GAL4 , tinD-GAL4 , UAS-tin#2 , {tin-ABD}T003-1B1; tinEC40 , UAS-p35 were as described previously ( Reim et al . , 2012; Reim et al . , 2005; Zaffran et al . , 2006 ) . In addition , the following strains were used: aop1 = aopIP ( Nüsslein-Volhard et al . , 1984; Rogge et al . , 1995 ) , UAS-aop . ACT-IIa ( Rebay and Rubin , 1995 ) , bib1 ( Lehmann et al . , 1983 ) , edlL19 = Df ( 2R ) edl-L19 ( edl and some neighboring genes deleted ) and UAS-edl-X ( both from Y . Hiromi; Yamada et al . , 2003 ) , P{lacW}edlk06602 ( Baker et al . , 2001; Török et al . , 1993 ) , Egfrf2 ( Clifford and Schüpbach , 1994 ) , UAS-EgfrDN . B-29-77-1;UAS-EgfrDN . B-29-8-1 ( Buff et al . , 1998 ) , htlYY262 ( Gisselbrecht et al . , 1996 ) , kuze29-4 ( Rooke et al . , 1996 ) , PBac{lbe-GFP . FPTB}VK00037 ( A . Victorsen and K . White ) , mam8 ( Lehmann et al . , 1983 ) , mid1 ( Buescher et al . , 2004 ) , midE19-GFP ( Jin et al . , 2013; from M . Frasch ) , pntΔ88 ( Scholz et al . , 1993 ) , pntMI03880 ( PntP2-specific; harbors a gene-trap cassette with an artificial splice acceptor followed by stop codons upstream of the pntP1 transcription start site; Venken et al . , 2011 ) , UAS-pntP2VP16-2 ( Halfon et al . , 2000; originally from C . Klämbt ) , UAS-pntP1-3 and UAS-pntP2-2 ( Klaes et al . , 1994 ) , PBac{pnt-GFP . FPTB}VK00037 ( R . Spokony and K . White; Boisclair Lachance et al . , 2014 ) , pyr18 and ths759 ( Klingseisen et al . , 2009 ) , rho7M43 ( Jürgens et al . , 1984 ) , rhoL68 ( Salzberg et al . , 1994 ) , rhoEP3704 ( Bidet et al . , 2003 ) , UAS-rho ( ve . dC ) ( de Celis et al . , 1997 ) , spi1 = spiIIAIIA14 ( Nüsslein-Volhard et al . , 1984 ) , StarB0453 ( Chen et al . , 2008; from F . Schnorrer ) , tinCΔ4-GAL4 ( Lo and Frasch , 2001; from M . Frasch ) , Df ( 2R ) Exel7157 , and about 180 additional deficiencies spanning chromosome 2 ( except where noted , all stocks available from the Bloomington Stock Center ) . Flies expressing edl+ from a transgene were generated anew by standard P-element transgenesis using the previously described edl[+t18] rescue construct ( named AF1 in Yamada et al . , 2003; provided by Y . Hiromi ) . Line P{edl . AF1}BS12 carrying an insertion on chromosome three was used in this study . Unless noted otherwise , y w or S-18a-13b-16c . 1 control ( Hollfelder et al . , 2014 ) flies were used as wild type controls . Mutant lines were maintained over GFP- or lacZ-containing balancer chromosomes to allow recognition of homozygous embryos . Flies were raised at 25°C , except for UAS/GAL4-driven overexpression at 29°C . Novel EMS-induced mutants were obtained from our screen for embryonic heart and muscle defects and mapped to a particular gene through extensive complementation testing analogous to the previously described procedure ( Hollfelder et al . , 2014 ) . Many alleles were mapped by unbiased complementation tests with a set of chromosome 2 deficiencies and subsequent non-complementation of lethality and embryonic phenotype by previously described alleles . Df ( 2R ) edl-S0520 was mapped by non-complementation of lethality with Df ( 2R ) Exel7157 , Df ( 2R ) edl-L19 and Df ( 2R ) ED3636 , but the cardiac phenotype was only reproduced in trans with Df ( 2R ) Exel7157 , Df ( 2R ) edl-L19 and edlk06602 . Novel alleles of Egfr and Star were mapped using a candidate gene approach . Several EMS alleles and the unmutagenized S-18a-13b-16c . 1 control were analyzed by sequencing of overlapping PCR products covering the coding sequence and splicing sites of the candidate gene as described ( Hollfelder et al . , 2014 ) . Details about the mutations are provided in Supplementary file 1-Table S1 . The area deleted by Df ( 2R ) edl-S0520 and its approximate break points were determined by iterative PCR amplification tests . The insertion of P{lacW}edlk06602 near the edl transcription start site was confirmed by PCR using primers binding to the 5' P end and adjacent genomic DNA . Although the integrity of the both P element ends could be confirmed by PCR , no genomic edl sequences expected next to the 3' P end could be amplified using several primer pairs shown to amplify control DNA . This indicates that P{lacW}edlk06602 is associated with a deletion in edl . Details of the deletion mapping are listed in Supplementary file 2-Table S2 . The mid180-GFP reporter constructs were generated according to a similar lacZ construct published by Ryu et al . ( 2011 ) . The forward primer 5'-EcoRI-CGTGCCTCCCACTTCAGGGCGG-3' and the backward primer 5'-BamHI-TTAATTTCATTTTTCACTCTGCTCACTTGAGATTCCCCTGCTTTGTCTGCGGCATTTCCGCTTCT-3' were used to amply DNA from y w flies . The predicted ETS binding site matching the antisense sequence of published ETS binding motifs ( Halfon et al . , 2000; Hollenhorst et al . , 2011; underlined ) was mutated in mid180-mETS-GFP by replacing the invariable TCC core ( bold ) with AAA in the backward primer . Amplicons were cloned into EcoRI/BamHI of pH-Stinger-attB ( Jin et al . , 2013 ) , sequenced and inserted into the attP2 landing site via nos-driven ΦC31 integrase . Embryo fixations , immunostainings for proteins and RNA in situ hybridizations were carried out essentially as described ( Knirr et al . , 1999; Reim and Frasch , 2005 ) , except for stainings with anti-dpMAPK , for which the formaldehyde concentration was doubled and embryos were rehydrated from methanol and stained immediately after fixation . VectaStain Elite ABC-HRP kit ( Vector Laboratories ) and tyramide signal amplification ( TSA , PerkinElmer Inc . ) were used for detection of RNA and certain antigens ( as indicated ) . The following antibodies were used: guinea pig anti-Doc2+3 ( 1:2000 , TSA ) and anti-Doc3+2 ( 1:1000 ) ( Reim et al . , 2003 ) , rabbit anti-H15/Nmr1 ( 1:2000 ) , guinea pig anti-H15/Nmr1 ( 1:2000 ) , rabbit anti-Mid/Nmr2 ( early stages: 1:250 , TSA; late stages: 1:1000 direct ) and rabbit anti-PntP1 ( 1:250 , TSA ) ( all from J . Skeath; Alvarez et al . , 2003; Leal et al . , 2009 ) , rabbit anti-Mef2 ( 1:1500 ) ( from H . T . Nguyen ) , rat anti-Odd ( 1:600 , TSA ) ( Kosman et al . , 1998 ) , rabbit anti-Eve ( 1:3000 ) ( Frasch et al . , 1987 ) , rabbit anti-Tin ( 1:750 ) ( Yin et al . , 1997 ) ( all from M . Frasch ) , rabbit anti-Zfh1 ( 1:2000 ) ( from R . Lehmann; Broihier et al . , 1998 ) , mouse anti-dpMAPK ( Sigma , 1:500 , TSA ) , rabbit anti-β-galactosidase ( Cappel , 1:1500 ) , rabbit anti-GFP ( Molecular Probes , 1:2000 and Rockland , 1:1000 ) , mouse anti-GFP 3E6 ( Life Technologies , 1:100 , TSA ) , anti-cleaved-Caspase-3 ( Asp175 , Cell Signaling Technology , 1:100 , TSA ) , sheep anti-Digoxigenin ( Roche , 1:1000 , TSA ) , monoclonal mouse antibodies anti-β-galactosidase 40-1a ( 1:20 direct or 1:50 with TSA ) , anti-Seven-up 5B11 ( 1:20 , TSA ) and anti-Wg 4D4 ( 1:30 , TSA ) ( all from Developmental Studies Hybridoma Bank , University of Iowa ) , fluorescent secondary antibodies ( 1:200 ) ( Jackson ImmunoResearch Laboratories and Abcam ) , biotinylated secondary antibodies ( 1:500 ) and HRP-conjugated anti-rabbit and anti-mouse IgG ( 1:1000 ) ( Vector Laboratories ) . TUNEL staining was performed as described ( Reim et al . , 2003 ) using the Millipore ApopTag S7100 kit in combination with TSA . Digoxigenin-labeled antisense riboprobes against mid , edl , rho and pntP2 were used for whole mount in situ hybridizations . The mid probe was generated as described previously ( Reim et al . , 2005 ) . T7 promoter-tagged edl , rho and pntP2 ( isoform-specific exons ) templates for in vitro transcription were generated by PCR ( primers edl: CAATCGTGAAAGAGCGAGGGTC , T7-TGACGAGCAGAACTAAGGACTAGGC , edlintron: GCACCGACGACTCAACTTCCTG , T7-GCTGCGATTGCGATTACAAACAAG , pnt: CCAGCAGCCACCTCAATTCGGTC , T7-GCGTGCGTCTCGTTGGGGTAATTG , rho: ATGGAGAACTTAACGCAGAATGTAAACG , T7-TTAGGACACTCCCAGGTCG ) from DNA of wild-type flies or flies carrying UAS-rho ( ve . dC ) or UAS-pntP2 , respectively . Embryos were mounted in Vectashield ( Vector Laboratories ) . Images were acquired on a Leica SP5 II confocal laser scanning microscope and projected using Leica LAS-AF and ImageJ . | Organs contain many different kinds of cells , each specialised to perform a particular role . The fruit fly heart , for example , has two types of muscle cells: generic heart muscle cells and ostial heart muscle cells . The generic cells contract to force blood around the body , whilst the ostial cells form openings that allow blood to enter the heart . Though both types of cells carry the same genetic information , each uses a different combination of active genes to perform their role . During development , the cells must decide whether to become generic or ostial . They obtain signals from other cells in and near the developing heart , and respond by turning genes on or off . The response uses proteins called transcription factors , which bind to regulatory portions of specific genes . The sequence of signals and transcription factors that control the fate of developing heart muscle cells was not known . So Schwarz et al . examined the process using a technique called a mutagenesis screen . This involved triggering random genetic mutations and looking for flies with defects in their heart muscle cells . Matching the defects to the mutations revealed genes responsible for heart development . Schwarz et al . found that for cells to develop into generic heart muscle cells , a signal called epidermal growth factor ( EGF ) switches on a transcription factor called Pointed in the cells . Pointed then turns on another transcription factor that switches off the genes for ostial cells . Conversely , ostial heart muscle cells develop when a protein called ‘ETS-domain lacking’ ( Edl ) interferes with Pointed , allowing the ostial genes to remain on . The balance between Pointed and Edl controls which type of heart cell each cell will become . Many cells in other tissues in fruit flies also produce the Pointed and Edl proteins and respond to EGF signals . This means that this system may help to decide the fate of cells in other organs . The EGF signaling system is also present in other animals , including humans . Future work could reveal whether the same molecular decision making happens in our own hearts . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2018 | Diversification of heart progenitor cells by EGF signaling and differential modulation of ETS protein activity |
As microtubule-organizing centers of animal cells , centrosomes guide the formation of the bipolar spindle that segregates chromosomes during mitosis . At mitosis onset , centrosomes maximize microtubule-organizing activity by rapidly expanding the pericentriolar material ( PCM ) . This process is in part driven by the large PCM protein pericentrin ( PCNT ) , as its level increases at the PCM and helps recruit additional PCM components . However , the mechanism underlying the timely centrosomal enrichment of PCNT remains unclear . Here , we show that PCNT is delivered co-translationally to centrosomes during early mitosis by cytoplasmic dynein , as evidenced by centrosomal enrichment of PCNT mRNA , its translation near centrosomes , and requirement of intact polysomes for PCNT mRNA localization . Additionally , the microtubule minus-end regulator , ASPM , is also targeted co-translationally to mitotic spindle poles . Together , these findings suggest that co-translational targeting of cytoplasmic proteins to specific subcellular destinations may be a generalized protein targeting mechanism .
A centrosome consists of a pair of centrioles embedded in a protein-dense matrix known as the pericentriolar material ( PCM ) . The PCM functions as a major microtubule organizing center in animal cells ( Gould and Borisy , 1977 ) as it serves as a platform onto which γ-tubulin ring complexes ( γ-TuRCs ) , the main scaffold mediating microtubule nucleation , are loaded ( Moritz et al . , 1995; Zheng et al . , 1995 ) . At the onset of mitosis , centrosomes rapidly expand their PCM . This process , termed centrosome maturation , is essential for proper spindle formation and chromosome segregation ( Woodruff et al . , 2014 ) . Centrosome maturation is initiated by phosphorylation of core PCM components , such as Pericentrin ( PCNT ) and Centrosomin ( Cnn ) , by mitotic kinases PLK1/Polo and Aurora kinase A ( Conduit et al . , 2014a; Joukov et al . , 2014; Kinoshita et al . , 2005; Lee and Rhee , 2011 ) . These events then trigger the cooperative assembly of additional PCM scaffold proteins ( e . g . PCNT , CEP192/SPD-2 , CEP152/Asterless , CEP215/CDK5RAP2/Cnn or SPD-5 ) into an expanded PCM matrix that encases the centrioles ( Conduit et al . , 2014b; Hamill et al . , 2002; Kemp et al . , 2004 ) , culminating in the recruitment of additional γ-TuRCs and tubulin molecules that promote microtubule nucleation and render centrosomes competent for mediating the formation of bipolar spindles and chromosome segregation ( Conduit et al . , 2015; Gopalakrishnan et al . , 2011; Woodruff et al . , 2014 ) . Pericentrin ( PCNT ) is one of the first core PCM components identified to be required for proper spindle organization ( Doxsey et al . , 1994 ) . Importantly , mutations in PCNT have been linked to several human disorders including primordial dwarfism ( Anitha et al . , 2009; Delaval and Doxsey , 2010; Griffith et al . , 2008; Numata et al . , 2009; Rauch et al . , 2008 ) . Pericentrin is an unusually large coiled-coil protein ( 3336 amino acids in human ) that forms elongated fibrils with its C-terminus anchored near the centriole wall and the N-terminus extended outwardly and radially across PCM zones in interphase cells ( Lawo et al . , 2012; Mennella et al . , 2012; Sonnen et al . , 2012 ) . Recent studies showed that pericentrin plays an evolutionarily conserved role in mitotic PCM expansion and interphase centrosome organization , as loss of pericentrin activity in human , mice , and flies all results in failed recruitment of other PCM components to the centrosome and affects the same set of downstream orthologous proteins in each system ( e . g . CEP215 in human , Cep215 in mice , and Cnn in flies ) ( Chen et al . , 2014; Lee and Rhee , 2011; Lerit et al . , 2015 ) . In vertebrates , a key function of PCNT is to initiate centrosome maturation ( Lee and Rhee , 2011 ) and serve as a scaffold for the recruitment of other PCM proteins ( Haren et al . , 2009; Lawo et al . , 2012; Purohit et al . , 1999; Zimmerman et al . , 2004 ) . However , the mechanism underlying the timely synthesis and recruitment of a large sum of PCNT proteins to the PCM is as yet unresolved . Given its large size ( >3300 amino acids ) and the modest rate of translation elongation ( ~3–10 amino acids per second , Boström et al . , 1986; Ingolia et al . , 2011; Morisaki et al . , 2016; Pichon et al . , 2016; Wang et al . , 2016; Wu et al . , 2016; Yan et al . , 2016 ) , synthesizing a full-length PCNT protein would take ~10–20 min to complete after translation initiation . Notably , after the onset of mitosis , the PCM reaches its maximal size immediately before metaphase in ~30 min in human cells ( Gavet and Pines , 2010; Lénárt et al . , 2007 ) . Thus , the cell faces a kinetics challenge of synthesizing , transporting , and incorporating multiple large PCM proteins such as PCNT into mitotic centrosomes within this short time frame . We show here that pericentrin mRNA is spatially enriched at the centrosome during mitosis in zebrafish embryos and cultured human cells . In cultured cells , the centrosomal enrichment of PCNT mRNA predominantly occurs during early mitosis , concomitantly with the peak of centrosome maturation . We further show that centrosomally localized PCNT mRNA undergoes active translation and that acute inhibition of translation compromises the incorporation of PCNT proteins into the centrosome during early mitosis . Moreover , we find that centrosomal localization of PCNT mRNA requires intact polysomes , microtubules , and cytoplasmic dynein activity . Taken together , our results support a model in which translating PCNT polysomes are being actively transported toward the centrosome during centrosome maturation . We propose that by targeting actively translating polysomes toward centrosomes , the cell can overcome the kinetics challenge of synthesizing , transporting , and incorporating the unusually large PCNT proteins into the centrosome . Lastly , we find that the cell appears to use a similar co-translational targeting mechanism to synthesize and deliver another unusually large protein , the microtubule minus-end regulator , ASPM , to the mitotic spindle poles . Thus , co-translational protein targeting might be a mechanism widely employed by the cell to transport cytoplasmic proteins to specific subcellular compartments and organelles .
We found that pericentrin ( pcnt ) transcripts were localized to distinct foci in early zebrafish embryos , whereas those of three other core PCM components , cep152 , cep192 , and cep215 , showed a pan-cellular distribution ( Figure 1A ) . This striking pcnt mRNA localization was observed using two independent , non-overlapping antisense probes against the 5’ or 3’ portion of RNA ( Figure 1B ) . The specificity of in situ hybridization was further confirmed by the loss of signals in two frameshift maternal-zygotic pcnt knockout embryos ( MZpcnttup2 and MZpcnttup5 ) ( Figure 1B and Figure 1—figure supplement 1 ) , where the pcnt transcripts were susceptible to nonsense-mediated decay pathway . By co-staining with the centrosome marker γ-tubulin , we demonstrated that zebrafish pcnt mRNA is specifically localized to the centrosome ( Figure 1C ) . To test whether centrosomal localization of pcnt mRNA is conserved beyond early zebrafish embryos , we examined the localization of human PCNT mRNA in cultured HeLa cells using fluorescent in situ hybridization ( FISH ) . Consistent with our observation in zebrafish , human PCNT mRNA was also localized to the centrosome ( Figure 2 ) . Interestingly , this centrosomal enrichment of PCNT mRNA was most prominent during early mitosis ( i . e . prophase and prometaphase ) and declined after prometaphase . The signal specificity was confirmed by two non-overlapping probes against the 5’ or 3’ portion of the PCNT transcript ( Figure 2—figure supplement 1A ) . Furthermore , using an alternative FISH method , Stellaris single-molecule FISH ( smFISH ) against the 5’ or 3’ portion of the PCNT transcript , we observed highly similar centrosomal enrichment of PCNT mRNA during early mitosis , with near single-molecule resolution ( Figure 2—figure supplement 1B ) . Similar smFISH results were observed in both HeLa and RPE-1 cells ( data not shown ) . Together , these results indicate that PCNT mRNA is specifically enriched at the centrosome during early mitosis in cultured human cells . We speculate that the seemingly constant presence of zebrafish pcnt mRNA at the centrosome of early blastula-stage embryos is due to the fast cell cycle without gap phases at this stage ( ~20 min per cycle ) . We next tested whether centrosomal localization of pcnt mRNA also takes place in differentiated tissues in vivo . We focused on the retinal neuroepithelia of 1-day-old zebrafish because at this developmental stage , retinal neuroepithelial cells in different cell cycle stages can be readily identified based on the known patterns of interkinetic nuclear migration ( e . g . mitotic cells at the apical side of retina ) ( Baye and Link , 2007 ) . Again , we observed that zebrafish pcnt mRNA was enriched at the centrosome of mitotic , but not of non-mitotic , neuroepithelial cells ( Figure 2—figure supplement 2 ) . We thus conclude that centrosomal enrichment of pericentrin mRNA is likely a conserved process in mitotic cells . Interestingly , the timing of this unique centrosomal accumulation of PCNT mRNA in cultured cells ( Figure 2 ) overlaps precisely with that of centrosome maturation ( Khodjakov and Rieder , 1999; Piehl et al . , 2004 ) . These observations raise the intriguing possibility that PCNT mRNA might be translated near the centrosome to facilitate the incorporation of PCNT proteins into the PCM during centrosome maturation . To determine whether PCNT mRNA is actively translated near the centrosome , we developed a strategy to detect actively translating PCNT polysomes by combining PCNT smFISH and double immunofluorescence to label PCNT mRNA , and the N- and C-termini of PCNT protein simultaneously ( Figure 3A ) . Given the inter-ribosome distance of approximately 260 nucleotides on a transcript during translation ( Wang et al . , 2016 ) and the large size of PCNT mRNA ( 10 knt ) , a single PCNT transcript can be actively translated by as many as 40 ribosomes simultaneously . Therefore , up to 40 nascent polypeptides emerging from a single PCNT polysome can be visualized by anti-PCNT N-terminus immunostaining . By combining this immunostaining strategy with PCNT smFISH , multiple nascent PCNT polypeptides can be visualized on a single PCNT mRNA . Furthermore , the signals from antibody staining are determined by the location of the epitopes . Therefore , the translating nascent PCNT polypeptides , with the C-terminus not yet synthesized , would only show positive signals from anti-PCNT N-terminus immunostaining ( and be positive for PCNT smFISH ) , whereas fully synthesized PCNT protein would show signals from both anti-PCNT N- and C-terminus immunostaining ( and be negative for PCNT smFISH because of release of the full-length protein from the RNA-bound polysomes ) . Using this strategy , we detected nascent PCNT polypeptides emerging from PCNT mRNA near the centrosome during early mitosis ( Figure 3B , top panel , PCNT N+/C-/PCNT smFISH+ ) . As an important control , we showed that colocalization of PCNT mRNA with anti-PCNT N-terminus signals was lost after a brief treatment of cells with puromycin ( Figure 3B , bottom panel ) , under a condition confirmed to inhibit translation by dissociating the ribosomes and releasing the nascent polypeptides ( Figure 3—figure supplement 1 , Wang et al . , 2016; Yan et al . , 2016 ) . Next , we developed a methodology to quantify the effect of puromycin treatment on the colocalization of PCNT mRNA and anti-PCNT N-terminus signals in three dimensional ( 3D ) voxels rendered from confocal z-stacks . Given that the mean radius of a mitotic centrosome is ~1 µm ( Figure 3—figure supplement 2 ) , we specifically quantified the fraction of PCNT mRNA between 1 and 3 µm from the center of each centrosome—that is , the RNA close to , but not within , the centrosome—with anti-PCNT N-terminus signals in early mitotic cells , with or without the brief puromycin treatment . Consistent with the results shown in Figure 3B , upon the short puromycin treatment , the fraction of PCNT mRNA with anti-PCNT N-terminus signals was significantly reduced , with many PCNT mRNA no longer bearing anti-PCNT N-terminus signals ( Figure 3C ) . Furthermore , we observed that PCNT mRNA molecules near the centrosome were often positive for both anti-PCNT N-terminus and anti-ribosomal protein S6 signals in both HeLa and RPE-1 cells during early mitosis ( Figure 3—figure supplement 3 ) . Together , these results indicate that during early mitosis , a population of PCNT mRNA is undergoing active translation near the centrosome . In addition to the loss of anti-PCNT N-terminus signals from PCNT mRNA , surprisingly , the brief puromycin treatment led to the population of PCNT mRNA shifting away from the centrosome ( Figure 4A ) . Similarly , when zebrafish embryos were injected with puromycin at the one-cell stage , pcnt transcripts became diffused throughout the cell ( Figure 4—figure supplement 1 ) . Because puromycin dissociates ribosomes and nascent polypeptides , these observations suggest that PCNT/pcnt mRNAs in human and zebrafish are enriched near the centrosome by tethering to the actively translating ribosomes . To further test the dependency of centrosomal enrichment of PCNT mRNA on intact , actively translating polysomes , we treated the cultured cells with either emetine , which stabilizes polysomes by irreversibly binding the ribosomal 40S subunit and thus ‘freezing’ translation during elongation ( Jiménez et al . , 1977 ) , or harringtonine , which disrupts polysomes by blocking the initiation step of translation while allowing downstream ribosomes to run off from the mRNA ( Huang , 1975 ) . We found that PCNT mRNA localization patterns in emetine- and harringtonine-treated cells resembled those observed in vehicle- ( control ) and puromycin-treated cells , respectively ( Figure 4A ) . Congruent with the detection of nascent PCNT polypeptides near the centrosome ( Figure 3 ) , these data support the model that centrosomal enrichment of PCNT mRNA relies on centrosomal enrichment of polysomes that are translating PCNT mRNA . We often observed that the two centrosomes in early mitotic cells were asymmetric in size where more PCNT mRNA was enriched near the larger centrosome ( Figure 4—figure supplement 2 ) . Because the microtubule nucleation activity is often positively correlated with the centrosome size , we speculated that centrosomal enrichment of pericentrin mRNA/polysomes might be a microtubule-dependent process . We thus tested if the localization of pericentrin mRNA would be perturbed when microtubules were depolymerized . We found that in both zebrafish and cultured human cells , pcnt/PCNT mRNA was no longer enriched around the centrosome upon microtubule depolymerization ( Figure 4B and C , Figure 4—figure supplement 3 ) . In contrast , a cytochalasin B treatment , which disrupts the actin cytoskeleton , had no effect on the centrosomal enrichment of PCNT mRNA ( Figure 4—figure supplement 3A ) . These results suggest that microtubules , but not actin filaments , serve as ‘tracks’ on which pericentrin mRNA/polysomes are transported . Given that cytoplasmic dynein is a common minus-end-directed , microtubule-based motor that transports cargo toward the microtubule minus end ( i . e . toward the centrosome ) , we next tested whether centrosomal localization of PCNT mRNA is a dynein-dependent process . We treated the cells with ciliobrevin D , a specific small molecule inhibitor of cytoplasmic dynein ( Firestone et al . , 2012 ) and quantified the effect of this treatment on the centrosomal localization of PCNT mRNA . We found that PCNT mRNA was no longer enriched at the centrosome upon the ciliobrevin D treatment ( Figure 4D ) . Together , these results indicate that centrosomal enrichment of pericentrin mRNA during early mitosis is a translation- , microtubule- and dynein-dependent process . To determine the functional significance of translation of centrosomally localized PCNT mRNA during early mitosis , we compared centrosomal PCNT levels shortly before and after mitotic entry ( i . e . late G2 vs . early M phase ) . We arrested cultured human cells from progression out of late G2 phase using the CDK1 inhibitor RO-3306 ( Vassilev et al . , 2006 ) . CDK1 is largely inactive during G2 and becomes activated at the onset of mitosis ( Gavet and Pines , 2010; Jackman et al . , 2003 ) . In the presence of RO-3306 , cells can be held at late G2 phase , and upon inhibitor washout , cells can be released into mitosis . Because cell cycle synchronization is rarely 100% homogeneous in a cell population , we decided to quantify the amount of centrosomal PCNT at the single-cell level using anti-PCNT immunostaining of individual cells . To confidently identify late G2 cells in RO-3306-treated population , we used a RPE-1 cell line stably expressing Centrin-GFP ( Uetake et al . , 2007 ) and categorized the cells as ‘late G2’ if ( 1 ) their two centrosomes ( with two centrin dots per centrosome ) were separated by >2 µm—a sign indicating the loss of centrosome cohesion that occurs during late G2 to M transition ( Bahe et al . , 2005; Fry et al . , 1998; Mardin et al . , 2011 ) and ( 2 ) their DNA was not condensed . We identified the cells as early M phase cells ( i . e . prophase or prometaphase ) 25 min after RO-3306 washout by observing DNA morphology . Using this strategy , we found that approximately twofold more PCNT proteins were incorporated into the centrosomes in early mitotic cells as compared to late G2 cells ( Figure 5A ) . Importantly , the numbers of PCNT mRNA did not significantly differ between late G2 and early M phases , even though there was an approximately fourfold increase from G1 to late G2 phases ( Figure 5B ) . Therefore , these results indicate that the increase in centrosomal PCNT protein levels when cells progress from G2 to M phases ( e . g . , the 25-min period after RO-3306 washout ) is due to upregulation of translation and not to altered mRNA abundance . To independently assess the impact of translation during early mitosis on PCNT incorporation into the centrosomes , we disrupted this process by pulsing the RO-3306 synchronized cells with puromycin to inhibit translation for 2 min , followed by immediate fixation and anti-PCNT immunostaining . As previously shown , this condition inhibits translation acutely and dissociates PCNT nascent polypeptides from PCNT mRNA-containing polysomes , including those near the centrosome ( Figure 3 ) . We found that in the puromycin-treated cells , ~30% fewer PCNT molecules were incorporated into the PCM than in the control cells during prophase/prometaphase ( Figure 5C ) . These results indicate that active translation during prophase/prometaphase is required for efficient incorporation of PCNT into the mitotic centrosomes; disruption of this process , even just briefly , significantly affects the PCNT level at the centrosomes . Collectively , these results indicate that active translation of PCNT mRNA during early mitosis contributes to the optimal incorporation of PCNT proteins into the mitotic PCM and that this is most plausibly achieved by co-translational targeting of the PCNT mRNA-containing polysomes to the proximity of the mitotic centrosomes . To determine if the cell uses a similar co-translational targeting strategy to target other large proteins to the centrosome , we examined the distribution of CEP192 , CEP215/CDK5RAP2 , and ASPM mRNA in cultured human cells . We found that while CEP192 and CEP215 mRNA did not show any centrosomal enrichment during early mitosis ( data not shown ) , ASPM mRNA was strongly enriched at the centrosome during prometaphase and metaphase in both RPE-1 and HeLa cells ( Figure 6 and Figure 6—figure supplement 1 ) . Furthermore , upon a short puromycin treatment , ASPM mRNA became dispersed throughout the cell , indicating that centrosomal enrichment of ASPM mRNA also requires intact polysomes as in the case with PCNT mRNA . ASPM ( and its fly ortholog Asp ) is not a PCM component per se , but a microtubule minus-end regulator ( Jiang et al . , 2017 ) and a spindle-pole focusing factor ( Ito and Goshima , 2015; Ripoll et al . , 1985; Tungadi et al . , 2017 ) . It is highly enriched at the mitotic spindle poles , particularly from early prometaphase to metaphase ( Ito and Goshima , 2015; Jiang et al . , 2017; Tungadi et al . , 2017 ) . Therefore , these data demonstrate another example of spatiotemporal coupling between active translation and translocation of polysomes to the final destination of the protein being synthesized .
In this study , we also developed a strategy of visualizing active translation . We took advantage of the large size of PCNT mRNA and combined PCNT smFISH and immunofluorescence against the N- or C-terminal epitopes of PCNT nascent polypeptides to detect which PCNT mRNA molecules were undergoing active translation ( Figure 3 ) . This imaging-based method allowed us to determine whether the PCNT was being newly synthesized ‘on site’ or the PCNT was made somewhere within the cell and then transported/diffused to the centrosome because only the former would show positive signals for N- , but not C-terminus immunostaining of the synthesized protein , and these signals would be sensitive to the puromycin treatment . However , detecting nascent PCNT polypeptides by anti-PCNT N-terminus antibody staining relies on multiple copies of polypeptides tethered to the translating ribosomes for generating detectable fluorescent signals . Therefore , this method is biased toward detecting the translating PCNT polysomes at later stages of translation elongation , when multiple ribosomes have been loaded and multiple copies of PCNT polypeptides are available for antibody detection . This method , however , would likely fail to detect anti-PCNT N-terminus signals on the mRNA that just started to be translated . We speculate that this could explain why not all centrosomally localized PCNT mRNAs showed anti-PCNT N-terminus signals , although most of these PCNT mRNAs would shift away from the centrosome upon the puromycin or harringtonine treatment ( Figure 4 ) . Translation of PCNT mRNA near the centrosome is further supported by the co-localization of ant-PCNT N-terminus , anti-ribosomal protein S6 , and PCNT smFISH signals near the centrosome during early mitosis in two different human cell lines ( Figure 3—figure supplement 3 ) . Together , these multiple lines of evidence— ( 1 ) co-localization of anti-PCNT N-terminus but not anti-PCNT C-terminus signals with PCNT mRNA , ( 2 ) puromycin-sensitive anti-PCNT N-terminus/PCNT mRNA signals , ( 3 ) polysome-dependent centrosomal enrichment of PCNT mRNA , and ( 4 ) co-localization of PCNT N-terminus/PCNT mRNA signals with a ribosomal protein—strongly support the conclusion that PCNT mRNA is locally translated near the centrosome during early mitosis . How are the polysomes actively translating PCNT or ASPM targeted to the centrosome ? In the case of PCNT , previous studies have shown that PCNT protein is transported to the centrosome through its interaction with cytoplasmic dynein ( Purohit et al . , 1999; Young et al . , 2000 ) , specifically through the dynein light intermediate chain 1 ( LIC1 ) ( Tynan et al . , 2000 ) . Moreover , the LIC1-interacting domain in PCNT is mapped within ~550 amino acids located in the N-terminal half of PCNT ( Tynan et al . , 2000 ) . Based on these findings , we propose a model in which the partially translated PCNT nascent polypeptide starts to interact with the dynein motor complex once the LIC1-interacting domain in the N-terminal half of PCNT is synthesized and folded , as early stages of protein folding can proceed quickly and co-translationally ( Fedorov and Baldwin , 1997; Komar et al . , 1997; Ptitsyn , 1995; Roder and Colón , 1997 ) . Subsequently , this nascent polypeptide-dynein interaction allows the entire polysome , which is still actively translating PCNT mRNA , to be transported along the microtubule toward the centrosome ( Figure 7 ) . Alternatively , it is also possible that the coupling of the polysome to the motor complex is mediated through the ribosome-dynein interaction . If this was the case , additional components/adaptors would need to be involved in the interaction to differentiate the ribosomes translating PCNT mRNA from the ones translating other transcripts . One of the above mechanisms ( i . e . via interaction through the nascent chain or ribosome itself ) may also be used to mediate the co-translational targeting of ASPM mRNA/polysomes to the mitotic spindle poles . Mapping the binding domains on both the motor and cargo sides , identifying the cargo adapter ( s ) that mediates the interaction , and testing the roles of mitotic kinases that are known to be involved in centrosome maturation such as Aurora A and PLK1 ( Glover et al . , 1998; Hannak et al . , 2001; Petronczki et al . , 2008 ) are important next steps to dissect the mechanisms underlying this co-translational protein targeting process . Among the mitotic kinases that could be directly involved in this process , PLK1 is an attractive candidate for the following reasons: Phosphorylation of human PCNT at S1235 and S1241 by PLK1 is required for the recruitment of several other PCM proteins for centrosome maturation ( Lee and Rhee , 2011 ) . In addition , inhibition of PLK1 activity also reduces PCNT levels at mitotic centrosomes ( Haren et al . , 2009; Lee and Rhee , 2011; Santamaria et al . , 2007 ) . Notably , these two PLK1 phosphorylation sites , S1235 and S1241 , are highly conserved among the vertebrates and are located within the putative LIC1-binding domain that interacts with cytoplasmic dynein ( Tynan et al . , 2000 ) . It is thus tempting to speculate that PLK1-dependent phosphorylation of these two conserved residues might be required for mediating the PCNT-dynein interaction and thus initiating co-translational targeting of PCNT to centrosomes . Our finding that new PCNT is delivered co-translationally to the centrosome during centrosome maturation also raises an important question of when and how PCNT interacts with other PCM components that are also required for centrosome maturation such as CEP192 and CEP215 ( Barr et al . , 2010; Choi et al . , 2010; Gomez-Ferreria et al . , 2007; Joukov et al . , 2014; Kim and Rhee , 2014; Zhu et al . , 2008 ) . For example , vertebrate PCNT and CEP215 interact and depend on each other for localizing to mitotic centrosomes ( Buchman et al . , 2010; Haren et al . , 2009; Kim and Rhee , 2014; Lawo et al . , 2012 ) . However , zebrafish Cep215 and human CEP215 may not be targeted to centrosomes co-translationally because their transcripts do not show centrosomal enrichment during early mitosis ( Figure 1 and data not shown ) . It thus remains unclear when and where the PCNT-CEP215 interaction occurs and if this interaction takes place co- and/or post-translationally . Determining if the translating PCNT polysomes contain CEP215 proteins could be the first step to distinguish these possibilities . Clearly , it will be important to elucidate how co-translational targeting of PCNT ( and possible other PCM components ) fits in with the current model of centrosome maturation that involves the interplay of several other PCM proteins . What might be the biological significance of co-translational targeting of unusually large proteins such as PCNT or ASPM to the centrosome during mitosis ? In the case of PCNT , we propose three nonexclusive possibilities . First , since PCNT has been placed upstream as a scaffold to initiate centrosome maturation ( Lee and Rhee , 2011 ) and to help recruit other PCM components , including NEDD1 , CEP192 , and CEP215/CDK5RAP2 ( Lawo et al . , 2012; Lee and Rhee , 2011 ) , it is critical to have optimal amounts of PCNT incorporated at the centrosome early during mitosis . Because dynein-mediated cargo transport is relatively fast , typically ranging from 0 . 5 to 3 µm per second in vivo ( Schlager et al . , 2014; Yang et al . , 2007 ) , it seems that PCNT protein molecules can be transported from anywhere in the cell to the centrosome in seconds , regardless whether they are in a polysome or not . However , dynein cargos in cells are likely powered by several dynein motors at a time ( Kardon and Vale , 2009 ) and the large PCNT protein requires 10–20 min to synthesize . Therefore , mechanistically coupling translation and translocation of polysomes containing multiple copies of nascent PCNT polypeptides could help the cell not only use the dynein motor pool economically but also enhance transport efficiency . Second , generating PCNT proteins elsewhere in the cell might be deleterious . For example , non-centrosomal accumulation of PCNT might recruit other PCM components to the unwanted locations , resulting in ectopic PCM assembly , as PCNT overexpression induces a marked increase in centrosome size and the recruitment of other PCM proteins ( Lawo et al . , 2012; Loncarek et al . , 2008 ) . Co-translational targeting of PCNT on defined microtubule tracks through the dynein motor can help confine most full-length PCNT proteins to the centrosome . Consistent with this view , we observed that if microtubules were depolymerized before mitosis , not only was less PCNT incorporated into mitotic centrosomes , a portion of PCNT also became dispersed throughout the cytoplasm as small PCNT puncta ( Figure 4—figure supplement 3B ) . This result implies that full-length PCNT synthesized in the cytoplasm was not incorporated into centrosomes efficiently without the microtubule-mediated , co-translational protein targeting . Third , co-translational targeting of nascent PCNT polypeptides might be an integrated part of mitotic PCM expansion . Akin to the co-translational targeting of membrane and secreted proteins to the endoplasmic reticulum ( ER ) , where the translating nascent polypeptides undergo protein folding and post-translational modifications in the ER lumen ( Bergman and Kuehl , 1979; Chen et al . , 1995 ) , co-translational targeting of nascent PCNT polypeptides might promote their proper folding and complex formation near the PCM , thereby facilitating integration into the expanding PCM during early mitosis . Another possible mechanism by which co-translational targeting may facilitate PCNT integration into the PCM is through the process of liquid-liquid phase separation . The centrosome is a membrane-less organelle in which the PCM has liquid-like properties . Emerging evidence suggests that such an organelle may be formed by phase separation of compartments into ‘biomolecular condensates’ ( Banani et al . , 2017 ) . Indeed , purified SPD-5 , a key mitotic PCM component with extensive coiled-coil domains in C . elegans , can phase separate into spherical condensates that recruit microtubule nucleating proteins , tubulin , and form microtubule asters , mimicking the properties of in vivo PCM ( Woodruff et al . , 2017 ) . In addition , ribonucleoprotein granules can also phase separate into dynamic liquid droplets in vitro ( Lin et al . , 2015; Patel et al . , 2015 ) . Given that PCNT is a large protein with numerous coiled-coil domains and is targeted to mitotic PCM as a large ribonucleoprotein complex ( polysome ) , it will be fascinating in the future to determine whether co-translational targeting of PCNT polysomes to the centrosome could be part of a phase-separation process that promotes the integration of newly synthesized PCNT proteins into the expanding PCM . Our data also underscore the importance of active translation of PCNT mRNA during early mitosis for the centrosome to gain the optimal level of PCNT because ( 1 ) during the G2/M transition , PCNT mRNA levels remain largely constant , but the centrosomal PCNT protein levels increase ~two fold in 25 min after the onset of mitosis; ( 2 ) inhibiting translation briefly during early mitosis—for example , 2 min of puromycin treatment in prophase or prometaphase—is sufficient to substantially reduce the amount of PCNT proteins incorporated at centrosomes ( Figure 5 ) . It is still unclear how the translation activation of PCNT mRNA is regulated during early mitosis . Previous studies show that translation is globally repressed during mitosis ( Bonneau and Sonenberg , 1987; Fan and Penman , 1970; Pyronnet et al . , 2000 ) , and this global translation repression is accompanied by the translation activation of a subset of transcripts through a cap-independent translation initiation mediated by internal ribosome entry sites ( IRESes ) ( Cornelis et al . , 2000; Marash et al . , 2008; Pyronnet et al . , 2000; Qin and Sarnow , 2004; Ramírez-Valle et al . , 2010; Schepens et al . , 2007; Wilker et al . , 2007 ) . However , a recent study has challenged this view of IRES-dependent translation during mitosis and instead finds that canonical , cap-dependent translation still dominates in mitosis as in interphase ( Shuda et al . , 2015 ) . Therefore , to elucidate the mechanism underlying the translation upregulation of PCNT mRNA during early mitosis , determining if this process is a cap- and/or IRES-dependent process might be a first logical step . In addition , our recent study has linked GLE1 , a multifunctional regulator of DEAD-box RNA helicases , to the regulation of PCNT levels at the centrosome ( Jao et al . , 2017 ) . Since all known functions of GLE1 are to modulate the activities of DEAD-box helicases in mRNA export and translation ( Alcázar-Román et al . , 2006; Bolger et al . , 2008; Bolger and Wente , 2011; Weirich et al . , 2006 ) , it is worth elucidating whether translation upregulation of PCNT mRNA during mitosis is regulated through the role of GLE1 in modulating certain DEAD-box helicases involved in translation control such as DDX3 ( Chen et al . , 2016; Lai et al . , 2008; Soto-Rifo et al . , 2012 ) . Protein targeting to subcellular localization via mRNA localization has been widely used in many other biological contexts . For example , in Drosophila and Xenopus oocytes , segregation of cell fates and embryonic patterning are driven by asymmetrically distributed fate determinants in the form of localized mRNA ( Bashirullah et al . , 1998; Deshler et al . , 1998; Ephrussi et al . , 1991 ) . In Saccharomyces cerevisiae , mating type switching is regulated by targeting ASH1 mRNA to the bud tip , where Ash1 protein is translated and acts as a repressor of mating type switching ( Long et al . , 1997; Takizawa et al . , 1997 ) . In fibroblasts , localizing β-actin mRNA to the leading edge , coupled to its local translation , promotes local actin assembly and directional migration ( Hill et al . , 1994; Sundell and Singer , 1991 ) . Similarly , in neurons , many mRNAs are axonally and dendritically enriched; local translation of a subset of these mRNAs allows synapse-restricted responses to environmental cues ( Lin and Holt , 2007; Sutton and Schuman , 2006; Wu et al . , 2005 ) . However , unlike the co-translational targeting of PCNT and ASPM mRNA to the centrosome described here , in most of the above examples , the mRNAs are transported in a translation-repressed state before arriving their destinations . For the proteins targeted to ER for the secretory pathway , translation is also arrested before the mRNA-ribosome-nascent chain complex reaches the destined membrane , where co-translational translocation of the polypeptide into the ER resumes ( Cross et al . , 2009; Keenan et al . , 2001 ) . A similar ER-like co-translational translocation mechanism is also used for importing a subset of mitochondrial proteins ( Verner , 1993; Yogev et al . , 2007 ) . Therefore , in contrast to all the above examples , we have described a new version of co-translational protein targeting mechanism in which mRNA targeting and translation take place simultaneously . In support of this new protein targeting mechanism , a recent study using a live translation reporter shows that reporter mRNA can be actively translated while being transported in neurons ( Wu et al . , 2016 ) . An important next step is to determine how widely this new mode of protein targeting is employed and how it contributes to a broad context of spatially restricted gene expression . In summary , the work presented here shows that incorporating PCNT into the PCM during centrosome maturation is at least in part mediated by upregulation of PCNT translation during the G2/M transition and the co-translational targeting of translating PCNT polysomes toward the centrosome during early mitosis . Efforts so far on elucidating the mechanism underlying centrosome maturation has focused for the most part on the interplay of protein-protein interactions and post-translational modifications ( e . g . phosphorylation ) of different PCM components . However , our study suggests that a spatiotemporal coupling between the active translation machinery and the motor-based transport may represent a new layer of control over centrosome maturation . Our work also suggests that spatially restricted mRNA localization and translation are not limited to early embryos or specialized cells ( e . g . polarized cells such as neurons ) . We anticipate that co-translational protein targeting to subcellular compartments beyond the centrosome may prove to be a recurrent cellular strategy to synthesize and deliver certain cytoplasmic proteins to the right place at the right time . This regulatory process might represent an underappreciated , universal protein targeting mechanism , in parallel to the evolutionarily conserved co-translational targeting of secreted and membrane proteins to the ER for the secretory pathway .
Wild-type NHGRI-1 fish ( LaFave et al . , 2014 ) were bred and maintained using standard procedures ( Westerfield , 2000 ) . Embryos were obtained by natural spawning and staged as described ( Kimmel et al . , 1995 ) . All animal researches were approved by the Institutional Animal Care and Use Committee , Office of Animal Welfare Assurance , University of California , Davis . Disruption of zebrafish pcnt was done by the CRISPR-Cas technology as described ( Jao et al . , 2013 ) . In brief , to generate guide RNA ( gRNA ) targeting pcnt , two complementary oligonucleotides ( sequences in Supplementary file 2 ) corresponding to a target sequence in the exon 2 of pcnt were annealed and cloned into pT7-gRNA plasmid to generate pT7-pcnt-gRNA . pcnt gRNA was generated by in vitro transcription using the MEGAshortscript T7 kit ( AM1354 , Thermo Fisher Scientific , Waltham , MA ) with BamHI-linearized pT7-pcnt-gRNA as the template . Capped , zebrafish codon-optimized , double nuclear localization signal ( nls ) -tagged Cas9 RNA , nls-zCas9-nls , was synthesized by in vitro transcription using the mMESSAGE mMACHINE T3 kit ( AM1348 , Thermo Fisher Scientific ) with XbaI-linearized pT3TS-nls-zCas9-nls plasmid as the template . Microinjection of the mix of pcnt gRNA and nls-zCas9-nls RNA into zebrafish embryos ( F0 ) was performed as described ( Jao et al . , 2012 ) . Pipettes were pulled on a micropipette puller ( Model P-97 , Sutter Instruments , Novato , CA ) . Injections were performed with an air injection apparatus ( Pneumatic MPPI-2 Pressure Injector , Eugene , OR ) . Injected volume was calibrated with a microruler ( typically ~1 nl of injection mix was injected per embryo ) . Injected F0 embryos were raised and crossed with wild-type zebrafish to generate F1 offspring . Mutations in F1 offspring were screened by PCR amplifying the target region ( primer sequences are in Supplementary file 3 ) , followed by 7 . 5% acrylamide gel electrophoresis to detect heteroduplexes and sequencing . Two frameshift mutant alleles of pcnt , pcnttup2 and pcnttup5 , were used in this study ( Figure 1—figure supplement 1 ) . Maternal-zygotic pcnt mutant embryos were generated by intercrosses of homozygous pcnttup2 or pcnttup5 fish . To inhibit protein synthesis in blastula-stage zebrafish embryos , one-cell stage embryos from wild-type NHGRI-1 intercrosses were injected with ~1 nl of Injection Buffer alone ( 10 mM HEPES , pH 7 . 0 , 60 mM KCl , 3 mM MgCl2 , and 0 . 05% phenol red ) or with 300 µM puromycin in Injection Buffer . The embryos were fixed and analyzed after they developed to the two-cell stage . HeLa cells ( ATCC CCL-2 , a gift from Susan Wente , Vanderbilt University , Nashville , TN , or a HeLa cell line stably expressing scFv-sfGFP-GB1 and NLS-tdPCP-tdTomato , a gift from Xiaowei Zhuang , Howard Hughes Medical Institute , Harvard University , Cambridge , MA; Wang et al . , 2016 ) and RPE-1 cells ( a gift from Irina Kaverina , Vanderbilt University ) or Centrin-GFP RPE-1 cells ( a gift from Alexey Khodjakov , Wadsworth Center , New York State Department of Health , Rensselaer Polytechnic Institute , Albany , NY; Uetake et al . , 2007 ) were maintained in Dulbecco’s Modification of Eagles Medium ( 10–017-CV , Corning , Tewksbury , MA ) and Dulbecco’s Modification of Eagles Medium/Ham’s F-12 50/50 Mix ( 10–092-CV , Corning ) , respectively . All cell lines were supplemented with 10% fetal bovine serum ( FBS ) ( 12303C , lot no . 13G114 , Sigma-Aldrich , St . Louis , MO ) , 1x Penicillin Streptomycin ( 30–002 CI , Corning ) , and maintained in a humidified incubator with 5% CO2 at 37°C . To inhibit cytoplasmic dynein activities , the cells were treated with 50 µM ciliobrevin D for 1 hr 25 min at 37°C . Cell lines used in this study were not further authenticated after obtaining from the sources . All cell lines were tested negative for mycoplasma using a PCR-based testing with the Universal Mycoplasma Detection Kit ( 30–1012K , ATCC , Manassas , VA ) . None of the cell lines used in this study were included in the list of commonly misidentified cell lines maintained by International Cell Line Authentication Committee . In situ hybridizations of zebrafish embryos were performed as described ( Thisse and Thisse , 2008 ) . In brief , the DNA templates for making in situ RNA probes were first generated by RT-PCR using Trizol extracted total RNA from wild-type zebrafish oocytes as the template and gene-specific primers with T7 or T3 promoter sequence ( sequences in Supplementary file 3 ) . Digoxygenin-labeled antisense RNA probes were then generated by in vitro transcription and purified by ethanol precipitation ( sequences in Supplementary file 1 ) . Blastula-stage embryos were fixed in 4% paraformaldehyde in 1x PBS with 0 . 1% Tween 20 ( 1x PBS Tw ) overnight at 4°C , manually dechorionated , and pre-hybridized in hybridization media ( 65% formamide , 5x SSC , 0 . 1% Tween-20 , 50 µg/ml heparin , 500 µg/ml Type X tRNA , 9 . 2 mM citric acid ) for 2–5 hr at 70°C , and hybridized for ~18 hr with hybridization media containing diluted antisense probe at 70°C . After hybridization , embryos were successively washed with hybridization media , 2x SSC with 65% formamide , and 0 . 2x SSC at 70°C , and finally washed with 1x PBS Tw at 25°C . Embryos were then incubated for 3–4 hr with blocking solution ( 2% sheep serum , 2 mg/ml BSA , 0 . 1% Tween-20 in 1x PBS ) at 25°C , and incubated ~18 hr with blocking buffer containing anti-digoxigenin-alkaline phosphatase antibody ( 1:5000 dilution ) at 4°C . Embryos were washed successively with 1x PBS Tw and AP Buffer ( 100 mM Tris , pH 9 . 5 , 100 mM NaCl , 5 mM MgCl2 , 0 . 1% Tween-20 ) before staining with the NBT/BCIP substrates ( 11383213001/11383221001 , Roche Diagnostics ) in AP Buffer . For combined RNA in situ hybridization and immunofluorescence to label both the RNA and centrosomes in zebrafish embryos , the RNA in situ hybridization process was performed as described above until the antibody labeling step: The embryos were incubated for ~18 hr with blocking solution ( 2% sheep serum , 2 mg/ml BSA , 0 . 1% Tween-20 in 1x PBS ) containing anti-digoxigenin-peroxidase ( 1:500 dilution ) , anti-γ-tubulin ( 1:1000 dilution ) , and/or anti-phospho-Histone H3 ( 1:500 dilution ) antibodies at 4°C . Embryos were washed successively with 1x PBS Tw and then incubated for ~18 hr with blocking solution containing Alexa Fluor 568 anti-mouse secondary antibody ( 1:500 dilution ) . After secondary antibody incubation , embryos were washed successively with 1x PBS and borate buffer ( 100 mM boric acid , 37 . 5 mM NaCl , pH 8 . 5 ) with 0 . 1% Tween-20 . The RNA was visualized after tyramide amplification reaction by incubating embryos for 25 min in tyramide reaction buffer ( 100 mM boric acid , 37 . 5 mM NaCl , 2% dextran sulfate , 0 . 1% Tween-20 , 0 . 003% H2O2 , 0 . 15 mg/ml 4-iodophenol ) containing diluted Alexa Fluor 488 tyramide at room temperature . The reaction was stopped by incubating embryos for 10 min with 100 mM glycine , pH 2 . 0 at room temperature , followed by successive washes with 1x PBS Tw . In brief , the DNA templates for making in situ RNA probes were first generated by RT-PCR using Trizol extracted total RNA from human 293 T cells as the template and gene-specific primers with T7 or T3 promoter sequence ( sequences in Supplementary file 3 ) . Digoxygenin-labeled antisense RNA probes were then generated by in vitro transcription and purified by ethanol precipitation ( sequences in Supplementary file 1 ) . Cells were fixed for ~18 hr with 70% ethanol at 4°C , rehydrated with 2x SSC ( 0 . 3 M NaCl , 30 mM trisodium citrate , pH 7 . 0 ) containing 65% formamide at room temperature , pre-hybridized for 1 hr with hybridization media ( 65% formamide , 5x SSC , 0 . 1% Tween-20 , 50 µg/ml heparin , 500 µg/ml Type X tRNA , 9 . 2 mM citric acid ) at 70°C , and hybridized for ~18 hr with hybridization media containing diluted antisense probes at 70°C . Cells were then successively washed with hybridization media , 2x SSC with 65% formamide , and 0 . 2x SSC at 70°C , and finally washed with 1x PBS at room temperature . For tyramide signal amplification , cells were washed with 1x PBS , incubated for 20 min with 100 mM glycine , pH 2 . 0 , and washed with 1x PBS at room temperature . Cells were then incubated for 1 hr with blocking buffer ( 2% sheep serum , 2 mg/ml BSA , 0 . 1% Tween-20 in 1x PBS ) at room temperature , and incubated ~18 hr with blocking buffer containing anti-digoxigenin-peroxidase antibody ( 1:500 dilution ) at 4°C . Cells were washed successively with 1x PBS , borate buffer ( 100 mM boric acid , 37 . 5 mM NaCl , pH 8 . 5 ) with 0 . 1% Tween-20 , and incubated for 5 min in tyramide reaction buffer ( 100 mM boric acid , 37 . 5 mM NaCl , 2% dextran sulfate , 0 . 1% Tween-20 , 0 . 003% H2O2 , 0 . 15 mg/ml 4-iodophenol ) containing diluted Alexa Fluor tyramide at room temperature . Cells were washed successively with 1x quenching buffer ( 10 mM sodium ascorbate , 10 mM sodium azide , 5 mM Trolox in 1x PBS ) and 1x PBS at room temperature . Coverslips were mounted using ProLong Antifade media ( P7481 , Life Technologies ) . Sequential IF and smFISH were performed according to the manufacturer’s protocol ( LGC Biosearch Technologies , Petaluma , CA ) with the following modifications: IF was performed first . Cells were fixed for 10 min in 4% paraformaldehyde in 1x PBS , washed twice with 1x PBS , and permeabilized with 0 . 1% Triton X-100 in 1x PBS for 5 min at room temperature . Cells were washed once with 1x PBS and incubated with 70 µl of diluted primary antibody in 1x PBS for 1 hr at room temperature . Cells were washed three times with 1x PBS and incubated with 70 µl of diluted secondary antibody in 1x PBS for 1 hr at room temperature . Cells were washed three times with 1x PBS and post-fixed for 10 min in 3 . 7% formaldehyde in 1x PBS at room temperature . For the smFISH process , cells were washed with Wash Buffer A , incubated with 67 µl of Hybridization Buffer containing 125 nM DNA probes labeled with Quasar 670 ( sequences in Supplementary file 1 ) for 6 hr at 37°C . Cells were then incubated with Wash Buffer A for 30 min at 37°C , Wash Buffer A containing 0 . 05 µg/ml DAPI for 30 min at 37°C , and Wash Buffer B for 3 min at room temperature . Coverslips were mounted using ProLong Antifade media ( Life Technologies ) and sealed with clear nail polish before imaging . Cells were fixed for 10 min in 4% paraformaldehyde in 1x PBS , washed twice with 1x PBS , and permeabilized with 0 . 5% Triton X-100 in 1x PBS for 5 min at room temperature . Cells were incubated with blocking solution ( 2% goat serum , 0 . 1% Triton X-100 , and 10 mg/ml of bovine serum albumin in 1x PBS ) for 1 hr at room temperature , incubated with blocking solution containing diluted primary antibody for 1 hr at room temperature . Cells were washed three times with 1x PBS and incubated with blocking solution containing diluted secondary antibody for 1 hr at room temperature . Cells were washed with 1x PBS and nuclei were counterstained with 0 . 05 µg/ml of DAPI in 1x PBS for 20 min at room temperature before mounting . S phase cells were detected by using the Click-iT EdU Imaging Kit ( Life Technologies ) according to the manufacturer’s instruction . In brief , Centrin-GFP RPE-1 cells were grown on 12-mm acid-washed coverslips and pulse labeled with 10 µM 5-ethynyl-2’-deoxyuridine ( EdU ) for 30 min at 37°C . The cells were then fixed for 10 min with 4% paraformaldehyde in 1x PBS at room temperature , washed twice with 1x PBS , and permeabilized for 20 min with 0 . 5% Triton X-100 in 1x PBS . Cells were then washed twice with 1x PBS and incubated with a Click-iT cocktail mixture containing Alexa Fluor 488 or 594 azide for 30 min in the dark at room temperature . Embryos subjected to in situ hybridization were mounted in a 35-mm glass-bottom dish ( P35G-1 . 5–10 C , MatTek , Ashland , MA ) in 0 . 8% low melting point agarose and imaged using a stereo microscope ( M165 FC , Leica , Wetzlar , Germany ) with a Leica DFC7000 T digital camera . Confocal microscopy was performed using either a Leica TCS SP8 laser-scanning confocal microscope system with 63x/1 . 40 or 100x/1 . 40 oil HC PL APO CS2 oil-immersion objectives and HyD detectors in resonant scanning mode , or a spinning disk confocal microscope system ( Dragonfly , Andor Technology , Belfast , UK ) housed within a wrap-around incubator ( Okolab , Pozzuoli , Italy ) with Leica 63x/1 . 40 or 100x/1 . 40 HC PL APO objectives and an iXon Ultra 888 EMCCD camera for smFISH and live cell imaging ( Andor Technology ) . Deconvolution was performed using either the Huygens Professional ( Scientific Volume Imaging b . v . , Hilversum , Netherlands ) ( for images captured on Leica SP8 ) or the Fusion software ( Andor Technology ) ( for images captured on Andor Dragonfly ) . To quantify the RNA distribution within the cell in 3D voxels , we used Imaris software ( Bitplane , Belfast , UK ) to fit the protein signal as surfaces and the mRNA signal as spots of different sizes in deconvolved images of each confocal z-stack . The intensity of the mRNA signal in each spot is assumed to be proportional to the amount of mRNA in each spot and is used in lieu of mRNA units . The outline of the cell was obtained either from a transmitted light image or from the background in the pre-deconvolved image and was used to restrict fitting of both mRNA and protein signals to the cell of interest . The distance from each mRNA spot to each centrosome’s center of mass was calculated and the mRNA signal was ‘assigned’ to the closest centrosome . The mRNA spots were binned by distance to the centrosome and the intensities of the spots in each bin were added as a measure of the amount of mRNA at that distance . This was calculated for each cell and then averaged over all the cells for each condition . Thus , the graphs show average mRNA as a function of distance ( binned in 0 . 5 µm intervals ) . To quantify PCNT intensities at the centrosome , we put the surfaces of the anti-PCNT signals fit on the deconvolved images over the original images and used the statistics function in Imaris ( Bitplane ) to obtain the intensity sum of the original images within the fit volume . A HeLa cell line stably expressing scFv-sfGFP-GB1 and NLS-tdPCP-tdTomato was transfected with the SunTag/PP7 reporter plasmid pEF-24xV4-ODC-24xPP7 ( Wang et al . , 2016 ) using Lipofectamine 3000 transfection reagent ( Life Technologies ) according to the manufacturer’s instruction . 12–18 hr after transfection , the medium was changed to 10% FBS/DMEM without phenol red before imaging . Statistical analysis was performed using the GraphPad Prism 7 . Each exact n value is indicated in the corresponding figure or figure legend . Significance was assessed by performing an unpaired two-sided Student’s t-test , as indicated in individual figures . The experiments were not randomized . The investigators were not blinded to allocation during experiments and outcome assessment . | Before a cell divides , it creates a copy of its genetic material ( DNA ) and evenly distributes it between the new ‘daughter’ cells with the help of a complex called the mitotic spindle . This complex is made of long cable-like protein chains called microtubules . To ensure that each daughter cell receives an equal amount of DNA , structures known as centrosomes organize the microtubules during the division process . Centrosomes have two rigid cores , called centrioles , which are surrounded by a matrix of proteins called the pericentriolar material . It is from this material that the microtubules are organized . The pericentriolar material is a dynamic structure and changes its size by assembling and disassembling its protein components . The larger the pericentriolar material , the more microtubules can form . Before a cell divides , it rapidly expands in a process called centrosome maturation . A protein called pericentrin initiates the maturation by helping to recruit other proteins to the centrosome . Pericentrin molecules are large , and it takes the cell between 10 and 20 minutes to make each one . Nevertheless , the cell can produce and deliver large quantities of pericentrin to the centrosome in a matter of minutes . We do not yet know how this happens . To investigate this further , Sepulveda , Antkowiak , Brust-Mascher et al . used advanced microscopy to study zebrafish embryos and human cells grown in the laboratory . The results showed that cells build and transport pericentrin at the same time . Cells use messenger RNA molecules as templates to build proteins . These feed into protein factories called ribosomes , which assemble the building blocks in the correct order . Rather than waiting for the pericentrin production to finish , the cell moves the active factories to the centrosome with the help of a molecular motor called dynein . By the time the pericentrin molecules are completely made by ribosomes , they are already at the centrosome , ready to help with the recruitment of other proteins during centrosome maturation . These findings improve our understanding of centrosome maturation . The next step is to find out how the cell coordinates this process with the recruitment of other proteins to the centrosome . It is also possible that the cell uses similar processes to deliver other proteins to different parts of the cell . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2018 | Co-translational protein targeting facilitates centrosomal recruitment of PCNT during centrosome maturation in vertebrates |
Life-long lack of growth hormone ( GH ) action can produce remarkable extension of longevity in mice . Here we report that GH treatment limited to a few weeks during development influences the lifespan of long-lived Ames dwarf and normal littermate control mice in a genotype and sex-specific manner . Studies in a separate cohort of Ames dwarf mice show that this short period of the GH exposure during early development produces persistent phenotypic , metabolic and molecular changes that are evident in late adult life . These effects may represent mechanisms responsible for reduced longevity of dwarf mice exposed to GH treatment early in life . Our data suggest that developmental programming of aging importantly contributes to ( and perhaps explains ) the well documented developmental origins of adult disease .
Epidemiological studies of individuals born or conceived during the ‘Dutch famine’ led to an appreciation of the impact of early-life events on adult health , and this important concept was formalized as the ‘Barker hypothesis’ ( Barker , 1997 ) . Low birth weight infants born to undernourished mothers were found to have increased risk for hypertension , cardiovascular disease and diabetes in later life ( Huxley et al . , 2002; Xiao et al . , 2010 ) . Furthermore , recent studies have shown that maternal over-nutrition and obesity are also strongly associated with adult metabolic dysfunction in the offspring ( Curhan et al . , 1996; Levin and Govek , 1998 ) . This evidence linked early growth status and nutrient signals to the development of diseases in adult life , and now this concept has been formulated as the Developmental Origins of Health and Diseases ( Hanson and Gluckman , 2014; Sinclair et al . , 2007 ) . It should be noted that the impact of reduced nutrient availability , and the resulting slower growth and development is not always detrimental . Moderate reduction of food or protein intake during pregnancy of female rats and mice lead to improved metabolic status and extended longevity of offspring ( Hales et al . , 1996; Jennings et al . , 1999; Ozanne and Hales , 2004 ) . We have shown that a modest reduction of nutrient availability during the pre-weaning period produced by increasing the number of pups in a litter increased longevity of UM-HET3 mice ( Sun et al . , 2009a ) . Likely , the mechanisms of the effects of both under- and over-nutrition on adult disease , aging and longevity include alternations in hormonal signaling . Hepatic responsiveness to GH , the resulting changes in plasma IGF-1 levels , insulin levels and insulin sensitivity represent one possible mechanisms . There is considerable evidence that hormonal signals are among key determinants of mammalian longevity , and the role of the somatotropic axis ( GH-IGF-I axis ) in the control of aging is particularly well documented ( Bartke et al . , 2013 ) . Robust and reproducible extension of both female and male longevity characterizes mice lacking GH ( Brown-Borg et al . , 1996; Flurkey et al . , 2001 ) , GH receptors ( Coschigano et al . , 2003 ) or GH-releasing hormone , the key stimulator of GH release ( Sun et al . , 2013 ) . Longevity is also extended in mice with reduced tissue levels of bioavailable IGF-1 ( Conover and Bale , 2007 ) . Importantly , there is increasing evidence that the somatotropic axis similarly influences longevity in other mammalian species including humans ( Guevara-Aguirre et al . , 2011; Milman et al . , 2016; van der Spoel et al . , 2016 ) . Panici et al . reported that GH ( but not thyroxine ) treatment during early postnatal period had negative effect on the lifespan of male Ames dwarf mice ( Panici et al . , 2010 ) implying that the actions of GH during development can alter the trajectory of aging and that regulation of longevity may be difficult to uncouple from growth and adult body size . In contrast , an earlier study showed that early-life GH replacement had no effect on longevity in Snell dwarf mice despite dramatic effects on growth and development ( Vergara et al . , 2004 ) . Notably , Snell dwarf mice share the same endocrine defects with Ames dwarf mice and also live much longer than their normal siblings ( Flurkey et al . , 2001 ) . The discrepancy between the results of these studies is likely due to the difference in the protocols including onset and timing of early-life GH treatments and genetic background . Interestingly , in an earlier study using the dwarf rats , GH therapy during development was shown to extend longevity ( Sonntag et al . , 2005 ) , but the interpretation of these findings was complicated by the unexpectedly normal lifespan of these mutants . In view of the discrepancies between the available results and the potentially important implications of the impact of GH signaling during development on longevity , we re-examine this issue in both sexes of normal and long-lived mutant mice . To search for mechanisms linking early GH treatment with reduced longevity , a separate cohort of Ames dwarf mice was treated for six weeks with GH or saline starting at the age of one week . When these animals reached adulthood ( 18 months of age ) , they were used for metabolic profiling and assessment of the hepatic stress signaling , inflammation gene expression and xenobiotic detoxification pathways . Herein , we show that the relatively brief period of GH treatment during the defined early ‘developmental window’ can have a dramatic impact on the lifespan and age-associated characteristics .
To evaluate the effect of early-life GH treatment on longevity , we examined the survival of Ames dwarf ( Prop1df/df ) mice and normal littermate control mice employing two treatment protocols in which GH or vehicle ( saline ) was administered starting at the first or second postnatal weeks . Consistent with the previous reports ( Bartke et al . , 2001; Brown-Borg et al . , 1996 ) , saline treated dwarf mice had dramatically extended lifespan relative to N control mice ( p<0 . 0001; log-rank test , both sexes ) . Being subjected to the GH treatment between the postnatal first and seventh week , the median lifespan of GH treated dwarf mice ( sexes combined ) was decreased by 165 days ( or 16% ) relative to that of saline-injected dwarf mice ( 839 days for GH-dwarf mice vs . 1004 days for saline-dwarf mice ) ( Figure 1A; left panel ) , ( p=0 . 0382 ) based upon log-rank test . Intriguingly , there was no statistically significant ( log-rank test ) effect of the same GH treatment regimen on the lifespan of control mice . Analysis of each sex separately showed that median lifespan in male Prop1df/df mice was shortened by 204 days ( 20%; from 1011 to 807 days ) by GH treatment between postnatal first and seventh week ( Figure 1B , C ) , with the overall survival being significantly decreased ( p=0 . 008 ) . There was no significant treatment effect on overall or median survival in female dwarf mice , indicating sex dimorphism in response to the early GH exposure . 10 . 7554/eLife . 24059 . 003Figure 1 . Effects of early-life GH treatment on longevity starting at the first postnatal week . ( A ) Sex pooled Kaplan-Meier survival curves for each treatment and genotype: Ames dwarf ( Prop1df/df ) and Littermate control mice treated with either vehicle ( saline ) or GH; each point represents a single mouse . N = 41 for control mice groups with Saline; N = 41 for control mice groups with GH; N = 31 for dwarf mice groups with Saline and N = 36 for dwarf mice groups with GH . ( B ) Experimental scheme detailing administration time of GH and vehicle treatment between postnatal first and seventh week . ( C ) Male survival curves . ( D ) Female survival curves . DOI: http://dx . doi . org/10 . 7554/eLife . 24059 . 003 To determine the effect on the maximal longevity , quantile regression method was employed to compare the proportion of live mice in each group at the age at which only upper percentiles ( that is , 25th or 10th ) of the population remained alive ( Wang et al . , 2004 ) . As shown in the Figure 1A , this ( weeks 1–7 ) protocol in dwarf mice led to a significant decrease in maximum lifespan ( p=0 . 0462 for 25th percentile and p=0 . 0400 for 10th percentile ) relative to vehicle controls . Cox Proportional Hazzard ( PH ) models show that GH treatment mice have significantly higher hazards , compared to Saline , in males except in littermate control mice in week 1 data . Supplementary file 1 indicates significantly larger Hazzard Ratios ranging from 2 . 45 to 6 . 7 for males . Linear regression models show very consistent results with Cox PH model results . As illustrated in Supplementary file 1 , except for littermate control males , GH treated male mice have significantly lower adjusted mean life span compared to saline group male mice . In the setting of the second of the employed treatment protocols ( GH or vehicle between postnatal second and eighth week ) , the median lifespan of the GH-treated dwarf mice ( sexes combined ) was decreased by 199 days ( 19 . 5%; from 1019 to 821 days ) relative to that of vehicle-treated dwarf mice ( log-rank test , p<0 . 0229 ) . In dwarf mice , the median lifespan was decreased by 22% ( p=0 . 011 ) in males and 19 . 6% ( p=0 . 048 ) in females . Pooled and sex-specific data are summarized in Supplementary file 1 . In contrast with the results of treatment earlier in life ( weeks 1–7 ) , GH treatment between weeks 2 and 8 significantly shortened the longevity of littermate control mice ( log-rank test , p=0 . 0002 ) , with the median lifespan reduced by 12 . 6% . Littermate male GH-treated mice had a 19 . 6% decrease in mean lifespan relative to the saline group ( Figure 2 ) ( log-rank test , p<0 . 0001 ) ( Supplementary file 1 ) . The longevity of GH-treated littermate female mice seemed slightly decreased compared to the saline group , but this apparent difference was not statistically significant ( Figure 2 ) . 10 . 7554/eLife . 24059 . 004Figure 2 . Effects of early-life GH treatment on longevity starting at the second postnatal week . ( A ) Sex pooled Kaplan-Meier survival curves for each treatment and genotype: Ames dwarf ( df ) and Littermate control mice treated with either vehicle ( saline ) or GH; each point represents a single mouse . N = 36 for control mice groups with Saline; N = 26 for control mice groups with GH; N = 32 for dwarf mice groups with Saline and N = 29 for dwarf mice groups with GH . ( B ) Experimental scheme detailing administration time of GH and vehicle treatment between postnatal second and eighth week . ( C ) Male survival curves . ( D ) Female survival curves . DOI: http://dx . doi . org/10 . 7554/eLife . 24059 . 004 To further evaluate the effects of early-life GH treatment on the aging process , we calculated aging rates per age interval of the dwarf mice and control mice in the two GH treatment protocols . We employed a non-parametric method , explained and applied previously ( Koopman et al . , 2016 ) . In the first week GH treatment protocol ( between postnatal first and seventh week ) , GH treated dwarf mice had aging rates that increased at younger ages to a lower level relative to saline-treated dwarf mice ( Figure 3A ) . There was no clear effect on aging rates of normal control mice using the same GH treatment regime . These patterns are consistent with the effects on lifespan and survival discussed above . In the second week of the GH protocol ( between postnatal second and eighth week ) , a similar pattern was observed in GH treated dwarf mice with increased aging rates at younger ages relative to saline controls ( Figure 3B ) . In contrast with first week results , GH treated control mice had aging rates that increased at younger ages but did not reach the same level as the saline treated mice in later life . These patterns are consistent with the effects on lifespan and survival discussed above . Moreover , these effects of GH are consistent with the effect of reduced GH signaling in other mutant mice , which have aging rates that increase at the advanced ages to a higher level relative to control mice ( Koopman et al . , 2016 ) . 10 . 7554/eLife . 24059 . 005Figure 3 . Age-dependent aging rates of the groups of mice in the first week ( A ) and the second week ( B ) GH treatment groups . The aging rates describe the increases in the mortality rates with age and are expressed in deaths per 10 , 000 mice per day , which equals the change in mortality rate per day . Data from both sexes were pooled for each genotype . ( C and D ) Body weight of Ames dwarf and control mice subjected to early-life 6 weeks of GH/vehicle treatment . Time points represent mean ± SE weight of each group . ( N = 15 mice/group ) . Data from both sexes are pooled for each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 24059 . 005 Together , the reduction in lifespan of these long-lived Prop1df/df mice by early-life GH treatment indicates that alteration in hormonal signaling during critical developmental windows is pivotal for lifespan determination and potential susceptibility to age-related illness . As expected , early-life GH treatment led to a significant somatic growth of dwarf mice in comparison to the saline-treated dwarf mice ( p<0 . 01; Figure 3C and D ) . This increase in body weight persisted after the treatment was stopped although the growth almost leveled off , as expected . Notably , significant increase in body length was also observed in GH-treated dwarf mice indicating that the body weight gain was associated with skeletal growth ( Figure 4A ) . Moreover , absolute heart , kidney and liver weight were increased in GH-treated dwarf mice compared to the vehicle-injected group ( Figure 4C; p<0 . 05 ) whereas brain weight was not affected ( Figure 4B ) . In contrast , the weight of the subcutaneous white adipose tissue ( WAT ) was dramatically decreased by the GH treatment in dwarf mice ( Figure 4C ) . 10 . 7554/eLife . 24059 . 006Figure 4 . Impact of early-life GH treatment on physiological characteristics . ( A ) Tail lengths; the weights of ( B ) Brain , ( C ) Heart , Kidney and Liver and ( D ) Subcutaneous White Adipose Tissues weights presented as absolute values . ( E ) VO2 values plotted as hourly averages representing either dark or light periods . ( F ) RQ = respiratory quotient values . each bar represents means ± SEM for 8–10 mice per group . *p<0 . 05 , **p<0 . 01 , Data from male mice are presented . DOI: http://dx . doi . org/10 . 7554/eLife . 24059 . 006 To search for mechanisms linking early GH treatment with reduced longevity , a separate cohort of Ames dwarf mice was treated for six weeks with GH or saline starting at the age of one week . To further evaluate the long-term metabolic consequences of this early GH treatment , we assessed the metabolic alterations related to energy expenditure ( EE ) in these mice at the age of 18 months by means of indirect calorimetry . Similar to our previous reports ( Westbrook et al . , 2009 ) , O2 consumption rate ( VO2 ) , which declines during normal aging , was found increased in the adult dwarf mice when compared with age-matched littermate control mice ( Figure 4E ) . Remarkably , this upregulation of VO2 in 18-month-old dwarf mice was almost completely abolished by the transient early GH exposure ( Figure 4E ) . Furthermore , we examined the relative metabolic fuel utilization by measuring the respiratory quotient ( RQ ) , which is a dimensionless ratio comparing the volume of carbon dioxide to the volume of oxygen consumed over a given time ( RQ = VCO2/VO2 ) ( Johnston et al . , 2006 ) . The values of RQ for mice usually range from 1 . 0 which means that carbohydrate is the primary source of energy to around 0 . 7 which means that energy comes from fat oxidation ( Even and Nadkarni , 2012 ) . The dwarf mice were found to have a lower RQ value than control mice confirming our previous observations ( Westbrook et al . , 2009 ) while early-life GH treatment significantly increased RQ value ( Figure 4F ) . These data together suggest that early-life GH therapy dampens the metabolic flexibility in Ames dwarf mice by reducing their capacity of greater utilization of fat as energy source . Improved insulin sensitivity and increased metabolic homeostasis are the shared features of long-lived GH-related mutant mice ( Bartke , 2011; Bartke et al . , 2013 ) . As expected , fasting plasma glucose and insulin concentrations were significantly reduced in dwarf mice . Importantly , the early-life GH exposure increased both circulating insulin and glucose to the levels measured in normal littermate controls at the ages of 20 months ( Figure 5A ) . Similarly , the adiponectin level , a marker of insulin sensitivity , was higher in dwarf mice than in littermate control mice and GH treatment dramatically suppressed this upregulation ( p<0 . 01; Figure 5A ) . Similar pattern was also observed in the circulating ketone bodies and LDL levels . Interestingly , early GH administration did not affect the plasma triglycerides levels in dwarf mice , and there were no alterations in nonesterified fatty acids ( NEFA ) or cholesterol level in either genotype ( p=0 . 51 ) . 10 . 7554/eLife . 24059 . 007Figure 5 . Metabolic alterations in responses to early GH treatment . ( A ) Various plasma parameters from male Ames dwarf ( Prop1df/df ) and Littermate control male mice ( N ) subjected to early-life GH treatment . Saline-treated-control mice ( Black Bar ) , Saline-treated-dwarf mice ( White Bar ) and GH-treated-dwarf mice ( Grey Bar ) , a , b values that do not share a superscript letter are statistically significant ( p<0 . 05 ) . Data represent the means ± SEM . Insulin tolerance test ( ITT ) measured at ( B ) 6 months and ( C ) 18 months of age in both male and female mice . Mice were i . p . injected with 1 IU porcine insulin per kg of BW . N = 8 mice per group; *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 24059 . 007 Further , by employing insulin tolerance test ( ITT ) , we evaluated the sensitivity of blood glucose levels to the action of insulin in a cohort of dwarf mice at 6 months of age . Unexpectedly , both male and female early GH-treated dwarf mice displayed similar sensitivity to insulin as saline-treated groups ( Figure 5B ) ; with the reduction of blood glucose levels in response to the insulin challenge paralleling between the two groups . We repeated the ITT experiment when dwarf mice were 18-month-old . As seen in Figure 5B , male GH-treated dwarf mice exhibited dampened sensitivity to insulin in comparisons to the saline-treated groups ( * , p>0 . 05; at time points measured ) ; whereas female GH-treated dwarf mice had similar response to that measured in the saline groups ( Figure 5B ) . Collectively , we found that a six-week period of early-life GH exposure partially normalized ( ‘rescued’ ) many phenotypic characteristics of dwarf mice . These data further support the notion that the early-life hormonal milieu can induce a long-lasting effect on metabolic homeostasis and healthspan later in life . Recent evidence has shown that the stress responsive pathways including RAS-MAPK and PI3K-Akt play a critical role in lifespan extension in flies and worms ( Longo and Fabrizio , 2002; Slack et al . , 2015 ) . Our previous studies have shown that Erk and Akt signaling are diminished , in response to multiple forms of stress , in cells and tissues from long-lived GH deficient mice ( Sun et al . , 2011 , 2009b ) . To identify signaling pathways responsible for the decrease of lifespan in Ames dwarf mice by the early-life GH , we first evaluated the effects of this treatment on several protein kinases which were shown to be affected in our previous reports ( Sun et al . , 2011 , 2009b ) . In contrast to the littermate control mice , dwarf mice had lower hepatic phosphorylation of MAPKs , including the ERK1/2 , P38 and Akt kinases , each of which is known to participate in cellular stress responses ( Figure 6A ) . However , the livers of GH-treated dwarf mice had elevated levels of phosphorylated EKR1/2 , P38 and Akt ( ser473 ) which were indistinguishable from those measured in control mice . Interestingly , Akt phosphorylation on Thr308 was not altered ( Figure 6A ) . We next examined the effect of early GH treatment on the transcriptional regulation of immediate early genes ( IEGs ) . IEG including Egr-1 , Fra1 , Fos , and Jun are rapidly and transiently induced in response to stress or mitogen ( Criswell et al . , 2005; Sukhatme et al . , 1988 ) . Importantly , Akt and MAPK signaling have been found to regulate IEG induction ( Murphy et al . , 2004 ) . By qRT-PCR , hepatic IEGs mRNA levels were found to be significantly lower in saline-treated dwarf mice than in littermate control mice ( Figure 6B ) whereas early-life GH reversed this suppression and even augmented Egr-1 and Fra-1 mRNA levels above the values detected in control mice . Consistently , hepatic IGF-I mRNA levels , GH sensitive markers , followed a similar pattern ( Figure 6B ) . 10 . 7554/eLife . 24059 . 008Figure 6 . Hepatic cellular stress responsive pathways . ( A ) Representative Western blots for phosphorylated and total forms of P38 , Erk1/2 and Akt protein in liver lysates of male Ames dwarf ( Prop1df/df ) and Littermate control male mice ( N ) subjected to early-life GH or saline treatment . ( B ) Expression of Egr1 , Fos , Fra-1 and IGF-I mRNA levels in the liver of GH treated dwarf and normal mice . Data are normalized to Gapdh values and expressed as a ratio ( fold change ) to the level seen in saline injected mice . Saline-treated-control mice ( Black Bar ) , Saline-treated-dwarf mice ( White Bar ) and GH-treated-dwarf mice ( Grey Bar ) , N = 8 male mice per group; *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . a , b values that do not share a superscript letter are statistically significant ( p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24059 . 008 One of the hallmarks of aging is a state of chronic low-grade inflammation in multiple tissues . To determine the early-life GH effects on tissue inflammation status , we examined gene expression of several indices of inflammation in white adipose tissue ( WAT ) , liver , and brain ( cerebral cortex ) by qRT-PCR . As shown in Figures 7A and 20-month-old dwarf mice exhibited lower mRNA levels of several proinflammatory cytokines including IL-6 , IL-1β , MCP-1 , TNF-α and Socs3 particularly in WAT and liver . Importantly , early GH treatment of dwarf mice increased expression of the inflammatory cytokines in adipose and hepatic tissues to the level of the age-matched littermate control mice . Interestingly , brain tissues ( cortex ) , in contrast to other tissues , showed no such effect . These data suggest that GH signals during the early critical period have long-lasting effects on inflammation status in some tissues i . e . WAT and liver but not in other tissues such as cortex . To further understand how the early-life GH regulates tissue inflammation tone , we examined its effect on stress-responsive and metabolic-sensitive signaling molecules in these two tissues . Consistent with the data of inflammatory cytokines expression , GH-treated dwarf mice exhibited elevated hepatic JNK phosphorylation and even a larger increase in NF-kB activation in livers ( Figure 7B ) . Moreover , both P-JNK and P-IkB levels were also increased in WAT from GH injected dwarf mice . Together , these data clearly showed that early transient GH exposure influences the inflammatory status in late life in sensitive tissues including WAT and liver . 10 . 7554/eLife . 24059 . 009Figure 7 . Long-term effects on tissue inflammatory markers . ( A ) Effects of early GH treatment on the expression of various inflammatory indices ( IL-6 , IL-1β , MCP-1 , TNF-α , Socs3 and iNOS in WAT , liver , and cerebral cortex by qRT-PCR . Data are normalized to GAPDH or actin values and expressed as a ratio ( fold change ) to levels of mRNA in control mice . ( B ) Representative Western blots for phosphorylated and total forms of JNK and IkB in liver and eWAT lysates of dwarf subjected to GH or saline treatment . N = 8 male mice per group; *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 24059 . 009 The liver is a major organ to detoxify and eliminate xenobiotics and endobiotics which plays a key role in the metabolic homeostasis of the organism ( Österreicher and Trauner , 2012 ) . Recent studies in GH-related mutant mice have linked the activation of xenobiotic signaling with delayed aging and increased lifespan ( Amador-Noguez et al . , 2007; Steinbaugh et al . , 2012 ) . To investigate the effects of early-life GH exposure on the expression of hepatic xenobiotic genes , we measured hepatic mRNA levels of a set of phase I and phase II xenobiotic detoxification genes through qRT-PCR . Consistent with the previous reports ( Amador-Noguez et al . , 2004 ) , the hepatic expression of these xenobiotic genes was greatly upregulated in saline-treated dwarf as compared to the littermate control mice as shown in Figure 6A ( p<0 . 001 ) . Early-life GH treatment dramatically suppressed the elevation of these genes including Cyp2b9 , Cyp2b13 , Hao3 , FMO3 and Sth2 ( two-tailed t-test; p<0 . 001 ) ( Figure 8A ) . There was no such effect on Gpadh mRNA , the housekeeping control gene . These data indicate that early GH signaling plays a key role in setting-up the xenobiotic regulation pattern in later life . 10 . 7554/eLife . 24059 . 010Figure 8 . Alterations in Xenobiotic Detoxification Genes ( XDG ) and FXR . ( A ) GH treatment during early-life has a dramatic effects on hepatic XDG expression in male Ames dwarf mice . Typical XDE mRNAs such as Cyp2b9 , Cyp2b13 , Hao3 , FMO3 and Sth2 were measured using real-time RT-PCR . Data are normalized to GAPDH or actin values and expressed as a ratio ( fold change ) to levels of mRNA in control male mice . ( B ) Representative Western blots for FXR protein in liver lysates of dwarf and control mice subjected to GH or saline treatment . a , b values that do not share a superscript letter are statistically significant ( p<0 . 05 ) . N = 8 mice per group; *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 24059 . 010 Several reports have pointed out that farnesoid X receptor ( FXR ) is a crucial factor in the regulation of xenobiotic detoxification genes in mice ( Amador-Noguez et al . , 2007; Lee et al . , 2010 ) . We found that FXR protein level was increased in the dwarf livers ( Figure 8B ) despite the lack of difference in the mRNA levels of FXR between the two genotypes . Interestingly , as shown in Figure 8B , early-life GH treatment almost completely suppressed the upregulation of hepatic FXR protein in dwarf mice , indicating that the regulatory impact of the FXR activity on hepatic xenobiotic genes is dependent on GH signaling pathways during early-life .
Epidemiological evidence inspired numerous studies linking early-life nutritional environment to the risks of later obesity , diabetes , hypertension , and cardiovascular diseases ( Hanson and Gluckman , 2014 ) . The results of these studies raise a question of how the factors and signals during the early life may influence mammalian aging . We believe that this question is of fundamental importance from both individual and public health perspectives . In an attempt to address this issue , in the current study , we have examined the role of GH signaling during development and obtained new evidence that transient GH exposure during pre- and early post-weaning period can remarkably shorten longevity of long-lived hypopituitary mice . The striking effects of GH treatment of juvenile mice on their aging trajectory provide a novel opportunity to investigate the role of early-life hormonal environment in longevity and aging at the mechanistic level . Although the existence of critical developmental time window is known in other contexts , the effects of the early-life events on aging process have not been well studied . Earlier studies in rodents have shown that reduced growth in utero followed by rapid postnatal catch-up growth shortened lifespan ( Jennings et al . , 1999; Ozanne and Hales , 2004 ) . Interestingly , Ozanne et al . also found that offspring that were nursed by dams fed a low protein diet , had increased lifespan ( Ozanne and Hales , 2004 ) . Along with these studies , our previous observations have shown that a 50% increase in litter size limited to the first three weeks of life in genetically heterogeneous male mice resulted in a significant increase of longevity ( Sun et al . , 2009a ) . Intriguingly , Holzenberger and colleagues reported that alterations of the nutritional status during the suckling period modified the growth trajectories and metabolic plasticity in late life through the regulations of the GHRH-GH-IGF-axis in mice ( Kappeler et al . , 2009 ) . It is noteworthy that a series of studies in rats conducted by Vickers and his colleagues have shown that GH treatment from postnatal day 3 to day 22 could ameliorate or reverse the detrimental consequences of maternal under-nutrition indicating that hormonal intervention during this critical period can override prenatal predispostion ( Gray et al . , 2014 , 2013; Li et al . , 2015; Reynolds et al . , 2013 ) . A recent report has shown that disrupting the GH signaling during adulthood by induced disruption of the GHR gene did not affect the median lifespan of the mice although it led to a modest increase in the maximal lifespan of the females ( Junnila et al . , 2016 ) . These findings strongly support the critical role of developmental GH signaling in lifespan regulations ( Junnila et al . , 2016 ) . Consistent with these evidences , our study pinpoints the periods between the first and the eighth week of postnatal life as a bona fide critical developmental time window in which longevity and aging rate could be programmed by GH signals . Using GH deficient dwarf mice in which hormonal signaling , growth and adult body size are drastically altered and longevity is greatly increased , our experiments provide compelling evidence that transient hormonal alterations during early critical developmental periods has crucial and long-lasting effect on lifespan . An important advantage of using these mutant mice and their genetically normal siblings in this type of studies is the opportunity to make direct comparison of phenotypic characteristics in individuals that developed concurrently in the same uterus and were raised together by the same dam in an identical environment and yet have vastly different longevity phenotypes . The observation that early GH treatment can shorten the longevity of littermate control mice is novel and potentially important . It can be interpreted as additional evidence that GH actions during development predispose to accelerating aging . Thus , increasing GH action during early post-natal life leads to a reduction of longevity even if GH signals are not altered after eight weeks of age , which is more than 90% of the animal lifespan . However , these intriguing findings have to be interpreted with caution . Unlike GH deficient dwarf mice , the effect of exogenous GH is complicated by the impact of this treatment on endogenous GH activity and action . It is noteworthy that longevity was shortened by exogenous GH only in N male and in only one of the two experiment protocols . As indicated in multiple studies of long-lived mice with mutations affecting the somatotropic axis and mice on long-term calorie restriction , decreased insulin and glucose levels together with improved insulin sensitivity are associated with slow-aging most likely mechanistically ( Bartke , 2011 ) . Notably , these somatotropic mutant mice have markedly increased oxygen consumption ( VO2 ) per unit of body weight and concomitantly reduced respiratory quotient ( RQ ) ( Westbrook et al . , 2009 ) . Strikingly , our new data have shown that these characteristic features are reversed by a few weeks of exposure to GH during early postnatal period . Our findings suggest that the advantageous metabolic features of these long-lived mice are established during the early postnatal life and are particularly susceptible to the alterations in hormonal environments . Taken together , this evidence suggests that that first few weeks of postnatal life represent a critical window in which the set points related to aging and longevity are programed by the hormonal milieu . Another important implications of the present findings is that it may be difficult or perhaps impossible to completely uncouple the extended longevity of these animals from the reductions in somatic growth and adult body size . The mechanisms through which aging and longevity is programmed by the early-life GH signals remain unclear . In current studies , we focused on the possible involvement of the stress responsive pathways , inflammation signaling and xenobiotic detoxification gene expression . Our results suggest that the effects of the transient early GH action on gene expression and consequent functional alterations in various tissues persist into mid- and late adult life and may indeed be permanent . A number of lines of evidence suggest that stress responsive pathways play a critical role in delayed aging and extended longevity ( Miller , 2009 ) . It is worthwhile to note that long-lived mutant mice were characterized by decreased hepatic phosphorylation of stress related kinases , including the ERK , P38 and Akt , each of which is known to participate in cellular stress responses ( Hsieh and Papaconstantinou , 2006; Sun et al . , 2011 , 2009b ) . In the present study , we have shown that dwarf mice subjected to early postnatal GH administration lost these tissue-specific molecular characteristics , mirroring the change in metabolic features . Chronic low-grade inflammation contributes to the development of age-related diseases and degenerative pathology during aging . Activation of proinflammatory signaling pathways including JNK and NF-kB activity has been recognized as an important pathophysiological mechanism in mediating the age-associated inflammatory processes ( Cai and Liu , 2012; Hirosumi et al . , 2002; Tuncman et al . , 2006; Zhang et al . , 2013 ) . In the current study , administration of GH to dwarf mice during the critical developmental windows increased these proinflammatory markers in the livers and WAT indicating activation of JNK and NF-kB kinases in these tissues . In agreement with these observations , Sadagurski et al . found that early-life GH administrations significantly increased or normalized the indices of hypothalamic inflammation in the middle-aged Ames dwarf mice ( Sadagurski et al . , 2015 ) . Together , these data suggest that activation of proinflammatory pathways including JNK and NF-kB may contribute to the loss of metabolic advantages and decrease of longevity in dwarf mice subjected to early-life GH exposure . Previous studies by us and others suggest that a generalized up-regulation of many hepatic mRNAs for xenobiotic detoxification enzymes ( XDE ) represents a shared signature of long-lived mice and may be a general mechanism of slow-aging ( Amador-Noguez et al . , 2004; Shore and Ruvkun , 2013; Sun et al . , 2013 ) . The physiology and regulation of hepatic XDEs during early life is not fully understood . Here , in this study , we found that early-life GH administration had a major influence on the elevated expression of these XDE genes in adult Ames dwarf mice . The up-regulation of these typical XDE genes such as Cyp2b9 , Cyp2b13 , Hao3 , FMO3 and Sth2 was dramatically attenuated in the livers from GH-treated dwarf mice . The molecular mechanisms underlying activation of xenobiotic metabolism pathways remain poorly understood . Some studies indicate that the bile acid receptor , also known as farnesoid X receptor ( FXR ) plays an important role in the regulations of xenobiotic gene expressions ( Amador-Noguez et al . , 2007; Makishima et al . , 1999; Pineda Torra et al . , 2003 ) . In the present study , we observed that early-life GH supplement completely abolished the elevation of hepatic FXR protein levels in Ames dwarf mice . This implies that regulation of xenobiotic metabolism pathway by the FXR activity is at least partially mediated by early-life GH signals . In summary , using GH deficient dwarf mice in which hormonal signaling , growth and adult body size are drastically altered and longevity is greatly increased , our studies provide compelling evidence that transient hormonal alterations during early critical developmental periods has crucial and long-lasting effect on lifespan and longevity-related characteristics . We speculate that mammals during the early development stage respond to their early-life hormonal environment and set anatomical , physiological and biochemical trajectories that shape their future life course including aging and longevity . Hence , understanding the role of hormonal levels and nutrient availability in the first few weeks of postnatal life on preservation of optimal health in old age and mortality risks has potential to make a significant impact on healthcare management and devising interventions that could improve quality of life for older individuals .
Groups of Ames dwarf ( Prop1df/df ) and littermate control mice ( both males and females ) were subjected to treatment with porcine GH ( pGH ) via s . c . injection ( 6 μg/g bw/d ) , given in equally divided doses 2×/d starting at the age of 1 or 2 weeks and continuing for 6 wk ( dwarf-GH treated ) . On Saturdays and Sundays , animals were injected only once with a full dosage following a previous protocol ( Panici et al . , 2010 ) . In our study , normal siblings ( df/+ or N ) of Ames dwarf mice were used as controls . These control mice are heterozygous for the Prop1df mutation and phenotypically not distinguishable from wild type mice . After 6 wk of treatment , the animals were set aside for a longevity study , and mice in these groups were not exposed to any other manipulations except for recording of body weight . Separate cohorts of animals produced using the same protocol were evaluated for metabolic measurements at 6 months of age , for additional metabolic measurements starting at 18 months of age and euthanized around 20 months of age for tissue collections . Animal protocols were approved by the Animal Care and Use Committee of Southern Illinois University and the University of Alabama at Birmingham . Aging rates were calculated non-parametrically as described previously ( Koopman et al . , 2016 ) . First , mortality rates were calculated per group of mice per age interval of 200 days by dividing the number of mice that died by the number of days lived by all mice in the age interval of interest . If only one mouse died in the last age interval , this age interval was excluded . Aging rates were derived from the mortality rates per group of mice . The aging rate in each age interval was calculated as the absolute difference between the mortality rate in this age interval and the mortality rate in the subsequent age interval divided by the difference in age between both age intervals . Plasma was obtained from blood collected by cardiac puncture from isoflurane anesthetized animals at sacrifice and used for measurement of insulin using Mouse Insulin ELISA Kits ( Crystal Chem , Downers Grove , IL ) . Following the manufacturer’s protocol , total ketone bodies and non-esterified free fatty acids ( NEFA ) were measured using colorimetric assays from Wako Chemicals ( Richmond , VA ) ; glycerol was measured using kits from Sigma and triglycerides using kits from Pointe Scientific ( Canton , MI ) , respectively . Adiponectin and resistin levels were assayed using Mouse Adiponectin/Resistin ELISA Kits ( Linco Research , St . Charles , MO ) . Leptin levels were evaluated using Mouse Leptin ELISA Kits ( Crystal Chem Inc . , Downers Grove , IL ) . TNF-α and IL-6 were measured using Mouse TNF-α/IL-6 ELISA Kits ( Biosource , Camarillo , CA ) . Plasma FFAs were assayed using optimized enzymatic colorimetric assays ( Roche , Penzberg , Germany ) . Blood was taken from the tail to measure blood glucose using a glucometer ( AgaMatrix , Salem , NH ) . Mice fasted for 16 hr underwent GTT by i . p . injection with 1 g of glucose per kg of body weight ( BW ) . Blood glucose levels were measured at 0 , 15 , 30 , 45 , 60 , and 120 min using a PRESTO glucometer ( AgaMatrix , Salem , NH ) for GTT . Non-fasted mice were injected i . p . with 1 IU porcine insulin ( Sigma , St . Louis , MO ) per kg of BW . Blood glucose levels were measured at 0 , 15 , 30 , and 60 min for ITT . The data for ITT are presented as a percentage of baseline glucose . P values were calculated by unpaired , two-tailed Student’s t-tests to compare the specific time points . Mice were subjected to indirect calorimetry ( PhysioScan Metabolic System from AccuScan Instruments , Columbus , OH ) as described before ( Westbrook et al . , 2009 ) . This system uses zirconia , infrared sensors and light beams arrays to monitor oxygen ( VO2 ) , carbon dioxide ( VCO2 ) , and spontaneous locomotor activity , inside respiratory chambers in which individual mice were tested . All comparisons are based on animals studied simultaneously in eight different chambers connected to the same O2 , CO2 and light beam sensors in an effort to minimize the effect of environmental variations and calibration on data . After a 24 hr acclimation period , mice were monitored in the metabolic chambers for 24 hr with ad libitum access to standard chow ( Laboratory Diet 5001 ) and water . Gas samples were collected and analyzed every 5 min per animal , and the data were averaged for each hour . Quantitative real-time PCR was performed using a Rotor-Gene 3000 system ( Corbett Research , San Francisco , CA ) with a QuantiTect SYBR Green RT-PCR kit ( Biorad ) as described ( Sun et al . , 2011 ) . In brief , tissues were homogenized with RNA extraction buffer ( TRIZOL reagent; Life Technologies , CA ) to yield total RNA following the manufacturer’s instructions . Total RNA was reverse transcribed with poly-dT oligodeoxynucleotide and SuperScript II . After an initial denaturation step ( 95°C for 90 s ) , amplification was performed over 40–45 cycles of denaturation ( 95°C for 10 s ) , annealing ( 60°C for 5 s ) , and elongation ( 72°C for 13 s ) . Amplification was monitored by measuring the fluorometric intensity of SYBR Green I at the end of each elongation phase . Glyceraldehyde-3-phosphate dehydrogenase ( GAPDH ) or beta-actin expression was quantified to normalize the amount of cDNA in each sample . The change in threshold cycle number ( ΔCt ) was normalized to the GAPDH reference gene by subtracting ΔCtGAPDH from ΔCtgene . The effect of treatment ( ΔΔCt ) was calculated by subtracting ΔCtnormal from ΔCtTg . Fold induction was determined by calculating 2ΔΔCt . Tissues were homogenized in 0 . 5 ml ice-cold T-PER tissue protein extraction buffer ( Thermo Scientific , Rockford , IL ) with protease and phosphatase inhibitors ( Sigma , St . Louis , MO ) . 40 µg of total protein were separated electrophoretically according to size by SDS–polyacrylamide gel electrophoresis using Criterion XT Precast Gel ( Bio-Rad , Hercules , CA ) , and blotted with the antibodies . For visualization of specific bands in the chemiluminescence assays , the membrane was exposed to X-OMAT film; for chemifluorescence the membrane was incubated with ECF ( enhanced chemifluorescence ) substrate and a digital image was generated with the Molecular Dynamics Storm system . Quantification of immunoblot signals was performed using the ImageQuant software package ( Molecular Dynamics , Sunnyvale , CA ) . The following antibodies were obtained for immunoblotting: p38 MAPK , phospho-p38 MAPK ( Thr180/Tyr182 ) , ERK , phospho-ERK ( Thr202/Tyr204 ) , JNK , phospho-JNK ( Thr183/Tyr185 ) , phospho-Akt ( Ser473 ) and Akt , from Cell Signaling Technology ( Beverly , MA ) ; Nrf2 antibody from Novus Biologicals ( Littleton , CO ) ; β-actin , inhibitor from Sigma-Aldrich Corp . ; and goat anti-rabbit and goat anti-mouse antibodies from Santa Cruz Biotechnology , Inc . ( Santa Cruz , CA ) . Statistical analyses were performed to test the effects of treatments on lifespan of animals , separately in N and dwarf groups . To examine the effect of treatment on lifespan in each gender , the analyses were done separately for males and females , and also on combined data . Cox proportional hazards survival models were fitted to compare hazards of treatments in each group , both separately and combined by gender . As a part of survival analysis , in addition to Cox proportional hazards survival models , log-rank tests were also performed to test equivalent hypotheses . To assess the effect of treatment on maximum lifespan using quantile regression , Fisher’s Exact test was implemented as described previously ( Wang et al . , 2004 ) . Since there were no censored values in our data , the associations between treatment and lifespan were also tested by fitting linear regression models in addition to survival models . Sex was used as a covariate when data were analyzed combined for males and females . | For decades , research has shown that early-life events that happen when animals , including humans , are developing as embryos can later influence how those animals look and behave as adults . However , little is known about if the hormones that drive an animal’s growth shortly after its birth have long-lasting effects too . For example , do growth hormones influence how quickly an animal will age , how healthy it is during adulthood , and how long it will go on to live ? Sun et al . now show that increasing the levels of growth hormones in young mice for just six weeks can have a long-lasting effect on the animals’ lifespans . The experiments involved normal mice and dwarf mice , which are smaller and live for longer . From when they were one week old until they were seven weeks old , the mice were given either growth hormone or salty water as a negative control . As expected , the growth hormone helped the animals to grow longer and heavier . However , Sun et al . also found that this treatment significantly shortened the lives of the male dwarf mice , but not the female dwarf mice or the normal mice . Indeed , male dwarf mice given the growth hormone lived lives that were 20% shorter than those of male dwarf mice given the negative control , and further analyses suggested that they aged faster too . The biochemical processes that occur within a living animal in order to keep it alive are collectively referred to as the animal’s “metabolism” . Further experiments showed that the metabolism of adult dwarf mice that had been exposed to growth hormone at a young age was different from the dwarf mice that had been given the negative control instead . These metabolic changes could help to explain why exposure to growth hormone at an early stage in life can affect an animal into its adulthood . The next step will be to work out , in molecular detail , how exposure to growth hormone in early life has a long-lasting effect on aging and lifespan . Such studies might help scientists to understand more about how a person’s experiences during childhood could affect them in later life , including how it affects their risk of developing age-related diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2017 | Longevity is impacted by growth hormone action during early postnatal period |
The Lon AAA+ protease ( LonA ) is a ubiquitous ATP-dependent proteolytic machine , which selectively degrades damaged proteins or native proteins carrying exposed motifs ( degrons ) . Here we characterize the structural basis for substrate recognition and discrimination by the N-terminal domain ( NTD ) of LonA . The results reveal that the six NTDs are attached to the hexameric LonA chamber by flexible linkers such that the formers tumble independently of the latter . Further spectral analyses show that the NTD selectively interacts with unfolded proteins , protein aggregates , and degron-tagged proteins by two hydrophobic patches of its N-lobe , but not intrinsically disordered substrate , α-casein . Moreover , the NTD selectively binds to protein substrates when they are thermally induced to adopt unfolded conformations . Collectively , our findings demonstrate that NTDs enable LonA to perform protein quality control to selectively degrade proteins in damaged states and suggest that substrate discrimination and selective degradation by LonA are mediated by multiple NTD interactions .
The Lon AAA+ protease ( LonA ) , previously known as the protease La , is an ATP-dependent protease distributed in prokaryotes and eukaryotes ( Charette et al . , 1981 ) . It forms a homo-hexamer to execute its biological function ( Park et al . , 2006; Goldberg et al . , 1994 ) . LonA belongs to the AAA+ ( ATPases associated with various cellular activities ) superfamily and contains an N-terminal domain ( NTD ) , a middle ATPase domain with conserved Walker motifs for ATP hydrolysis , and a C-terminal protease domain ( CTD ) with a serine-lysine catalytic dyad in the active site ( Rotanova et al . , 2004; Botos et al . , 2004 ) . LonA is responsible for degrading damaged or unfolded proteins , as well as native proteins bearing specific recognition elements , known as degradation tags or degrons ( Higashitani et al . , 1997; Wohlever et al . , 2014 ) . How can LonA discriminate its substrates from other non-substrate proteins in a cell ? Substrate recognition of LonA is thought to be mediated by its NTD ( Ebel et al . , 1999; Roudiak and Shrader , 1998; Rudyak and Shrader , 2000; Iyer et al . , 2004; Melnikov et al . , 2008; Adam et al . , 2012; Cheng et al . , 2012 ) . The substrates are unfolded and translocated in an ATP-dependent process into a secluded chamber formed by the ATPase and protease domains . Finally , the unfolded substrates inside the chamber can be degraded into small peptide fragments ( Gottesman , 2003; Baker and Sauer , 2006 ) . The ATP-dependent translocation process of substrates into the hexameric chamber has been well characterized ( Lin et al . , 2016; Su et al . , 2016 ) . Structures of different N-terminal fragments of LonA had been reported ( Li et al . , 2005; Duman and Löwe , 2010; Li et al . , 2010; Chen et al . , 2019 ) , including Escherichia coli ( EcLon ) , Mycobacterium avium complex ( MacLon ) , and Bacillus subtilis ( BsLon ) . However , no structural study has been conducted to analyze substrate interactions by the NTD either in isolation or in the context of full-length LonA . Here we address the question of substrate recognition and discrimination by the NTD of LonA using nuclear magnetic resonance ( NMR ) spectroscopy . To understand how LonA selectively recognizes protein substrates being in damaged states , we have used a thermal stable LonA isolated from Meiothermus taiwanensis ( termed MtaLonA henceforth ) , which allows temperature cycling experiments to be used for switching the populations of protein substrates in the folded and unfolded states . In this work , five various protein substrates have been employed for NMR experiments , which include ( 1 ) Ig2 ( domains 5 and 6 of the gelation factor from Dictyostelium discoideum [Hsu et al . , 2009] , hereafter abbreviated as Ig2 ) with an immunoglobulin ( Ig ) fold forming a dimer that is natively folded up to 40°C; ( 2 ) α-casein , which is an intrinsically disordered protein with no tertiary structure; ( 3 ) native lysozyme with four disulfide bridges and reduced lysozyme forming loose and flexible aggregates; ( 4 ) Ig2D5 ( domain 5 of the gelation factor from D . discoideum ) , which is used for characterization of substrate conformation selection by the NTD; and ( 5 ) a degron-tagged Ig2D5 designed to identify the residues of degrons that can be recognized and bound by the NTD . Our results show that the presence of MtaLonA NTD is required to degrade damaged proteins and native proteins with degrons , but it does not play an active role in mediating the degradation by LonA of an intrinsically disordered substrate , α-casein . Here we report the crystal structure of an N-terminal fragment of MtaLonA determined at a 2 . 1 Å resolution , demonstrating that it is similar to the structures of both EcLon and MacLon NTDs , but not to that of BsLon . MtaLonA NTD appears to tumble independently of the hexameric core complex , indicating that the NTD is attached to the hexameric chamber by a flexible linker thus rendering it possible for detailed chemical shift perturbation ( CSP ) mapping of substrate binding in the context of full-length MtaLonA . We further structurally characterize the NTD-mediated interactions with unfolded proteins , protein aggregates , and degron-tagged proteins . This work suggests that the flexibly linked NTDs can help survey , discriminate , and selectively capture damaged unfolded protein species or native protein substrates carrying exposed degradation sequence motifs .
To evaluate the importance of the NTD in the proteolysis activity of MtaLonA , we constructed a MtaLonA variant , AAAP , of which the NTD ( residues 1–241 ) was removed ( Figure 1A ) , based on the structure of B . subtilis LonA . Substrate degradation activity of AAAP was evaluated using two protein substrates , namely α-casein and Ig2 . Native α-casein is an intrinsically disordered substrate of LonA . Ig2 is an all β-stranded protein that becomes partially unfolded at 55°C ( Figure 1—figure supplement 1 ) and only then is susceptible to proteolysis by MtaLonA ( Higashitani et al . , 1997 ) . Full-length MtaLonA degrades both α-casein and damaged Ig2 ( Van Melderen et al . , 1996 ) . Compared to the full-length MtaLonA , AAAP exhibited a slightly reduced degradation activity against α-casein ( Figure 1B ) , while its proteolytic activity for thermally unfolded Ig2 was severely impaired ( Figure 1C ) . These results showed that , despite the lack of the NTD , AAAP retains the proteolytic activity for intrinsically disordered protein substrates like α-casein . However , the NTD is essential for MtaLonA to recognize and degrade the thermally damaged substrate , Ig2 , indicating the NTD may mediate specific interactions with damaged Ig2 , but not α-casein . We obtained crystals of the N-terminal fragment NN206 of MtaLonA by in situ proteolysis during crystallization ( Figure 2A ) . The structure was refined to 2 . 1 Å resolution with a free R factor of 24 . 9% ( crystallographic R factor is 20 . 9%; Table 1 ) . NN206 forms a bilobal fold: ( 1 ) the N-terminal lobe , which consists of one α-helix and six β-strands , forming three β-sheets ( β1/β3/β4/β5 , β2/β5/β4 , and β1/β6/β5 ) ; ( 2 ) the C-terminal lobe , which is exclusively α-helical , consisting of a five-helix bundle ( Figure 2B ) . The two lobes are jointed together by an 18-residue linker , which is similar to that of both EcLon and MacLon ( Li et al . , 2005; Duman and Löwe , 2010; Li et al . , 2010; Chen et al . , 2019 ) . Based on the result , we also crystallized and solved the structure of the longer N-terminal fragment NN291 ( Figure 2A ) . The structure of NN291 , refined to 1 . 7 Å resolution ( Table 1 ) , is very similar to that of NN206 except for a longer C-terminal helix by four more residues , that is , residues 207–210 ( Figure 2C ) . Despite the lack of spontaneous proteolytic fragmentation in the crystals , no electron density was seen past residue A210 ( Figure 2D ) . Interestingly , the C-terminus of E . coli N-terminal fragment forms an extended 40-residue helix ( Li et al . , 2010 ) . Therefore , we further investigated the structural feature of the corresponding residues 207–243 of MtaLonA using solution-state NMR spectroscopy . Most of the well-dispersed cross-peaks in two-dimensional 15N-1H HSQC NMR spectra of NN206 and NN243 can be well superimposed , indicating that the folded domain structure corresponding to residues 1–206 is not perturbed by the C-terminal extension , residues 207–243 ( Figure 2—figure supplement 1A ) . Nonetheless , NN243 exhibited extra intense and poorly dispersed cross-peaks around 7 . 5–8 . 5 ppm along the proton dimension that is indicative of a highly disordered peptide segment , which most likely corresponds to the C-terminal extension of NN243 . Together , our crystallographic and NMR analyses suggest that the NTD of MtaLonA forms a bilobal globular structure ( residues 1–206 ) followed by a C-terminal flexible extension ( residues 211–243 ) . LonA forms a large homo-hexameric complex resulting in a very high molecular weight of 0 . 5 MDa . To verify the dynamic feature of the NTD , we compared the backbone amide 15N-1H NMR correlations of protonated NN206 and those of full-length MtaLonA ( with perdeuteration of the non-exchangeable hydrogens ) recorded at 55°C . With the aid of perdeuteration labeling and transverse relaxation optimized spectroscopy ( TROSY ) , we observed over 200 well-resolved 15N-1H correlations of full-length hexameric MtaLonA at an apparent molecular weight of 0 . 5 MDa . Importantly , most of these correlations were very similar to those observed in the NN206 spectrum ( Figure 2E ) and a systematic chemical shift offset was due to deuterium isotope effects . These results indicated that the NTD adopts the same structure as the isolated NTD while loosely tethered to the near-megadalton hexameric LonA core via a long flexible linker leading to an independently fast tumbling motion to yield highly favorable NMR relaxation properties and thus high-quality multidimensional NMR spectra ( Figure 2—figure supplement 1B and C ) . We further applied multidimensional heteronuclear NMR experiments to assign the observed cross-peaks in 1H–15N TROSY-HSQC spectrum of NN206 ( BMRB code: 50697 ) and the NMR-derived secondary structure of NN206 , based on the Chemical Shift Index ( Hafsa et al . , 2015 ) , was highly similar to its solved X-ray structures ( Figure 2—figure supplement 1D ) . Together , these results suggested that in a MtaLonA complex , each of the six NTDs may connect to the hexameric core by a flexible ~40-residue linker . Therefore , each 24-kDa NTD may tumble independently of the 350-kDa hexameric core ( Figure 2F ) . To directly observe substrate discrimination mediated by the NTD , we treated NMR samples by one thermal cycle with temperature increasing from 32°C to 55°C and then returning to 32°C . With temperature rising , the exposed hydrophobic areas of thermally unfolded Ig2 was increased , as detected by SYPRO Orange ( Figure 1—figure supplement 1 ) , indicating that the population of Ig2 in unfolded states is increased by thermal denaturation . By contrast , thermal cycling of isolated NN206 exhibited no temperature effects on its protein structure monitored by NMR spectroscopy ( Figure 3—figure supplement 1 ) , which could be explained by the high melting temperatures ( Tm ) of full-length MtaLonA and NN206 ( 68 . 0°C and 85 . 5°C , respectively; Figure 1—figure supplement 1 ) . Upon addition of a twofold molar excess of well-folded Ig2 at 32°C , no significant CSPs were observed ( Figure 3A ) , where only the NN206 was isotopically enriched . By increasing temperature from 32°C to 55°C for the same NMR sample , dramatic variations in the 2D spectral features were detected and the resonances underwent a prominent decrease in intensity ( Figure 3B ) , indicating that the NTD interacts with thermally unfolded Ig2 . With temperature dropping from 55°C to 32°C , linewidth of the expected bound-state resonances was too broad to be detected in the presence of binding events ( Figure 3C ) , suggesting a certain portion of Ig2 still stays in the unfolded or aggregated state . In contrast , the addition of α-casein resulted in little to no effect on the cross-peaks of residues 1–206 of the NTD ( Figure 3—figure supplement 2A and B ) , thus precluding specific interaction mode between the intrinsically disordered protein α-casein and MtaLonA NTD . To clarify whether overall broadening of the cross-peak signals may be due to protein aggregation or precipitation , we performed size-exclusion chromatography ( SEC ) experiments to examine the NMR sample treated with thermal cycling . The analysis revealed that more than half of Ig2 protein became unfolded or aggregated through one thermal cycle while the NTD was still well-folded ( Figure 3D ) . The NTD-Ig2 aggregate complex showed a dissociable SEC profile , suggesting the NTD-substrate interaction is quite dynamic by nature and the fast substrate dissociation may facilitate the translocation process . NMR analysis indicates that thermally damaged Ig2 induces significant CSPs and broadened resonances in the NTD N-lobe , which comprises one α-helix and three β-sheets β1/β3/β4/β5 , β2/β5/β4 and β1/β6/β5 . The residues of MtaLonA NTD with the pronounced CSPs are part of the interaction interface located mainly at helix α1 , loop L1 , L2 , L3 , and β-sheet β2/β5/β4 ( Figure 3E and Figure 3—figure supplements 3 and 4A ) , which consist primarily of hydrophobic residues and are decorated by a number of polar residues ( Figure 3—figure supplement 4B and C ) . To characterize the binding sites of MtaLonA NTD interaction with damaged Ig2 , we chose to mutate two exposed hydrophobic residues P22 and M85 located at the center of β-sheet β2/β5/β4 ( Figure 3—figure supplement 4B ) . We found that NN206* , containing two specific mutations P22A and M85A ( hereafter NN206* ) , may influence the substrate-binding interaction . Spectral analysis showed chemical shift differences between NN206 and NN206* can be negligible ( Figure 3—figure supplement 4D and E ) , suggesting that they have very similar structures , with the exception of the region surrounding the point substitutions . The NMR sample containing both 15N-labeled NN206* and unlabeled Ig2 was also treated with one thermal cycle . NN206* showed much weak interactions with thermal damaged Ig2 ( Figure 3F–H and Figure 3—figure supplement 4F ) , confirming that this binding event is mainly contributed by the specific hydrophobic residues with high hydrophobicity in the N-lobe . Collectively , our results showed that the NTD of LonA selectively interacts with protein substrates via critical hydrophobic residues of its N-lobe , demonstrating that the NTD enables LonA to perform protein quality control by selectively interacting with proteins in damaged or unfolded states . To further examine how the NTD is involved with protein aggregation , we probed the interactions between the scrambled lysozyme aggregate and MtaLonA NTD by NMR spectroscopy . Native lysozyme is a 14 . 3-kDa protein ( 129 amino acids ) comprising four disulfide bridges while reduced lysozyme forms loose and flexible aggregates ( Yang et al . , 2015 ) . TCEP ( Tris ( 2-carboxyethyl ) phosphine; nonthiol-based reducing agent ) drives the formation of amorphous lysozyme aggregates . Moreover , there is no cysteine residue found in full-length MtaLonA , suggesting TCEP does not affect the structure of MtaLonA NTD . Addition of TCEP-treated denatured lysozyme leaded to substantial spectral changes of NN206 resonances recorded at 32°C , while NN206* only showed marginal CSPs ( Figure 3—figure supplement 5A ) . The structural mapping showed that the affected residues were also located mainly in the N-lobe shown in Figure 3—figure supplement 5B and C . Furthermore , we also examined the degradation activity of MtaLonA against either native or denatured lysozyme , respectively . A gel-based assay showed that full-length MtaLonA can efficiently degrade denatured-reduced lysozyme , but not native lysozyme ( Figure 3—figure supplement 5D ) . Accordingly , AAAP , without the NTD , was inactive against either the scrambled lysozyme aggregate or native lysozyme ( Figure 3—figure supplement 5E ) . Together , these results highlighted the key role of NTD , via the N-lobe region , in discriminating and recognizing protein aggregates before subsequent translocation and degradation . LonA can also degrade natively folded proteins with degradation tags such as Sul20 degron ( Higashitani et al . , 1997; Gur and Sauer , 2008 ) . Previously , the NTD of E . coli LonA was shown to be required for degron binding by chemical cross-linking ( Wohlever et al . , 2014 ) . To understand how the NTD directly interacts with degrons , a 20-residue peptide ( the C-terminal 20 residues of SulA; namely Sul20 ) was synthesized and used for NMR titration experiments . Addition of unlabeled Sul20 caused significant spectral changes of NN206 resonances ( Figure 4A ) and the structural mapping showed that the affected residues were again located at β‐sheet β2/β5/β4 , loop L1 , L3 , and helix α1 of the N-lobe subdomain ( Figure 4B and Figure 4—figure supplement 1A ) . Mutations of P22 and M85 located at the center of the hydrophobic surface to moderately hydrophobic Alanine leaded to much weak binding of NN206* with Sul20 peptide ( Figure 4—figure supplement 1B and C ) , suggesting that specific hydrophobic residues of beta-stranded β2/β5/β4 located at the N-lobe of MtaLon NTD play a critical role in degron recognition . To characterize the effect of MtaLon NTD on degrons further , we appended Sul20 to the C-terminus of a protein substrate Ig2D5 ( domain 5 of the gelation factor from D . discoideum , hereafter abbreviated as Ig2D5 ) and investigated the conformational changes experienced by Sul20 upon binding the NTD . The NMR spectra of Ig2D5 and Ig2D5-S20 were well superimposed , indicating that the folded domain structure corresponding to Ig2D5 is not perturbed by the C-terminal Sul20 ( Figure 4C ) . Furthermore , the NMR spectrum of Ig2D5-S20 exhibited additional cross-peaks that well corresponded to Sul20 and the NMR-derived secondary structure of Sul20 was flexible random-coil ( Figure 4—figure supplement 1D ) . Measurement of the local backbone flexibility showed that elevated [1H]–15N nuclear Overhauser effect ( NOE ) values for most residues were from Ig2D5 ( Figure 4D ) , along with the large increase in chemical shift dispersion . Lower [1H]–15N NOEs for Sul20 residues were due to the increased flexibility on ps-ns timescale . By addition of unlabeled NN206 to isotopically enriched Ig2D5-S20 , we observed chemical shift changes and broadened resonances of C-terminal Sul20 ( Figure 4C and E ) , suggesting that residues Y19 and H20 are strongly affected by the binding event based on the resonance broadening beyond detection . The secondary structure of Sul20 remained random coil in the presence of MtaLonA NTD ( Figure 4—figure supplement 1D ) . The NMR spectrum of Ig2D5-S20 in the presence of an equimolar amount of unlabeled NN206* showed a much less pronounced chemical shift effect ( Figure 4—figure supplement 1E and F ) than wild-type NN206 binding to Ig2D5-S20 , indicating that Sul20 tag can be recognized by the hydrophobic patches of N-lobe of MtaLon NTD . Furthermore , we generated a Y19A/H20A double mutant in Ig2D5-S20 construct ( hereafter abbreviated as Ig2D5-S20-Y19A/H20A ) . To investigate an effect attributed to Y19A/H20A double mutant , a gel-based assay was performed at 42°C while Ig2D5 was still at its native state . The results showed that the degradation by the full-length MtaLonA against Y19A/H20A double mutant of Sul20 was much less efficient ( Figure 4—figure supplement 2A ) . The NMR spectra of Ig2D5-S20 and Ig2D5-S20-Y19A/H20A were well superimposed and only the chemical shifts of residues Y19A and H20A were prominently changed ( Figure 4—figure supplement 2B ) . By addition of 15N-labeled NN206 to 15N-labeled Ig2D5-S20-Y19A/H20A , careful comparison of chemical shifts revealed that no obvious changes , suggesting that Y19A/H20A double mutant indeed significantly reduces the binding with NN206 ( Figure 4F and Figure 4—figure supplement 2C ) . To understand further selective recognition of protein substrate in the unfolded states by the NTD , we asked whether the unfolding-refolding process of substrate may be modulated by interacting with the NTD . To investigate how the conformations of substrates are influenced in the presence of NTDs , we constructed a single-domain protein substrate Ig2D5 for NMR experiments due to its high-quality NMR spectrum ( Figure 5—figure supplement 1A ) . About 96% of the backbone resonances of Ig2D5 can be assigned by multidimensional heteronuclear NMR experiments . Ig2D5 is a β-sheet protein with Ig-like fold and the Tm of substrate Ig2D5 is similar to Ig2 ( Figure 5—figure supplement 1B and C ) . Here we applied NMR spectroscopy to investigate the unfolding-refolding equilibrium of Ig2D5 induced by temperature cycling while the structure of MtaLon NTD remained stable . The NMR samples containing isotopically enriched Ig2D5 in the absence and presence of unlabeled NN206 were heated from 32°C to 60°C and then cooled to 32°C . Based on the intensity changes of resonances corresponding to native Ig2D5 , we estimated that isolated Ig2D5 was largely unfolded at 60°C ( Figure 5A and B; coloured green ) and ~72% of Ig2D5 was folded after one thermal cycle ( Figure 5C and Figure 5—figure supplement 1D ) . However , in the presence of equimolar NN206 , the folded ratio of Ig2D5 was reduced to ~15% after thermal treatment ( Figure 5C; coloured blue ) . Similar results were obtained using equimolar full-length MtaLonA with catalytic mutant S678A ( LonA* ) ( Figure 5C; coloured orange ) . However , on addition of equimolar AAAP* ( residues 242–793 with catalytic mutant S678A ) , ~75% of Ig2D5 were folded under the same condition ( Figure 5C; coloured black ) . The results revealed that the unfolded state of Ig2D5 was engaged with MtaLon NTD and the temperature-induced unfolding-refolding of substrate was significantly affected in the presence of MtaLon NTD ( Figure 5D ) . Accordingly , we also examined how the unfolding-refolding of Ig2D5 could be affected by NN206* . After one thermal cycle , about 75% of Ig2D5 was folded in the presence of equimolar NN206* ( Figure 5E and Figure 5—figure supplement 2A–C ) . Upon addition of eightfold excess of NN206* , about 60% of Ig2D5 was folded ( Figure 5—figure supplement 2D ) , suggesting that NN206* interacts with denatured protein substrates weakly . These results demonstrated that the conformations of Ig2D5 may be affected by the hydrophobic substrate-binding interaction and the thermally unfolded states of substrate proteins can be stabilized by engagement with the NTDs . Thus , by selectively interacting with hydrophobic residues exposed in the thermally unfolded states of substrate proteins , the NTDs prominently perturb temperature-induced unfolding-refolding process of substrates .
In this work , we used α-casein and Ig2 , which represent two types of substrates . Native α-casein is an intrinsically disordered protein; by contrast , the native Ig2 is a β-sheet protein with Tm at 46°C ( Figure 1—figure supplement 1 ) . We show that the construct AAAP devoid of the NTD ( residues 242–793 , ATPase and protease domains ) , which retains wild-type ATPase and peptidase activities in LonA , lacks the degradation activity against thermally denatured Ig2 while exhibits a wild-type-like activity against α-casein ( Figure 1 ) . Therefore , the NTD is specific for denatured or damaged protein substrates , whose hydrophobic core regions normally buried in the folded state become exposed , but not for intrinsically disordered substrates like casein , which contains mainly charged or polar residues and lack hydrophobic regions in the protein sequence . Indeed , it has been shown that LonA prefers substrates rich in hydrophobic residues ( Gur and Sauer , 2008 ) . Our crystallographic results of NN206 and NN291 are consistent with previous findings that the ~40 residue region between the NTD and ATPase–Protease domains is structurally flexible with multiple accessible protease sites ( Roudiak and Shrader , 1998; Duman and Löwe , 2010; Patterson et al . , 2004; Vasilyeva et al . , 2002 ) . This notion is also corroborated by our high-resolution NMR analyses on MtaLonA NTD both in isolation and as a part of the full-length protein , demonstrating that the independently fast tumbling motion of the NTDs in the ~0 . 5 MDa hexameric assembly of LonA . Moreover , their highly favorable NMR relaxation properties have allowed us to analyze the interactions of the NTD with unfolded proteins , protein aggregates , degron tags , and intrinsically disordered substrates . Our work provides atomic details of the NTD-mediated substrate discrimination by temperature-based switching of the native folded protein to its unfolded state in an NMR sample . NMR characterization indicates MtaLonA NTD does not interact with well-folded Ig2 , native lysozyme , or intrinsically disordered α-casein , whereas unfolded proteins , aggregates , and degron tags can elicit pronounced chemical shift changes to the N-lobe of its NTD , which is subsequently confirmed by structure-based mutagenesis . Collectively , these results suggest that LonA has two substrate-interacting modes: ( 1 ) the NTD-non-requiring mode for intrinsically disordered substrates lacking hydrophobic-rich region , which may be engaged directly with the pore-loops in the LonA assembly without involving the six NTDs; ( 2 ) the NTD-requiring mode for recognizing and trapping damaged/denatured substrates with exposed hydrophobic regions before their engagement with the pore-loops of the LonA assembly . The key role of NTD is to enable LonA to perform protein quality control to selectively capture and degrade proteins in damaged unfolded states . Compared with two‐lobe organization of LonA NTDs , bacterial FtsH proteases have compact N domains formed by a topology of β1-α1-β2-β3-β4-α2-β5 , while ClpXP and ClpEP proteases include N-terminal Zinc-binding domains . ClpAP and ClpCP proteases , as well as of ClpB chaperones , consist of globular α‐helical N domains which bear striking resemblances to the C-lobe of MtaLonA NTD ( Rotanova et al . , 2019 ) . All these N domains , although structurally very distinct , are thought to mediate the interactions with substrates or adaptor proteins . The periplasmic N-domain of FtsH contributes to oligomerization and is essential for modulation the activity of the hexamer in conjunction with the membrane proteins HflK and Hfl ( Scharfenberg et al . , 2015 ) . The N-domain of ClpX is required for recognition of adaptors and some substrates ( Baker and Sauer , 2012 ) . The role of ClpB NTD in protein disaggregation is well characterized by NMR spectroscopy ( Rosenzweig et al . , 2015 ) , demonstrating that ClpB chaperone recognizes exposed hydrophobic stretches in unfolded or aggregated client proteins via a substrate-binding groove in its NTD . In the presence of ClpB NTD , the stability of client proteins can be seriously affected , indicating that the binding of client proteins to ClpB NTD selectively stabilizes their unfolded conformations . Interestingly , intrinsically disordered α-casein can significantly interact with ClpB NTD , but not MtaLonA NTD . Taken together , the α-helical NTD of ClpB chaperones and the beta-stranded N-lobe of MtaLonA NTD comprise substrate-binding sites that play similar roles in specifically recognizing exposed hydrophobic stretches in unfolded or aggregated proteins . By analyzing the NTD of MtaLonA interactions with damaged Ig2 , the scrambled lysozyme , and Sul20 peptide , we conclude that these binding events are mediated by two hydrophobic patches , which comprise ( 1 ) L10 , V14 , I15 , P22 , V23 , and M85 ( termed hpI in the following ) ; ( 2 ) M75 , L77 , P78 , and L82 ( termed hpII ) ( Figure 6A ) . Our results also demonstrate that a double mutant NN206* ( P22 and M85 replaced by Alanine ) , resulted in decreased hydrophobicity of hpI , has a significant effect on the ability of MtaLonA NTD to bind damaged proteins , indicating that the hydrophobic patches can be essential for substrate recognition and discrimination . We also examine the exposed hydrophobic residues of β-sheet β2/β5/β4 , loop L1 , and loop L3 located at the N-lobes of other reported structures of LonA NTDs ( Li et al . , 2005; Duman and Löwe , 2010; Li et al . , 2010; Chen et al . , 2019 ) . The N-terminal fragment from B . subtilis LonA was previously reported to adopt a domain-swapped dimer where the N-lobe of one monomer is positioned next to the C-lobe of the other monomer ( Duman and Löwe , 2010 ) . In contrast , the structures of the N-terminal fragments from MtaLon , EcLon , and MacLon show that the N- and C-lobes are joined together via a short linker ( Figure 6—figure supplement 1A , B ) . Based on structural alignments ( Figure 6—figure supplement 1C ) , the N-lobe of EcLon has almost the same pattern of exposed hydrophobic residues as that of MtaLon NTD ( Figure 6B ) . Interestingly , the NTD of MacLon shows merged hp I and II , while the hpI of BsLon NTD is a little distant from hpII ( Figure 6C , D ) . The structural comparison reveals that , similarly to MtaLon NTD , all three reported structures exhibit exposed hp I and II with slight variations in their shapes and sizes , suggesting that the conserved hydrophobic patches at the N-lobes may be potentially responsible for conferring substrate selectivity toward damaged proteins and degrons . It has previously been known that LonA can recognize a degron-tagged protein by binding to its largely exposed hydrophobic residues . Here , our results show the C-terminal Sul20 has substantially low heteronuclear NOE values , reflecting this region undergoes rapid local motion and is solvent exposed in solution . Therefore , MtaLonA NTD can easily recognize and interact with the C-terminal aromatic and hydrophobic residues of Sul20 . This finding indicates that a pair of aromatic residues consisting of Y19 and H20 in Sul20 may play crucial roles in the hydrophobic interactions with residues P22 and M85 of MtaLonA NTD , which is verified with a double-mutant NN206* . The analysis of the interaction of NN206 or NN206* with Ig2D5-S20 or Ig2D5-S20-Y19A/H20A suggests that the exposed substrate regions with aromatic or hydrophobic side chains are main determinants for their engagement with NN206 , which forms a binding site with similar hydrophobic residues . The primarily hydrophobic or aromatic interactions with NN206 , however , do not lead to specific disorder-to-order transition of the substrate , forming for example an extended or helical structures in the bound state . Perhaps , specific backbone or side-chain interactions are not involved in substrate binding by NN206 . Rather , the N-lobe of NN206 engages non-specific interactions with non-contiguous patches of exposed hydrophobic residues of the substrates , thereby serving a role mainly to trap the substrates and keep them to the proximity of the AAA+ ring . Furthermore , by selectively binding to hydrophobic residues exposed in the thermally unfolded states of proteins , the NTDs prominently stabilize substrate’s unfolded conformation . These results show that the NTD–substrate interactions involve the hydrophobic residues on both partners , suggesting aromatic and hydrophobic side chains , which are usually buried in the native proteins , are important for the NTD to enable MtaLonA to selectively capture and degrade proteins in a damaged unfolded state or with degron tags . Our results directly demonstrate that thermally damaged proteins and degrons induce chemical shift changes and broadened resonances located mainly in the two hydrophobic patches of N-terminal lobe , but we do not find evidence that the C-lobe of the NTD is involved in substrate interaction . However , upon titration of thermally unfolded Ig2 into U-2H/15N-labelled full-length LonA* , affected residues were shown in both N- and C-lobes of MtaLonA* NTD at 55°C ( Figure 7A and B ) . Compared 1H–15N TROSY-HSQC spectra of protonated NN206 and highly deuterated MtaLonA* in the presence of thermally unfolded protein Ig2 , the signals from the C-lobe of MtaLonA* NTD showed notable line broadening , suggesting that other NTDs of LonA* hexamer may join to interact with substrates . It is likely that when a substrate like Ig2 becomes damaged or partially unfolded one or more Lon binding sites are exposed . The multiple NTDs in the ~0 . 5 MDa hexameric assembly of LonA can work as team players and show the avidity effect on substrate binding . When one NTD of LonA hexamer binds to and tethers a substrate nearby , it would be easier for other NTDs to make further substrate contacts . All together , these results reveal the free molecular tumbling of the NTD and its substrate-binding site , based on which a model involving multi-NTD interactions , may be proposed for substrate recognition . The flexibly linked NTD of the hexametric LonA is swaying from side to side and back and forth to survey , recognize , and trap substrates with exposed hydrophobic sequences ( Figure 7C ) . After initial binding , other NTDs may join to increase the avidity of the LonA–substrate interaction ( Figure 7D ) ; the engagement of a substrate protein by multiple NTDs also serves to facilitate effective substrate unfolding and translocation mediated by the coordinated movements of the pore-loops in the ATPase modules of the LonA chamber , powered by cycles of ATP binding and hydrolysis . Finally , the substrate polypeptide chain is pulled inside the chamber and undergoes proteolysis by protease modules of LonA .
Sul20 peptide ( sequence: ASSHATRQLSGLKIHSNLYH ) was synthesized by GenScript ( https://www . genscript . com/ ) at >95% purity . The plasmids expressing for full-length MtaLonA ( 1–793 residues ) or AAAP ( 242–793 residues ) with a C-terminal 6xHis-tag were transformed into E . coli BL21 ( DE3 ) cells ( Su et al . , 2016 ) . Site-directed mutagenesis was performed using the Quickchange kit ( Stratagene ) . NN206 ( 1–206 residues ) , NN243 ( 1–243 residues ) , and NN291 ( 1–291 residues ) were cloned into pET-modified vector with a tobacco etch virus ( TEV ) cleavage site . The three resultant plasmids , encoding the proteins NN206 , NN243 , and NN291 with the N-terminal His-tag , were transformed into E . coli BL21 ( DE3 ) cells . The target proteins synthesis was induced with 0 . 5 mM isopropyl-thio-β-D-galactoside ( IPTG ) at an absorbance at 600 nm ( OD600 ) ~0 . 6 at 28°C . The target proteins were purified with nickel-chelating resins ( Ni-NTA , Qiagen ) . The protein samples were collected and purified by a Superdex 200 ( GE Healthcare ) column ( Lin et al . , 2016 ) . MtaLonA and AAAP proteins were collected and further purified on a MonoQ ( GE Healthcare ) chromatography . For NN206 , NN243 , and NN291 purification , the N-terminal His-tagged proteins were cleaved by TEV protease for overnight at 4°C and then reloaded onto Ni-NTA to remove TEV protease . The flow-through fraction containing target proteins were collected and purified with Superdex 75 ( GE Healthcare ) chromatography . Ig2 ( domains 5 and 6 of the gelation factor ABP-120 of D . discoideum ) ( McCoy et al . , 1999 ) and Ig2D5-S20 ( Ig2 fused with Sul20 ) were cloned into pET28a ( + ) tev for generation of pET28a ( + ) tev-Ig2 and pET28a ( + ) tev-Ig2D5-S20 . The recombinant protein was induced by 0 . 5 mM IPTG at OD600 ~ 0 . 6 for 16 hr . The purifications of Ig2 , Ig2D5 , and Ig2D5-20 is the same as NN206 , NN243 , and NN291 . Isotopically labeled samples for NMR studies were prepared by growing the cells in minimal ( M9 ) medium . All NMR samples ( NN206 , NN206* , NN243 , Ig2 , Ig2D5 , and Ig2D5_S20 ) used in this study were protonated except the sample of the full-length MtaLonA was highly deuterated . The samples of full-length MtaLonA for NMR experiments were prepared by supplementing the growing medium with 1 g/l of 15NH4Cl and 4 g/l of 2H7/12C6-glucose in 99 . 8% 2H2O ( Sigma-Aldrich ) . Ig2 and α-Casein ( Sigma , USA ) were used as the substrates in these assays ( Lin et al . , 2016 ) . Four micromolar substrate proteins were incubated with 0 . 4 μM MtaLonA ( hexamer ) or AAAP or mutations at 55°C . The degradation reactions were halted by protein sample dye and heat inactivation at 95°C for 5 mins . The treated reaction mixtures were then analyzed by SDS–PAGE . Crystallizations of NN206 and NN291 were performed at 295 K by hanging drop vapor-diffusion method . For crystallization of NN206 , in situ proteolysis was performed ( Wernimont and Edwards , 2009 ) . One microliter of MtaLonA ( 10 mg/ml ) plus trypsin in ratio 1000:1 was mixed with 1 μl of 0 . 1 M Tris–HCl ( pH 7 . 5 ) , 0 . 2 M sodium acetate , and 30% PEG 4000 . The NN206 crystals appeared in the mixed drop after 3 days . The crystals of NN291 were grown by mixing with 1 μl of NN291 protein sample ( 15 mg/ml ) and 1 μl of well solution containing 0 . 1 M MES ( pH 5 . 8 ) and 0 . 8 M ammonium sulfate . Crystals grew to 0 . 2–0 . 3 mm over 2 weeks . Crystals of NN206 and NN291 were cryoprotected with 20% glycerol or 20% ethylene glycol before data collection . The data sets of NN206 and NN291 were collected at the beamlines BL-13C1 of National Synchrotron Radiation Research Center ( Taiwan ) and BL-1A of Photon Factory ( Japan ) , respectively . All data sets were processed by HKL2000 ( Otwinowski and Minor , 1997 ) . The structure of NN206 was solved by molecular replacement with the program Phaser ( McCoy , 2007 ) using the structures of the N-terminal fragment of B . subtilis LonA ( residues 4–116 , PDB code 3M65 ) and the N-terminal fragment of E . coli LonA ( residues 120–210 , PDB code 3LJC ) as the search models . The structure of NN291 was determined by molecular replacement using NN206 structure . All initial models were automatically rebuilt using the program AutoBuild ( Terwilliger et al . , 2008 ) . Subsequently , these structures were refined by manual refitting in Coot ( Emsley and Cowtan , 2004 ) and performance of refinement using the program Refmac5 ( Murshudov et al . , 2011 ) . Crystallographic and refinement statistics are listed in Table 1 . All structure figures were made using PyMOL ( Version 1 . 3 , Schrödinger , LLC ) . The atomic coordinates and structure factors for NN206 and NN291 have been deposited in the Protein Data Bank ( http://www . rcsb . org/ ) with the accession numbers 7CR9 and 7CRA , respectively . Thermal shift assay was carried out in qPCR 8-strip tubes ( Gunster Biotech ) using SYBRO orange ( Life Technologies ) as dye . For each reaction , 5 μM purified MtaLonA , NN206 , or Ig2 was mixed with assay buffer ( 50 mM NaPi , 2 mM β-Me , pH 6 . 5 ) and SYPRO Orange dye ( 5× final concentration ) in 45 μl total volume/well . One hundred micromolar Ig2D5 was mixed with assay buffer and SYPRO Orange dye ( 10× final concentration ) in 45 μl total volume/well . Samples were heated at a rate of 0 . 5°C per minute from 25°C to 95°C , and fluorescence signals were recorded with the CFX Connect Real-Time PCR Detection System ( Bio-Rad ) . The melting temperatures were analyzed using the derivative plots of the melting curve . NMR experiments were acquired on Bruker 800-MHz spectrometers ( Bruker BioSpin , Karlsruhe , Germany ) . The assignments of NN206 and Ig2D5-S20 backbone 15N , 1HN , 13Cα , 13Cβ , and 13C′ chemical shifts were obtained by non-TROSY versions of 15N-1H HSQC , HNCACB , HN ( CO ) CACB , HNCA , HN ( CO ) CA , and HNCO spectra ( Sattler et al . , 1999 ) . NMR samples containing 300 μM [U-15N]-labeled protonated NN206 in the absence and presence of 600 μM protonated Ig2 are prepared in the buffer: 50 mM sodium phosphate ( pH 6 . 5 ) with 5% D2O . For temperature cycling experiments , the first 2D 1H–15N TROSY-HSQC spectrum was acquired at 32°C; the same sample was heated to 55°C for recording the second spectra , and the temperature was decreased to 32°C for recording the third spectra . A significance level for CSP set by the average value plus one standard deviation and manual inspection of the affected residues establishes if this level needed further adjustment . For characterization of substrate conformation selection by MtaLonA NTD , 2D 1H-15N TROSY-HSQC NMR spectra of 50 μM [U-15N]-labeled protonated Ig2D5 titrated with 50 µM unlabeled protonated NN206 , MtaLonA , or AAAP proteins were recorded during a thermal cycle starting from 32°C to 60°C , and then back to 32°C . Seventy-two residues of Ig2D5 were selected for calculating folded ratio . The interaction of NN206 with Sul20 peptide was recorded with 50 μM [U-15N]-labeled NN206 titrated with unlabeled 50 and 150 μM unlabeled Sul20 peptide . The interaction of NN206 with TCEP-treated denatured lysozyme was recorded with 50 μM [U-15N]-labeled protonated NN206 titrated with 25 μM unlabeled denatured lysozyme . 1H-15N NOE data were recorded in an interleaved manner: one spectrum with 4 s recycle delay followed by 4 s saturation and another spectrum with no saturation and 8 s recycle delay . All NMR data were processed and analyzed by XWIN-NMR ( Bruker BioSpin ) , NMRPipe ( Delaglio et al . , 1995 ) , and NMRView ( Downing , 2004 ) . | There are many different types of protein which each have different roles in biology . Most proteins are surrounded by water and are folded so that their water-attracting regions are on the outside and more fat-like regions , which repel water , are on the inside . When a protein becomes damaged or is assembled incorrectly , some of the fat-like regions end up on the outside of the protein and become exposed to water . This can prevent the protein from performing its role and harm the cell instead . LonA proteases are responsible for dismantling and recycling these harmful proteins , as well as proteins that have been labelled for destruction . They do this by unfolding the unwanted protein and transporting it into an enclosed chamber made of six LonA molecules . Once inside the chamber , the target protein is broken down into smaller fragments that can be used to build other structures . LonA proteases contain a region called the N-terminal domain , or NTD for short , which is thought to be responsible for identifying which proteins need degrading . Yet it remained unclear how the NTD recognizes and binds to these target proteins . To answer this question , Tzeng et al . studied the detailed structure of a LonA protease that had been purified from bacteria cells . This revealed that the NTD of LonA contains two water-repelling regions which bind to fat-like segments on the surface of proteins that have become unfolded or tagged for destruction . Further experiments showed that the NTD is bound to the main body of LonA via a ‘flexible linker’ . This led Tzeng et al . to propose that the NTD sways around loosely at the end of LonA searching for proteins with exposed water-repelling regions . Once an NTD identifies and attaches to a target , the NTDs of the other LonA molecules then bind to the protein and help insert it into the chamber . Proteases are a vital component of all biological systems . Controlling protein destruction and recycling is a key factor in how cells divide and respond to a changing environment . This study provides new insights into how LonA operates in bacteria , which may apply to proteases more widely . This contributes to our knowledge of fundamental biology and may also be relevant in a range of diseases where protein recycling is defective or inefficient . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2021 | Molecular insights into substrate recognition and discrimination by the N-terminal domain of Lon AAA+ protease |
Horizontal gene transfer is a major force in bacterial evolution . Mobile genetic elements are responsible for much of horizontal gene transfer and also carry beneficial cargo genes . Uncovering strategies used by mobile genetic elements to benefit host cells is crucial for understanding their stability and spread in populations . We describe a benefit that ICEBs1 , an integrative and conjugative element of Bacillus subtilis , provides to its host cells . Activation of ICEBs1 conferred a frequency-dependent selective advantage to host cells during two different developmental processes: biofilm formation and sporulation . These benefits were due to inhibition of biofilm-associated gene expression and delayed sporulation by ICEBs1-containing cells , enabling them to exploit their neighbors and grow more prior to development . A single ICEBs1 gene , devI ( formerly ydcO ) , was both necessary and sufficient for inhibition of development . Manipulation of host developmental programs allows ICEBs1 to increase host fitness , thereby increasing propagation of the element .
Conjugative elements and phages are abundant mobile genetic elements in bacteria , capable of transferring DNA between cells ( Frost et al . , 2005 ) . Integrative and conjugative elements ( ICEs ) appear to be the most widespread type of conjugative element ( Guglielmini et al . , 2011 ) . ICEs are found integrated in a host genome . When activated , they excise and produce conjugation machinery that transfers the element DNA from the host cell to recipients ( Carraro and Burrus , 2015; Johnson and Grossman , 2015; Wozniak and Waldor , 2010 ) . ICEs often carry ‘cargo’ genes that are not necessary for transfer but confer a phenotype to host cells . In fact , ICEs ( conjugative transposons ) were first identified because of the phenotypes conferred by cargo genes ( Franke and Clewell , 1981 ) . Cargo genes include those encoding antibiotic resistances , metabolic pathways , and determinants of pathogenesis and symbiosis ( Johnson and Grossman , 2015 ) . Transfer of mobile elements between cells contributes to rapid evolution and spread of associated cargo genes and phenotypes ( Frost et al . , 2005; Treangen and Rocha , 2011 ) . Despite the benefits cargo genes can provide , the maintenance and transfer of mobile genetic elements requires host cellular resources and in some cases is lethal ( Baltrus , 2013 ) . Maintenance of a mobile genetic element in host cells requires balancing the costs and benefits to the host or a sufficiently high transfer frequency . Many mobile elements , especially ICEs , have been identified bioinformatically ( Bi et al . , 2012; Guglielmini et al . , 2011 ) . Many of these ICEs contain putative cargo genes . However , the phenotypes conferred by these genes cannot be inferred from sequence nor are they easily detected experimentally ( Cury et al . , 2017 ) . ICEBs1 , a relatively small ( ~20 kb ) ICE found in most strains of Bacillus subtilis , was identified bioinformatically ( Burrus et al . , 2002 ) and experimentally based on its regulation by cell-cell signaling ( Auchtung et al . , 2005 ) . Most of the ICEBs1 genes needed for conjugation are grouped together in an operon that is repressed until activating signals are sensed ( Figure 1 ) . Two pathways activate ICEBs1 , both of which lead to cleavage of the repressor ImmR by the protease and anti-repressor ImmA ( Auchtung et al . , 2007; Bose et al . , 2008; Bose and Grossman , 2011 ) . ICEBs1 contains the cell-cell signaling genes , rapI and phrI , which regulate ICEBs1 activation by sensing population density and the relative abundance of ICEBs1-containing host cells ( Auchtung et al . , 2005 ) . RapI is produced at high cell density and during the transition to stationary phase and stimulates the proteolytic cleavage of the repressor ImmR by the protease ImmA ( Bose and Grossman , 2011 ) . Overproduction of RapI stimulates activation of ICEBs1 in >90% of cells ( Auchtung et al . , 2005 ) . RapI activity ( and therefore ICEBs1 activation ) is inhibited by PhrI , a peptide that is secreted by cells that contain ICEBs1 . PhrI levels indicate the relative abundance of ICEBs1-containing cells in the population , preventing the activation and possible redundant transfer of ICEBs1 if most nearby cells already contain the element . ICEBs1 is also activated during the RecA-dependent DNA damage response ( Auchtung et al . , 2005 ) . Biofilms appear to be hotspots of horizontal gene transfer for bacteria growing in natural settings ( Madsen et al . , 2012; Molin and Tolker-Nielsen , 2003 ) . Undomesticated strains of B . subtilis form complex biofilms on agar plates and at the air-liquid interface in standing cultures ( Vlamakis et al . , 2013 ) . There is also extensive spore formation in B . subtilis biofilms ( Branda et al . , 2001; Vlamakis et al . , 2008 ) . In addition , during growth in a biofilm , ICEBs1 is naturally activated and transfers efficiently , generating on the order of 10 new ICEBs1-containing host cells ( transconjugants ) per donor cell under appropriate conditions ( Lécuyer et al . , 2018 ) . B . subtilis biofilms are held together by a matrix composed of secreted exopolysaccharides , protein fibers , and DNA ( Vlamakis et al . , 2013 ) . This matrix reinforces cell-cell contacts , likely promoting rapid spread of ICEBs1 by conjugation . Additionally , the conditions that promote biofilm formation ( high cell density ) also promote activation and transfer of ICEBs1 and sporulation ( Auchtung et al . , 2005; Grossman and Losick , 1988 ) . Although biofilm growth is clearly beneficial to conjugation , it is unknown how ICEBs1 impacts its host cells under these conditions . In this study , we describe a selective advantage provided by ICEBs1 to its host cells during growth in biofilms . This fitness benefit was due to inhibition of host biofilm and spore development . We identified the ICEBs1 gene devI ( formerly ydcO ) as necessary and sufficient to inhibit host development and provide a selective advantage to ICEBs1-containing cells . We also provide evidence indicating that devI likely inhibits the key developmental transcription factor Spo0A , reducing its ability to stimulate biofilm and sporulation gene expression . devI ( ydcO ) is conserved in other ICEBs1-like elements , indicating that manipulation of host development may be a conserved strategy among this family of mobile genetic elements . We postulate that manipulation of host pathways may be a common function of many of the as yet uncharacterized cargo genes in ICEs .
Biofilm formation is characteristic of many bacteria growing in natural settings , including B . subtilis . We used biofilm growth to determine if ICEBs1 affected the fitness of its host cells under conditions that naturally promote its spread . We performed competition experiments in biofilms using strains of undomesticated B . subtilis ( NCIB3610 plasmid-free ) with or without ICEBs1 . We observed highly efficient spread of ICEBs1 at low donor to recipient ratios ( Table 1 ) during growth in biofilms , similar to results reported previously ( Lécuyer et al . , 2018 ) . To measure mating , we mixed ICEBs1-containing cells ( potential donors ) with cells that did not contain ICEBs1 ( ICEBs10 , potential recipients ) and co-cultured the mix on standard biofilm-stimulating growth medium ( MSgg agar ) ( Branda et al . , 2001 ) . Since ICEBs1 induction is regulated by cell-cell signaling , we varied the initial frequency of ICEBs1+ cells between approximately 0 . 01 and 0 . 9 . We inserted unique selectable markers ( antibiotic resistances ) in the chromosomes of the donors and recipients as well as within ICEBs1 . After 4 days of growth at 30°C ( approximately 17 net doublings of the initial population ) , biofilms were disrupted and the number of transconjugants was determined by selective plating . We found that after four days of biofilm growth , the frequency of transconjugants ranged from ~0 . 4 to 0 . 6 of total cells in the biofilm for starting donor frequencies of ~0 . 01–0 . 5 ( Table 1 ) . The highest frequency of transconjugants was observed when the starting frequency of ICEBs1-containing cells was ~0 . 1 . Enhanced conjugation at low donor to recipient ratios is likely due to regulation of ICEBs1 by cell-cell signaling ( induction is inhibited by the presence of other potential donors ) and the higher likelihood of contacting potential recipients at low frequencies of donors . The high levels of ICEBs1 conjugation during growth in biofilms presented a challenge for quantifying the fitness of ICEBs1-containing host cells relative to ICEBs10 cells . Mating converts a large fraction of ICEBs10 cells to transconjugants ( ICEBs1-containing ) , reducing the ICEBs10 proportion of the population in a manner unrelated to host fitness . To measure the effect of ICEBs1 on host fitness , we blocked conjugative DNA transfer using the conEK476E mutation ( Berkmen et al . , 2010 ) . We then compared the proportion of ICEBs1-containing hosts to ICEBs10 cells without the confounding influence of conjugation . We found that cells containing ICEBs1 that is incapable of conjugation {ICEBs1 ( conEK476E ) } had a fitness advantage over cells lacking ICEBs1 during biofilm growth when they were initially present as a minority of the population . As before , we varied the initial frequency of ICEBs1-containing host cells in the inoculum between approximately 0 . 01 and 0 . 9 . To measure fitness we determined the frequency of ICEBs1-containing cells ( fICE ) and ICEBs10 ( fNULL ) cells in the initial inoculum and in mature biofilms ( 4 days of growth at 30°C ) by selective plating . The relative fitness of the ICEBs1-containing cells was calculated as the fold change in the ratio of fICE / fNULL over the course of the competition . We found that the fitness of ICEBs1-containing cells was dependent on their initial frequency in the population ( Figure 2A ) . The frequency-dependence was most likely due to regulation of ICEBs1 gene expression by the cell-cell signaling genes rapI-phrI or some other function of rapI . Cells with ICEBs1 had a selective advantage at low frequencies ( 0 . 01 or 0 . 1 ) when the element is most strongly activated . At high frequencies in the population ( 0 . 5 or 0 . 9 ) , when there is little or no activation , fitness of ICEBs1-containing cells was approximately neutral ( Figure 2A ) . The final growth yields of the populations were similar regardless of the frequency of ICEBs1-containing cells ( Figure 2C ) . We performed control competitions of two differentially marked ICEBs10 strains to verify that the enhanced fitness we observed was due to the presence of ICEBs1 rather than an inherent fitness difference associated with antibiotic resistances ( see Materials and methods , Source data 1 ) . There was a small cost associated with the marker used to select cells containing ICEBs1 ( median relative fitness 0 . 7 ± 0 . 09 ) , leading to a slight underestimate of the selective advantage to these cells . There is a large amount of sporulation in B . subtilis biofilms ( Branda et al . , 2001; Vlamakis et al . , 2008 ) . Consistent with this , we found that approximately 80% of viable colony-forming units ( CFUs ) in a mature biofilm after 3 days were from spores . The selective advantage to cells containing ICEBs1 growing in biofilms could be due to sporulation and/or biofilm development . We blocked sporulation using a mutation that causes a reduction in the amount of the transcription factor Spo0A that is required for spore formation . The spo0A∆Ps mutation is a deletion of the sigma-H-dependent promoter upstream of spo0A ( Siranosian and Grossman , 1994 ) . This mutation reduces production of Spo0A , and cells do not achieve the threshold concentration required to initiate sporulation ( Chung et al . , 1994 ) . spo0A∆Ps mutant cells formed biofilms that were morphologically similar to those formed by wild-type cells . In biofilms without sporulation , spo0A∆Ps mutant cells containing ICEBs1 ( JMJ788 ) had a selective advantage compared to spo0A∆Ps mutant cells without ICEBs1 ( JMJ786 ) ( Figure 2E ) . Notably , the median fitness for spo0A∆Ps mutant cells containing ICEBs1 at a low frequency in the population was approximately six . Thus , sporulation was not required for a fitness benefit to ICEBs1-containing cells in biofilms . Fitness of cells in biofilms can be affected by production of the biofilm matrix . For example , cells that ‘cheat’ by contributing less to biofilm matrix production reap the benefits of growing with other cells that bear the cost of matrix gene expression ( Dragoš et al . , 2018 ) . We showed that cells containing ICEBs1 ‘cheat’ by decreasing expression of biofilm matrix genes compared to cells without ICEBs1 ( see below ) . We found that cells containing ICEBs1 also had a frequency-dependent selective advantage during sporulation , in the absence of biofilms . We prepared mixtures of cells with and without ICEBs1 as described above . These mixtures were spotted onto a medium ( DSM agar ) that promotes high levels of sporulation . During growth on this medium , there are no complex colony features found in biofilms . As in the biofilm competitions , cells containing ICEBs1 had a frequency-dependent selective advantage during sporulation ( Figure 2B ) . At an initial frequency of approximately 0 . 01 , the median relative fitness of the ICEBs1-containing cells was approximately 14 ( 14 . 5 ± 4 . 3 ) . As in biofilms , the total growth yields of the populations were similar regardless of ICEBs1 host frequency ( Figure 2D ) . These results demonstrate that ICEBs1 confers a selective advantage to cells growing on DSM agar , outside the context of biofilms . This could be due to sporulation or growth under these specific conditions . We found that sporulation was required for the strong selective advantage during growth on sporulation medium ( DSM agar ) . The fitness benefit associated with the ICEBs1-containing cells at a low frequency in the population was greatly reduced in the spo0A∆Ps mutant ( no sporulation ) ( Figure 2F ) . The sporulation mutant with ICEBs1 had a median fitness of approximately 1 . 5 compared to approximately nine for wild-type . Based on these results , we conclude that the presence of ICEBs1 confers a frequency-dependent selective advantage during sporulation . Together , our results demonstrate that cells containing ICEBs1 have a frequency-dependent selective advantage in biofilms and during sporulation . This selective advantage is independent of the ability of ICEBs1 to actually transfer from one cell to another . Biofilm formation ( Hamon and Lazazzera , 2001 ) and sporulation ( Hoch , 1993; Sonenshein , 2000 ) are both regulated by the transcription factor Spo0A . Our results indicate that the presence of ICEBs1 could somehow be inhibiting the activity or activation of Spo0A . We hypothesized that some ICEBs1-encoded gene ( s ) inhibit host cell development . This inhibition could delay development and enable cells to continue growth for a small number of generations . This model derives from analogous phenotypes of mutants that do not enter the sporulation pathway ( Dawes and Mandelstam , 1970 ) . Mutants that delay the start of sporulation have a growth advantage as they are able to divide one or a few more times while other cells in the population stop growing and start to sporulate . We found that in mixed populations , sporulation was delayed in cells containing ICEBs1 compared to cells without ICEBs1 , in a frequency-dependent manner . As above , we used an ICEBs1 mutant that is incapable of conjugation {ICEBs1 ( conEK476E ) } . We started several replicate populations , each of which we sampled once at different times to create a time-course . ( This was done because sampling to quantify CFUs [spores and cells] disrupts and prevents monitoring a single population over time . ) Spore frequency was determined by measuring heat-resistant CFUs as a fraction of total CFUs for ICEBs1-containing and ICEBs1-cured strains that contained different antibiotic resistance markers to distinguish the strains . The fitness benefits provided by ICEBs1 were dependent on the relative abundance of ICEBs1-containing cells , indicating that the cell-cell signaling genes rapI-phrI in ICEBs1 were likely involved , either directly or indirectly . Other Rap proteins in B . subtilis are known to regulate development by inhibiting phosphorylation ( activation ) of the transcription factor Spo0A ( Sonenshein , 2000 ) . RapI , like other Rap proteins in B . subtilis , can inhibit the pathway needed to phosphorylate ( activate ) the transcription factor Spo0A , and overexpression of rapI in vivo inhibits sporulation ( Even-Tov et al . , 2016; Parashar et al . , 2013; Singh et al . , 2013 ) . However , it was unknown whether RapI regulates development in vivo under physiological conditions . Results described below demonstrate that the rapI-phrI system is required for the fitness advantage of ICEBs1-containing cells , but that this requirement is by virtue of causing induction of ICEBs1 gene expression . Another gene in ICEBs1 is both necessary and sufficient for the selective advantage of ICEBs1-containing cells during development . We found that an ICEBs1 gene of unknown function , devI ( ydcO ) , was necessary for the fitness advantage of ICEBs1 host cells . We constructed a deletion of devI in the locked-in-ICEBs1 strain . When started at a low frequency in the population ( ~0 . 01 ) the relative fitness of the devI mutant was approximately 3 . 5 ( Figure 5D ) , much less than that of the isogenic devI+ cells ( median fitness ~14 ) in biofilms with sporulation ( Figure 5A ) . Interestingly , the deletion of devI did not reduce fitness fully to neutral , indicating a possible role for other ICEBs1 genes . devI ( ydcO ) is predicted to encode an 86 amino acid protein . A search for conserved motifs and structural similarity between DevI ( YdcO ) and other proteins did not significantly inform our understanding of DevI function . However , devI ( ydcO ) homologs are found in other Bacillus species ( see below ) . We found that when expressed constitutively , devI alone , in the absence of all other ICEBs1 genes , was sufficient to inhibit sporulation and provide a selective advantage . We cloned devI under the control of Pxis at an ectopic locus ( lacA ) in a strain without ICEBs1 . In the absence of ICEBs1 ( and its repressor ImmR ) , Pxis is constitutively active ( Auchtung et al . , 2007 ) . Fitness was measured relative to a control strain that had Pxis with no gene downstream . Sporulation of the Pxis-devI strain was strongly inhibited under conditions that normally support robust sporulation , including in biofilms ( Figure 6 ) . During sporulation either with ( Figure 6A ) or without biofilm formation ( Figure 6B ) , the frequency of the Pxis-devI strain in the population rose from ~0 . 01 to ~0 . 05 , giving a relative fitness of ~5 . This is greater than the typical fitness conferred by ICEBs1 in biofilms , but less than that observed during sporulation without biofilms . We suspect these differences are due to constitutive expression of devI in the absence of ICEBs1's regulatory systems and the earlier onset of starvation on DSM agar compared to MSgg agar; cells that are unable to sporulate eventually die . Results described above demonstrated that devI is a robust inhibitor of sporulation . Sporulation is controlled by the transcription factor Spo0A ( reviewed in Hoch , 1993; Sonenshein , 2000} ) which both directly and indirectly regulates the expression of many genes needed for development , including biofilm formation ( Hamon and Lazazzera , 2001 ) . The results described below indicate that DevI most likely targets Spo0A , either directly or indirectly . We found that devI ( ydcO ) is conserved among Bacillus species and in many cases is located within what appear to be ICEs similar to ICEBs1 . We used NCBI BLAST to search for homologous protein sequences using both pBLAST ( protein database ) and tBLASTn ( translated nucleotide database ) . Homologs with 100% sequence coverage and greater than 70% identity to YdcO from B . subtilis NCIB3610 were found in dozens of other B . subtilis strains and in closely related species including B . licheniformis , B . atrophaeus , and B . amyloliquefaciens . Excluding Bacillus species from the searches to possibly identify more distantly related proteins with known functions produced no hits . We analyzed the sequence surrounding the devI ( ydcO ) homologs identified to determine if there is similarity to ICEBs1 . All of the devI ( ydcO ) homologs appear to be within mobile element regions similar to ICEBs1 , though some are clearly missing genes present in ICEBs1 . Although we cannot infer whether any of these regions are functional mobile elements , we suspect that the ability to inhibit host development may be a conserved strategy among ICEBs1-like elements and possibly other ICEs with cargo genes of unknown function .
We hypothesize that in addition to providing a fitness advantage to its host cell , delaying sporulation may also improve stability of ICEBs1 in the host during development . Sporulation involves the formation of an asymmetric division septum generating the larger mother cell and the smaller forespore ( Errington , 2001; Higgins and Dworkin , 2012 ) . Sporulation is induced when cells are at a high population density and running out of nutrients , conditions that also activate ICEBs1 ( Auchtung et al . , 2005; Grossman and Losick , 1988 ) . The plasmid form of ICEBs1 that is generated after excision from the chromosome is not known to have a mechanism for active partitioning and is more likely to remain in the larger mother cell if the cells do enter the sporulation pathway and divide asymmetrically . Therefore , the ability of ICEBs1 to delay the initiation of sporulation under conditions where the element is activated could help prevent loss of the element and maintain ICEBs1 in host cells . Mobile genetic elements employ various strategies to promote their maintenance during sporulation . Rates of curing during sporulation for various plasmids in Bacillus species vary widely and do not necessarily correlate with their stability during normal cell division ( Tokuda et al . , 1993; Turgeon et al . , 2008 ) . Mechanisms encoded by plasmids to promote their stability during growth and sporulation include the production of dynamic cytoskeletal filaments ( Becker et al . , 2006 ) and post-segregational killing of plasmid-cured pre-spores with toxin-antitoxin systems ( Short et al . , 2015 ) . Interestingly , even lytic phage genomes can be incorporated into spores ( first described in the 1960 s ) by co-opting the host’s chromosomal partitioning system ( Meijer et al . , 2005 ) . Mobile genetic elements , especially ICEs , are widespread in bacteria ( Frost et al . , 2005; Guglielmini et al . , 2011 ) . Many known mobile genetic elements encode cargo genes that confer easily recognizable phenotypes , notably antibiotic resistance . Other cargo genes provide less obvious phenotypes but still fundamentally alter the physiology of the host cell . A large ( 500 kb ) ICE was discovered in Mesorhizobium loti because its horizontal transfer conferred the ability to form nitrogen-fixing symbiotic rood nodules on Lotus plant species ( Sullivan and Ronson , 1998 ) . In many pathogens , cargo genes in mobile elements are largely responsible for virulence . For example , Vibrio cholerae is capable of a pathogenic lifestyle in human hosts due to the toxin-encoding phage CTXΦ ( Waldor and Mekalanos , 1996 ) . In the sporulating pathogen Bacillus anthracis , mobile genetic elements regulate both virulence and host development . Two plasmids , pXO1 and pXO2 , provide the genes for toxin synthesis and production of a protective capsule , respectively ( Green et al . , 1985; Mikesell et al . , 1983 ) . pXO1 also contains a regulatory gene , atxA , that regulates virulence factor production and inhibits host cell sporulation ( Dale et al . , 2018 ) . Co-regulation of virulence factors and sporulation is likely important during infection , as B . anthracis spores are thought to be more susceptible than vegetative cells to eradication by the immune system ( Mock and Fouet , 2001 ) . Mobile elements are also known to alter the host’s interaction with other horizontally acquired DNA , which has implications for the fitness and evolvability of the host . For example , the plasmid pBS32 in B . subtilis encodes an inhibitor of the host’s DNA uptake machinery , blocking natural transformation ( Konkol et al . , 2013 ) . Interestingly , genes with roles in defense against foreign DNA , CRISPR-Cas systems , are also identified within mobile elements ( Faure et al . , 2019; McDonald et al . , 2019; Millen et al . , 2012 ) . Competition between mobile elements not only shapes the repertoire of cargo genes in a given cell , but it may also protect the host from harmful elements . Many mobile genetic elements have been identified bioinformatically from genome sequences or discovered by means other than the phenotypes they provide ( Bi et al . , 2012; Guglielmini et al . , 2011; Johnson and Grossman , 2015 ) . Many elements lack obvious cargo genes , or at least lack cargo genes that have recognizable functions ( Cury et al . , 2017 ) . We suspect that many elements with uncharacterized cargo genes provide important traits to their hosts beyond the scope of the phenotypes currently attributed to mobile elements . Although mobile genetic elements can have remarkably broad host ranges , such as the Tn916-Tn1545 group of ICEs ( Clewell et al . , 1995; Roberts and Mullany , 2009 ) and the IncP-1 group of plasmids ( Popowska and Krawczyk-Balska , 2013 ) , cargo genes and their associated functions could be highly specific to certain hosts . Characterization of unknown cargo genes is likely to expand the diversity of traits currently attributed to mobile genetic elements . We speculate that many of these genes modulate normal host functions rather than provide entirely new phenotypes . Understanding cargo gene function is critical for understanding interactions between and co-evolution of mobile elements and their hosts .
Prior to competition experiments , cells were grown as light lawns for approximately 20 hr at room temperature on 1 . 5% agar plates containing 1% w/v glucose , 0 . 1% w/v monopotassium glutamate , and 1x Spizizen’s salts ( 2 g/l ( NH4 ) SO4 , 14 g/l K2HPO4 , 6 g/l KH2PO4 , 1 g/l Na3citrate-2H2O , and 0 . 2 g/l MgSO4-7H2O ) ( Harwood and Cutting , 1990 ) . Cells were resuspended from light lawns and grown at 37°C with shaking in S750 defined minimal medium ( Jaacks et al . , 1989 ) with 1% w/v glucose and 0 . 1% w/v monopotassium glutamate . Biofilms were grown at 30°C on MSgg agar plates ( as defined in Branda et al . , 2001 with the exception of tryptophan and phenylalanine , which we did not include ) . The sporulation medium used was DSM ( in liquid form or as plates solidified with 1 . 5% agar ) ( Harwood and Cutting , 1990 ) . MSgg agar and DSM agar plates were dried for 20–24 hr at 37°C prior to use . Antibiotics were used at the following concentrations for selection on LB agar plates: chloramphenicol ( 5 μg/ml ) , kanamycin ( 5 μg/ml ) , spectinomycin ( 100 μg/ml ) , tetracycline ( 12 . 5 μg/ml ) , and a combination of erythromycin ( 0 . 5 μg/ml ) and lincomycin ( 12 . 5 μg/ml ) to select for macrolide-lincosamide-streptogramin ( MLS ) resistance . The B . subtilis strains used are listed in Table 2 . The strain background used in all experiments was a derivative of the undomesticated strain NCIB3610 lacking its endogenous plasmid pBS32 ( NCIB3610 plasmid-free ) . ICEBs10 indicates the strain is cured of ICEBs1 . Standard techniques were used for cloning and strain construction ( Harwood and Cutting , 1990 ) . Some alleles were previously described and are summarized below . Variants of ICEBs1 that were blocked for transfer contained the conEK476E mutation derived from MMB1118 ( Berkmen et al . , 2010 ) . The spo0A∆Ps allele was derived from AG1242 ( Siranosian and Grossman , 1994 ) . The amyE::PspoIIA-lacZ cat allele was derived from KI938 ( Chung et al . , 1994 ) . Essentially identical alleles with PspoIIE and PspoIIG were also used . Cultures were started from resuspended light lawns ( described above ) diluted to an initial OD600 of 0 . 05 in S750 minimal medium . Cultures were grown to mid-exponential phase ( OD600 ~0 . 5 ) at 37°C with shaking . Cells were pelleted , resuspended in 1x Spizizen’s salts , and diluted to an OD600 of 0 . 01 . Donor and recipient strains were mixed at the indicated frequencies and 50 μl of the mixture was spotted onto the center of MSgg agar plates . Spots were allowed to dry at 30°C before flipping the plates . Plates were incubated at 30°C for 4 days . At the time of inoculation , the strain mixes were diluted and plated in duplicate on LB agar plates containing the appropriate antibiotics to determine the initial CFU/ml of the donor and recipient strains . After 4 days , the mature biofilms were scraped from the agar surface with sterile wooden sticks and resuspended in 5 ml 1x Spizizen’s salts , followed by mild sonication to disperse the cells . Cells were diluted and selectively plated to determine the final CFU/ml of transconjugants . Cells were prepared for competition experiments as described above for biofilm mating experiments . Strain mixtures at the indicated frequencies were spotted onto MSgg agar plates for biofilm competitions and DSM agar plates for sporulation competitions . Plates were incubated at 30°C for 4 days unless otherwise indicated . Biofilms/colonies were collected , disrupted , and plated as described above . For time-course competitions , two replicate biofilms/colonies were collected at each of the indicated times . Sporulation frequency was determined by selective plating before and after a heat treatment at 85°C for 20 min to enumerate total CFUs and CFUs derived from heat-resistant spores . Relative fitness of ICEBs1-containing cells over ICEBs10 cells was determined as ( pf/ ( 1-pf ) ) / ( pi/ ( 1-pi ) ) , where pf , pi are the frequencies of ICEBs1-containing cells in the final and initial populations , respectively . Control competitions between ICEBs1-cured cells were performed to determine the fitness associated with the lacA::{Pveg-mTagBFP mls} marker ( JMJ574 ) used in ICEBs1-containing cells relative to the lacA::{Ppen-mApple2 kan} marker used in ICEBs1-null cells ( JMJ550 ) . When the mls-marked cells were started at a frequency of 0 . 01 , their relative fitness was 0 . 7 ± 0 . 09 ( average and standard deviation from three independent experiments and a total of 9 biofilms ) . Expression of sporulation genes was measured in cultures grown from single colonies in liquid DSM at 37° with shaking . Cells were harvested at the indicated timepoints . For β-galactosidase assays , growth was stopped by the addition of toluene ( ~1 . 5% final concentration ) . β-galactosidase specific activity ( [∆A420 per minute per ml of culture per OD600] x 1000 ) was measured as described ( Miller , 1972 ) after pelleting cell debris . Biofilm gene expression was measured in cultures grown from single colonies in liquid MSgg at 37° with shaking . For RT-qPCR assays , cells were harvested directly into ice-cold methanol ( 1:1 methanol to culture volume ) and pelleted . RNA was isolated using Qiagen RNeasy PLUS kit with 10 mg/ml lysozyme . iScript Supermix ( Bio-Rad ) was used for reverse transcriptase reactions to generate cDNA . Control reactions without reverse transcriptase were performed to assess the amount of DNA present in the RNA samples . RNA was degraded by adding 75% vol of 0 . 1 M NaOH and incubating at 70°C for 10 min , followed by neutralization with an equal volume of 0 . 1 M HCl . qPCR was done using SSoAdvanced SYBR master mix and CFX96 Touch Real-Time PCR system ( Bio-Rad ) . Primers used to measure epsB were oJJ363 ( 5’-CGGAACAATATCGCACCATTC-3’ ) and oJJ364 ( 5’-CGCTGCACTGAACGATTTAC-3’ ) . Primers used to quantify tasA were oJJ367 ( 5’-GGATCACTTGCGATCAAAGAAG-3’ ) and oJJ368 ( 5’-CTTCAAACTGGCTGAGGAAATC-3’ ) . Primers used to measure the control locus gyrA were oMEA128 ( 5’-TGGAGCATTACCTTGACCATC-3’ ) and oMEA129 ( 5’-AGCTCTCGCTTCTGCTTTAC-3’ ) . The relative transcript copy numbers ( as indicated by the Cp values measured by qPCR ) of epsB and tasA were normalized to gyrA after subtracting the signal from control reactions without reverse transcriptase . | Many bacteria can ‘have sex’ – that is , they can share their genetic information and trade off segments of DNA . While these mobile genetic elements can be parasites that use the resources of their host to make more of themselves , some carry useful genes which , for example , help bacteria to fight off antibiotics . Integrative and conjugative elements ( or ICEs ) are a type of mobile segments that normally stay inside the genetic information of their bacterial host but can sometimes replicate and be pumped out to another cell . ICEBs1 for instance , is an element found in the common soil bacterium Bacillus subtilis . Scientists know that ICEBs1 can rapidly spread in biofilms – the slimly , crowded communities where bacteria live tightly connected – but it is still unclear whether it helps or hinders its hosts . Using genetic manipulations and tracking the survival of different groups of cells , Jones et al . show that carrying ICEBs1 confers an advantage under many conditions . When B . subtilis forms biofilms , the presence of the devI gene in ICEBs1 helps the cells to delay the production of the costly mucus that keeps bacteria together , allowing the organisms to ‘cheat’ for a little while and benefit from the tight-knit community without contributing to it . As nutrients become scarce in biofilms , the gene also allows the bacteria to grow for longer before they start to form spores – the dormant bacterial form that can weather difficult conditions . Mobile elements can carry genes that make bacteria resistant to antibiotics , harmful to humans , or able to use new food sources; they could even be used to artificially introduce genes of interest in these cells . The work by Jones et al . helps to understand the way these elements influence the fate of their host , providing insight into how they could be harnessed for the benefit of human health . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2021 | A mobile genetic element increases bacterial host fitness by manipulating development |
Craniofacial characteristics are highly informative for clinical geneticists when diagnosing genetic diseases . As a first step towards the high-throughput diagnosis of ultra-rare developmental diseases we introduce an automatic approach that implements recent developments in computer vision . This algorithm extracts phenotypic information from ordinary non-clinical photographs and , using machine learning , models human facial dysmorphisms in a multidimensional 'Clinical Face Phenotype Space' . The space locates patients in the context of known syndromes and thereby facilitates the generation of diagnostic hypotheses . Consequently , the approach will aid clinicians by greatly narrowing ( by 27 . 6-fold ) the search space of potential diagnoses for patients with suspected developmental disorders . Furthermore , this Clinical Face Phenotype Space allows the clustering of patients by phenotype even when no known syndrome diagnosis exists , thereby aiding disease identification . We demonstrate that this approach provides a novel method for inferring causative genetic variants from clinical sequencing data through functional genetic pathway comparisons .
Genetic disorders affect almost 8% of people ( Baird et al . , 1988 ) , about a third of whom will have symptoms that greatly reduce their quality of life . While there are over 7000 known inherited disorders , only a minority of patients with a suspected developmental disorder receive a clinical , let alone a genetic , diagnosis ( Hart and Hart , 2009 ) . A genetic diagnosis allows more specific therapeutic interventions to be investigated and can aid the identification of primary vs secondary symptoms . The introduction of whole genome and exome sequencing into modern clinical medicine will be instrumental in raising the current low rate of genetic diagnoses for ultra-rare diseases . Nevertheless , tools to accurately assign functional and disease relevance to sequence variants are substantially lacking . Projects that apply next generation sequencing to patients in clinical settings fail to report genetic diagnoses for approximately 80% of cases ( de Ligt et al . , 2012 ) . The difficulty lies in identifying the causal variant in an individual patient: even when ignoring experimental error , each individual carries approximately 4 million differences , in the case of whole genome sequencing , relative to the reference genome . Computational analyses currently are able only to interpret the ∼2500 variants that alter protein sequence at evolutionarily conserved positions and ∼400 very rare variants that are likely to be causal for pathogenic processes ( Abecasis et al . , 2012 ) . Notably , of the ∼10% of the genome that is functional all except the 1 . 2% that is protein-coding is often disregarded ( Weischenfeldt et al . , 2013 ) . Therefore , the prediction of causal inherited variants in an individual can result in high false positive and high false negative rates . The most powerful approach to associate a particular gene with an ultra-rare disease is to identify multiple unrelated individuals with the disorder whose genomes harbor deleterious alleles in a shared gene , regulatory element or pathway ( Schuurs-Hoeijmakers et al . , 2012 ) . However , this approach relies on at least two individuals with the same disorder being available for comparison , an unlikely event given that these two individuals are selected for comparison from the roughly 100 million people affected by rare developmental disorders ( prevalence of less than 2 per 100 , 000 around the world ) ( Orphanet , 2013 ) . For the past 65 years , clinical geneticists have studied , diagnosed , and characterized developmental disorders on the basis of common characteristics among patients ( Rimoin and Hirschhorn , 2004 ) . When a given causal variant is ultra-rare , however , this presents substantial difficulties . Consequently , to realize the full potential of next generation sequencing in clinical diagnostics , phenotypic characterization must also become correspondingly high throughput and sensitive ( Hennekam and Biesecker , 2012 ) . The facial gestalt provides valuable information to identify similarities between patients because 30–40% of genetic disorders manifest craniofacial abnormalities ( Hart and Hart , 2009 ) . The utility of computer vision for diagnosis and phenotyping of dysmorphic disorders has been explored previously by several groups and with varying approaches ( Loos et al . , 2003; Hammond et al . , 2005; Hammond , 2007; Boehringer et al . , 2006; Dalal and Phadke , 2007; Vollmar et al . , 2008; Boehringer et al . , 2011 , reviewed in Hammond and Suttie , 2012; Baynam et al . , 2013 ) . The computational analysis of facial morphology using 3D imaging has been applied to conditions such as fetal alcohol syndrome ( Suttie et al . , 2013 ) , schizophrenia ( Buckley et al . , 2005; Hennessy et al . , 2006 , 2007 ) and autism ( Aldridge et al . , 2011 ) . While 3D imaging studies have shown high discriminatory power in terms of classification they have relied on specialized imaging equipment and patient cooperation . Previous work with 2D images has relied on manual annotation of images , controlling lighting , pose and expression to allow consistent analyses . These factors greatly limit the availability , and ultimately the potential widespread clinical utility of such approaches . We have adopted a complementary approach that takes advantage of the wealth of data available for human faces , an indirect result of the ubiquitous availability of cameras . To do so we provide a new representation ( 'Clinical Face Phenotype Space' ) , which is an application of computer vision and machine learning algorithms for analyzing craniofacial dysmorphisms from ordinary photographs . We have ensured that Clinical Face Phenotype Space is robust to spurious variations such as lighting , pose , and image quality which would otherwise bias analyses . The approach is fully automated and provides objective and consistent computational descriptions of facial gestalt . Our method both greatly narrows the search space for investigating known disorders and will increase the power of inferring causative variants in previously unidentified genetic disease .
We first collected a database of 2878 images , including 1515 healthy controls and 1363 pictures for eight known developmental disorders from publically available sources across the internet ( Table 1 , references for image sources are available from Supplementary file 1 ) . Manual checks were performed to exclude images where the face or an eye was not clearly visible , or where an expert clinician ( DRF ) could not verify the diagnosis . Manual annotation of facial features points was performed on all images to allow training and testing of an automated annotation algorithm . These initial requirements for manual intervention are dispensed with in the final automatic algorithm ( see below ) . 10 . 7554/eLife . 02020 . 006Table 1 . Composition of the databaseDOI:http://dx . doi . org/10 . 7554/eLife . 02020 . 006SyndromeNr imagesSyndromeNr imagesPublic images onlinePublished images Angelman205 PACS12 Apert203 BRAF35 Cornelia de Lange179 CFC1 Down199 Costello10 Fragile X164 ERF5 Progeria78 HRAS5 Treacher Collins103 KRAS12 Williams-Beuren232 MAP2K15 MAP2K24 Controls1515 MEK15 NRAS2 22q118 PTPN1119 Marfan18 RAF19 Sotos36 SHOC28 Turner12 SOS130The Gorlin Collection Aarskog19 Klippel-Trenaunay10 Achondroplasia12 Langer-Giedion14 Alagille8 Larsen11 Albright7 Lenz_Majewski17 Angelman13 Lymphedema-Lymphangiectasia-MR8 Apert49 Melnick_Needles17 Beckwith-Wiedemann11 Moebius9 Bloom9 Muenke15 BOF15 Myotonicdystrophy9 Cartilagehair13 Neurofibromatosis7 CHARGE12 Noonan29 Cherubism20 OAVdysplasia18 CleidoCranialdysostosis13 ODD21 Coffin-Lowry20 OFCD10 Costello9 OFD18 CriduChat17 OPD31 Crouzon16 Osteopetrosis2 Crouzonodermoskeletal5 Osteosclerosis5 Cutislaxa11 Otodental2 DeLange17 Poland4 Diastrophicdysplasia5 Prader–Willi16 Down8 Progeria14 Dubowitz12 Proteus6 Dyggve-Melchior-Clausen8 Rieger4 EEC6 Rothmund-Thomson13 Ehlers-Danlos17 Rubinstein-Taybi8 Ellis-vanCreveld3 Saethre-Chotzen25 FG11 Sclerosteosis4 FragileX27 SeckelMOD7 Frontometaphysealdysplasia12 SEDcongenita6 Gorlin91 Sotos16 Gorlin_Chaudry_Moss13 Stickler42 Greig7 TRP24 Hallermann-Streiff9 Waardenburg39 Incontinentiapigmenti4 Weaver13 Kabuki25 Williams-Beuren19 Klippel-Feil3 We proceeded to train a computer vision algorithm for automatic annotation of 36 feature points of interest across the face ( Figure 1A ) . Our approach takes advantage of a variety of facial detection algorithms ( OpenCV [Bradski , 2000] , Viola Jones [Viola and Jones , 2001] and Everingham [Everingham et al . , 2009] ) and custom learning ( consensus of exemplars [Belhumeur et al . , 2011] ) to accurately place feature points on a given face ( ‘Materials and methods’ ) . Across all images in our database , manual checking found that our algorithm detected and annotated 99 . 5% of tested faces correctly with accuracies in the range 6–60% of the width of an eye ( individual feature point accuracies are provided in Figure 1—figure supplement 1 ) . We used an Active Appearance Model ( 'Materials and methods' ) to calculate an average face within any set of images , representing consistent shape and appearance features within the group ( Figure 1B and animated morphs in Figure 2 ) . The average faces for each syndrome show that the algorithm effectively captures characteristic features of dysmorphic syndromes ( Figure 2—figure supplement 1 ) . For each feature point , the algorithm extracts a feature vector describing appearance of the surrounding patch . The algorithm then constructs a feature vector describing shape based on the relative pairwise distances between all feature points ( 'Materials and methods' ) . We next sought to compare the syndrome relevant information content of the feature descriptors to previous studies ( Hammond et al . , 2005; Boehringer et al . , 2006; Hammond , 2007; Vollmar et al . , 2008 ) . We found that classification analysis based on support vector machines provided similar accuracies to previous work , despite disparities in image variability ( average classification accuracy 94 . 4% , see Figure 4—figure supplement 1 , Figure 4—figure supplement 2 and 'Materials and methods' ) . 10 . 7554/eLife . 02020 . 008Figure 2 . Animated morphs of average faces from controls to syndromes . ( A ) Angelman , ( B ) Apert , ( C ) Cornelia de Lange , ( D ) Down , ( E ) Fragile X , ( F ) Progeria , ( G ) Treacher-Collins , ( H ) Williams-Beuren . Delineation of syndrome gestalt relative to controls with distortion graphs in Figure 2—figure supplement 1 . DOI:http://dx . doi . org/10 . 7554/eLife . 02020 . 00810 . 7554/eLife . 02020 . 009Figure 2—figure supplement 1 . Distortion graphs representing the characteristic deformation of syndrome faces relative to the average control face . Each line reflects whether the distance is extended or contracted compared with the control face . White—the distance is similar to controls , blue—shorter relative to controls , and red—extended in patients relative to controls . DOI:http://dx . doi . org/10 . 7554/eLife . 02020 . 009 It is important to emphasize that the analyzed images vary greatly , as there were minimal restrictions imposed on image selection placed by the two exclusion criteria ( both eyes visible and diagnosis verified by DRF ) . Photos were analyzed irrespective of the subject's age , gender , facial expression or ethnicity or the background scenery . Principal component analysis ( PCA ) of facial descriptor vectors illustrates that the main sources of variation among images are indeed lighting , pose , and facial expression , rather than phenotypic features ( Figure 1—figure supplement 2 ) . We next performed Metric Learning using a Large Margin Nearest Neighbor ( Weinberger and Saul , 2009 ) approach for the eight syndromes in the database . This approach linearly transformed the multidimensional space of PCA feature vectors to optimize the separation of syndromes: dimensions informative for dysmorphism phenotypes are expanded while uninformative dimensions are compressed ( thus changing the relative importance for clustering ) . We denote the resulting transformed 270 dimensional space as 'Clinical Face Phenotype Space' ( see 'Materials and methods' ) . Due to its design , Clinical Face Phenotype Space clusters patient faces based on diagnostically relevant phenotypic features , while tolerating spurious variation . Relative importance of spurious and phenotypic variation for clustering in Clinical Face Phenotype Space was tested using simulated faces ( 'Materials and methods' ) . For these faces feature dimensions that reflected known spurious variation such as lighting and head orientation were compressed and hence were of less relevance for clustering ( Figure 1—figure supplement 2 ) . For the eight syndromes with which Clinical Face Phenotype Space was created , we performed tests with supervised learning and clustering . A kNN-classifier applied within Clinical Face Phenotype Space was able to correctly classify images with an accuracy of 99 . 5% using the leave-one-out method . However , to avoid biases introduced by training data size , we also assessed the improvements in clustering by measuring the search space reduction ( hereafter referred to as the Clustering Improvement Factor or CIF , 'Materials and methods' ) . This estimates the factor by which the Clinical Face Phenotype Space improves the clustering of syndromes when compared with random chance ( to 95% confidence ) . On average , the clustering of the eight syndromes within the database was improved by 11 . 0-fold ( geometric mean of improved clustering , CIF range 9 . 1–23 . 5 , maximum possible mean 12 . 5; Figure 3 ) . 10 . 7554/eLife . 02020 . 010Figure 3 . Clinical Face Phenotype Space enhances the separation of different dysmorphic syndromes . The graph shows a two dimensional representation of the full Clinical Face Phenotype Space , with links to the 10 nearest neighbors of each photo ( circle ) and photos placed with force-directed graphing . The Clustering Improvement Factor ( CIF , fold better clustering than random expectation ) estimate for each of the syndromes is shown along the periphery . DOI:http://dx . doi . org/10 . 7554/eLife . 02020 . 010 Next , we tested and confirmed our hypothesis that Clinical Face Phenotype Space could be generalized to dysmorphic syndromes that were not used in the training . We had access to 75 syndromes from the Gorlin collection ( a kind gift curated and annotated by Professor Raoul Hennekam , Academic Medical Center , University of Amsterdam ) , which we supplemented with additional images of 22q11 , Marfan and Sotos syndromes . Furthermore , we collected images of patients with verified genetic mutations in PACS1 or in specific genes from the RAS/MEK pathway ( Supplementary file 1 references for image sources in 'Materials and methods' ) . The number of individuals within each syndrome varied between 2 and 223 . The search space reduction was on average 27 . 6-fold better than random chance ( CIF range 1 . 0–700 . 0 , maximum possible average CIF was 150 . 0; Figure 4A ) . That is to say , that among 2754 patients' faces associated with any of 90 syndromes Clinical Face Phenotype Space makes it 27 . 6-fold easier to make the correct diagnosis . This demonstrates that Clinical Face Phenotype Space is an effective approach to the identification of multiple individuals sharing ultra-rare , previously undocumented , genetic disorders . 10 . 7554/eLife . 02020 . 011Figure 4 . Clinical Face Phenotype Space is generalizable to dysmorphic syndromes that are absent from a training set . ( A ) Clustering Improvement Factor ( CIF ) estimates are plotted vs the number of individuals per syndrome grouping in the Gorlin collection or patients with similar genetic variant diagnoses . As expected , the stochastic variance in CIF is inversely proportional to the number of individuals available for sampling . The median CIF across all groups is 27 . 6-fold over what is expected by clustering syndromes randomly . That is to say , the CIF of a randomly placed set is 1 . The maximum CIF is fixed by the total number of images in the database and by the cardinality of a syndrome set: the theoretical maximal CIF upper bound is plotted as a red dotted line . The CIF for the minimum and maximum , Cutislaxa syndrome and Otodental syndrome , were 1 . 0 and 700 . 0 respectively . ( B ) Average probabilistic classification accuracies of each individual face placed in Clinical Face Phenotype Space ( class prioritization by 20 nearest neighbors weighted by prevalence in the database ) . The 8 initial syndromes used to train Clinical Face Phenotype Space are shown in color . For syndromes with fewer than 50 examples , accuracies were averaged across all syndromes binned by data set size ( i . e . , the average accuracy is shown for syndromes with 2–5 , 6–10 , 11–25 , and 26–50 images in the database , Supplementary file 1 ) . Classification accuracies increase proportional to the number of individuals with the syndrome present in the database . Accuracies using support vector machines with binary and forced choice classifications are shown in Figure 4—figure supplement 1 and Figure 4—figure supplement 2 . A simulation example of probabilistic querying of Clinical Face Phenotype Space is shown in Figure 4—figure supplement 3 . DOI:http://dx . doi . org/10 . 7554/eLife . 02020 . 01110 . 7554/eLife . 02020 . 012Figure 4—figure supplement 1 . SVM binary classification accuracies among the 8 syndromes in Table 1 . SVM classifier accuracies when tuned for equal false positive and false negative error rates . DOI:http://dx . doi . org/10 . 7554/eLife . 02020 . 01210 . 7554/eLife . 02020 . 013Figure 4—figure supplement 2 . SVM forced choice classification accuracies among the 8 syndromes in Table 1 . DOI:http://dx . doi . org/10 . 7554/eLife . 02020 . 01310 . 7554/eLife . 02020 . 014Figure 4—figure supplement 3 . Simulated example illustrating the Clustering Improvement Factor . A random scattering of 100 points in 2 dimensions is used as a background set ( black circles with white fill ) . The 20 red plus symbols ( within the red shaded area ) are a random set of points lying within the same limits as the background set and have a CIF of 0 . 9 . This is the actual degree of clustering of the red points with respect to the expectation of clustering them with 95% confidence ( E ( r ) = 5 . 6 ) . The filled green circles ( within the green shaded area ) are the red points shifted by +0 . 5 units in each dimension and have a CIF of 2 . 7 . The black points ( within the gray shaded area ) are the red plus symbol positions scaled by 0 . 5 and then shifted by +1 . 5 units in dimension 1 . The black points are non-overlapping with the background and represent the maximal CIF ( of 5 . 6 ) in this example . DOI:http://dx . doi . org/10 . 7554/eLife . 02020 . 01410 . 7554/eLife . 02020 . 015Figure 4—figure supplement 4 . Simulated example of probabilistic querying of Clinical Face Phenotype Space . ( A ) Visualization of a population of simulated faces in the first two Multi-Dimensional Scaling ( MDS ) modes . 7 classes of points ( simulated 'syndrome groups' ) are shown with different distributions and variances . A central 'query' face is indicated by the boxed cross . The 20 nearest neighbors of the query are encircled with a black border . ( B ) Inset bar graph shows diagnosis hypothesis ranked by class priority . The class priority ranking weights the dispersion and prevalence ( spread and number ) of a class in the Clinical Face Phenotype Space with the nearest neighbors to assign the most probable diagnosis hypotheses . In the example , the ranked diagnosis estimates of the query point would be class 7 , then class 6 , and thirdly class 4 . The scatter plot shows the individual similarity p0p1 estimates , reflecting their relative closeness in the space as compared to local neighborhood , for the 20 nearest neighbors of the query . The first nearest neighbor is estimated to be 2 . 6-fold closer to the query than the average based on the local density of neighbors . The dotted line indicates the average relative distance between points among the 20 nearest neighbors . ( C ) Inset bar graph shows the number of neighbors of the query per class . A scatterplot of dispersion vs cardinality , i . e . relative spread of points and what proportion of the total number of points belong to that class in the simulated space . Plots ( B ) and ( C ) allow objective assessment of the distribution of points shown in ( A ) , and aid the interpretation of classification confidence . DOI:http://dx . doi . org/10 . 7554/eLife . 02020 . 015 We proceeded to test if Clinical Face Phenotype Space recapitulates the modularity of genetic diseases , where clusters of phenotypically similar disorders reflect functional relationships among the genes involved ( see Oti and Brunner , 2007 for a review ) . We have shown that individuals with the same underlying genetic disease automatically cluster in Clinical Face Phenotype Space . We next tested whether disorders caused by mutations in different genes result in meaningful clusters in Clinical Face Phenotype Space . We selected disorders with a known genetic origin , using either gene associations from OMIM or publications describing the identification of causative genes ( see 'Materials and methods' ) . For each pair of genes , the shortest path in a protein–protein interaction network was obtained from Dapple ( Rossin et al . , 2011 ) , giving a protein interaction distance relevant to that gene pair . We compared genes underlying monogenic syndromes linked by 1 , 2 , or 3 path distances , with those with a path distance of 4 or that was unknown; unknown distances are those where no genes are associated with a syndrome , the syndrome is multigenic , or when DAPPLE has no known interaction documented , see 'Materials and methods' . For each pair of syndromes , an average Euclidean distance in Clinical Face Phenotype Space was calculated . The distance in Clinical Face Phenotype Space is significantly shorter between syndromes associated to genes with protein interaction distances of 1 , 2 , or 3 compared with syndromes with 4 or no known interactions ( p< 0 . 01 , p< 0 . 05 and p< 0 . 001 respectively , Figure 5 ) . This demonstrates that the distance in Clinical Face Phenotype Space partly recapitulates the functional relatedness of underlying developmental processes known to be disrupted in genetic diseases . 10 . 7554/eLife . 02020 . 016Figure 5 . Clinical Face Phenotype Space recapitulates features of functional gene links between syndromes . Protein–protein interaction distances of 1–3 for genetically characterized syndromes are associated with significantly shorter Euclidean distance ( arbitrary units ) between syndromes in Clinical Face Phenotype Space as compared to syndromes with distance 4 or no known interaction distance ( shown in orange ) ( Kruskal–Wallis tests with Bonferroni corrected p-values indicated as *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . The Spearman correlation across all distances was r = 0 . 09 , p<0 . 001 . The numbers of pairwise syndrome comparisons underlying each of the interaction distances are listed within the respective boxes . DOI:http://dx . doi . org/10 . 7554/eLife . 02020 . 016 Clinical Face Phenotype Space can provide clinical phenotyping and clustering to known genetic disorders that is objective and high-throughput . The method is , however , neither sufficiently accurate nor intended to determine diagnosis , yet it can help to narrow the diagnostic search space in an unprejudiced manner . A clinician could easily photograph a patient and immediately obtain clinically useful diagnostic hypotheses and matching cases . To this end , we implemented two primary methods to automatically and objectively query Clinical Face Phenotype Space . For any given image located in Clinical Face Phenotype Space , we obtain confidence ranked classifications to known disorders ( see 'Materials and methods' and Figure 4—figure supplement 4 ) . In addition , we objectively compare the image to others within the space . For any given query image , a probabilistic ranking of similar syndromes is obtained through nearest neighbor representation compared to random expectation of clustering among the 90 syndromes and 2754 faces . The classification confidence for a particular disorder depends on its location within the space , but also on the local densities of similar faces . We find that for the eight initial syndromes used to construct Clinical Face Phenotype Space , 93 . 1% ( range 81 . 0–99 . 2% ) are correctly classified as the top rank , cumulatively converging on 99 . 1% ( 95 . 8–100% ) by the 20th rank ( Figure 4B ) . Of syndromes not part of the Clinical Face Phenotype Space training , the classification accuracies positively correlated strongly with the number of instances in the database ( Figure 4B ) . For the 20 syndromes where the database held 5 or fewer examples ( Table 1 ) , we classify on average 20 . 3% correctly by the 6th rank ( exceeding 16 . 3-fold better than by chance alone ) . For individuals with a suspected ultra-rare or an undocumented novel disorder , we developed a metric , p0p1 , which assesses their similarity to others within Clinical Face Phenotype Space . The metric estimates the relative closeness of two faces given an average local density with the space: a p0p1 value exceeding 1 indicates a potentially new cluster , see 'Materials and methods' . The 2 PACS1 cases reported by Schuurs-Hoeijmakers et al . ( 2012 ) placed within Clinical Face Phenotype Space have a p0p1 value of 1 . 05 meaning that they are 5% closer to one another than the geometric mean of the distances to their 20 nearest neighbors . Taking into account that this is a local density estimate among 2754 faces in Clinical Face Phenotype Space , the search space to find them has been reduced ∼690 . 4-fold ( CIF , see 'Materials and methods' ) . The combination of syndrome clustering and de novo similarity metrics should aid the diagnosis of known syndromes and provides a means of clustering patients where no documented diagnosis exists .
We have developed our algorithm on normal-everyday 2D photographs and have focused on 36 facial feature points . Given the orders of magnitude lower dimensionality of our data as compared to a 3D imaging capture ( Hammond et al . , 2005 ) , we were initially concerned that this would be insufficient to capture facial phenotypes . However , we then demonstrated that the approach is able to describe and discriminate between syndromes with a comparable accuracy to previous studies ( Loos et al . , 2003; Hammond et al . , 2005; Hammond , 2007; Boehringer et al . , 2006; Dalal and Phadke , 2007; Vollmar et al . , 2008; Boehringer et al . , 2011 ) . The accessibility of normal 2D photographs ( as opposed to 3D imaging ) should outweigh any lower data resolution obtained from any one image and in future developments using multiple profile perspectives will allow 3D structure to be inferred . With accurate registration of a person's face from multiple images across time , from a family photo album for instance , it would capture not only the 3D structure but also the progression and development of dysmorphic gestalt . The automatic image analysis algorithm enables phenotypic metrics to be obtained with objective consistency from each image ( Figure 1 ) . Clinical Face Phenotype Space was instantiated using eight syndromes that were well populated in our database so as to be robust against spurious variation . In doing so , it has become a generalizable model for craniofacial dysmorphic variation ( Figure 5 ) . The high fidelity of the current Clinical Face Phenotype Space ( Figure 3 ) shows promise given that known deficiencies have yet to be addressed: ( 1 ) We used only single image examples of individuals . ( 2 ) The spectrum of phenotypes represented was limited . ( 3 ) The average image quality in the database was low . ( 4 ) The current 36 facial feature points only capture full frontal facial phenotypes , and thus miss valuable information from the full cranium and profile perspectives . Among the approaches that will be tested in future work are: increasing the number of feature points across the cranium , using profile images and taking advantage of multiple images of the same individual . Furthermore , we will be exploring performing explicit modelingmodeling of the 3D variation for 2D images ( Ramnath et al . , 2008 ) , other types of feature descriptors , alternative metric learning and dimensionality reduction approaches ( Simonyan et al . , 2013 ) . As Clinical Face Phenotype Space is developed and populated with more individuals , the predictive power to infer novel causative genetics would be expected to increase linearly until it asymptotically approaches a theoretical maximum . There are three anticipated primary applications for Clinical Face Phenotype Space in a clinical setting: narrowing the search space for documented developmental disorders , identifying multiple people that share an ultra-rare genetic disorders and aiding the inference of causative variants in clinical genetic sequencing ( Figure 4 ) . We envisage Clinical Face Phenotype Space becoming a standard tool to support clinical genetic counseling . Since any normal 2D image can be analyzed , this approach is available to any clinician worldwide with access to a camera and a computer . This can also reduce the need for patient inconvenience in a clinical setting because a family photo album could provide the required image ( s ) . A photograph will enable automatic digital phenotyping , and its placement in Clinical Face Phenotype Space will provide an unbiased list of candidate clinical hypotheses ( exemplified in Figure 6 ) . We anticipate that future developments of Clinical Face Phenotype Space will also identify sub-phenotypes or comorbidities . Where no known genetic disease or variant can be assigned , Clinical Face Phenotype Space can identify other patients with phenotypic similarities empowering the identification of ultra-rare genetic disorders . 10 . 7554/eLife . 02020 . 017Figure 6 . Class priority of diagnostic classifications for images . The full computer vision algorithm and Clinical Face Phenotype Space analysis procedure with diagnostic hypothesis generation exemplified by: ( A ) a patient ( Ferrero et al . , 2007 ) with Williams-Beuren . ( B ) Abraham Lincoln . The former US President is thought to have had a marfanoid disorder , if not Marfan syndrome ( Gordon , 1962; Sotos , 2012 ) . Bar graphs show class prioritization of diagnostic hypotheses determined by 20 nearest neighbors weighted by prevalence in the database . As expected , the classification of Marfan is not successfully assigned in the first instance as there were only 18 faces of individuals with Marfan in the database ( making this an example of a difficult case with the current database ) . However , the seventh suggestion is Marfan , despite this being among 90 different syndromes and 2754 faces . DOI:http://dx . doi . org/10 . 7554/eLife . 02020 . 017© 2007 Elsevier Masson SAS . All rights reserved2007Elsevier Masson SASThe patient figure in Figure 6 , part A is reproduced from Ferrero et al . , ( 2007 ) , European Journal of Medical Genetics with permission . In summary , we have presented an algorithmic approach that provides a critical advance in applying computer vision and machine learning techniques as a tool for clinical geneticists . The conjunction of a computer vision and machine learning algorithm with Clinical Face Phenotype Space makes this approach high-throughput , automatic , objective , and broadly accessible with existing digital photography and computers . Our ongoing research has begun to apply the Clinical Face Phenotype Space approach within large clinical sequencing collaborations . Computer vision for aiding diagnosis of developmental disorders in clinical genetics will be tenable and broadly applicable in the near future .
We built a database of publically available or scientifically published pictures of patients collected across the internet . We collected 100–283 images per syndrome for Angelman , Apert , Cornelia de Lange , Down , Fragile X , Progeria , Treacher Collins , or Williams-Beuren . Images were collected through publically available resources online accessible though search terms relating to each syndrome , primarily through support group pages and awareness event photographs . Source URLs were converted to shortened versions for the purposes of publication using TinyUrl ( http://tinyurl . com/ ) ( Supplementary file 1 ) . The links provided are expected to decay with time and should only be considered exemplars of database composition . Images were captured through screen shots and saved as PNG or JPEG file formats . The following two exclusion criteria were applied to the images: 1 . The face needed to be clearly visible and oriented so that both eyes were visible . 2 . The correct diagnosis was confirmed by an expert clinician ( DRF ) . DRF inspected each image to validate the supposed syndrome diagnosis; images not validated were discarded . Variation in lighting , image resolution , pose , and occlusions has only been restricted when it obscures the facial characteristics ( such as a hand covering the face ) . We also sought to avoid multiple images of the same individual at the same age in the database . No further restrictions were placed on variations in pose , facial expression , lighting , occlusions , or image quality . In the same manner , smaller numbers of images were collected for Marfan , 22q11 , Turner and Sotos syndromes ( Table 1 ) . Furthermore , we collected further published images of patients with confirmed genetic variants in genes of the RAS/MEK signaling pathways as well as in PACS1 ( Rauen , 1993 , 2006; Bertola et al . , 2007; Gripp et al . , 2007; Makita et al . , 2007; Rauen , 2007; Zampino et al . , 2007; Nystrom et al . , 2008; Schulz et al . , 2008; Tidyman and Rauen , 2008; Cordeddu et al . , 2009; Kratz et al . , 2009; Zenker , 2009; Allanson et al . , 2010; Wright and Kerr , 2010; Kleefstra et al . , 2011; Lepri et al . , 2011; Siegel et al . , 2011; Schuurs-Hoeijmakers et al . , 2012; Hopper et al . , 2013; Twigg et al . , 2013 ) . 3100 images were collected and manually annotated for training of the algorithms . Of these 2878 were successfully annotated by the automatic pipeline and are reported in the database counts of Table 1 . Original database , excluding the Gorlin collection , and previously published images ( which are available from the cited original publications ) can be requested by contacting CN ( christoffer . nellaker@dpag . ox . ac . uk; Ferry Q , Steinberg J , Webber C , FitzPatrick DR , Ponting CP , Zisserman A , Nellåker C , 2014 , Diagnostically relevant facial gestalt information from ordinary photos database ) . Requests will be assessed by a Data Access Committee ( DAC ) comprised of CPP , DRF , AZ , CN and Dr Zameel Cader of the Division of Clinical Neurology , University of Oxford . The DAC will make data available to researchers in good standing with the relevant institution and funding agencies ( i . e . , no known sanctions ) . The data are provided without copyright . Pipeline code was written in python 2 . 7 and uses the module Ruffus ( Goodstadt , 2010 ) for task management . The code is available through an open source MIT license at https://github . com/ChristofferNellaker/Clinical_Face_Phenotype_Space_Pipeline . The manner and method by which images were collected from publically available sources and stored were acceptable research practices and do not require special consent from a Research Ethics Committee . Advice from legal services , research ethics board members and the Information Commissioner's Office ( UK ) was sought in arriving at this conclusion . The computer vision algorithm analyses a 2D photograph for the location of a face , annotates the facial landmark points , and extracts feature vectors for subsequent machine learning applications . MATLAB ( MATLAB . R2011b Natick , Massachusetts: the MathWorks Inc . ) with OpenCV ( Bradski , 2000 ) was used to write scripts and functions for the algorithm . To identify a putative face in the photo , we used previously published algorithms ( Viola and Jones , 2001 ) . Within a box bounding the face , a pictorial structure model was used to identify 9 central facial feature points ( Everingham et al . , 2009 ) , which then were used to initialize the placement of an additional 27 feature points . The resulting facial mesh structure was fitted to the image using Active Appearance Models ( AAMs ) ( Cootes et al . , 1998 ) to generate average face visualizations ( Figure 2—figure supplement 1 ) . The placement of the 36 feature points was also further refined using custom written scripts based on consensus of exemplars ( Belhumeur et al . , 2011 ) ( see Methods ) . From the fitted constellation of facial landmarks two feature vectors were extracted . ( 1 ) The appearance as a concatenation of the pixel intensities of patches around the 9 inner facial feature points . ( 2 ) The shape vector was constructed as the normalized pairwise distances between all 36 facial feature points . Each image was converted to JPEG and submitted to the Facial Detection ( FD ) module of the algorithm . Face detection was achieved using the OpenCV ( Bradski , 2000 ) implementation of the Viola–Jones object detection framework ( using Haar like features and a cascade of classifiers ) ( Viola and Jones , 2001 ) . The output takes the form of a square bounding box delimiting the area of the picture where the face was found . Pictures containing the faces of healthy relatives ( or others ) were either discarded or cropped to only conserve regions with the patient face . Manual annotation of the 36 feature points was performed on 3100 of the images in the image collection . These were used as the ground truth reference point for all subsequent training and test sets for evaluations of automated facial landmark annotation accuracies ( Figure 1—figure supplement 1 ) . In the second step of the automatic algorithm detected faces were passed to a facial landmark annotation script ( Everingham et al . , 2009 ) ( FLA module ) , which annotates the face with an initial set of 9 well-characterized ( salient ) feature points . The 9 landmarks in that set were the medial and lateral canthi of the eyes , each subnasale , columella and the left and right vermilion border lateral midpoints . The FLA used the returned bounding box to approximate the location and size of the face to be annotated . Automatic annotation relies on a generative model of feature point position combined with a discriminative model of appearances . This joint model was based on the parts-based pictorial structure representation introduced by Fischler and Elschlager ( 1973 ) . For a given bounding box , the FLA module returns both a constellation of 9 landmarks and a corresponding confidence index computed via appearance mismatch with the model . We used this index to implement more robust , accurate and reliable annotation approaches ( Figure 1—figure supplement 1 ) as described below . Improved facial landmark annotation was performed with a custom script designed to refine the output from the FD-FLA modules in terms of annotation inaccuracies , false positive and false negative face detection . The images were transformed iteratively by mirror imaging , partial rotations ( ±45° ) , and by adding a frame around the image to produce 100 transformations of the original . Each image was subsequently submitted through the FD -FLA modules and returned a constellation of 9 points with associated confidence scores recorded . A consensus map is constructed by confidence weighted averaging of high confidence feature annotations thus reducing the number of spurious annotations and increasing annotation accuracy ( Figure 1—figure supplement 1 ) . From the improved 9 facial feature points , we expanded the feature detection to 36 feature points encompassing the points indicated in Figure 1—figure supplement 1 . We developed a computational module inspired by Belhumeur et al . ( 2011 ) to determine the localization of the 36 landmarks . Consensus of exemplar ( CoE ) relies on part base classifiers used to localize potential candidate points and a database of face exemplars used to introduce a shape prior in the search for the best constellation . While only a constellation of 9 feature points ( C9 ) is required to compute the appearance feature vector , the shape feature vector relies on a constellation of 36 feature points ( C36 ) covering the inner face in greater details along with its outlined ( Jaw line ) . Anatomical landmarks covered in C36 are shown in Figure 1—figure supplement 1 . We used the C9 obtained via the improved FLA module ( see previous section ) to initiate the automatic search for C36 . For each of the facial feature points in C36 , we delimited a region of interest ( ROI ) for the algorithm to consider using the following heuristic: 1000 exemplar faces are sample from among the controls in the database ( Table 1 ) . For each exemplar face i , we registered the C9i to the C9 of our query face using Procrustes algorithm . Next the sum of squares error was used to sort exemplars in order of accuracy with which C9i registered to C9 . The top 20 exemplars were used to map their C36i to the query face using the transform Ti obtained during registration . A consensus C36 for the query face was then derived by averaging all registered C36i . Finally , for each feature point in C36 , we defined a square ROI centered on the consensus point with two palpebral fissures ( PF ) length dimensions . The PF length was the average between right and left eyes and estimated based on C9 . The final C36 was derived from the ROIs by using a combination of part based detectors and a consensus of shape exemplars . Support Vector Machine ( SVM ) classifiers ( using Gaussian kernels ) were trained to recognize each feature point in C36 . We used a database of manually annotated control patients to obtain positive and negative training sets for each part classifier . The feature vector associated with a particular feature point was obtained by cropping a square patch centered on this point and describing its pixel content with a pyramidal histogram of gradients ( PHOG ) . Going back to the query face , each of the 36 ROIs was submitted to the corresponding part detector . From this we obtained a set of 20 potential candidates ( PC ) for each part within the ROI ( where candidates were sorted based on the classifier decision values ) . Next , we randomly sampled exemplars of C36 from the control database and registered them to the query face in order to enforce a shape prior . To avoid spurious outlier PC to drive the registration off , we randomly select PC to represent three randomly selected points from C36 . Exemplar C36i were registered via Procrustes algorithm to the query face using only these three PC . The registered C36i were scored by submitting its feature points to the part classifiers . We retained the top 20 C36i . Finally , each feature point of the final C36 for the query was derived independently as a consensus between probability density maps based on the PCs and the classifier decision values over the corresponding ROI . Using the set of 9 inner points annotated to the face by the FLA module and refined by the CoE module , four additional points were generated: left and right center of the eye , nasion ( between the eyes ) and center of the mouth . The points were then used to register the face , via an affine transformation T , to a canonical set of corresponding points ( face shape template ) . Next , circular patches were generated around the canonical points which were mapped back to the original picture using the inverse of T . This process creates 13 ellipses which were then used to crop the image content by bilinear interpolation . Extraction of the patch was performed as in Everingham et al . ( 2009 ) . The appearance feature vector was obtained by a concatenation of the pixel gray-scale content of the 13 cropped patches ( size of the feature vector = 1937 ) . While patch content could be further processed before concatenation ( gradients , HOG , PHOG , SIFT , Gabor Filter , Local binary Pattern , etc ) the gain in terms of discriminative power and relevant feature extracted was negligible compared to the computational cost in terms of memory consumption . The features vector was built from the constellation of the 36 landmarks annotations from the CoE module . We described the face shape as a vector d , the set of pair-wise distances through the constellation , resulting in a feature vector with 630 elements . To compare the different constellations , each was registered via Procrustes transformation to the average constellation of the AAM model ( canonical face mesh , see below ) . The vector of pair-wise distances was then normalized so that any distance variations were relative to the corresponding distances measured on the average control face template . Introduced in 1998 by Cootes et al . ( 1998 ) , the active appearance model ( AAM ) was designed to identify a set of facial landmarks on a given face . This task is achieved by iteratively modifying both the shape and the position ( location ) of a structured face mesh in which nodes represent target landmarks . We constructed a training set by manual annotation of 3100 patient images with the 36 landmarks ( Figure 1—figure supplement 1 ) . All constellations were registered together using an iterative Procrustes algorithm . We computed the average constellation and used it as a reference with which to build a canonical face mesh via Delaunay triangulation . Based on the obtained triangulation , we generated a face mesh dividing each patient's face into sub-regions . Using piecewise affine warping , we independently mapped the pixel content of each sub-region to the corresponding triangle in the canonical face mesh . We thus obtained a registered version of each patient's facial appearance . Shapes and appearance registrations were used in a principal component analysis ( PCA ) to generate both the shape and appearance models for use with the AAM . Using the AAM statistical models of shape and appearance that derive from the training of the AAM , we created a visual representation of canonical traits/phenotypes . Average faces were created from the images of patient with the same genetic disorder . AAM models were generated for each group , and after registration of the constellations ( annotation ) to each we derived the average shape constellation . A face mesh was generated from these constellations by triangulation ( Delaunay ) and the appearance of each individual was mapped to the average face mesh by piecewise affine warping . Thus , this average face mesh recapitulated the canonical phenotype of the syndromes . Furthermore , we created morphs between control and syndrome faces by registering appearance to the average shape of the controls ( Figure 2—figure supplement 1 ) . By computing the average across the syndrome group and the control group one can obtain the average appearance of the syndrome A0s and control A0c . Given that both A0s and A0c have identical dimensionality , we morphed appearance from one to the other with appearance for frame k where Ak = A0s + k/K* ( A0c − A0s ) and where K is the total number of frames . Similarly , we considered the average face mesh of the syndrome group M0s and the control group M0c to obtain the face mesh at frame k as Mk = M0s + k/K* ( M0c − M0s ) . Finally , for each frame k , both shape ( face mesh ) and appearance were combined by piecewise affine warping of Ak to Mk . We used SVMs to fine tune both appearance and shape descriptors . Differences in binary classification accuracies allow us to infer relative feature delineation capacity by our descriptors . We employed Libsvm to perform classification . The accuracies based on the raw feature vectors were comparable to accuracies reported in previous studies ( Boehringer et al . , 2006 , 2011 ) . SVM classifiers were trained on both shape and appearance feature vectors separately for each 36 pairwise combination of control and 8 syndromes ( Table 1 ) . Each binary classification was repeated 10 times with randomly generated positive/negative training and test sets with a 4:1 ratio . The linear kernel SVM classification accuracies using the final shape and appearance descriptors are shown in Figure 4—figure supplement 1 . We estimate a total accuracy when fusing information from shape and appearance feature vectors based SVMs by summing decision values returned by the shape and appearance classifiers respectively . In all classification experiments , we used both original and mirror versions of each image . Given a binary experiment to distinguish group G1 from G2 ( where G refers to a set of images of a syndrome ) , we randomly partitioned the sets into training and test sets: G1tr , G1te , G2tr , and G2te respectively . From these we generated the corresponding mirror image sets: G1trm , G1tem , G2trm , and G2tem . The SVM model was trained using G1tr + G1trm as the positive set and G2tr + G2trm as the negative set . Decision values were tuned to give equal error rates ( the number of false positives = number of false negatives ) . With the trained SVM models , we next submitted the sets G1te + G2te and G1tem + G2tem separately to the classifier . Thus , for each instance in the test set we obtained two decision values . The final classification was determined by the sum of the decision values . We next applied the classification problem to assigning a face as being one of the 8 syndromes ( Table 1 ) . Each of the 8 syndrome groups was randomly split into training and test sets with a ratio of 4:1 . We used the training sets to train a linear binary SVM classifier for each of the 28 pairs . Each image in the training set was submitted to all 28 classifiers . The decision value was returned by each classified and used as a probabilistic estimate P for the test instance to belong to the positive class . Thus , after presenting instance i to the binary classifier distinguishing syndrome j ( positive ) from syndrome k ( negative ) , we assigned i a vote of weight P for syndrome j and a vote of weight 1-P for syndrome k . After summation of the votes from the 28 classifiers , the instance i was labeled as belonging to the syndrome diagnosis with the highest probability . The confusion matrix averaged from 10 repeats of the forced choice experiment is shown in Figure 4—figure supplement 2 . We performed PCA on the shape and appearance feature vectors to reduce dimensionality from 2567 to a concatenation of 340 orthogonal vectors . This was then used to transform the feature space with Large Margin Nearest Neighbor ( LMNN , Weinberger and Saul , 2009 ) . LMNN is an optimization algorithm that uses a training set of pairs of vector labels ( xi; li ) and learns a Mahalanobis distance that maximizes the kNN classification over a training set . Note that even if the system only considers local information ( i . e . , number of intruders for each instance ) the final metric is global . The Mahalanobis distance was computed as dist ( xi; xj ) = ( xi − xj ) tLtL ( xi − xj ) where L is a linear matrix . It is equivalent to the Euclidean distance taken in the space after transformation by L . That is to say , LMNN linearly transforms dimensions in feature space to maximize the margins separating classes of labeled instances . This should , in principle expand dimensions with phenotypically relevant information and compresses dimensions uninformative for classification . To validate the characteristics of the transformation of spurious and phenotypic vectors in Clinical Face Phenotype Space , we performed a series of experiments based on projected 3D faces . We used the 3D facial model proposed by Blanz & Vetter ( Blanz , 2006 ) http://faces . cs . unibas . ch/bfm/ which allowed us to create faces with a direct control over shape , appearance , lighting , and facial pose . We synthesized 5 test faces at random moving along the first 15 components of both the shape and appearance models . For each face , we generated a set of 20 images for each combination of 5 head rotations and 4 lighting conditions . We use these simulated images to compare similarity measures in the raw feature space and in Clinical Face Phenotype Space ( Figure 3 , Figure 1—figure supplement 2 ) . We performed a reorientation of the raw feature vectors and Clinical Face Phenotype Space using PCA without dimensionality reduction in order to sort the dimensions by variation magnitude . This allowed us to assess the relative contributions of phenotypic variation and spurious variations to clustering of faces . The strongest influences on clustering would be expected to be encoded in the first modes of variation . Placing the synthetic faces in the reoriented spaces allowed us to describe the PCA signatures of phenotypic variations ( shape , appearance ) and spurious variation ( lighting , head pose ) . We used several dimensionality reduction methods and metrics for visualization and estimation of the properties of Clinical Face Phenotype Space . We developed an estimate of search space reduction to determine the improvements in clustering in Clinical Face Phenotype Space controlling for the composition of the database . Essentially , this calculates the degree to which intruders in a nearest neighbor search between instances of the same syndrome are excluded in Clinical Face Phenotype Space . This equates to a factor estimate of increased clustering , CIF ( details of the procedure are provided below ) . We used a 20 nearest neighbor linkage map to visualize Clinical Face Phenotype Space using force directed graphs implemented through Gephi ( Bastian et al . , 2009 ) . Protein–protein interaction data were obtained from DAPPLE ( Rossin et al . , 2011 ) . After conversion to Ensembl gene IDs , 126 , 586 interactions between 10 , 442 genes remained . We considered the data as a network of genes , with edges denoting an interaction . The shortest paths between two genes were computed using Dijkstra's algorithm ( Dijkstra , 1959 ) . We calculated the median pairwise Euclidean distance between syndromes in Clinical Face Phenotype Space . The correlation between these two data sets underlies Figure 5 . Clinical Face Phenotype Space distance between groups was tested using Kruskal–Wallis ( Kruskal and Wallis , 1952 ) tests with Bonferroni ( Bonferroni , 1935 , 1936 ) multiple testing correction . Next , we performed estimations of clustering of syndromes in face space . Initial tests using kNN-classifiers showed that the classification accuracies were heavily dependent on spread and cardinality of the syndrome in the database . We went on to develop an estimate of search space reduction , hereafter referred to Clustering Improvement Factor ( CIF ) , to determine the improvements in clustering in Clinical Face Phenotype Space controlling for the composition of the database ( a simulated example is provided in Figure 4—figure supplement 3 ) . We considered a syndrome with Np positive and Nn negative instances in the Clinical Face Phenotype Space . We defined the CIF asCIF=expected rank ( r ) of nearest positive match under random rankingobserved average rank ( r ) of nearest positive match=E ( r ) O ( r ) with the average taken across all instances of the syndrome . O ( r ) was calculated from the observations in the Clinical Face Phenotype Space . To compute E ( r ) , we used probability theory as follows . Under a random ranking for a given positive query , the other Np−1 positive instances are each placed independently among the Nn negative instances , with a uniform discrete probability distribution . We defined the random variable Ni as the number of negative instances ranked higher than the first positive instance , so Ni takes integer values 0≤Ni≤Nn . For a given positive query , the expected rank of the nearest positive match is the expected value of Ni+1 , denoted by E ( Ni ) +1 . To calculate E ( Ni ) , we used the definition of expectation:E ( Ni ) =∑j=0NnjPr ( Ni=j ) since Ni can only take non-negative integer values , for each possible value j between 0 and Nn , Pr ( Ni=j ) =Pr ( Ni≥j ) −Pr ( Ni≥j+1 ) Substituting this in the formula for E ( Ni ) , E ( Ni ) =∑j=0Nnj[Pr ( Ni≥j ) −Pr ( Ni≥j+1 ) ] . Rewriting the sum , E ( Ni ) =∑j=1Nn[jPr ( Ni≥j ) − ( j−1 ) Pr ( Ni≥j ) ]=∑j=1NnPr ( Ni≥j ) For a given number j , Pr ( Ni ≥ j ) is the probability that all positive instances were placed after j negative instances . For any given individual positive instance , such placement has probability 1−jNn+1 . Since placement of all positive instances is independent , this gives Pr ( Ni≥j ) = ( 1−jNn+1 ) Np−1 . Therefore , E ( Ni ) =∑j=1NnPr ( Ni≥j ) =∑j=1Nn ( 1−jNn+1 ) Np−1 Finally , this gives E ( r ) =1+E ( Ni ) =1+∑j=1Nn ( 1−jNn+1 ) Np−1 . We developed two methods to retrieve information about the neighborhood of a given face placed in the Clinical Face Phenotype Space . Firstly , we assigned a syndrome classification based on the identity of its k nearest neighbors in Clinical Face Phenotype Space . Based on the neighbors' labels a list of syndromes to which the new face could belong was created . The number of neighbors supporting each hypothesis was compared with the probability to see N instances of that syndrome when sampling k from the population of faces in Clinical Face Phenotype Space . Secondly , we estimate the relative similarity between specific faces given the density of points in a local region of Clinical Face Phenotype Space . This is calculated as p0p1=d0 , 1/d0d1 , where d0 , 1 is the similarity measure between the query and its neighbor , d0 is the average of similarities between the query and k = 20 neighbors and d1 is the average of similarities between the neighbor of the query and k of the neighbor's neighbors . Figure 4—figure supplement 4 illustrates the method and metrics using a simulated example . We see Clinical Face Phenotype Space as a means to facilitate collaborative investigations of genetic diseases between clinicians . Of course , sharing of data raises questions regarding ethics approval and data security . These questions are tightly linked to the debate of how clinical sequencing information should be treated in global health care systems . We anticipate that it would be suitable for future implementations of Clinical Face Phenotype Space to follow similar guidelines as for clinical sequencing data . | Rare genetic disorders affect around 8% of people , many of whom live with symptoms that greatly reduce their quality of life . Genetic diagnoses can provide doctors with information that cannot be obtained by assessing clinical symptoms , and this allows them to select more suitable treatments for patients . However , only a minority of patients currently receive a genetic diagnosis . Alterations in the face and skull are present in 30–40% of genetic disorders , and these alterations can help doctors to identify certain disorders , such as Down’s syndrome or Fragile X . Extending this approach , Ferry et al . trained a computer-based model to identify the patterns of facial abnormalities associated with different genetic disorders . The model compares data extracted from a photograph of the patient’s face with data on the facial characteristics of 91 disorders , and then provides a list of the most likely diagnoses for that individual . The model used 36 points to describe the space , including 7 for the jaw , 6 for the mouth , 7 for the nose , 8 for the eyes and 8 for the brow . This approach of Ferry et al . has three advantages . First , it provides clinicians with information that can aid their diagnosis of a rare genetic disorder . Second , it can narrow down the range of possible disorders for patients who have the same ultra-rare disorder , even if that disorder is currently unknown . Third , it can identify groups of patients who can have their genomes sequenced in order to identify the genetic variants that are associated with specific disorders . The work by Ferry et al . lays out the basic principles for automated approaches to analyze the shape of the face and skull . The next challenge is to integrate photos with genetic data for use in clinical settings . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"medicine",
"computational",
"and",
"systems",
"biology"
] | 2014 | Diagnostically relevant facial gestalt information from ordinary photos |
Analysis of the spatial distribution of endomembrane trafficking is fundamental to understand the mechanisms controlling cellular dynamics , cell homeostasy , and cell interaction with its external environment in normal and pathological situations . We present a semi-parametric framework to quantitatively analyze and visualize the spatio-temporal distribution of intracellular events from different conditions . From the spatial coordinates of intracellular features such as segmented subcellular structures or vesicle trajectories , QuantEv automatically estimates weighted densities that are easy to interpret and performs a comprehensive statistical analysis from distribution distances . We apply this approach to study the spatio-temporal distribution of moving Rab6 fluorescently labeled membranes with respect to their direction of movement in crossbow- and disk-shaped cells . We also investigate the position of the generating hub of Rab11-positive membranes and the effect of actin disruption on Rab11 trafficking in coordination with cell shape .
Modern light microscopy associated with fluorescence molecule tagging allows studying the spatial distribution of intracellular events . Unfortunately , fluorescent images are complex to analyze and additional software is needed to evaluate statistical differences between different conditions ( Meijering et al . , 2016; Tinevez et al . , 2017 ) . Automatic methods have the obvious advantage of being quicker and reproducible . However , most computational methods are based on the complex combination of heterogeneous features such as statistical , geometrical , morphological and frequency properties ( Peng , 2008 ) , which makes it difficult to draw definitive biological conclusions . Additionally , most experimental designs , especially at single-cell level , pool together data coming from replicated experiments of a given condition ( Schauer et al . , 2010; Merouane et al . , 2015; Biot et al . , 2016 ) , neglecting the biological variability between individual cells . Micro-patterning is now a well-established strategy to reduce morphological variability by imposing constraints on adhesion sites , which has been shown to influence the cytoskeleton geometry and transport carrier localization ( Théry et al . , 2005; Schauer et al . , 2010 ) . This technique opened the way to pairwise comparisons of conditions with a two-sample kernel density-based test by pooling together all data from each condition ( Duong et al . , 2012 ) . Unfortunately , it does not consider the sample-to-sample variability because all replicated experiments from a given condition are simply merged together . Additionally , the visualization of the kernel density maps enables to average several experiments but fails to identify specific locations of interest in the cell ( e . g . docking areas ) . Finally , assessing the dynamical behavior of labeled membrane structures , a fundamental task for trafficking analysis , remains out of scope in this framework . In this paper , we describe a method that we call QuantEv dedicated to the analysis of the spatial distribution of intracellular events represented by any static or dynamical descriptor ( e . g . detected points , segmented regions , trajectories . . . ) provided that the descriptors are associated with spatial coordinates . QuantEv offers a unifying frame to decipher complex trafficking experiments at the scale of the whole cell . It is typically able to detect subtle global molecular mechanisms when trajectory clustering fails . An overview of the approach is presented in Figure 1 . Our approach first computes 3D histograms of descriptors in a cylindrical coordinate system ( parameterized by radius r , angle θ and depth z ) with computational cell shape normalization , enabling comparisons between cells of different shape . Densities are obtained via adaptive kernel density estimation ( Silverman , 1986; Taylor , 2008 ) . Visualization through histograms and densities allows giving a clear biological interpretation of the experiments . We use the Earth Mover’s Distance ( Rubner et al . , 2000 ) and the Circular Earth Mover’s Distance ( Rabin et al . , 2011 ) to measure the dissimilarity between densities associated with different experimental conditions . A statistical analysis of these distances reliably takes into account the biological variability over replicated experiments . By computing weighted densities for each point in the cell as the reference center , QuantEv identifies the point that gives the most uniform angular distribution . This point may coincide with a biological structure of interest that would act as the events emitter or attractor . In the section Results , we describe the application of QuantEv to detect significant differences between molecular trafficking and phenotypes observed in cells with various shapes . The first application is concerned with the distribution of membranes labeled by GFP-Rab6 as a hallmark of vesicular carriers in unconstrained , crossbow- and disk-shaped cells . Rab6 proteins are transiently anchored to moving transport carriers from the Golgi apparatus located at the cell center to Endoplasmic Reticulum entry sites or to plasma membrane ( White et al . , 1999; Chavrier and Goud , 1999; Echard et al . , 2000; Opdam et al . , 2000; Grigoriev et al . , 2007; Bardin et al . , 2015 ) , both assumed to be located at the cell periphery . Cell shape imposes constraints on the cytoskeleton and consequently influences the spatial distribution of Rab6 transport carriers , as confirmed with kernel density maps ( Schauer et al . , 2010 ) . We apply QuantEv to visualize and quantify this influence and to localize regions in the cell associated with Rab6 trafficking stages . In addition , Rab6 positive membranes were reported to move from and toward the Golgi in apparent close proportions ( Grigoriev et al . , 2007 , Grigoriev et al . , 2011 ) , and yet these membrane associated proteins are believed to traffic in majority from the Golgi located at the cell center to the cell periphery ( White et al . , 1999; Chavrier and Goud , 1999; Echard et al . , 2000; Opdam et al . , 2000; Grigoriev et al . , 2007 ) where they should dissociate from membranes and recycle back to the cytosol . To investigate these apparently antagonist statements , we apply QuantEv on Rab6 trajectories to characterize the dynamical behaviors of these transport carriers . The second application focuses on the dynamics of mCherry-Rab11-positive membranes . Rab11 is known to be primarily localized to the Endosomal Recycling Compartment ( ERC ) , and it organizes spatially and temporally recycling from this compartment ( Ullrich et al . , 1996; Gidon et al . , 2012; Baetz and Goldenring , 2013; Boulanger et al . , 2014 ) . Here , we confirm by using QuantEv the hypothesis that the labeled transport intermediates are uniformly distributed around the ERC at the plasma membrane plane . Furthermore , we also investigate the progressive effect of actin disruption induced by Latrunculin A injection on the ERC localization with respect to time . We finally apply QuantEv to analyze the joined influence of actin disruption and cell shape on the radial distribution of Rab11 vesicles trafficking .
We applied the QuantEv approach to visualize the spatial distribution of Rab6-positive membranes in unconstrained , crossbow- and disk-shaped cells ( see Appendix 1—figure 1 ) and quantify their differences . To test the generic performance of QuantEv , these image sequences were acquired with two different 3D imaging modalities , a multi-point confocal microscopy and a wide field video microscopy . We compared the results obtained with QuantEv to those obtained with the more conventional kernel density ( KD ) maps ( Schauer et al . , 2010; Merouane et al . , 2015 ) . The KD approach concludes that the distribution of Rab6-positive membranes are clearly different between cells of different shapes ( see Figure 2a–c , p value = 0 when considering unconstrained cells versus crossbow-shaped cells , unconstrained cells versus disk-shaped cells , and crossbow-shaped cells versus disk-shaped cells ) . Unfortunately , it also leads to a significant difference when image sequences with the same cell shape are compared ( see Figure 2d ) . This demonstrates that the KD approach is too sensitive . Instead , QuantEv shows a uniform range of p values when cells with same shape are compared ( see Figure 2d ) while it leads to significant differences between radial , angular and in-depth distributions of Rab6 proteins from cells with different cell shapes ( see Figure 2e–f ) . The angular distribution of Rab6 proteins is different for the three cell shapes . It ranges from a completely uniform distribution for disk-shaped cells , to a less regular distribution for unconstrained cells and to a distribution oriented toward the three tips of the crossbow for crossbow-shaped cells . In-depth and radial distributions are similar for crossbow- and disk-shaped cells . In contrast , they are different from unconstrained cells . Unconstrained cells show diverse sizes with a strong tendency to spread . This explains why the in-depth distribution is flatter for the unconstrained cells than for the constrained cells . Interestingly , QuantEv is able to reflect these differences . QuantEv also highlights a distribution maximum for a radius at the two-thirds ( resp . five-sixth ) the distance between the Golgi region border and the cell periphery for both micro-patterns ( resp . unconstrained cells ) ( see Figure 2f–i ) . These maxima correspond to an accumulation of Rab6-positive membranes and identify the area where they enter a docking phase before switching to a tethering phase . The localization difference for these maxima between constrained and unconstrained cells is explained by a smaller adhesion area without micro-patterns , pushing the docking phase for vesicles closer to the cell periphery . Both radial and angular distributions unraveled by QuantEv represent a measurement of the environment constraints undergone by living cells . Rab6-positive membranes are trafficking from the Golgi located at the cell center to the cell periphery ( White et al . , 1999; Chavrier and Goud , 1999; Echard et al . , 2000; Opdam et al . , 2000; Grigoriev et al . , 2007 ) and at the same time move from and toward the Golgi in comparable proportions ( Grigoriev et al . , 2007; Grigoriev et al . , 2011 ) . To reconcile these two antagonist statements , we applied QuantEv as follows . Rab6 trajectories were classified into two categories ( Figure 3a–c ) : ( i ) vesicles moving toward the cell periphery; ( ii ) vesicles moving toward the Golgi . As shown in Figure 3d–f , the proportion of Rab6-positive membranes moving toward the cell periphery and toward the Golgi are close ( 0 . 531 versus 0 . 469 for unconstrained cells , 0 . 497 versus 0 . 503 for crossbow-shaped cells , 0 . 521 versus 0 . 479 for disk-shaped cells ) . However , the radial distributions shown in Figure 3d–f display two distinctive modes for vesicles moving toward the cell periphery and those moving toward the Golgi ( p value = 0 . 0002 for unconstrained cells , p value = 0 . 021 for crossbow-shaped cells , p value = 0 . 0008 for disk-shaped cells ) . Between the Golgi and the distribution maxima shown in Figure 2f , Rab6 vesicles are predominantly moving toward the cell periphery . Between these maxima and the cell periphery , they are in majority moving toward the Golgi , indicating that during their docking-tethering phase , the vesicles are predominantly moving toward the cell center . These two distinctive dynamical behaviors are consistent with the aforementioned antagonist statements . To go further in the analysis , we looked at the confinement ratio ( Figure 4a ) , the total path length and the lifetime of Rab6 trajectories , conventional dynamical measures used for particle tracking analysis . The combination of these measures with spatial localization is of high interest ( Applegate et al . , 2011; Tinevez et al . , 2017 ) and QuantEv provides a good framework to quantitatively analyze and visualize the distribution of these dynamical measures with respect to their intracellular localization . We focus on the radial distribution of Rab6 trajectories from unconstrained and constrained cells as the differences between trajectories moving toward cell periphery and trajectories moving toward Golgi lie in these distributions ( see Figure 3d–f ) . Rab6-positive membranes moving toward the cell periphery have a much more direct path than the ones moving toward the Golgi , except near the cell periphery ( see Figure 4b , e ) . Consistently , Rab6 positive membranes moving toward the Golgi have longer total path length and lifetime than the ones moving toward the cell periphery , especially when approaching the cell periphery ( see Figure 4c–e ) . In summary , this analysis clearly demonstrated that Rab6-positive membranes move predominantly and quite directly from the Golgi to the cell periphery until they enter a docking phase . Then , they mostly go back toward the cell center by following long and indirect trajectories . Rab11-positive recycling membranes originate their journey from the so-called endosomal recycling compartment ( ERC ) . We formulate the assumption that Rab11 positive membranes are uniformly distributed at the membrane plane around the ERC position within the cell , whatever the cell shape is . To test this hypothesis , we used images acquired at the membrane with TIRF microscopy showing Rab11 proteins ( see Appendix 1—figure 1 c–d ) . Most labeled membranes of the ERC are not located near the cell surface . However , for each TIRF sequence , one highly inclined wide field image was also acquired , enabling to visually define its location ( red disks in Figure 5a ) . To test our assumption , the QuantEv uniformity analysis is applied by considering intensity on segmented regions . The results are shown in Figure 5a ( blue disks ) . To have a line of comparison , we also plot the cell centers as green disks in Figure 5a . Interestingly , the blue disk is close to the red disk for all image sequences except one ( second line , middle image in Figure 5a ) . The blue disk is also closer to the red disk than the green disk in seven out of eight image sequences ( see Figure 5a–b ) . Although the point that gives the most uniform angular distribution does not strictly coincide with the manually identified ERC , it is sufficiently close to indicate that the Rab11-positive membranes are quite uniformly distributed around the ERC position at the membrane plane whatever the cell shape is . This indicates that the ERC corresponds to the organizing hub of the Rab11 carrier vesicles . Applying the QuantEv uniformity analysis at each time step of a sequence allows studying the location stability of the particle emitter or attractor . To test if the estimated ERC location is stationary over time , we computed the Euclidean distance between the reference point estimated at time t = 0 and the points estimated for the next frames . In untreated cells , this distance remains stable ( see Figure 6 green line ) . We analyzed cells treated with Latrunculin A , which inhibits actin polymerization ( see Appendix 1—figure 1 e–f ) . We show that the ERC location is moving away as the drug is affecting the cell ( see Figure 6 blue and orange lines ) , enlightening the role of cytoskeleton in stabilizing the cellular localization of the ERC . We then acquired image sequences of unconstrained , crossbow-shaped and disk-shaped cells at 10 and 15 min after Latrunculin A addition , and we extracted Rab11 trajectories . The confinement ratio of Rab11 tracks is decreasing with time ( see Figure 7 ) , which is consistent with actin cytoskeleton being involved in Rab11 vesicle trafficking , as already reported ( Schafer et al . , 2014 ) . The radial distribution of Rab11 vesicles is constantly shifting from the cell periphery to the cell center for unconstrained , crossbow- and disk-shaped cells ( see Figure 8a ) . However , before and at drug injection time , we observe significant differences in radial distributions between the three tested conditions ( p value = 0 . 0023 , see Figure 8b ) . After Latrunculin A treatment , we progressively observe no difference between the radial distributions , as the actin organization is drastically perturbed . Together , these quantifications allow us to conclude that exocytosis/recycling vesicle trafficking is dependent on both cell shape and actin organization .
This article presents a computational framework taking into account cell variability to quantify the distribution of fluorescently labeled proteins . Using dynamical descriptors , detailed insight into dynamical processes is also unraveled and the uniformity analysis allows to localize an organizing region for the observed biological objects . Additionally to the input image , the user has to define three other inputs that depend on the biological application . First , the user has to decide which coordinate system to use . If the imaged cells are flat as in this study ( see Appendix 1—figure 1a–c ) , a cylindrical coordinate system is well suited while a spherical coordinate system will fit better rounded cells . If the user is not familiar with cylindrical or spherical coordinate systems , a classical Cartesian system is also available , even though less suited to intracellular spatial distribution . Finally , QuantEv also allows to analyze the spatial distribution with respect to a reference point or to membrane borders ( Heride et al . , 2010 ) . Once the reference coordinate system is chosen , the user has to define a reference point , typically the particle emitter or attractor , and a reference direction in order to fairly compare cells . For example , in this study , the direction between the Golgi and the cell center were used to define a reference direction for unconstrained and disk-shaped cells while the crossbow principal axis was used for crossbow-shaped cells . As intensity is proportional to the amount of proteins in fluorescence microscopy , using intensity observed in segmented areas is potentially more informative than binary segmentation masks . However , because of phenomena such as photobleaching , phototoxicity , shading , uneven illumination etc . , appropriate normalization procedures within and between images need to be applied . If the user is able to correct for these phenomena , it is preferable to use intensity as weights in QuantEv analysis . Otherwise , intensity weights should be avoided . Given its genericity , QuantEv can easily be applied to any intracellular event and gives useful insights about their spatial distribution across conditions . From these observations , the user can then apply more sophisticated analyses such as mechanistic models of dynamics ( Ponti et al . , 2004; Jaqaman et al . , 2008 ) or generative models ( Li et al . , 2012; Johnson et al . , 2015a , Johnson et al . , 2015b ) . QuantEv analysis conclusions can also be the starting point of a new modeling . We demonstrate with the help of QuantEv that the distributions of Rab6-positive membranes from unconstrained , crossbow- and disk-shaped cells are statistically different . QuantEv also enables to identify the locations where Rab6-positive membranes enter their docking phase . By considering the directions of the moving Rab6-positive membranes , QuantEv allows demonstrating that these membranes first move predominantly and directly toward the cell periphery before reaching their docking phase . They then go back to the cell center in an undirected and long fashion . This intriguing result showing statistically bi-directional movements of Rab6 was reported before . The Rab6-positive vesicles generated at the Golgi membranes are predestined to the cell periphery , in order to deliver their exocytic cargo ( Grigoriev et al . , 2007; Grigoriev et al . , 2011 ) , which should favor a centrifuge directionality . Our data reconciles this two apparently opposed observations and show for the first time , that a majority of Rab6 vesicles reverses their movement only toward close docking-fusion sites and only during this ultimate phase of docking-fusion . QuantEv also demonstrates that Rab11 positive membranes are uniformly distributed around the ERC at the plasma membrane plane . This shows that the ERC represents an organizing hub for the Rab11 carrier vesicles . By applying the QuantEv uniformity analysis along time , we exhibit how the ERC location is affected by actin disruption caused by Latrunculin A injection . The radial distribution analysis of Rab11-positive membranes combined with Latrunculin A injection reveals a dual regulation by cell shape and actin organization on Rab11 trafficking at the plasma membrane , and more generally on the exocytosis/recycling vesicle distribution . In conclusion , QuantEv has the potential to become a very popular analysis method for dynamics and intracellular event analysis as ( i ) it is publicly available; ( ii ) it is fully automated and semi-parametric; ( iii ) it provides results that are easy to biologically interpret; ( iv ) it performs a statistical analysis that takes into account the biological variability over the replicated experiments of a same condition and is efficient with small and large amounts of data . On a biological prospective beyond the two particular models presented here , QuantEv will be of great interest for studies where quantitative and statistical analysis of intracellular membrane or particle behaviors are required , depending on physical and external constraints . For instance , in single-cell experiments performed in microfluidics devices , QuantEv will efficiently provide automation and diversity of statistical analyses in ‘one shot’ , for a relatively small amount of data . Applying QuantEv in multi-cellular systems , in which cell-cell constraints necessarily affect molecular distribution and particle movements will also be of great interest . Finally , in vivo imaging of single-cell intracellular processes in a very confined and constrained environment will benefit from the generic aspect of the QuantEv sensing and measuring of particle spatial distributions , dynamical measures with respect to intracellular localization and cell to cell variability . An Icy plugin and a tutorial are available at http://icy . bioimageanalysis . org/plugin/QuantEv . A QuantEv analysis module is available on TrackMate and a QuantEv track processor is available in Icy .
In the first dataset , we use cell lines stably expressing fluorescently tagged proteins in order to minimize the cell-to-cell variability in fluorescence signal . HeLa cells stably expressing fluorescently tagged GFP-RAB6A were previously generated in the Lab at Institut Curie ( Teber et al . , 2005 ) . They were maintained in DMEM supplemented with 10% fetal bovine serum . Cells were then spread onto fibronectin Cytoo chips ( Cytoo Cell Architect ) 4 to 5 hr before imaging . Cell adhesion on micro-patterns both constrains the cells in terms of lateral movement and averages their size and shape ( disk-shaped and crossbow-shaped , Cytoo Cell Architect , 1100μ m2 ) . As a control of patterning effect , the same cell line was grown under the same culture conditions , and spread on regular glass coverslips , 4 to 5 hr before imaging . For a second set of experiments , wild-type RPE1 cells ( hTERT RPE-1 obtained from ATCC collection ) were grown in Dulbecco’s Modified Eagle Medium , Nutrient Mixture F-12 ( DMEM/F12 ) supplemented with 10% ( vol/vol ) FCS in six-well plates . RPE1 cells were transiently transfected with plasmids coding for Rab11a-GFP , and Langerin-mCherry using the following protocol: 2 μg of each DNAs , completed to 100 μL with DMEM/F12 ( FCS free ) were incubated for 5 min at room temperature . 6 μL of X-tremeGENE 9 DNA Transfection Reagent ( Roche ) completed to 100 μL with DMEM/F12 ( FCS free ) were added to the mix and incubated for further 15 min at room temperature . The transfection mix was then added to RPE1 cells grown 1 day before and incubated further at 37oC overnight . Cells were then spread on regular coverslips or onto fibronectin Cytoo chips ( Cytoo Cell Architect ) for 4 hr at 37oC with F-12 ( with 10% ( vol/vol ) FCS , 10 mM Hepes , 100 units/ml of penicillin and 100 ug/ml of Strep ) before imaging . When specified , 2 mM Latrunculin A ( Sigma ) was dissolved to 0 . 02 mM in F-12 DMEM . 300 μL of culture medium with Latrunculin A ( 600 nM ) was added to establish a final Latrunculin A concentration of 3 μM . All cell lines were routinely tested for mycoplasma , using PCR or the MycoAlert Mycoplasma Detection Assay . For Rab6-positive membranes in unconstrained cells , videos were recorded with an epifluorescence video automated system composed of a Ti Eclipse inverted microscope equipped with a 100x objective Plan NA ( 1 . 4 ) and a piezo stage for 3D acquisitions ( Nikon , S . A , France ) . The fluorescence was collected using a 512 × 512 EM-CCD ( Evolve , Photometric , USA ) and driven through the Metamorph software ( Molecular Devices ) . 18 series of 120 Z image stacks of 10 frames were recorded at a rate of about 1 stack/s . The volume rendering of one image from this dataset is shown in Appendix 1—figure 1a . For Rab6-positive membranes on micro-patterns , the 488 nm laser of a spinning-disk confocal microscope ( Ti Eclipse , Nikon , S . A , France equipped with spinning disk system , a 100x/1 . 4 oil objective and CoolSnap HQ2 CCD , from Roper Scientific S . A . R . L , France ) was used to acquire 3D 380 × 380 × 8 stacks at a rate of one stack per second . 18 image sequences with crossbow-shaped cells and 22 image sequences with disk-shaped cells were acquired . The system was driven by the Metamorph software ( Molecular Devices ) . The volume rendering of two images from this dataset are shown in Appendix 1—figure 1b–c . For the Rab11 dataset , live-cell imaging was performed using simultaneous dual color Total Internal Reflection Fluorescence ( TIRF ) microscopy . All imagings were performed in full conditioned medium at 37oC and 5% CO2 unless otherwise indicated . Simultaneous dual color TIRF microscopy sequences were acquired on a Nikon TE2000 inverted microscope equipped with a 100x TIRF objective ( NA = 1 . 49 ) , an azimuthal TIRF module ( Ilas2 , Roper Scientifc ) , an image splitter ( DV , Roper Scientific ) installed in front of an EMCCD camera ( Evolve , Photometrics ) that can be bypassed or not , depending on the experimental conditions , as indicated in the text , and a temperature controller ( LIS ) . GFP and m-Cherry were excited with a 488 nm and a 561 nm laser , respectively ( 100 mW ) . The system was driven by the Metamorph software ( Molecular Devices ) . Four selected image projections from this data set are shown in Appendix 1—figure 1d–g . We use two datasets in this study that are publicly available on the iMANAGE database at https://cid . curie . fr/iManage/standard/login . html with username public and password Welcome ! 1 in the project entitled QuantEv-Data . Before applying QuantEv , the intracellular events have to be identified and localized . The Rab6 proteins are extracted from each image sequence by using the C-CRAFT method ( Pécot et al . , 2015 ) with default parameters , except the p value that ranges from 0 . 0025 to 0 . 35 depending on the noise level , available on Icy ( de Chaumont et al . , 2012 ) . The Rab11-positive membranes on micro-patterns are segmented at each time point with the ATLAS algorithm ( Basset et al . , 2015 ) with default parameters , except the p value that ranges from 0 . 05 to 0 . 45 depending on the noise level . In both cases , a variance stabilization transform ( Boulanger et al . , 2010 ) is performed to take into account the Poisson-Gaussian nature of the noise in the CCD sensors . As unconstrained cells are more mobile than cells on micro-patterns , the image sequences showing Rab11-positive membranes in unconstrained cells are not in focus . To correct this phenomenon , a deconvolution method ( Lefkimmiatis et al . , 2012 ) is first applied to the image sequences . The Rab11 positive membranes are then segmented at each time point with the Bernsen local thresholding method ( Bernsen , 1986 ) ( radius equal to 15 pixels ) . Finally , Rab6 and Rab11 trajectories are estimated with the multiple hypothesis tracking method developed by Chenouard et al . ( 2013 ) with default parameters , available on Icy ( de Chaumont et al . , 2012 ) , the combinatorial optimization tracking method developed by Sbalzarini and Koumoutsakos ( 2005 ) with default parameters , available on ImageJ ( Schneider et al . , 2012 ) and the hybrid approach TrackMate ( Tinevez et al . , 2017 ) that first connects detected points into short tracks and then links the resulting tracks together , with default parameters , available on Fiji ( Schindelin et al . , 2012 ) . To identify the trajectories estimated with different methods , we use the gated distance ( Chenouard et al . , 2014 ) defined between two trajectories θ1 and θ2 as: ( 1 ) d ( θ1 , θ2 ) =∑t=0T min ( ||θ1 ( t ) −θ2 ( t ) ||2 , ϵ ) , where ϵ is the gate . For each image sequence , the gated distance is computed between the trajectories estimated with the three different methods with ϵ=5 pixels . Only the trajectories for which the gated distance is inferior to 2 pixels for at least two methods are used for the analysis . The localization of events needs to be defined on a common coordinate system to compare the experiments . We propose to use the cylindrical coordinate system where only a reference point such as the event emitter or attractor and a reference direction have to be specified by the user . To fairly compare experiments with different cell shapes , we define appropriate distances to obtain normalized densities , that is independent from the cell shape . We illustrate the importance of shape normalization in Appendix 2 . More formally , let us define Ω the 3D cell support and ∂Ω the 3D cell surface . Let us consider a set of N sample points associated with intracellular events S={ ( ri , θi , zi , wi , dθi , dzi ) , i∈[1 , N]} , where ( ri , θi , zi ) denote the spatial cylindrical coordinates . The weight wi enables to take into account features associated to events such as intensity , track length , confinement ratio . . . wi can typically be a function of fluorescence intensity , proportional to the number of molecules observed at a given location . The distance dθi is equal to the Euclidean distance between the coordinate system origin O∈Ω projected on plane zi ( Ozi ) and the point Pθi , zi∈∂Ω with angle θi at plane zi , such that dθi=||Pθi , zi−Ozi||22 . The distance dzi is equal to the Euclidean distance between the coordinate system origin O and the point Pri , θi∈∂Ω with radius ri and angle θi such that dzi=||Pri , θi−O|| . These two distances allow estimating normalized densities that are independent from cell shapes . We propose to estimate three densities defined as follows:f ( r ) =1Zr , θ∑i=1N Gσ^r ( ri−r ) widθi , ( 2 ) f ( θ ) =1Zr , θ∑i=1N Hκ^ ( θi−θ ) widθi , f ( z ) =1Zz∑i=1N Gσ^z ( zi−z ) widzi , where Gσ^ ( ⋅ ) is a Gaussian kernel with bandwidth σ^ , Hκ^ is a von Mises kernel with concentration κ^ such that Hκ^ ( θ ) =eκ^cosθ2πI0 ( κ^ ) and I0 ( ⋅ ) is the Bessel function of order 0 . The bandwidths σ^r and σ^z are estimated with the Silverman’s rule of thumb ( Silverman , 1986 ) and κ^ is estimated using the robust rule of thumb proposed by Taylor ( 2008 ) . The normalization constants are defined as follows: ( 3 ) Zr . θ=N∑i=1Nwidθi , Zz=N∑i=1Nwidzi . In case the distribution of dynamical features such as confinement ratio or lifetime with respect to the track localization is to be studied , the weighted densities are defined differently than in the previous section . In this case , histograms are first computed as the averaged dynamic features for each bin . A density estimation is then estimated from the histograms . Let us consider a set of T tracks associated with spatial coordinates T={ ( ri , θi , zi , mi ) , i∈[1 , T]} , where ( ri , θi , zi ) denote the spatial cylindrical coordinates of the median point of the trajectory i and mi is a dynamic feature associated to track i . The histograms corresponding to the averaged dynamic features for each bin are defined as:hr ( br ) =∑iT𝟙br[ri]mi∑iT𝟙br[ri] , ( 4 ) hθ ( bθ ) =∑iT𝟙bθ[θi]mi∑iT𝟙1bθ[θi] , hz ( bz ) =∑iT𝟙bz[zi]mi∑iT𝟙bz[zi] , where br∈[1 , Br] is a radius bin and Br is the total number of radius bins , bθ∈[1 , Bθ] is a polar bin and Bθ is the total number of polar bins , bz∈[1 , Bz] is an in-depth bin and Bz is the total number of in-depth bins , and 1br[ri] is equal to 1 if ri is defined in bin br and equal to 00 otherwise . Densities are then estimated from the histograms as follows:fd ( r ) =∑i=1Br Gσ^r ( hr ( bi ) −r ) , ( 5 ) fd ( θ ) =∑i=1Bθ Hκ^ ( hθ ( bi ) −θ ) , fd ( z ) =∑i=1Bz Gσ^z ( hz ( bi ) −z ) , where Gσ^ ( ⋅ ) is a Gaussian kernel with bandwidth σ^ , Hκ^ is a von Mises kernel with concentration κ^ . The bandwidths σ^r and σ^z are estimated with the Silverman’s rule of thumb ( Silverman , 1986 ) and κ^ is estimated using the robust rule of thumb proposed by Taylor ( 2008 ) . Quantitative comparison between different conditions is mandatory to analyze biological data . In most computational biology studies , data from different experiments corresponding to the same condition are pooled together ( Schauer et al . , 2010; Merouane et al . , 2015 ) . This usual procedure enables to add statistical power when comparing two conditions . Therefore , it is especially useful when few data are available . Unfortunately , pooling data together presents two main drawbacks . First , if large amounts of data are available , the opposite problem arises and the statistical tests may become significant for every comparison ( Olivier and Walter , 2015 ) . One solution is to down sample the data , but the amount of down sampling becomes another issue . Second , pooling data together for one condition partially hides the variability between the replicated experiments for this condition . As an example , let us consider a study aimed at analyzing the effects of a drug on a sample of normal individuals . To evaluate the drug efficiency , a comparison between normal individuals and individuals that were administered the drug is conducted . Let us assume that the drug is effective on half the individuals . Consequently , normal individuals are compared to a mix of normal individuals and individuals with the drug effects . This comparison should not be statistically significant as the drug is not efficient on all individuals . However , the effects on the individuals for which the drug is efficient might hide the fact that it is not efficient on all individuals if all the data are pooled together . In what follows , we propose to compute a distance between all experiments instead of a distance between conditions . The idea is demonstrated in Appendix 3 and validated on synthetic image sequences ( see Appendix 1—figure 1c–d and Appendix 3 ) . We propose to compute the earth mover’s distance ( also known as the Kantorovich-Rubinstein or the first order Wasserstein distance ) between every replicate of every condition to apply a statistical test . This transport-based distance demonstrated its efficiency for other studies on cell phenotypes ( Wang et al . , 2013; Basu et al . , 2014 ) . The discrete Earth Mover’s Distance ( EMD ) between two unidimensional distributions is simply defined as the sum of the absolute differences between their cumulated distribution functions ( Rubner et al . , 2000 ) : ( 6 ) EMD ( f1 , f2 ) =∑i=1K |F1 ( i ) −F2 ( i ) | , where F1 and F2 are the cumulated distribution functions of f1 and f2 . Although the EMD depends on the number of bins K , EMD proportions are kept intact when the number of bins is high enough as shown in Appendix 4 . For the angular distribution , the Circular Earth Mover’s Distance ( CEMD ) ( Rabin et al . , 2011 ) is defined as: ( 7 ) CEMD ( f1 , f2 ) =mink∈{1 , ⋯ , K}∑i=1K |Qk1 ( i ) −Qk2 ( i ) | , with ( 8 ) Qk ( i ) ={∑j=ki f ( j ) ifi≥k , ∑j=kK f ( j ) +∑j=1i f ( j ) ifi<k . The EMD and CEMD enable to compute a distance between two single experiments for the radial , angular and in-depth densities . The distances between the replicates of one condition and the replicates of the other condition ( s ) give an idea about the difference between the conditions . However , a baseline distance is also needed to state if the difference is random or significant . Therefore , two distances are defined for each experiment and each density: Considering more than two groups does not change the intra-condition distance and only expands the inter-condition distance to more than one group . We define as the condition difference the difference between the inter-condition distance and the intra-condition distance . If the condition difference is high , the conditions are different . A statistical test is applied on the difference distance to state if the observed conditions are significantly different . A non-parametric statistical test is better suited as there is no underlying model for the condition difference . In addition , a negative condition difference implies that the current experiment is closer to the replicated experiments of the other condition than the replicated experiments of the same condition . Consequently , the condition difference has to be positive if the conditions are different . For those two reasons , we propose to use the one-sided non-parametric Wilcoxon signed-rank test on the condition differences for all experiments to state if two conditions are statistically different . In case we focus on the intracellular events assumed to be uniformly distributed around a given biological object , for example the events emitter , QuantEv allows us to estimate a location for this traffic-organizing component . This source location is then defined as the reference point with the most uniform angular distribution . It is established that the maximum entropy corresponds to the most uniform distribution . Consequently , the reference point O∗ is defined as the location that maximizes the entropy: ( 9 ) O∗=maxO∈Ω−∑i=1Nf ( θi ) logf ( θi ) . The most straightforward way to find this point is to estimate the entropy map that gives , for each point in Ω , the entropy value computed with the current point used as the reference center . We also propose to use the bisection method to speed up the computation ( about sixty times faster than the entropy map computation , see Appendix 5—table 1 ) . A uniformity analysis conducted on simulations is presented in Appendix 6 . The entropy criterion can be extended to detect multiple organizing components if needed . The jar file of the QuantEv Icy plugin is available at http://icy . bioimageanalysis . org/plugin/QuantEv . The jar file of the QuantEv track processor is available at http://icy . bioimageanalysis . org/plugin/ QuantEv _ ( track_processor ) . The source codes can be extracted from the jar files . The QuantEv analysis module for TrackMate is available on GitHub ( Pécot , 2018; copy archived at https://github . com/elifesciences-publications/QuantEvForTrackMate ) . These codes are released under the GNU Affero General Public License v3 . 0 . | Proteins are the workhorses of the body , performing a range of roles that are essential for life . Often , this requires these molecules to move from one location to another inside a cell . Scientists are interested in following an individual protein in a living cell ‘in real time’ , as this helps understand what this protein does . Scientists can track the whereabouts of a protein by ‘tagging’ it with a fluorescent molecule that emits light which can be picked up by a powerful microscope . This process is repeated many times on different samples . Finally , researchers have to analyze all the resulting images , and conduct statistical analysis to draw robust conclusions about the overall trajectories of the proteins . This process often relies on experts assessing the images , and it is therefore time-consuming and not easily scalable or applied to other experiments . To help with this , Pécot et al . have developed QuantEV , a free algorithm that can analyze proteins’ paths within a cell , and then return statistical graphs and 3D visualizations . The program also gives access to the statistical procedure that was used , which means that different experiments can be compared . Pécot et al . used the method to follow the Rab6 protein in cells of different shapes , and found that the conformation of the cell influences where Rab6 is located . For example , in crossbow-shaped cells , Rab6 is found more often toward the three tips of the crossbow , while its distribution is uniform in cells that look like disks . Another experiment examined where the protein Rab11 is normally placed , and how this changes when the cell’s skeleton is artificially disrupted . Both studies help to gain an insight into the behavior of the cellular structures in which Rab6 and Rab11 are embedded . Following proteins in the cell is an increasingly popular method , and there is therefore a growing amount of data to process . QuantEV should make it easier for biologists to analyze their results , which could help them to have a better grasp on how cells work in various circumstances . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"cell",
"biology",
"tools",
"and",
"resources"
] | 2018 | A quantitative approach for analyzing the spatio-temporal distribution of 3D intracellular events in fluorescence microscopy |
Behavioral plasticity is widespread in swarming animals , but little is known about its underlying neural and molecular mechanisms . Here , we report that a neuropeptide F ( NPF ) /nitric oxide ( NO ) pathway plays a critical role in the locomotor plasticity of swarming migratory locusts . The transcripts encoding two related neuropeptides , NPF1a and NPF2 , show reduced levels during crowding , and the transcript levels of NPF1a and NPF2 receptors significantly increase during locust isolation . Both NPF1a and NPF2 have suppressive effects on phase-related locomotor activity . A key downstream mediator for both NPFs is nitric oxide synthase ( NOS ) , which regulates phase-related locomotor activity by controlling NO synthesis in the locust brain . Mechanistically , NPF1a and NPF2 modify NOS activity by separately suppressing its phosphorylation and by lowering its transcript level , effects that are mediated by their respective receptors . Our results uncover a hierarchical neurochemical mechanism underlying behavioral plasticity in the swarming locust and provide insights into the NPF/NO axis .
Swarming occurs in a wide variety of animal taxa , including insects , fish , birds , and mammals . Individuals benefit from swarming in many aspects , including food searching , territory selection , and defense ( Okubo , 1986; Weaver et al . , 1989 ) . Typically , to maintain the required fission–fusion dynamics , swarming animals exhibit striking behavioral plasticity of different types ( Snell-Rood , 2006; Szyf , 2010 ) . Biochemical changes in the levels of neuromodulators , such as monoamines , neuropeptides , and neurohormones , are able to induce behavioral variation thus mediate behavioral plasticity ( Freudenberg et al . , 2015; Godwin et al . , 2015; Zupanc and Lamprecht , 2000 ) . Nevertheless , the molecular basis by which neural factors orchestrate behavioral plasticity in swarming animals is poorly understood in detail . Neuropeptides , a group of chemically diverse neural modulators , affect a broad range of physiological and behavioral activities ( Lieberwirth and Wang , 2014; Nässel , 2002 ) . Accumulating evidence shows that neuropeptides serve as conserved neuronal signals that modulate animal behaviors in social contexts ( Lieberwirth and Wang , 2014; Nilsen et al . , 2011 ) . These peptides exert their actions by binding to specific membrane receptors , most of which are G-protein-coupled receptors ( Quartara and Maggi , 1997 ) . The binding initiates a second-message cascade unique for each receptor and results in a distinct molecular response ( Hökfelt et al . , 2003 ) . It has been revealed that neuropeptides can induce plasticity in a series of behavioral processes , including sensory detection ( Shankar et al . , 2015 ) , signal integration ( Grammatopoulos , 2012 ) , and behavioral responsiveness ( Ruzza et al . , 2015 ) by acting either individually or in concert with other neuromodulators ( Dölen et al . , 2011; Flores et al . , 2015; Maroun and Wagner , 2016 ) . Therefore , neuropeptides and their downstream components may act as vital parts of the regulatory network underlying behavioral plasticity in swarming animals . The migratory locust , Locusta migratoria , exhibits two interconvertible phases , the solitarious phase ( S-phase ) and the gregarious phase ( G-phase ) , the latter of which is characterized by swarming behavior ( Ariel and Ayali , 2015 ) . Locust behaviors in the two phases significantly differ , most notably in the interaction among individuals and in locomotor activity ( Uvarov , 1977 ) . S-phase locusts are sedentary and repel their conspecifics , whereas G-phase individuals are highly active and attract their conspecifics ( Simpson et al . , 1999 ) . The behavioral transition between two phases is promoted by either isolating G-phase locusts ( that is , solitarization ) or , in the opposite direction , by forced crowding of S-phase locusts ( that is , gregarization ) , the key step in seeding locust swarming ( Pener and Simpson , 2009 ) . Behavioral solitarization occurs faster than behavioral gregarization in the migratory locust . The attraction index and locomotor activity of locusts continuously decrease within 16 hr after isolation . By contrast , these behaviors do not increase until 32 hr after crowding , but are far below the level of gregarious controls even after crowding for 64 hr ( Guo et al . , 2011 ) . The locust brain undergoes strong neurochemical reconfiguration during behavioral phase transition; for instance , the contents of several neurotransmitters that mediate synaptic plasticity show significant change ( Rogers et al . , 2004; Ma et al . , 2011 , 2015 ) . Recently , we have found that several neuropeptide genes are differentially expressed between the central nervous systems of G-phase and S-phase locusts ( Hou et al . , 2015 ) , suggesting possible modulatory roles for these neuropeptides in the behavioral phase transition . Here , we show that two related neuropeptides , NPF1a and NPF2 , act as crucial neural modulators in the phase-related locomotor plasticity of the migratory locust . We uncover a potentially important connection between the atypical neurotransmitter NO and the two NPFs , a connection mediated by NOS . We therefore suggest that the actions of NPFs ( or their homolog NPY ) may be mediated , partly through NOS and NO , in other organisms .
We have previously shown that 15 neuropeptide-encoding genes are differentially expressed in the brains of G-phase and S-phase locusts ( Hou et al . , 2015 ) . Here , we extend our work to explore which of these neuropeptides are closely tied to the behavioral phase transition . qPCR analysis ( Figure 1 and Figure 1—figure supplement 1 ) revealed that the mRNA levels of four neuropeptide encoding genes , namely , AKH/Corazonin related peptide ( ACP ) , Insulin-like peptide ( ILP ) , NPF1a , and NPF2 , significantly changed in the phase transition , that is , during solitarization or gregarization or both . During gregarization , the mRNA levels of ACP and ILP steadily increased , whereas those of NPF1a and NPF2 rapidly decreased . During solitarization , the transcript levels of ILP and NPF1a significantly changed compared to those of ACP and NPF2 . 10 . 7554/eLife . 22526 . 003Figure 1 . Levels of transcripts encoding the neuropeptides NPF1a , NPF2 , ACP and ILP change during the G/S phase transition in the migratory locust . qPCR was performed to determine the transcript levels of 15 neuropeptide-encoding genes in locust brains in the time course of the isolation of gregarious ( G-phase ) locusts or the crowding of solitarious ( S-phase ) locusts . Four neuropeptide genes displayed clear expression changes during isolation or crowding or both ( in the case of NPF1a and ILP ) . Raw data measuring the mRNA levels of the four neuropeptide genes are shown in Figure 1—source data 1 . For the transcript levels of the other 11 neuropeptide genes , see Figure 1—figure supplement 1 . The data are presented as mean ± s . e . m . Significant differences at different times are denoted by letters ( n = 4 samples per timepoint , 8 animals/sample , one-way ANOVA , p<0 . 05 ) . *indicates a significant difference between typical G-phase ( 0 hr after isolation ) and typical S-phase ( 0 hr after crowding ) locust brains ( Student’s t-test , *p<0 . 05 , **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 00310 . 7554/eLife . 22526 . 004Figure 1—source data 1 . mRNA levels of the four neuropeptide-encoding genes during isolation and crowding processes . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 00410 . 7554/eLife . 22526 . 005Figure 1—figure supplement 1 . The transcript levels of 11 neuropeptides do not change during the G/S phase transition in the migratory locust . Isolation is shown in blue; crowding is shown in red . The data are presented as mean ± s . e . m . , *indicates a significant difference between typical G-phase ( 0 hr after isolation ) and typical S-phase ( 0 hr after crowding ) locust brains ( n = 4 samples , 8 locusts/sample , one-way ANOVA for multi-group comparisons , Student’s t-test for two-group comparison , *p<0 . 05 , **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 005 To assess whether these four neuropeptides are involved in the behavioral phase transition , we performed a behavioral screen in the G-phase locusts using transcript knockdown or peptide injection . The behavioral phase state was then assessed in an arena assay and measured by Pgreg , which is calculated using a binary logistic regression model that retains three variables: attraction index , total distance moved , and total duration of movement ( Guo et al . , 2011 ) . Pgreg varies between 0 ( in the fully S-phase behavioral state ) and 1 ( in the fully G-phase behavioral state ) . We performed RNAi-mediated transcript knockdown to reduce the levels of ACP and ILP , which show higher transcript levels in G-phase locust brains ( Figure 1 , lower ) . We found that knockdown of either ACP or ILP transcript did not significantly change the Pgreg values of G-phase locusts ( Figure 2—figure supplement 1 ) . On the other hand , we injected synthetic peptides to increase the concentrations of NPF1a and NPF2 , which display lower transcript levels in G-phase locust brains ( Figure 1 , upper ) . G-phase locusts that were injected with NPF1a or NPF2 peptide behaved in a way that became considerably more solitarious , in a dose-dependent manner , when compared to control locusts ( Figure 2A and Figure 2—figure supplement 2A ) . Co-injection of both NPF1a and NPF2 peptides into G-phase locusts enhanced the reduction of Pgreg compared to that seen following the injection of either NPF peptide alone ( Figure 2A ) . Moreover , injection of NPF1a peptide provoked a faster inhibitory effect on the Pgreg values of locusts than that caused by NPF2 peptide injection ( Figure 2B and Figure 2—figure supplement 2B ) . However , G-phase locusts that were injected with either dsNPF1a or dsNPF2 or with a mixture of these constructs did not show significant behavioral changes relative to control locusts ( Figure 2—figure supplement 3 , left ) . 10 . 7554/eLife . 22526 . 006Figure 2 . Perturbations of NPF1a or NPF2 peptide levels or of their transcript levels leads to changes in locomotor activity related to the G/S phase transition . Locust behaviors are measured by the term Pgreg , which is a combined assessment of movement and inter-insect attraction ( indicated as attraction index , see Figure 2—figure supplement 2 ) . Pgreg = 0 represents a fully S-phase behavioral state; Pgreg = 1 represents a fully G-phase behavioral state . ( A ) and ( B ) Dose- and time-dependent changes in the median Pgreg of G-phase locusts after injection of NPF1a and NPF2 peptides , separately and together . For detailed Pgreg distributions and statistics , see Figure 2—figure supplement 2 ( n ≥ 18 locusts , Mann–Whitney U test , p<0 . 05 ) . ( C ) Pgreg in S-phase locusts 48 hr after transcript knockdown of NPF1a , or NPF2 , or both ( n ≥ 20 locusts , Mann–Whitney U test , p=0 . 020 , 0 . 064 and 0 . 017 , respectively ) . Lines indicate median Pgreg . Significant differences are denoted by letters . ( D ) Pgreg in crowded S-phase locusts after transcript knockdown of NPF1a , or NPF2 , or both ( n ≥ 20 locusts , Mann–Whitney U test , p=0 . 024 , 0 . 039 and 0 . 037 , respectively ) . Locusts were forced into a crowd 32 hr after dsRNA injection , and their behaviors were measured after 16 hr of crowding ( that is 48 hr after dsRNA injection ) . ( E ) and ( F ) Total distance moved ( TDM ) and total duration of movement ( TDMV ) 4 hr after injection of NPF1a or NPF2 or both peptides in G-phase locusts ( 5 μg/individual ) . The data are presented as mean ± s . e . m . Significant differences are denoted by letters ( n ≥ 18 locusts , one-way ANOVA , p<0 . 05 ) . ( G ) and ( H ) Total distance moved ( TDM ) and total duration of movement ( TDMV ) 48 hr after transcript knockdown of NPF1a or NPF2 or both genes in S-phase locusts ( n ≥ 20 locusts ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 00610 . 7554/eLife . 22526 . 007Figure 2—figure supplement 1 . Transcript knockdown of ACP or ILP does not significantly affect behavioral phase state in G-phase locusts . ( A ) Behavioral tests ( measured as Pgreg ) after transcript knockdown of ACP or ILP in G-phase locusts ( n ≥ 27 locusts , Mann–Whitney U test , p=0 . 053 and 0 . 042 for dsACP and dsILP injection , respectively ) . ( B ) and ( C ) Efficiency and specificity of ACP and ILP transcript knockdown in G-phase locusts . RNAi effects were examined by qPCR 48 hr after dsRNA injection . The data are presented as mean ± s . e . m . ( n = 4 samples , 6–8 locusts/sample , Student’s t-test , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 00710 . 7554/eLife . 22526 . 008Figure 2—figure supplement 2 . Injection of NPF1a or NPF2 peptide into G-phase locusts induces S-phase-like behaviors in a dose- and time-dependent manner . ( A ) Dose-dependent effects on the Pgreg values of G-phase locusts 4 hr after injection of NPF1a and NPF2 peptides , separately and together . Significant differences between treatments are denoted by letters ( n ≥ 20 locusts , Mann–Whitney U test , p=0 . 002 , 0 . 006 and 0 . 000 for 0 . 1 , 1 and 5 μg NPF1a peptide injections; p=0 . 031 , 0 . 008 and 0 . 001 for 0 . 1 , 1 and 5 μg NPF2 peptide injections; p=0 . 004 , 0 . 000 and 0 . 000 for 0 . 1 , 1 and 5 μg NPF1a and NPF2 co-injections , respectively ) . ( B ) Time-dependent effects on the Pgreg values of G-phase locusts after injection of NPF1a and NPF2 peptides , separately and together . Significant differences are denoted by letters ( n ≥ 18 locusts , Mann–Whitney U test , p=0 . 002 , 0 . 006 and 0 . 000 for 1 , 2 and 4 hr NPF1a peptide injections; p=0 . 031 , 0 . 008 and 0 . 001 for 1 , 2 and 4 hr NPF2 peptide injections; p=0 . 004 , 0 . 000 and 0 . 000 for 1 , 2 and 4 hr NPF1a and NPF2 co-injections , respectively ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 00810 . 7554/eLife . 22526 . 009Figure 2—figure supplement 3 . Transcript knockdown of NPF1a or NPF2 in G-phase locusts and peptide injection of NPF1a or NPF2 in S-phase locusts do not affect phase-related behaviors . Attraction index ( AI ) represents the extent by which the tested animals are attracted by the stimulus group ( AI = total duration in stimulus area − total duration in the opposite of stimulus area ) . ( A ) Pgreg of G-phase locusts 48 hr after transcript knockdown of NPF1a and NPF2 , separately and together ( n ≥ 18 locusts , Mann–Whitney U test , p=0 . 426 , 0 . 584 and 0 . 636 , respectively ) . Lines indicate median Pgreg . n . s . indicates no significant difference . ( B ) Attraction index of G-phase locusts 48 hr after transcript knockdown of NPF1a and NPF2 , separately and together ( n ≥ 18 locusts , one-way ANOVA ) . ( C ) Pgreg of S-phase locusts 4 hr after injection of NPF1a or NPF2 or both peptides ( n ≥ 18 locusts , Mann–Whitney U test , p=0 . 381 , 0 . 939 and 0 . 475 , respectively ) . ( D ) Attraction index of S-phase locusts 4 hr after injection of NPF1a or NPF2 or both peptides ( n ≥ 18 locusts , one-way ANOVA ) . ( E ) and ( F ) Total distance moved ( TDM ) and total duration of movement ( TDMV ) 48 hr after transcript knockdown of NPF1a or NPF2 or both genes in G-phase locusts ( n ≥ 18 locusts , one-way ANOVA ) . ( G ) and ( H ) Total distance moved ( TDM ) and total duration of movement ( TDMV ) 4 hr after injection of NPF1a or NPF2 or both peptides in S-phase locusts . ( n ≥ 18 locusts , one-way ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 00910 . 7554/eLife . 22526 . 010Figure 2—figure supplement 4 . Efficiency and specificity of NPF1a and NPF2 transcript knockdown . The S-phase locusts were injected with dsNPF1a or dsNPF2 or dsGFP . RNAi effects were examined by qPCR 48 hr after injection . The data are presented as mean ± s . e . m . ( n = 4 samples , 6–8 locusts/sample , Student’s t-test , **p<0 . 01; ***p<0 . 001 ) . n . s . indicates no difference in a paired comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 01010 . 7554/eLife . 22526 . 011Figure 2—figure supplement 5 . Perturbation of NPF1a or NPF2 peptide , or of their transcript levels , do not change attraction index related to the G/S phase transition . ( A ) Effects on attraction index of G-phase locusts 4 hr after injection of NPF1a or NPF2 or both peptides . The data are presented as mean ± s . e . m . ( n ≥ 20 locusts , Student’s t-test ) . n . s . means not significant . ( B ) Effects on attraction index of S-phase locusts 48 hr after transcript knockdown of NPF1a or NPF2 or both genes ( n ≥ 20 locusts , Student’s t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 011 We validated the roles of two NPFs in the behavioral change in S-phase locusts by transcript knockdown of NPF1a and NPF2 individually or together . S-phase locusts that were injected with dsNPF1a displayed a significant behavioral change in the direction of G-phase , whereas injection of dsNPF2 did not significantly change the Pgreg values ( Figure 2C and Figure 2—figure supplement 4 ) . However , the S-phase locusts that were injected with either dsNPF1a or dsNPF2 were more gregarious than the controls in response to crowding stimuli , and these effects were strengthened by the dual-knockdown of the NPF1a and NPF2 transcripts ( Figure 2D ) . Furthermore , peptide injection of NPF1a or NPF2 or their mixture in S-phase locusts did not affect their behavioral phase states ( Figure 2—figure supplement 3 , right ) . Behavioral parameter analysis demonstrated that locust locomotor activity , including total duration of movement and total distance moved , were strongly suppressed by the treatments that increased the levels of NPF1a or NPF2 peptide in G-phase locusts , but enhanced by dsNPF1a or dsNPF2 injection in S-phase locusts ( Figure 2E–H ) , while the attraction index was not significantly changed by these treatments ( Figure 2—figure supplement 5 ) . Thus , NPF1a and NPF2 play important roles in the locust behavioral phase transition by modulating locomotor activity . Bioinformatically , we obtained two locust sequences with high similarity to the Drosophila NPFR gene ( Supplementary file 1 ) . They were named LomNPFR and LomNPYR , based on their phylogenetic relationship with homologs in other species ( Figure 3—figure supplement 1B ) . Competitive binding experiments indicated that NPF1a peptide displayed much higher affinity to HEK 293 T cells expressing NPFR protein ( IC50 = 24 nM ) than did NPF2 peptide ( IC50 = 355 nM ) ( Figure 3A and Figure 3—figure supplement 2 ) , whereas NPF2 displayed much higher affinity to NPYR-expressing cells ( IC50 = 64 . 5 nM ) than did NPF1a ( IC50 = 380 nM ) ( Figure 3B ) . 10 . 7554/eLife . 22526 . 012Figure 3 . Receptors for NPF1a and NPF2 are involved in transmitting the effects of these neuropeptides on locomotor activity . ( A ) Competitive inhibition of TAMRA-NPF1a binding to HEK 293 T cells transfected with pcDNA3 . 1-NPFR vector ( n = 6 ) . ( B ) Competitive inhibition of TAMRA-NPF2 binding to HEK 293 T cells transfected with pcDNA3 . 1-NPYR vector ( n = 6 ) . ( C ) and ( D ) Time course patterns of NPFR and NPYR transcript levels during the G/S locust phase transition ( isolation , shown in blue; crowding , shown in red ) . The data are presented as mean ± s . e . m ( n = 4 samples per timepoint , 8 locusts/sample , one-way ANOVA , p<0 . 05 ) . Detailed expression levels of the two NPF receptors are shown in Figure 3—source data 1 . ( E ) and ( F ) Total distance moved ( TDM ) and total duration of movement ( TDMV ) 48 hr after transcript knockdown of NPFR or NPYR or both genes in S-phase locusts . Significant differences are denoted by letters ( n ≥ 19 locusts , one-way ANOVA , p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 01210 . 7554/eLife . 22526 . 013Figure 3—source data 1 . Transcript levels of NPFR and NPYR during the G/S locust phase transition . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 01310 . 7554/eLife . 22526 . 014Figure 3—figure supplement 1 . Phylogenetic relationship of NPF or NPY precursors and their receptors in different species . ( A ) Phylogenetic relationship of NPF and NPY precursors . The NPF precursors were obtained from Acyrthosiphon pisum , Aedes aegypti , Anopheles gambiae , Apis mellifera , Aphis gossypii , Bombyx mori , Culex pipiens , Drosophila melanogaster , Danaus plexippus , Nilaparvata lugens , Nasonia vitripennis , Schistocerca schistostatin , and Locusta migratoria; mammalian NPY precursors were obtained from Ovis aries , Mus musculus , and Homo sapiens . Most insect NPF peptides share an identical phenylalanine ( F ) instead of tyrosine ( Y ) at their C-terminus . The migratory locust genome contains two NPF1 precursors ( NPF1a and NPF1b ) and an NPF2 precursor . NPF1b , which encodes an 86 aa peptide , is hardly detected in the locust brain because of its extremely low expression level ( Hou et al . , 2015 ) . ( B ) Phylogenetic relationship between insect NPF/Y receptors and human NPY recepetors . The receptors are obtained from A . pisum , A . aegypti , A . florea , B . mori , Culex quinquefasciatus , D . melanogaster , T . castaneum and L . migratoria , together with mammalian NPY receptors from H . sapiens . The locust NPFR is close to 'type two' mammalian NPYR ( Homo Y2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 01410 . 7554/eLife . 22526 . 015Figure 3—figure supplement 2 . Overexpressions of ( A ) NPFR and ( B ) NPYR in HEK 293 T cells validated by western blot . DNA fragments encoding either NPFR ( 492 aa ) or NPYR ( 641 aa ) , followed by a Flag-tag , were inserted to a pcDNA3 . 1-expressing vector . Red arrow indicates target protein . The total protein of cells that were transiently transfected with pcDNA3 . 1-NPFR or pcDNA3 . 1-NPYR was used for western blot analysis . The HEK 293 T cells transfected with pcDNA3 . 1 were used as a control . Mouse monoclonal antibody against Flag ( CoWin , 1:5000 ) was used to validate the expression of the two receptors . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 01510 . 7554/eLife . 22526 . 016Figure 3—figure supplement 3 . Transcript knockdown of NPFR or NPYR in S-phase locusts induces G-phase-like behaviors without affecting attraction index . ( A ) and ( B ) Efficiency and specificity of NPFR and NPYR transcript knockdown examined by qPCR . dsNPFR or dsNPYR or dsGFP was microinjected into the locust brains . The effects of RNAi were analyzed 48 hr after dsRNA injection . The data are presented as mean ± s . e . m . ( n = 4 samples , 6–8 locusts/sample , Student’s t-test , *p<0 . 05 ) . ( C ) Effects on Pgreg of S-phase locusts 48 hr after transcript knockdown of NPFR or NPYR or both genes . Lines indicate median Pgreg . Significant differences are denoted by letters ( n ≥ 19 locusts , Mann–Whitney U test , p=0 . 0005 , 0 . 0015 and 0 . 0002 for dsNPFR , dsNPYR , or dsNPFR and dsNPYR injection , respectively ) . ( D ) Attraction index of S-phase locusts 48 hr after transcript knockdown of NPFR or NPYR or both genes . The data are presented as mean ± s . e . m . n . s . indicates not significant ( n ≥ 19 locusts , Student’s t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 016 The mRNA level of NPFR increased greatly within 1 hr after isolation of G-phase locusts , whereas it showed no change during locust crowding ( Figure 3C ) . By contrast , the transcript level of NPYR responded to both isolation and crowding , with an obvious increase during isolation and a significant reduction during crowding ( Figure 3D ) . Transcript knockdown of either NPFR or NPYR facilitated the transition from S-phase traits towards G-phase traits by influencing the locomotor activity of locusts ( Figure 3E , F and Figure 3—figure supplement 3 ) . Moreover , the dual-knockdown of NPFR and NPYR significantly strengthened the enhancement of both total distance moved and total duration of movement caused by knockdown of either transcript individually ( Figure 3E , F ) . These results suggest that these two NPF receptors are essential for the regulation of phase-related locomotor activity . To explore how NPF1a and NPF2 regulate locomotor plasticity during the G/S phase transition , we analyzed RNAseq-based transcriptomic differences in three comparisons: G-phase and S-phase locusts ( comparison 1: C1 ) ; co-injection of NPF1a and NPF2 peptides in G-phase locusts with control injection ( comparison 2: C2 ) ; co-injection of dsNPF1a and dsNPF2 in S-phase locusts with control injection ( comparison 3: C3 ) . We identified a total of 221 , 317 , and 313 differentially expressed genes in the three comparisons , respectively ( Figure 4—figure supplement 1A ) , and 32% of these genes were annotated ( Figure 4—figure supplement 1B ) . Numerous differentially expressed genes encoding catalytic and binding activities were clearly enriched in each treatment ( Figure 4—figure supplement 1C ) . A number of genes displayed altered transcription patterns ( Figure 4A ) that are consistent with locust behavioral change caused by the manipulation of NPF1a and NPF2 levels , as shown in Figure 2 . The transcript levels of these genes were different between the typical G-phase and S-phase locusts ( C1 ) . Moreover , their transcript levels changed oppositely in the two treatments: co-injection of NPF1a and NPF2 peptides in G-phase locusts ( C2 ) and dual-knockdown of NPF1a and NPF2 transcripts in S-phase locusts ( C3 ) . Among these genes , we found that several genes encode important signaling molecules . Using qPCR , the expression patterns of two genes , adenylate cyclase ( AC2 ) and NOS , were confirmed in all three comparisons ( Figure 4B and Figure 4—figure supplement 1D ) . The two genes showed high transcript levels in the brains of G-phase locusts . Moreover , their transcript levels were significantly lower after the co-injection of NPF1a and NPF2 peptides in G-phase locusts , and were increased by dual-knockdown of NPF1a and NPF2 transcripts in S-phase locusts ( Figure 4B and Figure 4—figure supplement 1D ) . 10 . 7554/eLife . 22526 . 017Figure 4 . Cluster analysis of RNA-Seq data leads to the identification of nitric oxide synthase ( NOS ) as a downstream component of the NPF1a and NPF2 pathway . ( A ) Cluster analysis of differentially expressed genes in the transcriptome . Several important genes ( highlighted in yellow ) involved in signaling in other organisms display expression patterns that correlate with behavioral change after the manipulation of NPF1a and NPF2 peptides or transcript levels . Logarithmic fold alteration of treatment versus control is shown in the heat map . Yellow and blue colors indicate up- and downregulation , respectively ( n = 3 samples per treatment , 10 animals/sample ) . For detailed gene-expression data , please see Figure 4—source data 1 . ( B ) Transcript levels of NOS in the brains after co-injection of NPF1a and NPF2 peptides in G-phase locusts or transcript knockdown of both NPF1a and NPF2 in S-phase locusts ( n = 5 samples , 8 locusts/sample , Student’s t-test , *p<0 . 05 , different letters labeled in columns indicate a significant difference ) . ( C ) NO levels after injection of NPF1a and NPF2 peptides , separately and together , in G-phase locusts , or after transcript knockdown of NPF1a and NPF2 , separately and together , in S-phase locusts . The data are presented as mean ± s . e . m . Significant differences are denoted by letters ( n = 4 samples , 16 locusts/sample , one-way ANOVA , p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 01710 . 7554/eLife . 22526 . 018Figure 4—source data 1 . The effects of NPF1a and NPF2 on the expression of annotated genes in the brains of fourth-instar locusts . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 01810 . 7554/eLife . 22526 . 019Figure 4—figure supplement 1 . Transcriptomic profiles influenced by NPF1a and NPF2 in locust brains revealed by RNA-seq . ( A ) Venn diagram of differentially expressed genes ( DEGs ) in the transcriptome . ( B ) The numbers of up- and downregulated genes in three comparisons: C1 ( G vs . S ) ; C2 ( G-NPF1a+NPF2 vs . G ) ; C3 ( S-dsNPF1a +dsNPF2 vs . S ) . ( C ) Molecular function analysis of DEGs in the transcriptome . Genes encoding catalytic and binding activities are obviously enriched in each treatment . ( D ) Verification of changes in the transcription of candidate genes based on the manipulation of NPF1a and NPF2 peptides or their transcript levels in the locust brains . Data are presented as mean ± s . e . m . ( n = 5 samples , 8 locusts/sample , Student’s t-test , *p<0 . 05 , different letters labeled in columns indicate a significant difference ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 01910 . 7554/eLife . 22526 . 020Figure 4—figure supplement 2 . cAMP levels after artificial manipulation of NPF1a or NPF2 peptide or their transcript levels . The data are presented as mean ± s . e . m . n . s . indicates no significance ( n = 4 samples , 12–16 locusts/sample , one-way ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 020 AC2 catalyzes cAMP production and might activate the PKA pathway , whereas NOS catalyzes NO production resulting in the activation of NO signaling ( Mete and Connolly , 2003; Watts and Neve , 1997 ) . We therefore examined whether cAMP and NO levels could be influenced by the manipulation of NPF1a and NPF2 levels . NO concentration in brains decreased dramatically within 4 hr after injection of NPF1a or NPF2 or of the peptide mixture into G-phase locusts , and significantly increased after knockdown of NPF1a or NPF2 or both NPF transcripts in S-phase locusts ( Figure 4C ) . By contrast , there was no change in cAMP level 4 hr after manipulation of either NPF1a or NPF2 level ( Figure 4—figure supplement 2 ) . These data suggest that NO signaling may serve as a downstream pathway for both NPFs in the locust . The mRNA and protein levels of NOS were considerably higher in G-phase than in S-phase locust brains ( Figure 5A , B ) , and significantly changed during the G/S phase transition ( Figure 5C–E ) . Interestingly , NOS was present in both phosphorylated and non-phosphorylated forms ( Figure 5—figure supplement 1A–C ) . Phosphorylated NOS was more abundant in the brains of G-phase locusts than in those of S-phase locusts ( Figure 5B ) . Dephosphorylation of NOS by λ-phosphatase significantly reduced NOS activity and the NO level ( Figure 5—figure supplement 1D , E ) . During the G/S phase transition , the level of NOS phosphorylation decreased or increased within 4 hr after solitarization or gregarization , respectively ( Figure 5D , E ) . These changes occurred much faster than the alterations in NOS mRNA level , which did not change until 16 hr after solitarization or gregarization ( Figure 5C ) . In addition , NO levels in the locust brains continuously decreased during solitarization , but sharply increased 32 hr after gregarization ( Figure 5F ) . The changes in NO levels are tightly linked to the G/S behavioral phase transition . 10 . 7554/eLife . 22526 . 021Figure 5 . NOS transcript levels and phosphorylation states and NO levels differ in G-phase and S-phase locust brains . ( A ) NOS mRNA levels in the brains of G-phase and S-phase locusts ( n = 4 samples , 8 locusts/sample , Student’s t-test , *p<0 . 05 ) . ( B ) NOS protein levels in the brains of G-phase and S-phase locusts . The upper band detected by anti-uNOS indicates phosphorylated NOS ( p-NOS , see Figure 5—figure supplement 1 ) ( n = 3 samples , 12 locusts/sample , Student’s t-test , *p<0 . 05 ) . ( C ) Time course of NOS mRNA levels during the G/S phase transition ( n = 4 samples/timepoint , 8 locusts/sample , one-way ANOVA , p<0 . 05 , isolation shown in blue; crowding shown in red ) . ( D ) and ( E ) Time course of NOS protein levels during the G/S phase transition ( n = 3 samples , 10–12 locusts/sample , phosphorylated NOS data are represented by triangles; total NOS data are represented by dots ) . The protein level is referenced to β-tubulin . ( F ) Time course of NO levels during the G/S phase transition . All data are presented as mean ± s . e . m . Significant differences are denoted by letters ( n = 4 samples , 16 locusts/sample , one-way ANOVA , p<0 . 05 ) . Raw data showing the changes in NOS mRNA level , NOS protein level and NO level are shown in Figure 5—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 02110 . 7554/eLife . 22526 . 022Figure 5—source data 1 . Time-course changes in NOS mRNA level , NOS protein level and NO level during the G/S phase transition . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 02210 . 7554/eLife . 22526 . 023Figure 5—figure supplement 1 . Reducing NOS expression and reducing NOS phosphorylation levels decrease NOS activity and NO level . ( A ) Western blot confirms the specificity of anti-uNOS , and two bands are detected . ( B ) The specificity of anti-uNOS was validated by transcript knockdown and Western blot analyses . The locust NOS is probably modified at the post-translational level . ( C ) NOS phosphorylation was confirmed using lambda phosphatase ( λPP , an enzyme that can remove protein phosphorylation ) . In Western blots , the upper band is removed , whereas the intensity of the lower band is increased by λPP treatment . ( D ) and ( E ) NOS activity and NO levels after the removal of NOS phosphorylation by λPP . The data are presented as mean ± s . e . m . ( n = 4 samples , 12–16 locusts/sample , Student’s t-test , *p<0 . 05; **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 023 We then conducted a series of molecular , pharmacological and behavioral experiments to investigate the function of NO signaling in the G/S locust phase transition . Knockdown of the NOS transcript or injection of the NOS inhibitor N-Nitro-L-arginine Methyl Ester ( L-NAME ) into G-phase locusts strongly suppressed the total duration of movement and total distance moved ( Figure 6A–D ) , thus resulting in S-phase-like behavior ( Figure 6—figure supplement 1A , C ) . By contrast , injection of S-phase locusts with the NO donor S-nitroso-N-acetyl-penicillamine ( SNAP ) enhanced the total duration of movement and total distance moved ( Figure 6E , F ) , and pushed locust behavioral change from S-phase to G-phase state ( Figure 6—figure supplement 1E ) . All manipulations ( including NOS transcript knockdown and injections of the two chemicals ) did not change the attraction index of tested locusts ( Figure 6—figure supplement 1B , D , F ) . Furthermore , both transcript knockdown and L-NAME injection significantly reduced NOS activity and NO levels in G-phase locust brains , whereas SNAP injection increased NO levels in S-phase locust brains without affecting NOS activity ( Figure 6—figure supplement 2 ) . 10 . 7554/eLife . 22526 . 024Figure 6 . Perturbations of NO levels by transcript knockdown and drug injection dramatically change G-phase and S-phase locust behaviors . ( A ) and ( B ) Total distance moved ( TDM ) and total duration of movement ( TDMV ) of G-phase locusts 48 hr after knockdown of the NOS transcript . All data are presented as mean ± s . e . m . ( n ≥ 23 locusts , Student’s t-test , *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . ( C ) and ( D ) Total distance moved ( TDM ) and total duration of movement ( TDMV ) of G-phase locusts 2 hr after injection of NOS inhibitor ( L-NAME ) . ( E ) and ( F ) Total distance moved ( TDM ) and total duration of movement ( TDMV ) of S-phase locusts 2 hr after injection of NO donor ( SNAP ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 02410 . 7554/eLife . 22526 . 025Figure 6—figure supplement 1 . Effects on Pgreg and attraction index after NOS transcript knockdown and drug treatments in G-phase and S-phase locusts . ( A ) and ( B ) Effects on Pgreg and attraction index 48 hr after transcript knockdown of NOS in G-phase locusts ( n ≥ 23 locusts , Mann–Whitney U test for Pgreg , p=0 . 005; lines indicate median Pgreg; Student’s t-test for attraction index analysis , the data are presented as mean ± s . e . m . ) . n . s . means no significant difference . ( C ) and ( D ) Effects on Pgreg ( n ≥ 23 locusts , p=0 . 001 ) and attraction index 2 hr after injection of NOS inhibitor ( L-NAME ) in G-phase locusts . ( E ) and ( F ) Effects on Pgreg ( n ≥ 23 locusts , p=0 . 036 ) and attraction index 2 hr after injection of NO donor ( SNAP ) in S-phase locusts . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 02510 . 7554/eLife . 22526 . 026Figure 6—figure supplement 2 . Effects on NOS activity and NO levels after NOS transcript knockdown and drug treatments in G-phase and S-phase locusts . ( A ) and ( B ) Effects on NOS activity and NO levels 48 hr after transcript knockdown of NOS into G-phase locusts . The data are presented as mean ± s . e . m . ( n = 4 samples , 12–16 locusts/sample , Student’s t-test , *p<0 . 05 ) . ( C ) and ( D ) Effects on NOS activity and NO levels 2 hr after injection of NOS inhibitor ( L-NAME ) into G-phase locusts ( n = 4 samples , Student’s t-test , *p<0 . 05 ) . ( E ) and ( F ) Effects on NOS activity and NO levels 2 hr after injection of NO donor ( SNAP ) into S-phase locusts ( n = 4 samples , Student’s t-test , *p<0 . 05 ) . n . s . means no significant difference . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 026 We have shown that NO levels were decreased by injection of either NPF1a or NPF2 and increased by knockdown of NPF1a or NPF2 transcripts ( Figure 4C ) . Next we asked whether the two NPFs suppress the NO signaling pathway . The mRNA and protein levels of NOS significantly decreased 4 hr after injection of NPF2 peptide into G-phase locusts ( Figure 7B , E ) . On the other hand , the mRNA and protein levels of NOS increased after knockdown of the NPF2 transcript in S-phase locusts ( Figure 7C , F ) . By contrast , no change in NOS mRNA level was observed in any treatments involving NPF1a ( Figure 7A , C ) . However , the level of phosphorylated NOS significantly decreased 1 hr after injection of NPF1a peptide into G-phase locusts ( Figure 7D ) and increased after knockdown of the NPF1a transcript in S-phase locusts ( Figure 7F ) . Injection of NPF1a or NPF2 peptide into G-phase locusts significantly decreased NOS activity and NO levels in a time-dependent manner , with NPF1a exhibiting an earlier inhibitory effect on NO signaling than NPF2 ( Figure 7G , H ) . Conversely , knockdown of NPF1a or NPF2 enhanced NOS activity in S-phase locusts ( Figure 7I ) , which is consistent with the changing patterns of NO levels in the same treatments ( Figure 4C ) . These data further verify the effects of NPF1a and NPF2 on NOS/NO signaling . 10 . 7554/eLife . 22526 . 027Figure 7 . Manipulations of NPF1a and NPF2 levels alter NOS activity and phosphorylation states in the brains of G-phase and S-phase locusts . ( A ) and ( B ) NOS mRNA levels after injection of NPF1a or NPF2 peptide into G-phase locusts . The data are presented as mean ± s . e . m . Significant differences are denoted by letters ( n = 4 samples , 8 locusts/sample , one-way ANOVA , p<0 . 05 ) . ( C ) NOS mRNA levels 48 hr after transcript knockdown of NPF1a or NPF2 in S-phase locusts ( n = 4 samples , one-way ANOVA , p<0 . 05 ) . ( D ) and ( E ) NOS protein levels after injection of NPF1a or NPF2 peptide into G-phase locusts ( n = 3 samples , 10–12 locusts/sample , one-way ANOVA , p<0 . 05 ) . ( F ) NOS protein levels 48 hr after transcript knockdown of NPF1a or NPF2 in S-phase locusts ( n = 3 samples , one-way ANOVA , p<0 . 05 ) . ( G ) NOS activity after injection of NPF1a or NPF2 peptide into G-phase locusts ( n = 4 samples , 12–16 locusts/sample , one-way ANOVA , p<0 . 05 ) . ( H ) NO levels after injection of NPF1a or NPF2 peptide into G-phase locusts ( n = 4 samples , 12–16 locusts/sample , one-way ANOVA , p<0 . 05 ) . ( I ) NOS activity 48 hr after transcript knockdown of NPF1a or NPF2 in S-phase locusts ( n = 4 samples , one-way ANOVA , p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 027 To understand the neural basis for the interactions between two NPFs and NO signaling in the regulation of phase-related locomotion , we localized NOS and the two NPF peptides in the locust brain by double immunofluorescence staining . NOS was extensively expressed in the cell bodies of neurons in the pars intercerebralis and in the Kenyon cells anterior to the calyces of mushroom bodies in each brain hemisphere ( Figure 8 and Figure 8—figure supplement 1 ) . The distribution of NPF1a peptide was similar to that of NOS . NPF1a and NOS were co-localized in two regions , namely , the pars intercerebralis ( Figure 8 , upper ) and the pars lateralis anterior to the calyces of mushroom bodies ( Figure 8—figure supplement 1 ) . However , NPF2 showed co-localization with NOS only in the cell body of neurons in the pars intercerebralis ( Figure 8 , lower ) . The co-localization of NPF1a and NPF2 with NOS in the pars intercerebralis of locust brain supports their linked action in phase-related behavioral changes . 10 . 7554/eLife . 22526 . 028Figure 8 . NOS and the two neuropeptides NPF1a and NPF2 co-localize in the pars intercerebralis of the locust brain . NPF1a and NOS also co-localize in the neurons of pars lateralis anterior to the calyces of mushroom in each hemisphere in the locust brain ( see Figure 8—figure supplement 1 ) . White arrows indicateNPF1a or NPF2 staining , yellow arrows show NOS staining , pink arrows indicate merged signal of NOS and NPF1a or NOS and NPF2 . Scale bars represent 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 02810 . 7554/eLife . 22526 . 029Figure 8—figure supplement 1 . NOS and the two neuropeptides NPF1a and NPF2 co-localize in the pars lateralis around the mushroom bodies in each hemisphere of locust brain . NC indicates negative control . White arrows indicate NPF1a staining , yellow arrows show NOS staining , pink arrows indicate merged signal of NPF1a and NOS . Scale bars represent 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 029 On the basis of the different binding activities between each NPF and the two receptors , we speculated that the two NPF receptors , NPFR and NPYR , are responsible for the distinct effects on NOS induced by NPF1a and NPF2 ( phosphorylated NOS levels were decreased by NPF1a injection whereas NOS transcript levels were reduced by NPF2 injection , as shown in Figure 7B , D ) . Knockdown of the NPFR transcript in S-phase locusts increased NOS phosphorylation level without affecting NOS transcript level ( Figure 9A , B ) , similar to the effect caused by NPF1a knockdown ( Figure 7A , D ) . By contrast , knockdown of the NPYR transcript led to increased NOS mRNA and NOS protein levels ( Figure 9C , D ) . Furthermore , we investigated whether NPF1a and NPF2 could affect NOS phosphorylation or transcript level in G-phase locusts in which the transcripts of NPFR or NPYR had been knocked down . We found that knockdown of the NPFR transcript relieved the inhibition of NOS phosphorylation caused by NPF1a administration ( Figure 9E , F ) , whereas knockdown of the NPYR transcript blocked NPF2-induced reduction in NOS mRNA and NOS protein levels in G-phase locusts ( Figure 9G , H ) . These data indicate that NPFR and NPYR mediate distinct effects of NPF1a and NPF2 on NOS phosphorylation and transcription , respectively , in the locust brain . 10 . 7554/eLife . 22526 . 030Figure 9 . Two receptors mediate distinct effects of NPF1a and NPF2 neuropeptides on NOS phosphorylation and on NOS transcript levels , respectively . ( A ) and ( B ) NOS mRNA levels ( n = 5 samples , 6–8 locusts/sample ) and NOS protein levels ( n = 3 samples , 10–12 locusts/sample ) 48 hr after transcript knockdown of NPFR in S-phase locusts . The data are presented as mean ± s . e . m . Significant differences are denoted by letters . n . s . means not significant ( Student’s t-test , *p<0 . 05 ) . ( C ) and ( D ) NOS mRNA levels ( n = 5 samples ) and NOS protein levels ( n = 3 samples ) 48 hr after transcript knockdown of NPYR in S-phase locusts . ( E ) and ( F ) NOS mRNA levels ( n = 4 samples ) and NOS protein levels ( n = 3 samples ) 4 hr after injection of NPF1a or NPF2 peptide in G-phase locusts pre-injected with dsNPFR . ( G ) and ( H ) NOS mRNA levels ( n = 4 samples ) and NOS protein levels ( n = 3 samples ) 4 hr after injection of NPF1a or NPF2 peptide in G-phase locusts pre-injected with dsNPYR . Detailed data describing NOS expression after injection of NPF1a or NPF2 peptide in G-phase locusts pre-injected with dsNPFR or dsNPYR are shown in Figure 9—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 03010 . 7554/eLife . 22526 . 031Figure 9—source data 1 . NOS mRNA levels and NOS protein levels after injection of NPF1a or NPF2 peptide into G-phase locusts pre-injected with dsNPFR or dsNPYR . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 031 To determine whether the NPF-induced NO reduction directly regulates phase-related locomotor plasticity , we conducted rescue experiments by administrating SNAP to enhance NO concentration in G-phase locusts pre-treated with NPF1a or NPF2 peptide . SNAP administration resulted in robust recovery of the Pgreg values , total duration of movement , and total distance moved for G-phase locusts in which Pgreg values had been reduced by injection of either NPF1a or NPF2 peptide ( Figure 10A ) . 10 . 7554/eLife . 22526 . 032Figure 10 . NPF1a , NPF2 and compounds that affect NO levels mediate effects on the locomotor behavior of G-phase and S-phase locusts . ( A ) Behavioral test after administration of NO donor ( SNAP ) to G-phase locusts pre-injected with NPF1a or NPF2 peptide . Significant differences are denoted by letters . For Pgreg analysis , lines indicate median value ( n ≥ 24 locusts; Mann–Whitney U test , p=0 . 0003 and 0 . 0001 for Pgreg NPF1a&SNAPvs . Pgreg NPF1a and Pgreg NPF2&SNAPvs . Pgreg NPF2 , respectively ) . For TDM and TDMV analysis , the data are presented as mean ± s . e . m . ( n ≥ 24 locusts , Student’s t-test , p<0 . 05 ) . ( B ) Behavioral test after administration of NOS inhibitor ( L-NAME ) in S-phase locusts pre-injected with dsNPFR or dsNPYR ( n ≥ 16 locusts , Mann–Whitney U test , p=0 . 022 and 0 . 042 for Pgreg dsNPFR&L-NAMEvs . Pgreg dsNPFR and Pgreg dsNPYR&L-NAMEvs . Pgreg dsNPYR , respectively ) . For TDM and TDMV analysis , the data are presented as mean ± s . e . m . ( n ≥ 16 locusts , Student’s t-test , p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22526 . 032 We then tested the effects of the NOS inhibitor L-NAME in S-phase locusts that had been pre-treated with dsNPFR or dsNPYR . Transcript knockdown of either NPFR or NPYR enhanced phase-related locomotor activity and thus promoted the behavioral shift from S-phase state towards G-phase state ( Figure 10B ) . However , L-NAME administration robustly abolished the increase in Pgreg values , total duration of movement , and total distance moved for test locusts induced by NPFR or NPYR transcript knockdown . These data suggest that NO signaling is an essential mediator for the effects of two NPFs and their receptors on phase-related locomotor plasticity in locusts .
We show that manipulating the levels of two NPFs by peptide injection or transcript knockdown significantly affects phase-related behaviors among four neuropeptides that had differential levels during locust phase transition . These changes in locomotor behavior can be fully overcome by pharmacological administration of compounds that affect NO levels . Notably , NO signaling displays marked effects on locomotor activity; and the time-course changes in NO levels coincide well with locust behavioral transitions during both isolation and crowding ( Guo et al . , 2011 ) , indicating NO is a decisive molecule for phase-related locomotion . The increased NO concentration may serve as a proximate cause of high locomotor activity in G-phase locusts , and decreased NO levels may lead to low locomotor activity in S-phase locusts . These data clearly suggest that the NPF/NO pathway plays a vital role in the modulation of phase-related locomotor plasticity . Our studies do not , however , preclude the regulatory roles of two other neuropeptides , ACP and ILP , in other phase-related characteristics or in long-term behavioral effects ( Pener and Simpson , 2009 ) . Numerous studies have suggested that NPF signaling can influence a broad range of physiological and behavioral activities in insects , for instance feeding , reproduction , learning , circadian activity and stress responses ( Nässel and Wegener , 2011; Lee et al . , 2006; Krashes et al . , 2009 ) . Similar functional roles of the NPF ( or NPFY ) system in locomotion have been observed in two model invertebrate species , Caenorhabditis elegans ( de Bono and Bargmann , 1998 ) and Drosophila melanogaster ( Wu et al . , 2003 ) . These findings , in conjunction with ours , raise the possibility that the NPF system might serve as a common neural signaling pathway that shapes locomotor plasticity in invertebrates . In addition , recent studies have reported that NPF ( referred to as NPF1a here ) can mediate food intake , body weight and male-specific reproduction processes in the adults of another locust species , Schistocerca gregaria ( Van Wielendaele et al . , 2013a , 2013b ) , indicating that NPF plays multiple roles in locust biology . Previous studies have suggested that phase differences in food choice are related to cryptic and aposematic strategies . Gregarious nymphs of S . gregaria are prone to consume more nutritionally imbalanced food and to accept food containing toxic chemicals more readily than do solitarious nymphs ( Simpson et al . , 2002 ) . This is partly due to the relative gustatory insensitivity to low-quality food in gregarious locusts ( Despland and Simpson , 2005 ) . A link between NPF signaling and food choice has been suggested in Drosophila ( Shen and Cai , 2001 ) . Thus , a possible involvement of the NPF system could be speculated in phase-related behavioral choices of food quality in the locust nymphs . Further functional analysis is required to confirm this possibility and to elucidate the mechanisms through which NPF regulates multiple phase-related behavioral characteristics . Our results indicate that NO signaling has a stimulatory role in locomotor activity in locusts . Several studies have suggested the significance of NO signaling in locomotor activity and behavioral plasticity in various animal species ( Del Bel et al . , 2005; Kyriakatos et al . , 2009 ) . For example , NO-initiating signaling has been shown to suppress aggression by promoting the tendency to flee in crickets ( Stevenson PA , 2016 ) and to increase oviposition digging rhythm so as to control egg-laying movements in desert locusts ( Newland and Yates , 2007 ) . The cGMP/protein kinase G ( PKG ) pathway ( the main downstream target of NO signaling ) is involved in the control of foraging and locomotor behavior in Drosophila ( Osborne et al . , 1997 ) , as well as in the regulation of labor division in honey bees ( Ben-Shahar , 2005 ) and ants ( Ingram et al . , 2005 ) . However , although there is a phase-dependent regulation of NO synthesis ( Rogers et al . , 2004 ) and a higher PKG activity in the anterior midline of brains of insects in the gregarious phase ( Lucas et al . , 2010 ) , significant effects of PKG on behavioral phase state could not be observed in the desert locust S . gregaria ( Ott et al . , 2012 ) . A reasonable explanation could be that the regulatory mechanisms of NO signaling in phase transition are species-specific , as is true of the roles played by several other neurotransmitters ( Ma et al . , 2011; Anstey et al . , 2009 ) . Another possibility is that NO regulates behavioral phase transition via a PKG-independent pathway in locusts ( Newland and Yates , 2008 ) . We provide clear evidence that two NPFs acts as brakes that sequentially modify NO levels to control locomotor plasticity . The regulatory role of the NPF-NO pathway in locomotor behavior is further supported by the overlap immunostaining of two NPFs and NOS in the pars intercerebralis , which is linked to the regulation of locomotor rhythm in insects ( Matsui et al . , 2008 ) . NO levels may reflect distinct physiological states and affect a wide variety of behaviors across species ( Collmann et al . , 2004; Davies , 2000; Del Bel et al . 2005; Cayre et al . , 2005 ) , yet how this molecule’s level responds to varied internal or external conditions remains unclear . To the best of our knowledge , this study is the first to show the link between NPF and NO signaling in shaping behavioral plasticity . We show that the sequential inhibitory effects of NPF1a and NPF2 on NO levels are attributed to their regulation of NOS phosphorylation and NOS gene transcription , respectively , indicating that these two NPF members are not redundant in regulating phase-related locomotion . Phosphorylation is known to be an important form of post-translational modification ( PTM ) for a broad range of proteins , including receptors , transcriptional factors and vital enzymes ( Kasuga et al . , 1982; Matsuzaki et al . , 2003; Bertorello et al . , 1991 ) . The phosphorylated proteins usually display changed spatial structures , subcellular locations and catalytic activity , and thus play key roles in rapid cellular signaling ( Aguirre et al . , 2002; Ho et al . , 2011; Hurley et al . , 1990 ) . Studies in mammals have shown that NOS activity is tightly regulated by phosphorylation . For instance , the phosphorylation of Ser1412 stimulates NOS activity whereas Ser847 phosphorylation inhibits enzyme activity ( Watts et al . , 2013; Komeima et al . , 2000 ) . NOS has also been suggested to be modified post-translationally in the locust embryo ( Stern et al . , 2010 ) . Here , we show that NOS is modified by phosphorylation in the locust brains . Even if the total NOS protein level were not influenced by the activities of NPFs , simply reducing NOS phosphorylation leads to significantly decreased NOS activity and thus results in lower NO level , suggesting that NOS activity in the locust largely depends on its modification by phosphorylation . Therefore , NPF1a may lower the NO level by directly reducing NOS phosphorylation , whereas NPF2 may lower the NO level by reducing NOS substrate for phosphorylation . Our results show that NPF1a-regulated NOS phosphorylation cycles quickly , whereas NOS expression may respond more slowly to NPF2 regulation . Thus , the distinct modes of changing NO levels that are regulated by the two NPF systems not only explain the more rapid behavioral effect of NPF1a when compared to that of NPF2 , but also emphasizes that downregulation of the NPF2 system is necessary in the G-phase to increase locomotion . We show that two NPF peptides and their receptors may play synergistic roles in regulating the dynamic changes in NO levels during the two time-course processes of phase transition . The continuous reduction of NO levels during isolation is tightly controlled by the decreased NOS phosphorylation that results from the upregulation of NPFR and NPYR . By contrast , the reduction of two NPFs contributes mainly to the overall enhancement of NOS phosphorylation and NO levels during crowding . Although phosphorylated NOS shows greater activity than the unphosphorylated protein in promoting NO production , as shown previously , the enhancement of NO levels upon crowding seems to be delayed relative to that of NOS phosphorylation , implying that the stimulation of NO levels during crowding is a complex process that might involve additional regulators beyond the enzyme activity . NO level is dependent upon the balance between its production and degradation ( Sansbury and Hill , 2014 ) . NO generation not only depends on NOS expression and its post-translational modification but also relies on the availability of the corresponding substrate ( e . g . , L-Arginine ) and cofactor ( e . g . , BH4 , FAD or FMN ) ( Li and Poulos , 2005 ) , whereas NO degradation may result directly from its reaction with reactive species ( e . g . , superoxide ) ( Channon , 2012 ) . Given this , modulations of the availability of these factors may responsible for the sluggish increase of NO level during locust crowding . Locomotor activity is a major phase-related behavior that changes in response to population density ( Wang and Kang , 2014 ) . The high locomotor activity of G-phase locusts is potentially beneficial for rapid aggregation , synchronous movement , and avoidance of predators or conspecific cannibalism during locust swarming ( Simpson et al . , 1999 ) . Therefore , the sequential modifications of NO levels resulting from NPF1a and NPF2 should allow dynamic locomotor adaptation to maintain locust swarming . Our previous studies have indicated that several other regulators , such as dopamine , serotonin and carnitines , are also involved in the modulation of phase-related locomotion in the migratory locust ( Wu , et al . , 2012; Ma et al . , 2015 ) . In addition , protein kinase A , a possible downstream factor of serotonin and dopamine , can regulate behavioral phase transition in the desert locust ( Ott et al . , 2012 ) . It has been shown that the NPF/NPFR pathway has a dominant suppressive effect on PKA-sensitized sugar aversion in Drosophila ( Xu et al . , 2010 ) . In our study , the expression level of AC2 , one of the enzymes catalyzing cAMP production and activating PKA , is also affected by alteration of NPF levels in locusts . Studies in mammals have shown that both dopaminergic transmission and PKA could enhance NO levels thus leading to distinct biological actions ( Wang and Lau , 2001; Yang et al . , 2011 ) . On the basis of these findings , we hypothesize that the two NPF systems may cooperate with the dopamine pathway to modulate locomotor activity during locust phase transition . We show that the NPF/NO pathway is not involved in the modulation of another major phase-related behavioral characteristic , conspecific attraction induced by odors , in the migratory locust . This finding is superficially inconsistent with previous results on the roles of NPFs or NO in fine-tuning of food odor-induced behavior and olfactory learning in mice and the fruit fly ( Rohwedder et al . , 2015; Sung et al . , 2014 ) . One possible explanation is that pheromone-induced olfactory behaviors that are related to the locust phase change may involve regulatory mechanisms that are different from those involved in food-odor-induced olfactory responses in locusts . And , the locust phase transition is a continuous process involving changes of various characteristics including behaviors , metabolism , immunity and body color ( Wang and Kang , 2014 ) . In addition to its significance in behavioral modulation , NO signaling is also able to affect a variety of physiological and pathological processes ( Bogdan , 2015; Calabrese et al . , 2004; Sansbury and Hill , 2014 ) . Thus , uncovering the long-term effects of the NPF/NO pathway on phase-related characteristics , such as disease resistance , energy metabolism and aging , will provide a more comprehensive understanding of the phase transitions that underlie locust swarming .
G-phase locusts were maintained in large well-ventilated cages ( 40 cm × 40 cm × 40 cm ) at a density of 500–1000 locusts per cage . S-phase locusts were reared individually in boxes ( 10 cm × 10 cm × 25 cm ) supplied with charcoal-filtered compressed air . Both colonies were maintained at 30 ± 2°C and under 14:10 light/dark photocycle regime . The locusts were fed with fresh wheat seedling and bran ( Guo et al . , 2011 ) . For solitarization , fourth-instar G-phase nymphs were separately raised under solitarious conditions as described above . After 0 , 1 , 4 , 16 , or 32 hr of isolation , locust brains were collected and snap frozen . For gregarization , two fourth-instar S-phase nymphs were reared in small cage ( 10 cm × 10 cm × 10 cm ) containing 20 G-phase locusts of the same developmental stage . After 0 , 1 , 4 , 16 , or 32 hr of crowding , locust brains were dissected and frozen in liquid nitrogen . All samples were stored at −80°C . Each sample contained a total of eight insects , including four male and four female insects . Four independent biological replicates were prepared for further experiments . Total RNA was extracted using the RNeasy Mini Kit ( Qiagen ) according to the manufacturer’s protocol . cDNA was reverse-transcribed from 2 μg of total RNA using M-MLV reverse transcriptase ( Promega ) . Gene-specific mRNA levels were assessed by qPCR using the SYBR Green kit on a LightCycler 480 instrument ( Roche ) . RP49 was used as internal reference . The primers used are shown in Supplementary file 2 . The dsRNAs of target genes were prepared using the T7 RiboMAX Express RNAi system ( Premega ) . dsRNA was microinjected into the brains of test insects ( 69 ng/locust for NPF1a , NPF2 , ACP , ILP , and NOS , 1 μg/locust for NPFR and NPYR ) . dsGFP-RNA was used as control in all RNAi experiments . The behaviors of test locusts were measured 48 hr after injection as described below . The concentrations of peptides and drugs that were used were determined according to described methods ( Newland and Yates , 2007; Van Wielendaele et al . , 2013b ) . The commercially synthesized peptides ( BGI , NPF1a peptide — YSQVARPRF-NH2; and NPF2 peptide — RPERPPMFTSPEELRNYLTQLSDFYASLGRPRF-NH2 ) were dissolved in ddH2O as stock solution ( 20 μg/μl ) . Working solutions of different concentrations ( 0 . 05 , 0 . 5 and 2 . 5 μg/μl ) were injected into the hemolymph in the heads of fourth-instar locusts using a microinjector ( 2 μl/locust ) . The arena behavioral assay was conducted 1 hr , 2 hr , and 4 hr following injection . L-NAME was dissolved with ddH2O to make a 1 mM stock solution . Working solution ( 100 μM ) was microinjected into the heads of G-phase locusts ( 2 μl/locust ) . SNAP was dissolved with DMSO to prepare 100 mM stock solution . Working solution ( 200 μM ) was microinjected into the heads of S-phase locusts ( 2 μl/locust ) . Locust behaviors were tested 2 hr after drug injection . The behavioral assay was performed in a rectangular Perspex arena ( 40 cm × 30 cm × 10 cm ) containing three chambers . The left chamber ( 7 . 5 cm × 30 cm × 10 cm ) contained 30 fourth-instar G-phase locusts as a stimulus group , and the right chamber was empty ( 7 . 5 cm × 30 cm × 10 cm ) . Locusts behaviors were recorded for 300 s by an EthoVision video tracking system and analyzed according to the binary logistic regression model constructed in our previous work ( Guo et al . , 2011 ) . Details are as follow: Pgreg = eη/ ( 1+eη ) ; η = −2 . 11 + 0 . 005 × AI ( attraction index ) + 0 . 012 × total distance moved +0 . 015 × total duration of movement; AI = total duration in stimulus area − total duration in the opposite of stimulus area; this parameter represents the extent to which the tested animals are attracted by the stimulus group . TDMV ( total duration of movement ) and TDM ( total distance moved ) indicate the locomotor activity levels . Pgreg indicates the probability that a locust is considered gregarious . Pgreg = 1 represents a fully gregarious behavior , whereas Pgreg = 0 represents a fully solitarious behavior . In the behavioral assay , 16–35 locusts were tested for each treatment according to the sample size reported in previous studies ( Ott et al . , 2012; Ma et al . , 2011 ) . Locusts that did not move during behavioral testing were excluded . The amino acid sequence of the Drosophila NPF receptor was used to search for NPF homologs in the locust genome database utilizing the tblastn algorithm . The phylogenetic relationship of NPFR and NPYR of insects and human was analyzed using MEGA software . HEK 293 T cells ( RRID: CVCL-0063 ) were purchased from the American Type Culture Collection ( ATCC , CRL-3216 , the identity has been authenticated using STR profiling ) and cultured in low glucose DMEM ( Life Technology ) supplemented with 10% fetal bovine serum . Cells were routinely tested for mycoplasma every 6 months . For the competition binding assays , HEK 293 T cells transiently transfected with pcDNA3 . 1-NPFR or pcDNA3 . 1-NPYR ( with a Flag-tag encoding sequence following target gene ) were washed with 1 X PBS and added into 96-well plates ( 2 × 105 cells/well ) coated with poly-L-lysine ( 0 . 1 mg/mL ) . Cells were then incubated with 25 μL TAMRA-NPF1a or TAMRA-NPF2 ( 10 nM ) in the presence of increasing concentrations of unlabeled ligands in a final volume of 100 μL of binding buffer ( PBS containing 0 . 1% bovine serum albumin ) . Nonspecific binding was determined by the addition of 25 μL unlabeled ligand . Mixtures were incubated at 30°C for 2 hr . Fluorescence intensity was measured using a fluorimeter ( Molecular Devices ) after washing twice with binding buffer . The HEK 293 T cells transfected with pcDNA3 . 1 were used as a control . The binding displacement curves were analyzed using the non-linear logistic regression method . Western blotting was carried out to validate the protein expressions of NPFR and NPYR in HEK 293 T cells using the mouse monoclonal antibody against Flag ( CoWin , 1:5000 ) . The brains of fourth-instar G-phase locusts were collected 4 hr after injection of the mixture of NPF1a and NPF2 peptides or ddH2O ( a total of 5 μg ) . Similarly , the brains of fourth-instar S-phase locusts were collected 48 hr after injection of the mixture of dsNPF1a and dsNPF2 or dsGFP . Each sample contained 10 brains ( 5 males and 5 females ) . Three independent replicates were performed for each treatment . Total RNA was isolated as previously described , and RNA quality was confirmed by agarose gel . cDNA libraries were prepared according to Illumina’s protocols . Raw data were filtered , corrected , and mapped to locust genome sequence using Tophat software . The number of total reads was normalized by multiple normalization factors . Transcript levels were calculated using the reads per kb million mapped reads criteria . The difference sbetween the test and control groups were represented by P values . Differentially expressed genes with significance levels at p<0 . 05 in each comparison were enriched . In addition , unsupervised hierarchical clustering was performed using Clustal 3 . 0 , which employs uncentered Pearson correlation and average linkage; results are presented by Java Treeview software . The RNA-seq data have been deposited in the Sequence Read Archive database of the National Center for Biotechnology Information ( NCBI ) ( accession no . SRP092214 ) . Locust brains ( 10–12 individuals/sample ) were collected and homogenized in 1 X PBS buffer ( 0 . 1 M phosphate buffer , 0 . 15 M NaCl , pH 7 . 4 ) containing the phosphatase inhibitor PhosSTOP ( Roche ) and a proteinase inhibitor ( CoWin ) . Total protein content was examined using the BicinChoninic Acid ( BCA ) Protein Assay Kit ( Thermo ) . The extracts ( 100 μg ) were reduced , denatured , and electrophoresed on 8% SDS-PAGE gel and then transferred to polyvinylidene difluoride membrane ( Millipore ) . The membrane was then cut to two pieces and incubated separately with a specific antibody against the target protein of ~130 KD or a reference protein of ~55 KD overnight at 4°C ( affinity-purified polyclonal rabbit antibody against uNOS , Sigma-Aldrich , 1:200; Rabbit polyclonal antibody against tubulin , CoWin , 1:5000 ) . Goat anti-rabbit IgG was used as secondary antibody ( CoWin , 1:10000 ) . Protein bands were detected by chemiluminescence ( ECL kit , CoWin ) . The band intensity of the Western blot was quantified using densitometry in Quantity One software . For determination of NOS phosphorylation , 200 μg brain extracts were incubated with λ phosphatase and 1 X NEB buffer supplemented with 1 mM Mncl2 for 1 hr at 30°C ( NEB ) . Control protein was treated under the same conditions without λ phosphatase . Western blot analysis was performed to confirm NOS phosphorylation in the locust brains . An enzyme-linked immunosorbent assay ( ELISA ) kit ( R and D Systems , Inc . ) was used to detect the relative cAMP level in locust brains . For NO content determination , Total Nitric Oxide Assay Kit ( Beyotime ) was used . Because NO molecules are unstable , the total NO levels in all test groups were assessed by detecting the content of nitrate and nitrite . The NOS Kit ( Nanjing Jiancheng Bioengineering Institute ) was used to detect the total NOS activity in locust brains . Protein concentrations were measured using the BicinChoninic Acid ( BCA ) Protein Assay Kit ( Thermo ) . All of these three measurements were performed according to the manufacturer’s instructions . Each measurement was from at least four biological replicates ( 12–16 locusts/replicate ) . Data were normalized to the protein concentration . Whole-mount double immunohistochemistry of locust brains was performed using affinity-purified polyclonal rabbit antibody against NPF1a or NPF2 ( AbMAX , China , 1:100 ) and monoclonal mouse antibody against uNOS ( Thermo , 1:200 , RRID: AB_325476 ) . Alexa Fluor-488 goat anti-rabbit IgG ( Cat . A-11008 , 1:500; Life Technologies ) and Alexa Fluor-568 goat anti-mouse IgG ( Cat . A-11019 , 1:1000; Life Technologies ) were used as secondary antibodies for NPFs and NOS staining , respectively . Fluorescence was detected using an LSM 710 confocal laser-scanning microscope ( Zeiss ) . Photos for both positive staining and negative controls were imaged under the same conditions . To validate the involvement of NPFR and NPYR in the regulation of NOS expression and phosphorylation by two NPF peptides , the brains of fourth-instar S-phase locusts were microinjected with dsNPFR or dsNPYR , and collected 48 hr after injection . For gregaria , the brains of fourth-instar locusts were microinjected with dsNPFR , dsNPYR or dsGFP followed by NPF1a or NPF2 treatment 4 hr before sample collection . Total RNA and protein in each treatment were extracted according to the Invitrogen TRIzol RNA and protein extraction protocol . qPCR and Western blot analysis were performed to examine the influence of NPFR and NPYR on NOS expression and phosphorylation . For G-phase locusts , synthesized NPF1a or NPF2 peptide ( 2 . 5 μg/μl ) was microinjected into the fourth-instar insects . Two hours later , SNAP ( 200 μM , 2 μl/locust ) was injected into the heads of the experimental insects . Control insects were treated with an equal amount of saline . The injected locusts were then raised under the gregarious condition and subjected to behavioral analysis 2 hr after injection of SNAP . For S-phase locusts , dsNPFR or dsNPYR was microinjected into the brains of fourth-instar S-phase insects . Forty-six hours after injection , the NOS inhibitor L-NAME was microinjected into the locusts pre-treated with dsNPFR or dsNPYR . The insects treated with dsGFP were used as a control . Tested insects were thus raised under the solitarious condition and subjected to behavioral analysis 2 hr after injection of L-NAME . For gene expression and enzyme activity analysis , we knew from the previous studies that a sample size of 6 animals per treatment was enough to detect significant differences among treatments ( Ott et al . , 2012; Yang et al . , 2014 ) . Therefore , 8–16 animals were examined in each experimental treatment . For behavioral measurement , we knew that 15 individuals per group was sufficient to detect reproducible differences between groups ( Ma et al . , 2011 ) . All of the experiments were performed with at least three independent biological replicates . Student’s t-test was used for two-group comparison . One-way ANOVO followed by Turkey’s post-hoc test was used for multi-group comparisons . Data that do not meet normal distribution were excluded in these statistics . Behavioral phase state analysis was performed using the Mann–Whitney U test because of its non-normal distribution feature . Differences were considered statistically significant at p<0 . 05 . Data were analyzed using SPSS 20 software and presented as mean ± s . e . m . except for the Pgreg values , which are shown as median values . | Migratory locusts are widespread throughout the Eastern Hemisphere , especially in Asia , Australia and Africa . Although usually solitary insects , locusts can also form swarms made up of millions of individuals , which can devastate crops . Swarming can be studied on a smaller scale in the laboratory by forcing locusts to crowd together . This causes the locusts to enter a so-called gregarious state in which they are more active and sociable , which in turn promotes swarming . Isolating individual locusts has the opposite effect , causing the insects to enter a solitary state in which they are less active . Chemicals in the locust brain called neuropeptides control phase transitions between solitary and gregarious behavior . Neuropeptides bind to specific proteins called receptors in the outer membranes of neurons and initiate unique signaling cascades inside cells . However , exactly how neuropeptides regulate the changes in locust behavior that affect swarming was not clear . Hou et al . now reveal the role that two related neuropeptides , NPF1a and NPF2 , play in this process . Crowding causes the levels of NPF1a and NPF2 in the locust brain to decrease , whereas isolating individual locusts causes the levels of two NPF receptors to increase . Both neuropeptides reduce levels of a molecule called nitric oxide in the brain . NPF1a does this by reducing the activity of the enzyme that produces nitric oxide , whereas NPF2 reduces the production of this enzyme . The reduction in nitric oxide in turn makes the locusts less active . Similar NPF neuropeptides had previously been shown to affect activity levels in other invertebrates , such as roundworms and fruit flies . This , combined with the results now presented by Hou et al . , suggests that the NPF/nitric oxide pathway may regulate activity in insects in general . Future work should investigate this possibility , as well as whether the NPF/nitric oxide pathway controls changes in other insect behaviors such as feeding and mating . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"neuroscience"
] | 2017 | The neuropeptide F/nitric oxide pathway is essential for shaping locomotor plasticity underlying locust phase transition |
The gaseous plant hormone ethylene regulates a multitude of growth and developmental processes . How the numerous growth control pathways are coordinated by the ethylene transcriptional response remains elusive . We characterized the dynamic ethylene transcriptional response by identifying targets of the master regulator of the ethylene signaling pathway , ETHYLENE INSENSITIVE3 ( EIN3 ) , using chromatin immunoprecipitation sequencing and transcript sequencing during a timecourse of ethylene treatment . Ethylene-induced transcription occurs in temporal waves regulated by EIN3 , suggesting distinct layers of transcriptional control . EIN3 binding was found to modulate a multitude of downstream transcriptional cascades , including a major feedback regulatory circuitry of the ethylene signaling pathway , as well as integrating numerous connections between most of the hormone mediated growth response pathways . These findings provide direct evidence linking each of the major plant growth and development networks in novel ways .
Despite the importance of the plant hormone ethylene , we lack a comprehensive understanding of how its linear signaling pathway mediates many different morphological responses . The dynamic ethylene physiological response , a rapid growth inhibition independent of the master transcriptional regulator ETHYLENE INSENSITIVE3 ( EIN3 ) , followed by an EIN3-dependent sustained growth inhibition , calls for a temporal study of ethylene transcriptional regulation ( Binder et al . , 2004a ) . EIN3 has been shown to be necessary and sufficient for the ethylene response and accumulates upon a duration of exogenous ethylene gas treatment ( Guo and Ecker , 2003 ) . Although hundreds of ethylene response genes have been identified , because some of the targets of EIN3 are transcription factors ( e . g . ETHYLENE RESPONSE FACTOR1 [ERF1] ) , it is challenging to distinguish immediate early targets from those further downstream . To understand the dynamics of the EIN3-mediated ethylene transcriptional response , we performed a genome-wide study of the ethylene-induced EIN3 protein-DNA interactions using chromatin immunoprecipitation followed by sequencing ( ChIP-Seq ) and simultaneously determined the repertoire of target genes that are transcriptionally regulated by ethylene ( mRNA-Seq ) . Tracing the transcriptional cascade , we asked if EIN3-mediated genes contribute to a component of the ethylene transcriptional response . For a select number of EIN3 targets that are putative transcriptional regulators , DNA-binding motifs were identified using protein binding microarrays ( PBM ) and the enrichment for these motifs in the promoters of ethylene response genes was determined .
We performed ChIP-Seq using a native antibody that recognizes EIN3 ( Guo and Ecker , 2003 ) as well as mRNA-Seq in three-day-old dark grown seedlings during a timecourse of ethylene treatment ( Figure 1—figure supplements 1 , 2; Supplementary file 1A ) . By stringent analysis of the temporal ChIP-Seq data ( see ‘Materials and methods’ ) , we identified 1460 EIN3 binding regions in the Arabidopsis genome associated with 1314 genes ( Supplementary file 1B ) . We refer to genes associated with EIN3 binding regions as EIN3 candidate targets . In the sequences of EIN3 binding regions , we found significant enrichment of the consensus TEIL motif ( Hypergeometric p<10−87 ) ( Kosugi and Ohashi , 2000 ) , and de novo motif analysis identified the known EIN3 motif ( Figure 1—figure supplement 3 ) . We detected three previously described EIN3 targets using our stringent analysis ( Figure 1—figure supplements 3 , 4 ) ( Solano et al . , 1998; Konishi and Yanagisawa , 2008; Chen et al . , 2009; Zhong et al . , 2009; Boutrot et al . , 2010 ) . One example of a known target of EIN3 , EIN3-BINDING F-BOX PROTEIN 2 ( EBF2 ) , is shown in Figure 1A . EBF2 directs the proteolysis of EIN3 and exhibits ethylene-induced transcription ( Figure 1A ) , resulting in feedback regulation of the ethylene signaling pathway . Our study identified additional distal EIN3 binding in the EBF2 promoter region ( Figure 1A , Figure 1—figure supplement 4 ) . 10 . 7554/eLife . 00675 . 003Figure 1 . Dynamics of ethylene-induced EIN3 binding and transcription supports the role of EIN3 as an activator of the ethylene response . ( A ) Ethylene treatment results in an increase of EIN3 binding in three regions of the EBF2 promoter , corresponding to an increase in steady-state mRNA levels . Binding and transcription levels are indicated by reads per kilobase per million reads in sample ( RPKM ) . Gene model: green ( exon ) , red ( UTR ) , grey ( intron/transposon ) . ( B ) Patterns of EIN3 binding and expression of ethylene-regulated targets are strikingly evident over a timecourse of ethylene gas treatment . EIN3 binding increases with ethylene treatment to a maximum at 4 hr of ethylene treatment for all candidate targets . Each line in the heatmap represents the RPKM value for the representative EIN3 binding site ( left panel ) and transcript ( right panel ) . ( C ) ( Upper panel ) Equivalent numbers of genes are up- and down-regulated upon ethylene treatment . ( Lower panel ) Majority of EIN3 targets differentially expressed upon ethylene treatment are up-regulated . ( D ) A subset of EIN3 targets is transcriptionally regulated by ethylene ( EIN3-R ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 00310 . 7554/eLife . 00675 . 004Figure 1—figure supplement 1 . EIN3 antibody reproducibly enriches DNA in chromatin immunoprecipitation . ( A ) Enrichment of the known target of EIN3 , the promoter of ERF1 , using Dynabeads Protein A and Dynabeads Sheep anti-Rabbit IgG to collect protein-DNA complexes . The average fold change for two technical ChIP replicates with three QPCR technical replicates each is shown . ( B ) Reproducibility in the two biological replicates for EIN3 ChIP performed upon treatment of ethylene gas for 0 , 0 . 5 , 1 , and 4 hr . ( C ) Average RPKM of EIN3 binding sites 0 , 0 . 5 , 1 , and 4 hr of ethylene gas treatment . ( D ) EIN3 binding preferentially occurs in the promoter regions of genes ( 1000 bp upstream of the TSS ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 00410 . 7554/eLife . 00675 . 005Figure 1—figure supplement 2 . Ethylene-regulated genes are induced and repressed . ( A ) Reproducibility of RNA-Seq experiments . Ethylene-regulated have a higher reproducibility amongst replicates than non-replicates . ( B ) Genes that are both up- and down-regulated occur at different timepoints of ethylene treatment . ( C ) The majority of EIN3 targets exhibit increased binding upon ethylene treatment , however , changes in steady-state levels of mRNA do not occur for the majority of these targets . EIN3-NR genes are indicated by the blue rectangle . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 00510 . 7554/eLife . 00675 . 006Figure 1—figure supplement 3 . Binding of EIN3 to previously known targets . ( A ) De novo motif from the top 50 EIN3 binding sites with the best match to the known EIN3 motif ( E-value = 1 . 12 × 10−5 ) . EIN3 binding of the promoters of ( B ) ERF1 , ( C ) EDF1 , and ( D ) FLS2 . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 00610 . 7554/eLife . 00675 . 007Figure 1—figure supplement 4 . EIN3 ChIP-Seq identified an additional binding in the EBF2 promoter . ( A ) Binding of EIN3 to the EBF2 promoter increases upon ethylene gas treatment . EIN3 binding is strongest in the most proximal site to the TSS , and weakest in the most distal site to the TSS , which is not called by our stringent analysis . The most distal EIN3 binding site was not detected by our analysis but displays the characteristic ChIP-Seq binding pattern . ( B ) Alignment of motifs of the three binding sites in the EBF2 promoter . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 00710 . 7554/eLife . 00675 . 008Figure 1—figure supplement 5 . Functional categories are over-represented for EIN3 targets that are ethylene-regulated ( EIN3-R ) . Network was generated using BiNGO ( v . 2 . 44 ) using the GOSlim_Plants ontology , Benjamini and Hochberg p-value legend is indicated below . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 008 The majority of studies that exist in the literature have shown that EIN3 acts as an activator , and we observed this activation at the genome-wide level ( Figure 1B ) . We found that a majority of EIN3 candidate targets that are regulated by ethylene ( referred to as EIN3-R ) are induced ( 85% ) , Moreover , when compared to the regulation of all genes that respond to ethylene , we observed an over-representation of up-regulation of EIN3 candidate targets ( Figure 1B , C ) . Interestingly , many EIN3-R are transcription factors ( ∼14% ) ; EIN3 candidate targets are significantly enriched in gene ontology ( GO ) terms related to transcription factor regulation , confirming that EIN3 activates a transcriptional cascade ( Figure 1—figure supplement 5; Supplementary file 1C ) ( Maere , 2005 ) . Numerous studies have reported that transcription factor binding does not necessarily coincide with changes in transcription ( Macquarrie et al . , 2011; Menet et al . , 2012 ) , especially for master regulators targeting other transcription factors or other factors involved in chromatin state regulation . Only about 30% of the EIN3 binding sites were associated with transcriptional changes , but at least two-thirds were not ( Figure 1D , Figure 1—figure supplement 2 ) . EIN3 candidate targets that are not transcriptionally activated may require cofactors to induce a change in expression for a specific environmental response or developmental program . Quantitatively , the changes in EIN3 binding and steady-state transcription upon ethylene treatment do not correlate because the temporal transcription patterns are very diverse ( Figure 2—figure supplement 1 ) . However , relatively high levels of EIN3 occupancy in etiolated seedlings treated with ethylene indeed correspond to increases in steady-state levels of transcription ( Figure 2A ) . In fact , we were able to differentiate the characteristics of EIN3 candidate targets that exhibited a transcriptional response to ethylene from those that do not ( Figure 2A ) . EIN3 candidate targets that exhibit increased occupancy and increased levels of transcription ( EIN3-R ) are functional targets , enriched in gene families with specific functions , for example BZR , TIFY , and bHLH transcription factor families , which play a role in other hormone pathways ( p<0 . 05 ) ( Figure 2B ) . The highest percentage of hormone-associated genes occurs in EIN3 candidate targets that are ethylene-regulated ( EIN3-R ) ( Figure 2B , inset ) , and it is likely that these EIN3-R targets are direct and/or functional . Other EIN3 candidate targets may play roles in different developmental stages/tissue types , or may be under spatial regulation , requiring specific cofactors . 10 . 7554/eLife . 00675 . 009Figure 2 . The ethylene transcriptional response occurs in four distinct waves of transcriptional induction . ( A ) Ethylene-regulated EIN3 targets ( EIN3-R ) exhibit increased binding at transcription start sites ( TSS ) upon ethylene treatment ( black arrows ) in comparison to those not transcriptionally regulated by ethylene ( EIN3-NR and EIN3-ND ) . Each boxplot represents the distribution of EIN3 ChIP-Seq RPKMs near the TSS . ( B ) Distribution of gene families among EIN3-R targets reveals over-representation of gene families related to hormone responses function . ( Inset ) Percentage of hormone-related genes in EIN3 binding and transcription categories . ( C ) DREM paths representing waves of induction of steady-state levels of transcription by ethylene for genes that are regulated by EIN3 , implicating different modes of transcriptional regulation in the ethylene response . Right panels contain all genes for each wave . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 00910 . 7554/eLife . 00675 . 010Figure 2—figure supplement 1 . Quantitative correlation between EIN3 binding and ethylene-regulated expression . Linear regression revealed that there is no quantitative correlation of changes in EIN3 binding and ethylene-regulated steady-state levels of transcription . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 01010 . 7554/eLife . 00675 . 011Figure 2—figure supplement 2 . Temporal characterization of the ethylene transcriptional response . ( A ) The EIN3-modulated ethylene transcriptional response occurs in four waves with various levels of noise . A decrease in standard deviation correlates to an increase of hormone-related genes . ( B ) The kinetics of the ethylene transcriptional response on a linear scale . Approximately 35–50% of the ethylene-regulated genes are transcriptionally affected by the first 4 hr of ethylene treatment , for all ethylene transcriptional response genes and EIN3-modulated ethylene transcriptional response genes , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 01110 . 7554/eLife . 00675 . 012Figure 2—figure supplement 3 . Ethylene transcription and associated transcription factor regulation kinetics from DREM analysis . This dynamic regulatory map contains major bifurcation or convergence events in temporal expression data and determines the transcription factors potentially responsible for these events . EIN3-modulation is significant ( 10−10 ) in the expression trajectories labeled EIN3 ( teal , dark brown , purple , green-blue; first four endpoint trajectories from the top of the figure ) . The EIN3-associated trajectories represent the four waves of ethylene-regulated expression regulated by EIN3 , as further discussed in the main text . The TF associations are based on motifs from AGRIS and the PBM data in this paper . There are two caveats based on the TF association annotation: ( 1 ) It is possible the TF controlling these RNA is not represented in the database , and ( 2 ) Genes may not be regulated by the specific TF indicated , but instead by a homolog or other TF that has a similar target sequence not in the database . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 012 Projection of the dynamic EIN3 binding ( ChIP-Seq ) onto the transcriptional ethylene response ( mRNA-Seq ) using the Dynamic Regulatory Events Miner ( DREM ) ( Ernst et al . , 2007 ) revealed that the ethylene response occurs in four waves of transcription significantly regulated by EIN3 ( Pathway hypergeometric p<10−10 ) ( Figure 2C ) . These waves display distinct temporal transcription behaviors ( Hypergeometric p<0 . 001 ) , and the reduction of transcriptional noise occurs in successive temporal waves ( Figure 2C , Figure 2—figure supplement 2 ) . Genes that were enriched in specific biological functions within these four transcriptional waves include RNA binding/translation ( Wave 1 , Wave 3 ) , cell wall maintenance ( Wave 2 ) , and response to endogenous stimulus ( Wave 4 ) . The second wave is also enriched for genes involved in cell wall maintenance , and the expression of these genes steadily increases following 1 hr of ethylene treatment , consistent with kinetics of EIN3-dependent growth inhibition ( Binder et al . , 2004a; Vandenbussche et al . , 2012 ) . The four waves of the ethylene transcriptional response each contain a unique subset of EIN3 candidate targets . The first wave is highly variable , lower in steady-state levels of transcription , and it also contains the lowest percentage of EIN3 candidate targets and hormone-related genes ( Figure 2C ) . Previous ethylene growth rate inhibition studies have shown that low amounts of ethylene can result in adaptation and desensitization to subsequent ethylene stimulation ( Binder et al . , 2004a , 2004b ) . This first wave may serve as the immediate ethylene response , activating initial ethylene response genes as well as those that serve to desensitize the plant to subsequent ethylene stimulation , but this has yet to be shown . The next three waves of transcription are successively less variable and contain higher percentages of EIN3 candidate targets and hormone-related genes . The four waves of ethylene-induced transcription account for 50% of the transcriptionally ethylene-regulated EIN3 targets ( EIN3-R ) , and the remaining EIN3 candidate targets are distributed among other patterns of transcription that do not contain significant numbers of EIN3 candidate targets in each transcriptional trajectory ( Pathway hypergeometric p<10−10 ) ( Figure 2—figure supplement 3 ) . The expression kinetics and reduction of transcriptional noise we observe in the ethylene-induced waves may be tied to distinct mechanisms of transcriptional control , or they may reflect heterogeneity of the ethylene response in different tissues , which can be resolved using single cell analysis . From the temporal ethylene transcriptional response patterns , it appears that the initial early ethylene transcriptional response is noisy and less focused functionally . During sustained exogenous ethylene application , EIN3 accumulates , and the established ethylene transcriptional response is hormone-focused and less noisy , but feed-forward and feed-back mechanisms mentioned below may serve to establish this functional specificity . A recurring theme throughout this study is that the key players in the ethylene transcriptional response regulated by EIN3 are involved in plant hormone response pathways , and we anticipate a dense network of interconnections between the coregulated hormone pathways because hormones operate in concert , synergistically/antagonistically regulating growth and development . Although hormone pathway interconnections have been previously described by many groups ( Kaufmann et al . , 2009 , 2010; Sun et al . , 2010; Yu et al . , 2011 ) , here we show that these interconnections exist at many regulatory levels and that the targets of EIN3 may regulate genes in these responses . Among the EIN3 candidate targets , we observed the enrichment of hormone-related targets among many different categorical sets ( Figure 2B , inset ) . These EIN3 targets include downstream effectors of the ethylene response , key ethylene signaling players , and genes involved in other hormone pathways/responses . Many of the EIN3-modulated downstream effectors are members of the AP2/ERF transcription factor family , and as expected , these transcriptional initiators are up-regulated by ethylene ( Figure 3A , inset , green font ) . 10 . 7554/eLife . 00675 . 013Figure 3 . Functional classification of EIN3 candidate targets reveals genes involved in hormone responses . ( A ) Feedback ( ethylene signaling components , above ) of the ethylene response and feedforward ( downstream effectors , below ) . Downstream effectors in green are transcriptionally induced by ethylene . Known EIN3 targets are noted by asterisks; all other EIN3 candidate targets were discovered by this study . ( B and C ) EIN3 candidate targets are involved in hormone co-regulation . Node color represents hormone annotation , as indicated in B; large nodes are EIN3 candidate targets . Dark grey edges represent protein-protein interactions ( PPI ) and light grey edges are protein–DNA interactions ( PDI ) . Hormone annotation legend: abscisic acid ( ABA ) , brassinosteroid ( BR ) , cytokinin ( CK ) , ethylene ( ETH ) , gibberellin ( GA ) , auxin ( IAA ) , methyl jasmonate ( MJ ) , salicylic acid ( SA ) , >1 , more than one hormone . ( D ) EIN3-mediated ethylene co-regulation occurs at many different levels . PPIs are from the Arabidopsis Interactome Mapping Consortium , and EIN3 PDIs are from this study . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 01310 . 7554/eLife . 00675 . 014Figure 3—figure supplement 1 . Motifs of EIN3 targets that are transcriptionally regulated by ethylene were determined in vitro using protein binding microarrays . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 014 Given that EIN3 , the master regulator of the ethylene transcriptional response , acts at the culmination of the ethylene signal transduction pathway and is the transcriptional initiator of the ethylene response , one would expect a large number of downstream effectors to coordinate the transcriptional cascade and feedback regulators to maintain the circuitry in a homeostatic state as opposed to a feed-forward runaway response . Analysis of the ethylene-regulated EIN3 targets reveals a number of sites of ethylene signaling modulation of which the majority are negative regulators , supporting the idea that EIN3 is at the end of a signal transduction pathway , and that this regulatory logic dictates a negative feedback loop for homeostatic adaptable systems . More specifically , several negative regulators of the ethylene signaling pathway ( Kendrick and Chang , 2008 ) were targets of EIN3 ( Figure 3A ) , including three ethylene receptors ( ETHYLENE RESPONSE2 [ETR2] , ETHYLENE RESPONSE SENSOR1/2 [ERS1/2] ) , as well as REVERSION-TO-ETHYLENE SENSITIVITY1 ( RTE1 ) , CONSTITUTIVE TRIPLE RESPONSE1 ( CTR1 ) , and the previously mentioned EBF1/2 . The induction of ETR2 , ERS1/2 by ethylene was previously reported and has been suggested to restore ethylene receptor activity , resensitizing the plant to ethylene ( Binder et al . , 2004b; Vandenbussche et al . , 2012 ) . The negative regulation of ethylene signaling by EIN3 through induction of CTR1 and ETR2 is further supported by the literature ( Chen et al . , 2007 ) , suggesting that these proteins exhibit an increase in stabilization upon ethylene treatment ( Gao et al . , 2003 ) . The EIN3 candidate targets account for more than twice the proportion of hormone genes than in the genome ( 46% , Hypergeometric p=10−96 ) ( Figure 2B , inset ) ( Alonso et al . , 2003a; Nemhauser et al . , 2006; Peng et al . , 2008 ) . Many of the genes were involved in more than one hormone response , highlighting the extensive hormone co-regulation in Arabidopsis ( Figure 3B ) . Hormone co-regulation is evident in the protein-protein as well as the transcriptional regulator interactions and this network reveals interconnectivity suggestive of robust regulatory co-regulation ( Figure 3C ) . Many detailed examples of hormone co-regulation exist in the literature , but often the mechanism ( s ) of co-regulation is unknown . Previous ChIP-chip or ChIP-Seq studies from plants have also revealed cross-regulation within pathways involved in flowering and in roots ( Yant et al . , 2010; Iyer-Pascuzzi et al . , 2011; Winter et al . , 2011; Immink et al . , 2012 ) . The findings presented in our study suggest that ( 1 ) hormone co-regulation can occur through the binding of EIN3 , ( 2 ) EIN3 targets hormone pathways at multiple levels , and ( 3 ) some of these events are transcriptionally regulated by ethylene ( Figure 3D ) . Ethylene and jasmonate co-regulation occurs at the transcriptional level , sharing a complement of genes responsive to both hormones , for example RAP2 . 6L , ERF1 . EIN3 also targets four JAZ repressors , two of which are transcriptionally regulated by ethylene ( JAZ1 , JAZ6 ) . In general , ethylene and jasmonate are known to function synergistically and in the presence of jasmonate , JAZ1 proteins bound to EIN3 are degraded , relieving the EIN3 transcriptional activation ( Zhu et al . , 2011 ) . Here , the presence of an exogenous ethylene stimulus primes cells for a jasmonate response , by loading the promoters of jasmonate/ethylene response genes with EIN3 and JAZ proteins , poising the plant for a jasmonate-ethylene driven transcriptional program , as required for plant pathogen response . Reports of anticipatory binding in other organisms have been forth coming ( Macquarrie et al . , 2011; Lickwar et al . , 2012 ) . Ethylene and gibberellin co-regulation through EIN3 occurs at signal reception ( GID1B , GID1C ) and transcription ( PIF3 ) . The regulatory logic of EIN3 binding results in an up-regulation of the gibberellin response; GID receptors target DELLA repressors for degradation , which releases PIF3 from repression , resulting in the activation of the gibberellin transcriptional response . Additional support for feed-forward transcription is provided by over-representation of the PIF3 motif in the promoter sequences of the ethylene transcriptional response genes ( Supplementary file 1E , Figure 3—figure supplement 1 ) . Hormone co-regulation may also occur bidirectionally as a recent study reported negative regulation of ethylene by FUSCA3 ( FUS3 ) , known to regulate and be regulated by gibberellin and abscisic acid in embryonic and vegetative timing ( Lumba et al . , 2012 ) . FUS3 negatively regulates genes upstream and downstream of EIN3 ( EIN2 and ERF1 ) in leaf aging ( Lumba et al . , 2012 ) . Ethylene and auxin co-regulation occurs at both the level of transport and transcriptional response , as EIN3 modulates a regulator of auxin efflux ( PID ) and its upstream activator ( PBP1 ) , and at least seven auxin response proteins ( Supplementary file 1B ) . EIN3 also targets the auxin transporter ( AUX1 ) and an auxin signaling gene ( IAA29 ) , but these candidate targets are not responsive to ethylene in etiolated seedlings ( Supplementary file 1B ) . Ethylene has been reported to stimulate auxin transport through AUX1 away from the root apex , to decrease lateral root primordia ( Lewis et al . , 2011 ) . Therefore , it is likely these binding events have functional outcomes in specific tissue types or developmental programs not addressed in this study . The establishment of a transcriptional program tailored to result in a specific growth and development process requires multiple levels of transcriptional modulation . EIN3 was previously suggested to initiate a transcriptional cascade because it activates AP2/ERF transcription factors ERF1/EDF1 ( Solano et al . , 1998 ) . To determine additional candidate downstream effectors that may modulate the ethylene transcriptional response cascade , we used in vitro protein binding microarrays to generate DNA-binding motifs for 12 transcription factors that were ethylene-regulated targets of EIN3 ( see ‘Materials and methods’ ) . We then used the in vitro DNA-binding motifs to scan the promoter sequences of all ethylene transcriptional response genes ( Lam et al . , 2011 ) . EIN3 targets that may regulate a secondary transcriptional ethylene response include AP2/ERFs AT-ERF1 , ERF5 , and WRKY14/47 , PIF3 , NAC6 , and RAP2 . 2 , and the DNA-binding motifs of the aforementioned transcription factors are over-represented in the promoter regions of genes that are regulated by ethylene ( Hypergeometric p<10−5 ) ( Supplementary file 1E , Figure 3—figure supplement 1 ) . Future in vivo analyses of the targets of these transcription factors may help elucidate their contribution to the transcriptional cascade of the ethylene response . The extensive hormone co-regulation that occurs in waves of transcription leads to certain testable predictions regarding the key regulatory hubs and transcriptional cascades at a genome-wide level . Using a global approach , we are able to determine not only if one gene is a candidate target of EIN3 , but whether its homologs are targets as well . Transcription factor targeting of genes that are homologous , with overlapping and unique functions , can add diversity to the outputs of transcriptional programs ( Macquarrie et al . , 2011 ) . One of the most striking and surprising example we found was the direct regulation of the three homologs by EIN3 , HOOKLESS1 ( HLS1 ) and HLS1-LIKE HOMOLOG2 ( HLH2 ) , and to a lesser extent HLH1 ( Figure 4A , HLH1 in Figure 4—figure supplement 1 ) . This led us to experimentally test the functionality of all four members of the HLS1 gene family in etiolated seedling growth and development . HLS1 is a well-known signal integrator of ethylene , light , auxin , sugar , and brassinolide ( de Grauwe et al . , 2005; Hou et al . , 1993; Li et al . , 2004; Ohto et al . , 2006 ) and was previously hypothesized to be a target of ERF1 because of the presence of a GCC box motif in the HLS1 promoter region sequence ( Lehman et al . , 1996 ) . The binding of EIN3 to the promoters of HLS1 , HLH2 , and HLH1 increased upon ethylene treatment ( Figure 4—figure supplement 1 ) and is specific to EIN3 ( Figure 4B ) . The EIN3 binding sites in these promoters contain known EIN3 motifs ( Figure 4—figure supplement 1 ) . The functional significance of the HLS1 EIN3 binding site is supported by a previous study that identified two allelic mutations in the HLS1 promoter sufficient to yield a ‘hookless’ phenotype ( Lehman et al . , 1996 ) . Previous studies have also shown that ein2 is deficient in the accumulation of EIN3 protein ( Guo and Ecker , 2003 ) and HLS1 mRNA , ( Lehman et al . , 1996 ) . We also observed HLS1 steady-state transcript levels were significantly reduced in the ein3-1 eil1-1 mutant ( Figure 4—figure supplement 2 ) . 10 . 7554/eLife . 00675 . 015Figure 4 . EIN3 binding facilitates HLS1 ethylene-auxin hormone co-regulation . ( A ) ( Top panel ) EIN3 targets HLS1 and HLH2 . Temporal EIN3 binding and expression patterns are shown with known EIN3 targets as a control . HLH1 and HLH3 are not expressed in etiolated seedlings . ( B ) Binding of EIN3 to HLS1/HLH2 promoters is dependent on presence of EIN3 . ( C ) – ( F ) Mutations in HLS1 and its homologs reveal severe growth and developmental defects . ( C ) Tri-cotyledon phenotypes in apical hook of quadruple mutants . Images were taken at the same magnification . ( D ) HLS1 gene family has a role in embryo patterning . SEM image scale bar , 50 μm . ( E ) Adult three-week-old plants displayed dwarfed phenotypes similar to axr1 . ( F ) Quadruple mutants display floral defects similar to arf3/ettin . Inset and panels on the right show abnormal guard cell patterning . SEM scale bars , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 01510 . 7554/eLife . 00675 . 016Figure 4—figure supplement 1 . HLS1 , HLH1 , and HLH2 are targets of EIN3 . ( A ) EIN3 binding in HLS1 , HLH1 , HLH2 promoters . ( B ) EIN3 binding motifs in HLS1 , HLH2 reveal a consensus . ERF1 and EBF2 motifs are shown as a reference . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 01610 . 7554/eLife . 00675 . 017Figure 4—figure supplement 2 . HLS1 expression is decreased in ein3-1 , and ein3-1/eil1-1 mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 01710 . 7554/eLife . 00675 . 018Figure 4—figure supplement 3 . HLS1-like homologs ( HLHs ) are similar to HLS1 in protein sequence and domain structure . ( A ) Conservation of HLS1 and HLHs proteins . Amino acid sequence alignment of HLS1 and its three homologs are shown . Gaps are represented as ‘ . ’ . Shading indicates identical sequences ( black ) , conserved changes ( gray ) , similar residues ( light gray ) . ( B ) Phylogeny of HLS1 and HLHs and proteins from other organisms containing acetyltransferase domains . Amino acid sequences were aligned using Clustal , then a bootstrap 50% majority-rule consensus tree was constructed using PAUP . Abbreviations for species are as follows: Hs , Homo sapiens; Sc , Saccharomyces ceravisiae; Mm , Mus musculus; At , Arabidopsis thaliana; Bn , Brassica napus; Zm , Zea mays; Os , Oryza sativa; Ps , Pisum sativum; Ec , E . coli . ( C ) Location of T-DNA insertions in HLH genes . Boxes represent the exons of each HLH gene . Triangles represent the T-DNA alleles that are characterized in detail . Not all T-DNA insertion alleles in the HLH genes are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 01810 . 7554/eLife . 00675 . 019Figure 4—figure supplement 4 . Arabidopsis thaliana EIN3 , EIL1 , EIL3 , and Physcomitrella patens EIN3 DNA-binding motifs from protein binding microarray experiments . EIN3 ChIP-Seq DNA binding motif is shown for comparison . EIN3 ChIPSeq and protein binding motif alignment was performed with STAMP ( E-value = 1 . 59 × 10−6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00675 . 019 Ethylene and auxin co-regulate plant growth and development and it is likely that this co-regulation is mediated in part by EIN3 regulation of HLS1/HLHs . To understand this hormone co-regulation , we generated quadruple mutants for the HLS1 gene family and also further characterized their role as regulatory hub signal integrators ( Figure 4—figure supplement 3 ) . The pleiotropic phenotypes we observed support the role of the HLS1 gene family in auxin regulated plant growth and development ( Figure 4C–F ) . We observed severe defects in the embryonic patterning , etiolated seedlings , adult plant morphology , and floral morphology . The adult quadruple mutants display a dwarf phenotype , similar to the auxin mutant axr1 ( Leyser et al . , 1993 ) , and floral morphology of the quadruple mutants display two stigmas atop a gynoecium , similar to the arf3/ettin mutant floral phenotype ( Sessions and Zambryski , 1995 ) . Although HLS1 is known to be involved in the differential growth of the apical hook and is necessary for the accumulation of AUXIN RESPONSE FACTOR2 ( ARF2 ) DNA-binding protein ( Li et al . , 2004; Ohto et al . , 2006 ) , the biochemical function of these putative N-acetyltransferases remains to be determined . Using a genome-wide approach , we found that not only HLS1 , but other gene family members are targets of EIN3 and that the requirement of the HLS proteins for hormone responses extends beyond apical hook development to many other processes from embryo patterning to flowering , linking the regulation of growth and development by ethylene to many new biological processes in novel ways .
To date , few temporal transcription factor binding studies have been undertaken ( Hiroi , 2004; Ni et al . , 2009; Zinzen et al . , 2009 ) . Temporal protein–DNA interactions are often difficult to reconcile with gene expression profiles and the complexity of regulation that occurs transcriptionally is very challenging to characterize and interpret biologically . Here , by jointly analyzing the temporal expression and genome-wide binding data of one key transcription factor in response to hormone stimulus , we were able to reveal several important properties of the hormone responsive transcriptional program and identify new components in the signaling pathway . We found that upon a timecourse of ethylene treatment , EIN3 binding was induced , resulting in various transcriptional patterns and that the ethylene transcriptional response occurred in waves of transcription that were temporally distinct and could be attributed to different biological functions , variable in the amount of noise , and significantly regulated by EIN3 . EIN3 modulated genes were over-represented in hormone co-regulation , and the specific targets in the other hormone pathways , as reported in this study , suggest these ‘cross-talk’ events may involve multiple levels of regulation . Interestingly , feedback regulation of the ethylene response by EIN3 enabled the identification of the majority of known ethylene signaling pathway components . Moreover , the temporal resolution of steady-state levels of transcription confirmed the role of these genes in the ethylene transcriptional response . Other signaling networks utilize feedback regulation for overall system control/homeostasis , and this type of study may be used to identify novel components in signaling pathways ( Rosenfeld et al . , 2002; Amit et al . , 2007; Tsang et al . , 2007; Avraham and Yarden , 2011; Fang et al . , 2011; Feng et al . , 2011; Yosef and Regev , 2011 ) . The implication that EIN3 regulates the coordination of other hormone pathways is extensive because the transcriptional control by EIN3 is likely conserved in plants . EIN3 orthologs exist in poplar , soybean , rice , maize , moss , and multicellular algae , among many others ( www . phytozome . net ) and , in fact , we found that the Physcomitrella patens ( moss ) EIN3 protein binds a very similar motif sequence to that of the Arabidopsis thaliana EIN3 ( Figure 4—figure supplement 4 ) . The role of EIN3 in the coordination of the initiation of the ethylene transcriptional cascade , the negative feedback regulation of the ethylene signaling pathway , and the orchestration of other hormone pathways suggests that adaptable system homeostasis in plants requires robust hormone co-regulation .
The Arabidopsis thaliana ecotype Columbia ( Col-0 ) was the parent strain for these experiments . Genotypes used for this study include wild-type Col-0 , and mutants ein3-1 ( Chao et al . , 1997 ) , ein3-1/eil1-1 ( Alonso et al . , 2003b ) , hls1-1 ( hls1 ) ( Lehman et al . , 1996 ) , hlh1 , hlh2 , hlh3 ( Figure 4—figure supplement 3 ) . Three-day-old etiolated seedling tissue was used for these experiments unless otherwise noted . Seeds were sterilized and sown on Murashige and Skoog ( cat#LSP03 , Caisson ) media pH5 . 7 , containing 1% sucrose and 1 . 8% agar . After stratification for 3 days in the dark at 4°C , exposure to light for 2–4 hr to induce germination , seeds were dark-grown in hydrocarbon free air at 24°C for 3 days . Etiolated seedlings were subsequently treated with ethylene gas at 10 μl l−1 for 0 , 0 . 25 , 0 . 5 , 1 , 4 , 12 , and 24 hr . Etiolated seedlings were collected in the dark , immersed in 1% formaldehyde solution , and cross-linked under vacuum for 15 min . A final concentration of 125 mM glycine was used to quench the formaldehyde for 5 min under vacuum . Cross-linking under vacuum resulted in translucent etiolated seedling tissue . Tissue was liquid nitrogen ground and extraction of chromatin was performed as described in ( Lippman et al . , 2005 ) . Chromatin immunoprecipitation ( ChIP ) was performed as described in ( Lippman et al . , 2005 ) with modifications , including the use of the Bioruptor sonicator ( Diagenode , Belgium ) . Bioruptor settings used were: H , 25 cycles of 0 . 5 min on , 0 . 5 min off , with 5 min rests between every 5 cycles . Sonication was performed in a cooling water bath at 4°C . A small amount of chromatin ( 10 μl ) was evaluated for shearing; the size range of chromatin was 150–700 bp , the majority of fragments at 300–400 bp . Affinity-purified rabbit polyclonal antibodies capable of detecting the C-terminus of EIN3 were used in immunoprecipitation reactions . Details regarding the generation of EIN3 antibodies were previously described ( Guo and Ecker , 2003 ) . Prior to the experiments in this study , the amount of purified EIN3 antisera per immunoprecipitation reaction was optimized and 8 μl of purified EIN3 antisera was determined to yield the optimal enrichment of the ERF1 promoter , the known target of EIN3 ( data not shown ) . We then substituted Dynabeads Protein A ( Invitrogen , cat#100-1D ) and Dynabeads M-280 Sheep anti-Rabbit IgG ( Invitrogen , cat#112-04D ) for the salmon sperm DNA blocked Protein A agarose beads recommended in the protocol ( 4 ) , as to avoid sequencing of salmon sperm DNA . Immunoprecipitation and washing of Dynabeads were performed using the buffers in ( Lippman et al . , 2005 ) , otherwise Dynabeads were used as per the manufacturer’s instructions . Multiple pipetting steps were performed while washing the beads to reduce non-specific binding carryover . Resulting ChIP DNA was purified as in ( Lippman et al . , 2005 ) . Quantitative PCR revealed that relative ChIP enrichment for the promoter of ERF1 performed with the Dynabeads M-280 Sheep anti-Rabbit IgG was higher in comparison to Dynabeads Protein A ( Figure 1—figure supplement 1A ) . Thus , Dynabeads M-280 Sheep anti-Rabbit IgG was used in all subsequent experiments . Primers for the ERF1 promoter encompassing the EIN3 binding site , are as follows: F-GGGGGCATGTATCTTGAATC , R-TGCTGGATCAACTCAACAAAA . Actin primers were as in Mathieu et al . ( Mathieu et al . , 2003 ) . Enrichment was calculated using the Delta-Delta-Ct method with normalization to the reference Actin; fold change was calculated relative to the control for non-specific binding ( EIN3 ChIP performed in ein3-1 mutant ) . ChIP was performed in chromatin derived from wild-type Col-0 three-day-old etiolated seedlings treated with 0 , 0 . 25 , 0 . 5 , 1 , 4 , 12 , and 24 hr of ethylene . Two independent biological replicates were used in two replicates experiments for timepoints , 0 , 0 . 5 , 1 , 4 hr ethylene gas treatment . Single replicates exist for 0 . 25 , 12 , 24 hr of ethylene gas treatment . Total RNA was extracted from liquid nitrogen ground etiolated seedlings using the Qiagen RNeasy Plant Mini Kit with Qiashredder columns ( cat#74 , 904 ) , with DNaseI ( Qiagen , cat#79 , 254 ) treatment prior to RNA precipitation in sodium acetate and ethanol . Concentrations of RNA were determined using the ND-1000 spectrometer ( Nanodrop , Wilmington , DE ) . Experiments were performed in three biological replicates for timepoints , 0 , 0 . 25 , 0 . 5 , 1 , 4 , 12 , 24 hr ethylene gas treatment . Resulting ChIP DNA from two pooled ChIP reactions above was used to generate a sequencing library as per the Illumina ChIP-Seq manufacturer’s instructions . The Illumina Genome Analyzer II ( Illumina , San Diego , CA ) was used to sequence the single-read ChIP-Seq libraries as per manufacturer’s instructions , for 36–43 bps ( Supplementary file 1A ) . Raw sequencing data was analyzed using the Genome Analyzer Pipeline v . 1 . 4 . 0 . Reproducibility of the data is shown in Figure 1—figure supplement 1 . Although the general reproducibility of the data is lower than what was previously reported ( Kaufmann et al . , 2009; 2010 ) , it is clear that the reproducibility between biological replicates is much higher than that with respect to the control 0 hr ethylene gas treatment timepoint . We did not extend raw reads for calculation of reproducibility but instead determined the reproducibility of RPKM values between replicates . At least 80 μg total RNA was subject to polyA selection using the Poly ( A ) Purist MAG Kit ( Ambion , cat#AM1922 ) . PolyA RNA was subsequently concentrated by ammonium acetate ethanol precipitation and concentrations were determined using the Qubit fluorometer ( Invitrogen , Carlsbad , CA ) and the Quant-iT RNA Assay Kit ( Invitrogen , cat#Q33140 ) . 50–100 ng of polyA RNA was used in a strand-specific library preparation as per the SOLiD Total RNA-Seq Kit protocol ( Invitrogen , cat#4445374 ) and AMPure XP beads ( Agencourt , cat#A63881 ) were used for purification of cDNA and amplified DNA . Samples were barcoded for multiplexing using the SOLiD RNA Barcoding Kit ( Invitrogen , Module 1-16 cat#4427046 , Module 17-32 cat#4453189 , Module 33-48 cat#4453191 ) as per manufacturer’s instructions; final size selection was performed using AMPure XP beads instead of the PAGE purification recommended in the protocol . Size selected libraries were then purified using the MinElute Gel Extraction Kit ( Qiagen , cat#28 , 604 ) . Resulting concentrations of libraries were detecting using the Qubit fluorometer and Quant-iT dsDNA High-Sensitivity Assay Kit ( Invitrogen , cat #Q33120 ) . RNA libraries were sequenced for 50 bps on the SOLiD4 platform ( Life Technologies , Carlsbad , CA ) ( Supplementary file 1A ) . The Illumina GERALD module was used to align the sequenced reads to the Col-0 reference genome , version TAIR10 ( ftp://ftp . arabidopsis . org/ ) . The analysis variable for the ELAND alignment program was set to eland_extended , as read length was greater than 32 bases ( e . g . , 36–43 ) . Resulting aligned unique single copy reads were used in ChIP-Seq peak analysis ( Supplementary file 1A ) . Saturation analysis of the ChIP libraries was conducted using the spp software ( Kharchenko et al . , 2008 ) revealed that all samples were at least within 15% of saturation . Peak analysis was performed individually on each timepoint in each biological replicate using the corresponding 0 hr ethylene treated wild-type Col-0 EIN3 ChIP sample as a control . Two additional ethylene treated ( 4 hr ) wild-type EIN3 ChIP biological replicates were included in the analysis , with corresponding mutant ein3-1 ethylene treated ( 4 hr ) EIN3 ChIP samples as controls . Three software packages: spp ( Kharchenko et al . , 2008 ) , MACS ( Zhang et al . , 2008 ) , PeakSeq ( Rozowsky et al . , 2009 ) were originally used to identify peaks/regions of binding . Parameters for each software were as follows: MACS ( p=0 . 01 ) , spp ( FDR = 0 . 1 ) , PeakSeq ( FDR = 0 . 1 , mingap = 200 , minhit = 20 , minratio = 3 . 5 ) . Binding regions were merged when the maximum gap between two peaks was less than 200 bp determined by separate software packages . Subsequent analysis was performed in R . Overlapping peaks in one biological replicate in one timepoint by more than one software package were retained as binding regions . Because of the variation of the number of called peaks in each software and each timepoint , we used a majority vote to call peaks to identify all high stringency EIN3 candidate targets . PeakSeq results differed significantly from spp and MACS ( 12–76% ) , therefore only spp and MACS were ultimately used . Using this method , 1460 EIN3 binding regions were identified ( Supplementary file 1B ) . For each EIN3 binding region , the reads per kbp of binding site per million sample reads ( RPKM ) were calculated . Median normalization of the RPKM values between timecourse biological replicates was performed in R . Resulting RPKMs were log2 transformed with respect to the 0 hr ethylene treatment wild-type Col-0 EIN3 ChIP . Normalization with respect to an input genomic control did not produce distinctively different EIN3 binding pattern profiles ( data not shown ) . EIN3 binding regions were then associated to a gene if located within 5 kbp . The nearest expressed gene ( RPKM>1 ) was assigned if there were more than one gene within 5 kbp . If both genes were not expressed , the nearest gene was selected . Distance was determined from the binding region center to the gene feature using the TAIR10 annotation ( ftp://ftp . arabidopsis . org ) ( Figure 1—figure supplement 1 ) . EIN3 binding profiles of previously determined targets are shown in Figure 1—figure supplement 3 . Data from biological replicate 1 is shown; biological replicate 2 results were similar . Four of seven previously determined EIN3 targets were identified as EIN3 candidate targets in our dataset . Browser images of data were generated using AnnoJ ( Lister et al . , 2008 ) . ChIP browser images display read tracks normalized per library , the lowest number of reads for all ChIP samples was used as a minimum . This minimum number of reads was randomly selected from all other libraries for display , to effectively visualize enrichment among different samples . The trends in the data were reproducible statistically ( Figure 1—figure supplement 1 ) , and also evident in the visualization of data ( see example of EIN3 binding for both biological replicates in EBF2 promoter depicted in Figure 1—figure supplement 4 ) . Motif identification was performed with the matrix screening software Patser ( Hertz and Stormo , 1999 ) and the known EIN3 consensus motif ( TEIL ) from TRANSFAC previously determined using SELEX ( Kosugi and Ohashi , 2000 ) . ClustalW2 was used to align motifs ( www . ebi . ac . uk/Tools/msa/clustalw2/ ) . Consensus motif representation of the three EIN3 binding sites in the promoter of EBF2 is shown in Figure 1—figure supplement 4 . Gene ontology over-representation of selected groups of genes were visualized and determined using the Cytoscape v . 2 . 8 . 1 ( Shannon et al . , 2003 ) plugin BiNGO v . 2 . 44 ( Maere , 2005 ) ( Supplementary file 1C ) . The hypergeometric test was used with Benjamini and Hochberg multiple testing correction ( FDR = 0 . 05 ) . The GOSlim_Plants Ontology was used for Arabidopsis thaliana ( Figure 1—figure supplement 5 ) . EIN3 binding sites were ranked using the R package timecourse , which has been previously used to analyze microarray timecourse data . We used this R package because no available software to analyze timecourse data for ChIP-Seq data exists . The top 50 EIN3 binding regions were determined and the repeatmasked . De novo motif analysis of these top 50 EIN3 binding regions was performed using SOMBRERO ( Mahony et al . , 2005 ) , and alignment to known Arabidopsis motifs ( AGRIS , http://arabidopsis . med . ohio-state . edu/ ) was performed using STAMP ( Mahony and Benos , 2007 ) ( Figure 1—figure supplement 3 ) . Twelve transcription factors that are ethylene-regulated EIN3 targets were analyzed on protein binding microarrays ( PBMs ) . Details of the design and use of universal PBMs has been described elsewhere ( Berger et al . , 2006; Badis et al . , 2009; Berger and Bulyk , 2009 ) . Here , we used two different universal PBM array designs , designated ‘ME’ and ‘HK’ , after the initials of their designers ( Lam et al . , 2011 ) . Information about individual plasmids is available in Supplementary file 1D . We identified the DNA Binding Domain ( DBD ) of each TF by searching for Pfam domains ( Finn et al . , 2009 ) using the HMMER tool ( Eddy , 2009 ) . DBD sequences along with 50 amino acid residue ‘pads’ on either side were cloned as SacI–BamHI fragments into pTH5325 , a modified T7-driven GST expression vector . Briefly , we used 150 ng of plasmid DNA in a 15 μl in vitro transcription/ translation reaction using a PURExpress In Vitro Protein Synthesis Kit ( New England BioLabs ) supplemented with RNase inhibitor ( Invitrogen ) and 50 μM zinc acetate . After a 2 hr incubation at 37°C , 12 . 5 ml of the mix was added to 137 . 5 ml of protein-binding solution for a final mix of PBS/2% skim milk/0 . 2 mg per ml BSA/50 μM zinc acetate/0 . 1% Tween-20 . This mixture was added to an array previously blocked with PBS/2% skim milk and washed once with PBS/0 . 1% Tween-20 and once with PBS/0 . 01% Triton-X 100 . After a 1 hr incubation at room temperature , the array was washed once with PBS/0 . 5% Tween-20/50 mM zinc acetate and once with PBS/0 . 01% Triton-X 100/50 mM zinc acetate . Cy5-labeled anti-GST antibody was added , diluted in PBS/2% skim milk/50 mM zinc acetate . After a 1 hr incubation at room temperature , the array was washed three times with PBS/0 . 05% Tween-20/50 mM zinc acetate and once with PBS/50 mM zinc acetate . The array was then imaged using an Agilent microarray scanner at 2 mM resolution . Images were scanned at two power settings: 100% photomultiplier tube ( PMT ) voltage ( high ) , and 10% PMT ( low ) . The two resulting grid images were then manually examined , and the scan with the fewest number of saturated spots was used . Image spot intensities were quantified using ImaGene software ( BioDiscovery ) . The creation of a position frequency matrix ( PFM ) from a PBM experiment is non-trivial . For each TF , we therefore evaluated a panel of three algorithms and chose the PFM with the highest performance . For each TF , we ran each algorithm individually on both PBM experiments ( HK and ME array designs ) . The resulting PFMs were then used to score the probe sequences of the opposite array , and these predictions were evaluated based on their Pearson correlation with the actual intensities across all probes . Based on these evaluations , a final PFM was chosen for each TF from the six possible PFMs ( three algorithms times two array designs ) . We chose three algorithms based on their high performance on an independent PBM dataset ( data not shown ) . Two of the methods , BEEML-PBM ( Zhao and Stormo , 2011 ) , and FeatureREDUCE ( PWM modification of [Foat et al . , 2006] ) are based on biophysical models of TF-DNA interactions . The third algorithm ( PWM_align ) is an in-house method that aligns all 8mers with E-scores > 0 . 45 ( Berger et al . , 2008 ) using ClustalW ( Chenna , 2003 ) , and trims the resulting alignment by restricting to positions present in at least half of the sequences in the alignment . The presence of these motifs in the promoter region ( −1000bp ) of genes that were transcriptionally induced/repressed by ethylene was evaluated to find candidate transcription factors that may be involved in regulating the secondary ethylene transcriptional response . The matrix screening software Patser ( Hertz and Stormo , 1999 ) was used to scan the promoter region of all genes that were transcriptionally regulated by ethylene , with the PBM motifs ( Figure 3—figure supplement 1; Supplementary file 1E ) . The SOLiD Bioscope v . 1 . 3 software was used to align the reads to the Col-0 reference genome TAIR10 ( ftp://ftp . arabidopsis . org/ ) . Two perfect matches per location were allowed . Exonic expression was determined ( RPKM ) using mRNA-Seq reads mapping in exons in the direction of transcription . Genes were denoted as expressed if they contained RPKM values greater than one for at least one biological replicate in one timepoint . Differentially expressed genes were then called ( t-test p=0 . 05 , 50% difference from prior timepoint of ethylene gas treatment ) , and log2 normalized with respect to the 0 hr ethylene gas treatment control ( Figure 1—figure supplement 2 ) . Overlap of up- and down-regulated genes was ∼1% . EIN3 ChIP candidate targets were classified as ethylene regulated ( EIN3-R ) , non-ethylene-regulated ( EIN3-NR ) , and transcription not detected in etiolated seedlings ( EIN3-ND ) . The heatmap ( Figure 1—figure supplement 2 ) revealed that there is a singular binding pattern but various transcription profiles , as displayed in Figure 2 . Although the majority of EIN3 candidate targets were up-regulated by ethylene , consistent with the previously determined role of EIN3 as an activator , a subset of EIN3 candidate targets was repressed upon ethylene treatment; one instance of EIN3 as a repressor has been previously reported ( Chen et al . , 2009 ) . The correlation of EIN3 binding and ethylene-regulated transcription was calculated at from 4 hr of ethylene treatment ( 0 hr ethylene as a control ) , for all EIN3 and EIN3-R ( ethylene-induced ) targets . The R2 values were much less than 0 . 50 , suggesting a lack of correlation of EIN3 binding levels and ethylene-regulated steady-state transcription . The kinetics of transcription was determined for all genes that were transcriptionally regulated by ethylene , and EIN3-R , and reflects the previous growth inhibition study kinetics ( Figure 2—figure supplement 2 ) . The ethylene transcriptional response was further analyzed in context of the dynamic EIN3 binding data . To reconstruct the dynamic regulatory networks that were activated following ethylene treatment , we used the Dynamic Regulatory Events Miner ( DREM ) ( Ernst et al . , 2007 ) . DREM integrates time-series gene expression data with static transcription factor ( TF ) —gene interaction data to reconstruct these dynamic networks . DREM searches for bifurcation events; places in the time series data where the expression of one set of genes diverges from the expression of another set , and annotates these events with the TFs that can explain them . This allows us to assign a time of activation to static TF-gene interactions data . To obtain the static interaction data we extracted 11 , 355 TF-gene interactions from the AtRegNet AGRIS database ( Yilmaz et al . , 2010 ) . In addition , for this work we have extended DREM so that it can utilize temporal EIN3 binding profiles as well as allowing us to identify functional binding events ( those with direct impact on expression ) . This is done by changing the set of targets for EIN3 so that different binding values are used at each time point . For each EIN3 candidate target gene , the average RPKM values from two input control samples at 0 and 4 hr were used as a cutoff to determine whether it was bound by EIN3 or not at each time point . We ran the modified DREM algorithm using the mRNA-Seq data allowing for 3-way splits . We filtered out genes that did not change at least twofold ( up or down ) at any time point , and we used the default values for all other parameters . Four temporally distinct ( Hypergeometric p<0 . 001 ) EIN3-modulated waves of transcription ( Pathway hypergeometric p<10−10 ) were observed . There was a variable amount of noise and percentage of hormone-related genes in each wave of transcription ( Figure 2—figure supplement 2 ) . The comprehensive DREM analysis results are shown ( Figure 2—figure supplement 3 ) , including all observed patterns of ethylene transcriptional regulation . The over path significance was used to determine whether these waves were regulated by EIN3 . A stringent threshold ( 10−10 ) was used to identify groups of genes with a significant percentage ( >15% ) of EIN3 candidate targets . The most current protein-protein interaction network for Arabidopsis ( Arabidopsis Interactome Mapping Consortium , 2011 ) containing high throughput yeast two hybrid and literature curated data was used as the foundation for the hormone co-regulation network . The protein-DNA interaction network AtRegNet from AGRIS ( http://arabidopsis . med . ohio-state . edu/; 7918 nodes , 10 , 640 edges ) included high throughput data ( ChIP-chip and ChIP-Seq ) for several transcription factors including AGL15 , HY5 , GL3 , AtbHLH15 , WRKY53 , GL1 , E2F , and SEP3 as well as literature curated data ( Yilmaz et al . , 2010 ) . Transcription factor-DNA binding interactions from six additional studies were added , including TGA2 ( Thibaud-Nissen et al . , 2006 ) , AP1 ( Kaufmann et al . , 2010 ) , BES1 ( Yu et al . , 2011 ) , BZR1 ( Sun et al . , 2010 ) , FLC ( Deng et al . , 2011 ) in addition to our data . This generated a protein-DNA interaction network of 8531 nodes and 11 , 953 edges , which was then merged with the protein-protein interaction network . Protein–protein interaction and protein–DNA interaction edges were indicated by dark and light grey lines , respectively . To identify genes associated with a hormone signal or response ( e . g . , hormone-related ) , we used the annotation in the Arabidopsis Hormone Database ( Peng et al . , 2008 ) ( http://ahd . cbi . pku . edu . cn/ ) in addition to other datasets including relevant ethylene microrarray studies in etiolated seedlings ( Alonso et al . , 2003a; Nemhauser et al . , 2006 ) . Hormone annotation attributes were imported into Cytoscape ( Shannon et al . , 2003 ) and colored according to hormone . The amount of genes involved in hormone responses in the genome was 21% ( 5729/27 , 416 ) , whereas the amount of genes involved in our EIN3 target group was 46% ( Figure 1—figure supplement 2 , inset ) . We identified loss-of-function mutants and performed thorough genetic analyses of HLS1 and its homologs to characterize the effect , if any , these genes have on the ethylene response . Three HLS1 homologs ( HLHs ) exist in Arabidopsis genome . The protein sequences of the HLHs are homologous to the full-length protein ( Figure 4—figure supplement 3 ) . Like HLS1 , these homologs contain acetyltransferase domains at the N-terminal portion of the protein . Phylogenetic analysis of HLS1-like genes with acetyltransferase domain containing proteins from various organisms revealed that the HLS1 family of acetyltransferases form a unique plant-specific class ( Figure 4—figure supplement 3 ) . We isolated the bona fide loss-of-function mutants in the coding regions of the genes for all the HLH genes using the Salk T-DNA mutant collection ( Figure 4—figure supplement 3 ) ( Alonso et al . , 2003a ) . The single knockout mutants of the HLHs exhibited normal apical hook development and had no obvious developmental defects compared to wild type ( data not shown ) , indicating functional redundancy among HLS1 family members . | All multicellular organisms , including plants , produce hormones—chemical messengers that are released in one part of an organism but act in another . The binding of hormones to receptor proteins on the surface of target cells activates signal transduction cascades , leading ultimately to changes in the transcription and translation of genes . Ethylene is a gaseous plant hormone that acts at trace levels to stimulate or regulate a variety of processes , including the regulation of plant growth , the ripening of fruit and the shedding of leaves . Plants also produce ethylene in response to wounding , pathogen attack or exposure to environmental stresses , such as extreme temperatures or drought . Although the effects of ethylene on plants are well documented , much less is known about how its functions are controlled and coordinated at the molecular level . Here , Chang et al . reveal how ethylene alters the transcription of DNA into messenger DNA ( mRNA ) in the plant model organism , Arabidopsis thaliana . Ethylene is known to exert some of its effects via a protein called EIN3 , which is a transcription factor that acts as the master regulator of the ethylene signaling pathway . To identify the targets of EIN3 , Chang et al . exposed plants to ethylene and then used a technique called ChIP-Seq to identify those regions of the DNA that EIN3 binds to . At the same time , they used genome-wide mRNA sequencing to determine which genes showed altered transcription . Over the course of 24 hr , ethylene induced four distinct waves of transcription , suggesting that discrete layers of transcriptional control are present . EIN3 binding also controlled a multitude of downstream transcriptional cascades , including a major negative feedback loop . Surprisingly , many of the genes that showed altered expression in response to EIN3 binding were also influenced by hormones other than ethylene . In addition to extending our knowledge of the role of EIN3 in coordinating the effects of ethylene , the work of Chang et al . reveals the extensive connectivity between pathways regulated by distinct hormones in plants . The results may also make it easier to identify key players involved in hormone signaling pathways in other plant species . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology"
] | 2013 | Temporal transcriptional response to ethylene gas drives growth hormone cross-regulation in Arabidopsis |
To identify genetic and environmental factors contributing to the pathogenesis of non-alcoholic fatty liver disease , we examined liver steatosis and related clinical and molecular traits in more than 100 unique inbred mouse strains , which were fed a diet rich in fat and carbohydrates . A >30-fold variation in hepatic TG accumulation was observed among the strains . Genome-wide association studies revealed three loci associated with hepatic TG accumulation . Utilizing transcriptomic data from the liver and adipose tissue , we identified several high-confidence candidate genes for hepatic steatosis , including Gde1 , a glycerophosphodiester phosphodiesterase not previously implicated in triglyceride metabolism . We confirmed the role of Gde1 by in vivo hepatic over-expression and shRNA knockdown studies . We hypothesize that Gde1 expression increases TG production by contributing to the production of glycerol-3-phosphate . Our multi-level data , including transcript levels , metabolite levels , and gut microbiota composition , provide a framework for understanding genetic and environmental interactions underlying hepatic steatosis .
Non-alcoholic fatty liver disease ( NAFLD ) encompasses a wide spectrum of liver abnormalities , ranging from benign hepatocellular accumulation of lipids ( steatosis ) , through non-alcoholic steatohepatitis ( NASH ) , to fibrosis and cirrhosis in the absence of excessive consumption of alcohol and hepatitis viral infection . Advanced NAFLD can eventually progress to end-stage liver disease with increased risk of hepatocellular carcinoma ( HCC ) ( Kopec and Burns , 2011 ) . Population studies have shown that NAFLD is strongly associated with obesity , diabetes , and dyslipidemia ( Marchesini et al . , 2003 ) . As such , NAFLD can be viewed as the hepatic manifestation of the metabolic syndrome . With the increasing prevalence of obesity , diabetes , and metabolic syndrome , it is not surprising that NAFLD is rapidly becoming the most common form of chronic liver disease worldwide ( Ratziu et al . , 2010 ) . It is estimated that 20–30% of the population in Western developed countries is affected ( Vernon et al . , 2011 ) . Despite the high prevalence of this disease , its natural history and etiology is poorly understood . Although simple steatosis appears to be benign and non-progressive in the majority of individuals affected with NAFLD ( Teli et al . , 1995 ) , excessive fat accumulation in the liver is associated with organ pathology , including NASH and cirrhosis . NASH frequently progresses to fibrosis , cirrhosis , liver failure , and HCC , resulting in poor long-term prognosis . The factors determining the progressive phenotype of this complex disease remain largely unknown , although it is clear that subtle genetic variations and environmental factors ( such as diet and lifestyle ) play a role in determining the disease phenotype and progression ( Day , 2002; Anstee et al . , 2011 ) . According to the prevailing ‘two-hit hypothesis’ model ( Day and James , 1998 ) , it is believed that the first insult involves lipid accumulation in the hepatocytes , due to an imbalance in triglyceride homeostasis . Subsequently , steatosis increases the vulnerability of the liver to a ‘second hit’ , which promotes liver injury , oxidative stress , inflammation , and fibrosis . However , silencing of diacylglycerol acyltransferase 2 ( DGAT2 ) by an antisense oligo improved steatosis but worsened liver injury and fibrosis in mice ( Yamaguchi et al . , 2007 ) . DGAT2-knockdown mice exhibit increased fatty acid ( FA ) oxidation through CYP2E1 , leading to increased oxidative stress , inflammation , and tissue damage . These findings lead to the proposal that steatosis may be an adaptive response to protect the liver from lipotoxicity through partitioning toxic lipids into stable intracellular triglyceride stores . Hence , despite the fact that accumulation of TG was alleviated in DGAT2-knockdown liver , blocking this protective mechanism exacerbated lipotoxicity and led to more severe liver injury . These data suggest that TG accumulation per se is not the ‘first hit’ but the underlying inability to compensate for increased FA flux , which makes the liver prone to subsequent oxidative damage . It is proposed that the ‘second hit’ could involve diverse parallel pro-inflammatory signals derived from multiple sources ( Tilg and& Moschen , 2010 ) . So far , there exists no established therapy targeting NAFLD . Since most NAFLD patients suffer from obesity and insulin resistance , treatment options aim at weight reduction , control of dyslipidemia and improving insulin sensitivity through lifestyle changes and pharmacological agents , such as metformin , statins , fibrates , and thiazolidinediones ( Schreuder et al . , 2008 ) . Understanding the underlying genetic factors contributing to NAFLD would not only fill the void of knowledge but would also facilitate the design of effective strategies in treating and preventing this important disease . Considerable variations in the flux of TG and FAs occur in the liver in response to changing nutritional and hormonal status . Nevertheless , under normal physiological conditions , the liver does not accumulate considerable amounts of TG despite its active regulatory role in lipid trafficking . Processes that offset the balance of TG acquisition and disposition give rise to hepatic steatosis , which is the hallmark of NAFLD . These include increased FA uptake or de novo biosynthesis , reduced FA oxidation , and impaired lipoprotein production and secretion ( Farrell and Larter , 2006 ) . The FA moiety of hepatic TG is derived from three major sources: diet , de novo biosynthesis , and adipose tissue . In humans , it is estimated that 15% of hepatic TG comes directly from the diet , 26% from de novo lipogenesis , and 59% from adipose tissue in the form of non-esterified FAs ( Donnelly et al . , 2005 ) . Together , the adipose tissue and liver play a major role in hepatic TG accumulation and supply >85% of FAs for the biosynthesis of TG in the liver . Epidemiologic studies in human populations and animal studies indicate that genetics plays a substantial role in determining the susceptibility to the development of NAFLD ( Browning et al . , 2004; Guerrero et al . , 2009; Kahle et al . , 2013 ) . The coexistence of NASH and cirrhosis clusters within families supports the existence of a common genetic link ( Struben et al . , 2000; Willner et al . , 2001 ) . The broad-sense heritability of hepatic steatosis has been estimated to be ∼39% after adjustment for age , sex , race , and body mass index ( Schwimmer et al . , 2009 ) . Although genetic predisposition to obesity and diabetes undoubtedly explains a fraction of the genetic component , additional independent genetic factors clearly contribute to the susceptibility to NAFLD ( Browning et al . , 2004 ) . To date , only a small fraction of genes accounting for fatty liver disease have been identified and the molecular pathogenesis of NAFLD is poorly understood . To gain insight into the etiology of NAFLD , we have employed a ‘systems genetics’ strategy to generate a comprehensive view of the genetic architecture of NAFLD in mice ( Civelek and Lusis , 2014; Mato et al . , 2014 ) . We examined natural variations in the development of steatosis in a panel of inbred strains of mice fed a high-fat , high-sucrose ( HF/HS ) diet for 8 weeks . This diet has previously been shown to have a profound effect on obesity and insulin resistance depending upon the genetic background ( Parks et al . , 2013 ) . We observed a wide range in phenotypes associated with hepatic steatosis following feeding the HF/HS diet . To further understand the molecular basis of steatosis , we employed a multi-omic approach to characterize the steatotic phenotype . Using global expression analysis of liver and adipose transcriptomes , we identified pathways enriched in strains susceptible to steatosis . We also showed that the susceptibility to steatosis is highly dependent on genetic background and was able to identify genetic loci contributing to steatosis . A causal gene at one of the loci was identified as Gde1 . Furthermore , we identified metabolites and gut microbes , which are associated with steatosis . These multi-dimensional data provide a valuable resource for understanding the genetic–environmental interactions in the disorder .
NAFLD is often referred to as the hepatic manifestation of metabolic syndrome , as it is associated with obesity , dyslipidemia , and insulin resistance ( Lazo and& Clark , 2008 ) . The development of NAFLD is strongly influenced by both dietary and genetic factors . We previously showed that the increase in body weight and fat accumulation in response to a HF/HS diet in mice is highly dependent on the genetic background of individual strains ( Parks et al . , 2013 ) . To study the gene-by-diet effects on hepatic steatosis , a panel of 8-week-old male HMDP mice were fed a HF/HS diet for 8 weeks to induce obesity and steatosis ( Parks et al . , 2013 ) . Lipids were extracted and quantified in 478 individual livers from 113 strains of male mice . A wide spectrum of hepatic TG content was observed among the strains with more than 30-fold difference between the high and low responders ( Figure 1A and Supplementary file 1 ) . Hepatic TG content was significantly correlated with the liver weight ( r = 0 . 38 , p = 5 . 75 × 10−14 , Figure 1B ) . Contrary to the large variations in TG among strains , less than threefold difference in cholesterol and phospholipid levels was observed among the strains ( Supplementary file 1 ) . Modest correlations between these lipids with hepatic TG were observed . The TG content in the liver was positively correlated with hepatic total cholesterol ( TC ) content ( r = 0 . 35 , p = 1 . 18 × 10−12 , Figure 1C ) but negatively correlated with the levels of phospholipids ( r = −0 . 26 , p = 3 . 45 × 10−7 , Figure 1D ) . These data suggest that increased neutral lipids content , in particular TG , contributes significantly to the enlarged livers . 10 . 7554/eLife . 05607 . 003Figure 1 . Effects of genetic background on hepatic TG accumulation . ( A ) Hepatic TG levels in male mice after 8 weeks of HF/HS feeding . Results are presented as mean + SD . ( B–D ) Correlation of hepatic TG with liver weight ( B ) , hepatic total cholesterol ( TC ) ( C ) , and hepatic phospholipid ( D ) . r , biweight midcorrelation; p , p-value . DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 003 To assess liver damage , we measured alanine aminotransferase ( ALT ) enzyme activity in the plasma ( Figure 2A ) . Similar to hepatic TG content , we also observed a large variation in ALT activities among the strains . Plasma ALT activities showed a positive correlation with hepatic TG levels , implicating a role of increased hepatic TG levels in liver damage ( Figure 2B ) . ALT is an established biochemical and clinical marker for liver damage as this enzyme is released into the circulation when the integrity of the cell membrane of hepatocytes is compromised . We performed histologic examination of livers from a subset of strains including some with high TG and did not observe evidence of significant pathology other than lipid accumulation ( data not shown ) . Furthermore , mRNA levels of markers for B-lymphocytes ( CD45/PTPPRC ) and macrophages ( CD68 ) were not increased in steatotic liver samples ( Figure 2C , D ) . Similarly , markers for other immune cells , such as T cells ( Cd28 , Csf2 , Cd4 , Ccr5 , Gata3 Cxcr4 ) , B cells ( Pax5 , Cd70 , Cd79b ) , and leukocytes ( Cd33 , Cd52 , Cd53 , Cd44 , Prg2 ) were also not increased ( data not shown ) , suggesting the absence of significant immune cell infiltration and steatohepatitis in these mice after 8 weeks on the HF/HS diet . 10 . 7554/eLife . 05607 . 004Figure 2 . Plasma ALT activities and immune cell marker expression among inbred and recombinant inbred strains of mice . ( A ) Plasma alanine aminotransferase ( ALT ) activities in male mice after 8 weeks of HF/HS feeding . Results are presented as mean + SD . ( B–D ) Correlation of plasma ALT activities with hepatic triglyceride ( B ) , hepatic Ptprc ( Cd45r ) expression ( C ) and hepatic . Cd68 expression . r , biweight midcorrelation; p , p-value . DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 004 A substantial amount of hepatic TG is derived from FAs of extrahepatic sources , in particular , the white adipose tissue . We measured lipid and metabolites in the plasma and compared them to the hepatic TG content . Hepatic TG content was poorly correlated with plasma TG levels ( r = −0 . 13 , p = 0 . 0115 , Figure 3A ) , whereas it was positively correlated with plasma cholesterol levels ( r = 0 . 41 , p = 5 . 04 × 10−17 , Figure 3B ) . The correlation between hepatic TG levels and plasma free FAs ( FFAs ) levels was not significant ( r = 0 . 04 , p = 0 . 44 ) ; however , hepatic TG levels were correlated with plasma glycerol levels ( r = 0 . 20 , p = 0 . 0001 , Figure 3C ) , suggesting a link between liver steatosis and lipolysis in the adipose tissue . NAFLD is often associated with dyslipidemia ( Diehl et al . , 1988 ) and insulin resistance ( Marchesini et al . , 1999 ) in humans . Similar to the findings in humans , there was a robust association between hepatic steatosis and plasma insulin ( r = 0 . 47 , p = 4 . 51 × 10−21 , Figure 3D ) , glucose ( r = 0 . 23 , p = 1 . 26 × 10−5 , Figure 3E ) as well as insulin resistance ( HOMA-IR ) ( r = 0 . 45 , p = 2 . 18 × 10−20 , Figure 3F ) . 10 . 7554/eLife . 05607 . 005Figure 3 . Correlation of hepatic TG content with plasma metabolites and HOMA-IR . ( A–F ) Correlation of hepatic TG with plasma TG ( A ) , plasma cholesterol ( B ) , plasma glycerol ( C ) , plasma insulin ( D ) , plasma glucose ( E ) , and HOMA-IR ( F ) . r , biweight midcorrelation; p , p-value . DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 005 Increased adiposity has been linked to the incidence of NAFLD in humans . Consistent with this finding , there was a robust correlation between hepatic TG levels and adiposity ( r = 0 . 59 , p = 6 . 70 × 10−35 , Figure 4A ) . We dissected the individual fat depots and found that steatosis was associated with increased subcutaneous , gonadal and mesenteric fat mass but not with retroperitoneal fat content ( Figure 4B–E ) . The aforementioned p-values were not adjusted for multiple testing since the correlation analyses were performed based on knowledge of potential association between NAFLD and those clinical traits ( e . g . , insulin resistance , plasma lipids , and adiposity ) . Nevertheless , the correlations remained significant after Bonferroni correction for all measured traits . 10 . 7554/eLife . 05607 . 006Figure 4 . Correlation of hepatic TG content with adiposity and fat mass . ( A–E ) Correlation of hepatic TG with adiposity ( A ) , subcutaneous fat ( B ) , gonadal fat ( C ) , mesenteric fat ( D ) , and retroperitoneal fat ( E ) . r , biweight midcorrelation; p , p-value . DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 006 As liver and adipose tissues are primarily responsible for modulating liver TG accumulation , we performed transcriptomic analysis on these two tissues to gain insights into the pathogenesis of steatosis . Tables 1 , 2 show the top 50 most correlated genes with hepatic TG levels in the liver and adipose , respectively . None of the genes is in close proximity ( <1 . 5 Mb ) to each other on the same chromosome , and thus , it is unlikely that some were correlated due to shared linkage disequilibrium ( LD ) blocks with biologically linked genes . 10 . 7554/eLife . 05607 . 007Table 1 . Top 50 liver genes correlated with hepatic TG levelsDOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 007RankGene symbolrP1Cd360 . 6951 . 85E-172Mrpl160 . 6222 . 57E-133Enc10 . 6118 . 12E-1342010003K11Rik0 . 6002 . 84E-125Tceal80 . 5993 . 00E-126Cmpk10 . 5983 . 32E-127Avpr1a−0 . 5954 . 45E-128Hmgcl0 . 5907 . 60E-129Akap20 . 5811 . 86E-1110C1ra−0 . 5724 . 34E-1111Reep30 . 5621 . 16E-1012Skp1a0 . 5542 . 33E-1013Esd0 . 5513 . 11E-1014Hadh0 . 5503 . 48E-1015Syap10 . 5493 . 66E-1016Ermp10 . 5455 . 44E-1017Ang−0 . 5445 . 53E-1018Dak0 . 5436 . 24E-1019Matr3−0 . 5417 . 12E-1020Nudt90 . 5389 . 35E-1021Srsf5−0 . 5389 . 89E-1022Vps290 . 5371 . 04E-0923Ttc230 . 5371 . 05E-0924Entpd50 . 5361 . 14E-0925Chchd60 . 5351 . 27E-0926Plekha10 . 5341 . 32E-0927Mogat10 . 5311 . 76E-0928S100a100 . 5282 . 11E-0929Plin40 . 5272 . 36E-0930Anxa20 . 5252 . 73E-0931Srxn10 . 5242 . 96E-0932Cstb0 . 5233 . 36E-0933Cml1−0 . 5223 . 70E-0934Tpp10 . 5214 . 00E-0935Apoc20 . 5185 . 06E-0936F7−0 . 5156 . 32E-0937Wfdc20 . 5146 . 72E-0938Bche0 . 5146 . 72E-0939Mms19−0 . 5146 . 74E-0940Jun0 . 5137 . 05E-0941Lifr−0 . 5137 . 33E-0942Gjb1−0 . 5127 . 63E-0943Fabp20 . 5118 . 56E-0944Morc40 . 5109 . 22E-0945Rnf110 . 5109 . 33E-0946Egfr−0 . 5091 . 03E-0847Slc16a70 . 5071 . 12E-0848Gfm10 . 5051 . 32E-0849Chpt10 . 5051 . 33E-0850Rbbp4−0 . 5051 . 33E-0810 . 7554/eLife . 05607 . 008Table 2 . Top 50 adipose genes correlated with hepatic TG levelsDOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 008RankGene symbolrp1Nrbp2−0 . 6242 . 42E-132Cp−0 . 5984 . 15E-123Hoxa7−0 . 5918 . 42E-124Prkcb0 . 5851 . 62E-115Cstb0 . 5822 . 03E-116Cd530 . 5822 . 06E-117Smap20 . 5802 . 62E-118Sft2d10 . 5783 . 04E-119Btk0 . 5763 . 56E-1110Was0 . 5763 . 62E-1111Il7r0 . 5753 . 94E-1112Tmem53−0 . 5734 . 77E-1113Rgs100 . 5725 . 70E-1114Srp190 . 5716 . 27E-1115Gpc3−0 . 5706 . 53E-1116Bcl100 . 5706 . 72E-1117Gpr650 . 5706 . 82E-1118Tlr70 . 5697 . 18E-1119Efhd20 . 5697 . 26E-1120Actr30 . 5687 . 77E-1121Cd720 . 5688 . 02E-1122Dera0 . 5678 . 89E-1123Pip4k2a0 . 5669 . 63E-1124Hcst0 . 5669 . 64E-1125Tyms0 . 5669 . 78E-1126Cenpv−0 . 5669 . 79E-1127Plxnc10 . 5651 . 05E-1028Birc50 . 5651 . 07E-1029Ptpn180 . 5631 . 27E-1030Hoxa5−0 . 5621 . 37E-1031Fam105a0 . 5621 . 43E-1032Capza10 . 5621 . 44E-1033Nap1l3−0 . 5611 . 50E-1034Rgs180 . 5611 . 51E-1035Phtf20 . 5601 . 59E-1036Nckap1l0 . 5601 . 65E-1037Coro1c0 . 5601 . 67E-1038Coro1a0 . 5591 . 78E-1039Lrrfip10 . 5591 . 89E-1040C1qb0 . 5572 . 13E-1041Taok30 . 5562 . 28E-1042Ms4a6c0 . 5562 . 29E-1043Bco2−0 . 5562 . 33E-1044Adrb3−0 . 5562 . 34E-1045Arhgap90 . 5562 . 37E-1046Lrmp0 . 5562 . 47E-1047Fyb0 . 5552 . 50E-1048Lipa0 . 5552 . 63E-1049Cdt10 . 5552 . 65E-1050S100a40 . 5542 . 80E-10 In the liver , Cd36 had the highest correlation with TG levels ( r = 0 . 70 , p = 1 . 85 × 10−17 ) . CD36 is a multifunctional protein that enhances cellular FA uptake . Previous studies have shown that CD36-deficient mice are resistant to the induction of hepatic steatosis by alcohol and high-carbohydrate feeding ( Clugston et al . , 2014 ) . Besides Cd36 , many genes involved in lipid metabolism were also among the most correlated genes: Hmgcl ( HMG-CoA lyase ) , Hadh ( hydroxyacyl-CoA dehydrogenase ) , Mogat1 ( monoacylglycerol O-acyltransferase 1 ) , Plin4 ( perlipin 4 ) , Apoc2 ( apolipoprotein C-II ) , Fabp2 ( FA binding protein 2 ) , Slc16a7 ( monocarboxylic acid transporters ) , and Chpt1 ( diacylglycerol cholinephosphotransferase ) . Interestingly , expression of the proto-oncogene c-Jun was positively correlated with hepatic TG content ( r = 0 . 51 , p = 7 . 05 × 10−09 ) . Enhanced hepatic c-Jun levels were observed in NAFLD patients , which correlated with inflammation and the degree of hepatic steatosis ( Dorn et al . , 2014 ) . Increased c-Jun/AP-1 activation has been implicated in the progression of NAFLD ( Dorn et al . , 2014; Hasenfuss et al . , 2014 ) . In the adipose , hepatic steatosis was significantly correlated with genes associated with adiposity and inflammation . Prkcb ( protein kinase cβ ) , whose expression was significantly correlated with hepatic steatosis ( r = 0 . 58 , p = 1 . 62 × 10−11 ) , has been shown to be important in adipose tissue remodeling and FA metabolism ( Huang et al . , 2012 ) . Prkcb-null mice are protected against diet-induced obesity and the development of hepatic steatosis and insulin resistance ( Huang et al . , 2009 ) . Cd53 , an adipose inflammatory marker , was also highly correlated with hepatic steatosis ( r = 0 . 58 , p = 2 . 06 × 10−11 ) . Previous microarray profiling showed that Cd53 was significantly up-regulated in preadipocytes from obese human subjects ( Nair et al . , 2005 ) . Additionally , expression of many genes participating in immune response and inflammatory response was elevated in steatotic livers ( Bcl10 , Il7r , Tlr7 , and C1qb ) . Hcst ( hematopoietic cell signal transducer ) , previously shown to be a key hub gene in gene co-expression network that was significantly associated with serum TG levels ( Haas et al . , 2012 ) in humans , was also highly correlated with hepatic TG content ( r = 0 . 57 , p = 9 . 64 × 10−11 ) . The homeobox transcription factor Hoxa5 has previously been shown to be up-regulated after fat loss in human patients who have undergone bariatric surgery ( Dankel et al . , 2010 ) . In our study , Hoxa5 expression was negatively correlated with hepatic TG content ( r = −0 . 56 , p = 1 . 37 × 10−10 ) . Human GWAS studies have shown that a missense mutation in LRRFIP1 was associated with adiposity and inflammation ( Plourde et al . , 2013 ) . Lrrfip1 expression was significantly associated with hepatic steatosis in our study ( r = 0 . 56 , p = 1 . 89 × 10−10 ) . Adrb3 ( β3-adrenergic receptor ) activation induces white adipose remodeling and brown adipogensis ( Lee et al . , 2012 ) . In our study , Adrb3 expression was negatively correlated with hepatic TG levels ( r = −0 . 56 , p = 2 . 34 × 10−10 ) . Increased expression of Lipa ( lysosomal acid lipase ) , which is involved in lysosomal lipophagy and TG/cholesterol ester catabolism , was found to be associated with hepatic steatosis ( r = 0 . 56 , p = 2 . 63 × 10−10 ) . Enrichment analysis using the top 1000 hepatic genes correlated with hepatic TG levels ( Table 3 ) showed a significant enrichment of mitochondrial genes ( 1 . 57 fold , adjusted p = 1 . 99 × 10−5 ) . Among the 127 mitochondrial genes , the majority ( 100 genes ) were higher in steatotic livers , suggesting that altered mitochondria function is linked to the disease process of NAFLD . In addition , components of the extracellular matrix were enriched ( 2 . 91 fold , adjusted p = 5 . 89 × 10−3 ) and were also predominantly ( 13 out of 17 genes ) higher in steatotic livers ( Table 3 ) . Many of these genes are involved in wound healing and fibrosis , consistent with the observation of positive correlation between ALT and hepatic TG levels ( Figure 2B ) . Complement and the coagulation cascade were specifically enriched ( 3 . 03 fold , adjusted p = 3 . 93 × 10−3 ) and the genes were predominantly ( 15 out of 18 genes ) lower in steatotic livers . In the adipose tissue , mitotic cell cycle , actin polymerization , cytoskeleton organization , immune response , response to wounding , leukocyte activation and positive regulation of cytokine production , lysosome pathway , and B cell receptor signaling were all enriched ( Table 4 ) . These findings suggest that inflammation in adipose tissue likely plays a role in the development of NAFLD . Enrichment analysis using the top 500 positively or negatively regulated genes separately did not reveal any additional enriched pathways . 10 . 7554/eLife . 05607 . 009Table 3 . Pathway-enrichment analysis of the top 1000 hepatic genes correlated with hepatic TG levels , assessed with the DAVID database , and presented as total genes meeting that criterion in each pathway ( Count ) , along with Benjamini corrected p values ( Adj . p ) DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 009CategoryTermCountAdj . pFold enrichmentGOTERM_CC_FATGO:0005739 ∼ mitochondrion1271 . 99E-051 . 57KEGG_PATHWAYmmu04610:Complement and coagulation cascades183 . 93E-033 . 03GOTERM_CC_FATGO:0044420 ∼ extracellular matrix part175 . 89E-032 . 9110 . 7554/eLife . 05607 . 010Table 4 . Pathway-enrichment analysis of the top 1000 adipose genes correlated with hepatic TG levels , assessed with the DAVID database , and presented as total genes meeting that criterion in each pathway ( Count ) , along with Benjamini corrected p values ( Adj . p ) DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 010CategoryTermCountAdj . pFold enrichmentGOTERM_BP_FATGO:0000087 ∼ M phase of mitotic cell cycle369 . 69E-062 . 87KEGG_PATHWAYmmu04142:Lysosome251 . 73E-032 . 50GOTERM_BP_FATGO:0008064 ∼ regulation of actin polymerization or depolymerization195 . 97E-065 . 43GOTERM_BP_FATGO:0007010 ∼ cytoskeleton organization391 . 89E-021 . 87GOTERM_BP_FATGO:0006955 ∼ immune response501 . 80E-021 . 72GOTERM_BP_FATGO:0009611 ∼ response to wounding422 . 64E-021 . 78GOTERM_BP_FATGO:0045321 ∼ leukocyte activation347 . 64E-042 . 36KEGG_PATHWAYmmu04662:B cell receptor signaling pathway208 . 62E-043 . 08GOTERM_BP_FATGO:0001819 ∼ positive regulation of cytokine production131 . 85E-023 . 56 The expression of several genes identified in human NAFLD GWAS also showed significant correlation with hepatic TG level ( Table 5 ) ( Romeo et al . , 2008; Chambers et al . , 2011; Speliotes et al . , 2011; Kozlitina et al . , 2014 ) . In the liver , these included Gckr ( r = 0 . 19 , p = 0 . 044 ) and Lyplal1 ( r = 0 . 27 , p = 0 . 003 ) and in adipose these included Ncan ( r = 0 . 37 , p = 6 . 1 × 10−5 ) , Tm6sf2 ( r = −0 . 23 , p = 0 . 012 ) , and Trib1 ( r = 0 . 24 , p = 0 . 012 ) . 10 . 7554/eLife . 05607 . 011Table 5 . Correlation between human GWAS candidate gene expression in mouse liver and adipose tissue with hepatic TG levelDOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 011LiverAdiposerprpPnpla30 . 070 . 423−0 . 040 . 645Gckr0 . 190 . 044*0 . 140 . 156Ncan−0 . 100 . 3160 . 376 . 1 × 10-5*Tm6sf20 . 150 . 123−0 . 230 . 012*Lyplal10 . 270 . 003*−0 . 120 . 228Trib1−0 . 100 . 3130 . 240 . 012**Denotes p < 0 . 05 . To identify genetic loci associated with hepatic steatosis , we performed GWAS analysis on the hepatic TG contents with ∼200 , 000 high-quality SNPs spaced throughout the genome . The genome-wide significant threshold was set at 1 . 31 × 10−5 , which corresponds to a 1% false discovery rate ( FDR ) . We identified one genome-wide significant peak associated with hepatic TG content on chromosome 7 ( Figure 5A ) . This peak does not overlap with any loci identified for obesity and insulin resistance in the same cohort of mice ( Parks et al . , 2013 , 2015 ) . In addition , the peak locus on chromosome 7 still reached genome-wide significance when association mapping was conditioned using adiposity ( % fat ) or insulin resistance ( HOMA-IR ) as covariates ( Figure 6 ) , indicating that the effect of the chromosome 7 locus on hepatic steatosis is independent of obesity and diabetes . The peak SNP ( rs32519715 , p = 1 . 15 × 10−6 ) falls in a LD block containing 17 genes ( Figure 5B and Table 6 ) . Three suggestive peaks , on chromosomes 3 ( rs3708683; p = 5 . 47 × 10−5 ) , 9 ( rs36804270; p = 1 . 30 × 10−5 ) , and 11 ( rs13481015; p = 1 . 72 × 10−5 ) passed the genome-wide significant threshold of 5% FDR ( p = 1 . 34 × 10−4 ) . The peak on chromosome 9 also coincides with a genome-wide significant locus for insulin resistance ( data not shown ) and contains 13 genes within the LD block , whereas the peaks on chromosomes 3 and 11 contain 12 and 15 genes , respectively , within the associated LD blocks ( Table 7 ) . 10 . 7554/eLife . 05607 . 012Figure 5 . Association mapping of hepatic TG . ( A ) Manhattan plot showing the significance ( −log of p ) of all SNPs and hepatic TG . Genome-wide significance cut-off at 1% false discovery rate ( FDR ) is shown by the red line and cut-off at 5% FDR is shown in green . ( B ) Locus plot for genome-wide significant locus on chromosome 7 with approximate linkage disequilibrium block and candidate genes . DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 01210 . 7554/eLife . 05607 . 013Figure 6 . Association mapping of hepatic TG with adiposity or insulin resistance as covariates . Manhattan plot showing the significance −log10 ( p value ) of all SNPs and hepatic TG conditioned for percentage body fat ( A ) or HOMA-IR ( B ) as covariates . Genome-wide significance cut-off at 1% FDR is shown by the red line . DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 01310 . 7554/eLife . 05607 . 014Table 6 . Association p-values of candidate genes on chromosome 7 with the peak SNP for hepatic triglyceride levelsDOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 014GeneLiver p-valueLiver expressionAdipose p-valueAdipose expressionRps15aNAYesNAYesArl6ip10 . 011Yes0 . 294YesSmg1NANoNAYes4930583K01RikNANoNAYesSyt70 . 495No0 . 748YesItpripl2NAYesNAYesCoq77 . 18E-13Yes7 . 96E-14YesTmc7NAYesNAYesTmc5NANoNAYesGde13 . 21E-05Yes3 . 98E-08YesCcp110NANoNANo9030624J02RikNAYesNAYesKnop13 . 85E-07Yes0 . 043YesIqckNANoNANoGprc5b0 . 002No0 . 661YesGpr139NAYesNANoGp20 . 910No0 . 209No10 . 7554/eLife . 05607 . 015Table 7 . Candidate genes under the chromosomes 3 , 9 , and 11 lociDOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 015Chromosome 3Chromosome 9Chromosome 111110032F04RikNphp3Il12bIft80Uba5Ublcp1Smc4Acad11Rnf145Trim59Dnajc3Ebf1Kpna4AcppGm12159Gm1647Cpne4F630206G17RikArl14Mrpl3Clint1Ppm1lNudt16Lsm11B3galnt11700080E11RikThg1lNmd3Nek11Sox30SptssbAste1Adam19Otol1Atp2c1Nipal4Pik3r4Cyfip2ItkFam71b The genetic variations underlying complex traits , such as steatosis , most often affect gene expression levels rather than structural ( coding ) aspects ( Wang et al . , 2005; Hindorff et al . , 2009 ) . Most of the large differences in gene expression are due to local differences ( Orozco et al . , 2012 ) . Therefore , a useful approach to prioritizing candidate genes at a locus is to determine if the genes at the locus exhibit variation in expression that is controlled in cis . Such a variant is termed a cis-expression Quantitative Trait Locus ( eQTL ) . We therefore identified significant cis-eQTL for liver and adipose for the genes in the chromosome 7 locus ( Table 8 ) . In addition , we asked whether the expression levels correlated with the clinical trait of interest ( i . e . , hepatic TG levels ) , since that would be consistent with a causal relationship . Among the candidate genes in the chromosome 7 locus , only three genes ( Coq7 , Gde1 , and Knop1 ) have significant cis-eQTL and are also expressed in the liver ( Tables 6 , 8 ) . These three genes also have significant cis-eQTL associations in adipose tissue ( Table 8 ) . The expression variation of these three candidate genes showed a continuous spectrum across the strains , indicating that the expression variations are not bimodal ( data not shown ) . Hepatic TG levels correlated with Gde1 expression in both the liver ( r = 0 . 35 , p = 1 . 5 × 10−4 ) and adipose tissue ( r = −0 . 21 , p = 1 . 6 × 10−3 ) ( Figure 7A , C ) . Likewise , Knop1 expression correlated with hepatic TG levels ( r = 0 . 29 , p = 2 . 5 × 10−2 , Figure 7B ) but not in the adipose tissue ( r = 0 . 14 , p = 0 . 133 ) . On the other hand , hepatic TG levels did not correlate with Coq7 expression in the liver ( r = 0 . 09 , p = 0 . 368 ) or adipose tissue ( r = −0 . 14 , p = 0 . 142 ) . Gde1 ( also known as MIR16 ) encodes glycerophosphodiester phosphodiesterase 1 , a ubiquitously expressed enzyme involved in phospholipid metabolism , whereas Knop1 encodes a lysine-rich nucleolar protein . Neither of these two genes has previously been identified in studies related to TG or lipid metabolism . We also examined coding variants for genes at the chromosome 7 locus using the PROVEAN prediction tool ( Choi and Chan , 2015 ) . A number of genes exhibited missense variants but these tended to be neutral and not likely to cause deleterious effects on protein stability and function ( Table 9 ) . Only the Q117R substitution in Syt7 was predicted to be deleterious . Syt7 ( synaptotagmin VII ) belongs to a protein family , which mediates Ca2+−dependent vesicular trafficking and exocytosis ( Moghadam and& Jackson , 2013 ) . Ablation of Syt7 has been shown to decrease insulin and glucagon secretion in pancreatic cells ( Gustavsson et al . , 2008 , 2009 ) . 10 . 7554/eLife . 05607 . 016Table 8 . Significant cis-eQTL at chromosome 7 locusDOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 016GeneGene start positionSNP IDPositionLiver p-valueArl6ip1118118891rs306680411182669692 . 79E-24Coq7118509659rs324615101183500095 . 01E-24Gde1118688545rs325114191190705211 . 32E-06Knop1118842237rs322467451191608231 . 18E-07GeneGene start positionSNP IDPositionAdipose p-valueArl6ip1118118891rs306680411182669691 . 14E-36Coq7118509659rs324308511179610921 . 65E-18Gde1118688545rs315164251183727862 . 53E-12Knop1118842237rs325323701192889743 . 47E-10Gprc5b118972040rs486479261189184553 . 70E-09eQTL , expression Quantitative Trait Locus . 10 . 7554/eLife . 05607 . 017Figure 7 . Correlation of candidate gene expression with hepatic TG content . ( A–B ) Correlation of hepatic TG with expression levels of Gde1 ( A ) and Knop1 ( B ) in the liver . ( C ) Correlation of hepatic TG with expression levels of Gde1 in the white adipose tissue . r , biweight midcorrelation; p , p-value . DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 01710 . 7554/eLife . 05607 . 018Table 9 . Missense variants of candidate genes on chromosome 7DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 018GeneMissense variantsRps15aNoneArl6ip1NoneSmg1None4930583K01RikNoneSyt7Q117RItpripl2R204H , V240L , S420GCoq7A290GTmc7C73S , T358STmc5M11V , Q42L , V82D , A119T , V139I , P179S , A243G , R258K , V448I , E737DGde1NoneCcp110R180K , I199V , N248S , A326V , A332T , P427T , S439P , S445P , F624S , G746S9030624J02RikS3A , D23G , V122A , G175A , T660MKnop1I70M , V49AIqckV49A , S148NGprc5bNoneGpr139NoneGp2V79D , R483GThe effect of missense mutation was assessed by PROVEAN software . Neutral amino acid substitutions , which do not affect protein stability and function are shown in blue whereas deleterious mutations are labeled in red . Based on cis-eQTL analyses , the chromosome 3 locus contains 3 strong candidate genes: Smc4 , Kpna4 , and B3galnt1 . Expression of Smc4 in adipose tissue significantly correlated with hepatic TG content ( r = 0 . 51 and p = 9 . 8 × 10−9 ) . Smc4 encodes structural maintenance of chromosomes 4-like 1 , which is a core subunit of condensins I and II , large protein complexes involved in chromosome condensation and repair ( Onn et al . , 2007 ) . Hepatic expression of Kpna4 was correlated with liver TG content ( r = 0 . 23 and p = 0 . 01 ) . Kpna4 encodes the subunit alpha-3 of importin , which is a cytoplasmic protein that recognizes nuclear localization signals of protein to be imported into the nucleus . Adipose expression of B3galnt1 , encoding for the enzyme beta-1 , 3-galactosyltransferase 3 , was correlated with steatosis ( r = 0 . 30 and p = 0 . 001 ) . The finding of positive correlation between hepatic Gde1 expression and steatosis suggests that increased expression of Gde1 would promote hepatic TG accumulation . To directly assess the effect of Gde1 on hepatic TG in vivo , Gde1 was overexpressed in 8-week-old C57BL/6 mice ( Ad-Gde1 , 1 × 109 pfu per mouse , i . v . ) by adenoviral transduction ( Figure 8A ) . The control group ( Ad-LacZ ) received the same dose of adenovirus expressing LacZ . Mice were fed a HF/HS diet for 7 days after adenovirus injection and their hepatic lipids were measured . This regimen was chosen because gene expression by adenoviral transduction is only sustained for a short period of time ( a few weeks or less ) . Preliminary studies in mice showed that HF/HS diet induced a threefold increase in hepatic TG accumulation in 1 week ( data not shown ) . While there was no significant difference in body weight , the weight of livers from Gde1-overexpressing mice was 40% higher than that of the control ( Figure 8B , p = 4 . 7 × 10−5 ) . In addition , plasma TG , TC , and FFA were all elevated in Ad-Gde1 mice ( Figure 8C ) . MRI analysis showed that livers from Gde1-overexpressing mice contained significantly higher fat content ( Figure 8D , p = 0 . 0002 ) . Lipid analyses revealed that the increase in hepatic fat content was primarily due to increased accumulation of TG ( Figure 8E , p = 0 . 014 ) , as hepatic TC and phospholipid content were not significantly different from the control group ( Figure 8F , G ) . Hepatic genes involved in TG biosynthesis ( Fasn , Dgat2 , and Gpd1 ) were down-regulated in mice overexpressing Gde1 ( Figure 8H ) . 10 . 7554/eLife . 05607 . 019Figure 8 . Effects of Gde1 overexpression in mice by adenoviral transduction . C57BL/6 mice were injected with Ad-Gde1 ( 1 × 109 pfu per mouse , i . v . ) and fed with a HF/HS diet for 7 days . Control group received the same dose of Ad-LacZ . ( A ) Western-blot of liver homogenate using anti-GDE1 or anti-tubulin antibody . ( B ) Comparison of liver weight between Gde1-overexpressing mice and the control mice . ( C ) Differences in plasma triglyceride ( TG ) , TC , and free fatty acids levels between Gde1-overexpressing mice ( filled bars ) and the control mice ( empty bars ) . ( D ) Hepatic fat percentage in the two groups of mice was determined by MRI . ( E–G ) Liver lipids were extracted and quantified: triglyceride ( TG ) , TC , and phospholipids ( PL ) . ( H ) Expression of lipogenic genes was measured by qPCR and normalized to the level of the housekeeping gene 36B4 . Ad-LacZ ( empty bars ) and Ad-Gde1 ( filled bars ) Results are presented as mean + SD ( n = 7–8 ) * denotes p < 0 . 05 and ** denotes p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 019 Complementary to the overexpression studies , we also knocked down hepatic Gde1 expression in vivo by adenoviral expression of shRNA . The virus dose and diet treatment were identical to that described above for the overexpression studies . Both Gde1 protein and mRNA levels were ∼75% decreased in the mice received adenoviral shRNA ( Figure 9A , B ) . In contrast to the overexpression studies , knockdown of Gde1 led to a significant decrease in hepatic TG content ( Figure 9C , p = 0 . 011 ) , whereas TC and phospholipid levels were not different from the control group ( Figure 9D , E ) . Similarly , expression of lipogenic genes ( Fasn , Dgat2 , and Gpd1 ) was increased in Gde1-knockdown livers ( Figure 9F ) . These findings support a causal role for Gde1 in hepatic steatosis under the chromosome 7 locus . 10 . 7554/eLife . 05607 . 020Figure 9 . Effects of Gde1 knockdown in mice by adenoviral transduction . C57BL/6 mice were injected with Ad-shGde1 ( 1 × 109 pfu per mouse , i . v . ) and fed with a HF/HS diet for 7 days . Control group received the same dose of Ad-Ctl . ( A ) Equal amounts of liver protein were loaded in each lane and Western-blotted using anti-GDE1 or anti-actin antibody . ( B ) Comparison of Gde1 mRNA levels between Gde1-knockdown mice and the control mice . ( C–E ) Liver lipids were extracted and quantified: triglyceride ( TG ) , TC , and phospholipids ( PL ) . ( F ) Expression of lipogenic genes was measured by qPCR and normalized to the level of the housekeeping gene 36B4 . Results are presented as mean + SD ( n = 11–12 ) * denotes p < 0 . 05 and ** denotes p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 020 It is noteworthy that the direction of the effect of Gde1 expression with respect to hepatic TG levels in these studies is consistent with that observed in the HMDP in liver ( i . e . , a positive correlation ) . To identify metabolites associated with hepatic steatosis , we employed a metabolomic approach to measure 47 metabolites ( amino acids , amines , and other polar compounds , Supplementary file 2 ) in the plasma of mice after 8 weeks of HF/HS feeding . Correlation analysis revealed a significant negative relationship between hepatic TG levels and plasma levels of arginine ( r = −0 . 53 , p = 9 . 85 × 10−12 ) and its degradative metabolite ornithine ( r = −0 . 18 , p = 0 . 027 ) ( Figure 10A , B ) , whereas citrulline , another degradative metabolite of arginine , showed a positive correlation with hepatic TG content ( r = 0 . 18 , p = 0 . 034 , Figure 10C ) . Hepatic TG levels were positively correlated with plasma levels of trimethylamine-N-oxide ( TMANO , r = 0 . 18 , p = 0 . 034 , Figure 10D ) . Increased TMANO levels have previously been implicated in the susceptibility of strain 129S6 mice to diet-induced impaired glucose homeostasis and NAFLD ( Dumas et al . , 2006 ) . TMANO is an oxidative product of trimethylamine ( TMA ) , a metabolite of choline in animals . TMANO levels are regulated by both genetic and dietary factors and are strongly associated with atherosclerosis ( Bennett et al . , 2013 ) . The oxidation of TMA is catalyzed by the hepatic flavin-containing monooxygenase ( FMO ) family of enzymes with FMO3 having the highest catalytic activity . No significant correlation between FMO3 expression and hepatic TG content was observed; however , among the 5 members in the FMO family , FMO5 expression exhibited a significant correlation with hepatic TG content ( r = 0 . 46 , p = 2 . 71 × 10−7 ) . Hepatic TG content was also positively associated with plasma creatine ( r = 0 . 23 , p = 6 × 10−3 , Figure 10E ) and creatinine levels ( r = 0 . 25 , p = 6 × 10−3 , Figure 10F ) . 10 . 7554/eLife . 05607 . 021Figure 10 . Correlation of hepatic TG and polar metabolites in the plasma . Correlation of hepatic TG with plasma levels of arginine ( A ) , ornithine ( B ) , citrulline ( C ) , TMANO ( D ) , creatine ( E ) , and creatinine ( F ) . r , biweight midcorrelation; p , p-value . DOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 021 Gut microbiota has been implicated in certain human inflammatory and metabolic diseases , including obesity ( Ley et al . , 2006; Turnbaugh et al . , 2009 ) , inflammatory bowel disease ( Lepage et al . , 2011 ) , and type II diabetes ( Qin et al . , 2012 ) . Accumulating evidence also suggests that alterations in gut microbiota composition could contribute to the susceptibility and progression of NAFLD ( Mouzaki et al . , 2013; Zhu et al . , 2013 ) . To understand the relationship between microbiota composition and hepatic lipid accumulation in mice fed a HF/HS diet , we employed deep sequencing of a conserved region of the bacterial 16S ribosomal RNA gene to determine the cecal microbiota composition in 237 male mice among 100 different strains . There was no significant correlation between hepatic TG levels and overall microbial diversity , as determined by Shannon diversity index ( Mills and& Wassel , 1980 ) . For the bacterial families and genera detected at significant levels , none were significantly associated with hepatic TG levels ( 5% FDR ) ( Table 10 ) . However , associations between certain taxa were observed with other hepatic lipids ( Table 10 ) . Coprococcus and Oscillospira were positively associated with hepatic unesterified cholesterol levels , whereas negative associations were observed with Blatuia . Blautia and Allobaculum were negatively associated with TC . Coprococcus and Oscillospira were positively associated with phospholipids , whereas Ruminococcus showed negative association ( Table 10 ) . 10 . 7554/eLife . 05607 . 022Table 10 . Correlation between hepatic lipids and gut microbiotaDOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 022TGTCUCPLFamily: Clostridiaceae−0 . 1330 . 0350 . 1160 . 052 Erysipelotrichaceae0 . 1380 . 005−0 . 057−0 . 016 Lachnospiraceae0 . 066−0 . 1120 . 0000 . 045 Mogibacteriaceae0 . 033−0 . 055−0 . 092−0 . 089 Peptostreptococcaceae0 . 006−0 . 019−0 . 0030 . 096 Rikenellaceae0 . 0020 . 1420 . 125−0 . 047 Ruminococcaceae0 . 1380 . 1760 . 1690 . 030 S24-7−0 . 0920 . 003−0 . 111−0 . 166Genus: 02d06−0 . 146−0 . 062−0 . 050 . 012 Adlercreutzia0 . 0140 . 0010 . 1330 . 084 Akkermansia−0 . 0310 . 0290 . 0800 . 074 Allobaculum−0 . 083−0 . 205*−0 . 178−0 . 009 Anaeroplasma−0 . 019−0 . 0470 . 0010 . 018 Bifidobacterium−0 . 014−0 . 082−0 . 131−0 . 084 Blautia0 . 003−0 . 217*−0 . 244†−0 . 183 Clostridium−0 . 108−0 . 063−0 . 044−0 . 008 Clostridium . 10 . 0810 . 044−0 . 0230 . 015 Coprobacillus−0 . 008−0 . 075−0 . 138−0 . 130 Coprococcus−0 . 0160 . 1740 . 261†0 . 212* Dehalobacterium0 . 0860 . 0700 . 0660 . 070 Dorea0 . 0780 . 0700 . 0560 . 081 Lactobacillus0 . 0050 . 021−0 . 012−0 . 073 Lactococcus0 . 0850 . 0010 . 0480 . 065 Oscillospira0 . 1400 . 1270 . 261†0 . 206* R . Gnavus0 . 103−0 . 024−0 . 037−0 . 016 Roseburia0 . 063−0 . 090−0 . 132−0 . 151 Ruminococcus0 . 099−0 . 131−0 . 160−0 . 218* Sarcina−0 . 1090 . 0350 . 0670 . 014 SMB53−0 . 012−0 . 069−0 . 0290 . 052 Turicibacter−0 . 0350 . 0160 . 0190 . 013TG: triglyceride; TC: total cholesterol; UC: unesterified cholesterol; PL: phospholipids . *Denotes FDR < 0 . 05 . †Denotes FDR <0 . 01 .
Along with the increased prevalence of obesity and diabetes , NAFLD has become the most common cause of chronic liver disease in a number of Western countries including the United States ( Ratziu et al . , 2010; Vernon et al . , 2011 ) . The etiology and pathogenesis of NAFLD are still poorly understood , reflecting the genetic and environmental heterogeneity of the disease . The molecular basis of the metabolic response to dietary fat and its role in NAFLD development remains to be elucidated . Previous studies using inbred mouse strains have shown that genetic background affects lipid accumulation in the liver ( Hill-Baskin et al . , 2009; Millward et al . , 2009; Shockley et al . , 2009; Montgomery et al . , 2013 ) . In this study , we employed over 100 strains of inbred and recombinant inbred mice and demonstrated vast genetic variation in hepatic TG accumulation in fed with a HF/HS diet . We examined the relationship between NAFLD and insulin resistance , obesity , plasma metabolites , plasma lipoproteins , and gut microbiota . We also carried out global gene expression array analyses on two key metabolically relevant tissues for NAFLD ( liver and white adipose tissue ) and identified molecular pathways enriched in NAFLD . Finally , we identified genetic loci contributing to NAFLD in common inbred strains of mice and validated the role of a novel gene , Gde1 , at one of the loci . Our data substantially expand our knowledge of the genetic architecture of NAFLD and provide a rich resource for future biochemical and genetic studies . While there is strong evidence that genetic factors contribute significantly to the onset and progression of the disease , only a few genes have been identified by human GWAS studies ( Browning et al . , 2004; Romeo et al . , 2008; Chalasani et al . , 2010; Adams et al . , 2013; Kozlitina et al . , 2014 ) . Because genetic and environmental factors can be strictly controlled in mice , such models are valuable for dissecting the genetic and environmental contributions to complex disease traits . Nevertheless , classical linkage analysis approaches to dissect NAFLD in mice have been relatively unsuccessful . While a number of loci have been identified by quantitative QTL analyses , a major problem has been an inability to carry out fine genetic mapping to identify the responsible genes ( Rangnekar et al . , 2006; Kumazawa et al . , 2007; Minkina et al . , 2012 ) . In this study , we employed an innovative genome-wide association approach using >100 strains of inbred and recombinant inbred strains of mice to finely map genetic loci contributing to the development of steatosis . This approach circumvents the obstacles associated with human studies ( namely , environmental heterogeneity ) and mouse linkage analyses ( namely , poor mapping resolution ) . Furthermore , by integrating transcriptomic information for the liver and adipose tissue , we identified two high-confidence candidate genes ( Gde1 and Knop1 ) for hepatic steatosis on chromosome 7 . We pursued Gde1 using experimental perturbation based on its known role in lipid metabolism . Several lines of evidence indicate that Gde1 is a causal gene for hepatic steatosis in the chromosome 7 locus . First , the presence of cis-eQTL suggests that Gde1 expression is regulated locally , associating with the SNP haplotype . Second , hepatic expression of Gde1 is significantly correlated with the levels of hepatic TG . Third , overexpression of Gde1 promoted specifically TG accumulation in vivo , whereas hepatic cholesterol and phospholipid levels were unaffected . Forth , shRNA knockdown of Gde1 reduced hepatic TG accumulation . Gde1 encodes glycerophosphodiester phosphodiesterase 1 ( EC 3 . 1 . 4 . 46 ) , a broadly expressed integral membrane glycoprotein , which catalyzes the degradation of glycerophosphoethanolamine and glycerophosphocholine ( Okazaki et al . , 2010; Simon and& Cravatt , 2010 ) . Two forms of the protein ( ∼37 kDa and 43 kDa ) were detected in Western blots of endogenous and adenoviral overexpressed GDE1 , which is likely be due to variable glycosylation as previously documented ( Zheng et al . , 2000 ) . Its role in triglyceride metabolism has not been documented . Expression levels of TG biosynthetic pathways in the livers of Gde1-overexpressing mice were down-regulated , indicating that increased de novo lipogenesis due to increased lipogenic gene expression is unlikely the cause of TG accumulation . Increased expression of Cd36 in Gde1-overexpressing livers raises the possibility that Cd36 acts as a mediator of the effect of Gde1 through its lipid transport activity . One of the products of the Gde1 enzymatic reaction is glycerol-3-phosphate , which can be converted to phosphatidate by glycerol-3-phosphate acyltransferase ( GPAT ) and subsequently dephosphorylated by the phosphatidate phosphatase lipin to form diacylglycerol ( DAG ) . DAG can then be acylated by DGAT to form TG . We hypothesize that Gde1 affects the availability of glycerol-3-phosphate and modulates the flux of TG in the liver . The negative correlation of adipose Gde1 expression and hepatic TG may be due to a different role of glycerol-3-phosphate in the adipose tissue . Increased glycerol-3-phosphate production ( when Gde1 expression is high ) in the adipose tissue would promote FA re-esterification , leading to a decrease in FA release to the circulation . This reduction in FA supply to the liver may result in diminished TG synthesis . The opposite pattern of regulation of Gde1 implicates tissue-specific regulatory elements . Mouse ENCODE data ( Yue et al . , 2014 ) showed that there are differences in DNA hypersensitive sites in the gene region , suggesting that there may be differences in DNA accessibility and transcription factor binding in the liver and adipose tissue . The candidate genes under the chromosome 9 locus are likely to affect hepatic TG content via an insulin-dependent manner as the peak of association was diminished by co-mapping with obesity or insulin resistance . Insulin is a key hormone that drives lipogenesis and hepatic steatosis is often accompanied by hepatic insulin resistance . Our data showed that hepatic TG load has a robust association with plasma insulin levels and HOMA-IR ( Figure 6 ) . Impaired insulin signaling in the liver leads to the failure of insulin to suppress gluconeogenesis through the FoxO1 pathway , leading to hyperglycemia and ultimately diabetes ( Haas and Biddinger , 2009; Leavens and& Birnbaum , 2011 ) . Paradoxically , insulin-stimulated hepatic lipogenesis through SREBP-1c induction is not impaired in steatosis-associated insulin resistant livers ( Brown and Goldstein , 2008; Li et al . , 2010 ) . This selective insulin resistance leads to increased production of lipids and steatosis . The chromosome 3 signal was not affected by conditioning on percentage body fat or HOMA-IR , suggesting that the causal gene ( s ) determining hepatic TG content at this locus are unlikely to be mediated by pathways involving body fat or insulin sensitivity . Our observed enrichment of mitochondrial genes in steatotic livers suggests that disrupted mitochondrial bioenergetics may play a role in NAFLD pathobiology . This is in accordance with finding that chronic consumption of a HF diet-induced NAFLD with reduced mitochondrial oxidation and increased ROS production ( Mantena et al . , 2009 ) . Obesity-induced steatosis has also been linked to decreased hepatic ATP synthesis ( Chavin et al . , 1999 ) . These findings highlighted the importance of mitochondrial function in the pathogenesis of NAFLD . Variations in mitochondrial capacity and activity may contribute to the differences in susceptibility to HF/HS diet challenge . Strains with higher mitochondrial capacity would be more resistant to the development of NAFLD due to more efficient oxidation and disposal of surplus nutrients and lower production of ROS as a result of more efficient coupling between oxidation and phosphorylation . Determining the connection between the genetic determinants of mitochondrial bioenergetics and hepatic TG accumulation could provide important insights into the etiology of NAFLD . Metabolomics has been used to identify biomarkers for NAFLD ( reviewed in Dumas et al . , 2014 ) . Our analysis showed that hepatic steatosis was associated with low-plasma levels of arginine . The inverse relationship between arginine levels and hepatic TG content may be due to increased arginine degradation in steatotic mice . Previous studies have shown that an arginine-deficient diet-induced hepatic steatosis in rats ( Milner and& Hassan , 1981 ) . The levels of arginase 1 , which catabolizes L-arginine to urea and L-ornithine , were increased in steatotic livers from mice fed a high-fat diet ( Eccleston et al . , 2011 ) . Exposure to HF diet led to decreased hepatic levels of activated endothelial nitric oxide synthase ( eNOS ) , which converts L-arginine to nitric oxide ( NO ) and citrulline . In our study , gene expression of Arg1 and Nos3 did not show a significant correlation with hepatic TG content , suggesting the effect of HF diet on the activities of these two enzymes acts post-transcriptionally . Arginine plays an important role in maintaining the integrity of cell junctions and in the regulation of the epithelial barrier ( Marc Rhoads and& Wu , 2009 ) . It also modulates immune response and is essential for tissue healing ( Wu et al . , 2009 ) . Low levels of arginine may lead to increased gut bacterial and endotoxin translocation , which promotes liver inflammation and NAFLD progression . We observed that TMANO was positively correlated ( r = 0 . 18 , p = 0 . 034 ) with hepatic steatosis; however , the correlation was not significant after correction for multiple testing . Previous results from studies of 129S6 mice , a model of diet-induced fatty liver disease , showed that TMANO was a marker for NAFLD ( Dumas et al . , 2006 ) . Increased TMANO levels may indicate increased conversion of dietary choline into methylamines by microbiota , leading to reduced bioavailability of choline and mimicking the effect of a choline-deficient diet . Our study also identified Coprococcus and Oscillospira as potentially important microbes in lipid homeostasis , as the abundance of these microbes showed significant association with levels of cholesterol and phospholipids . Studies in humans showed a large reduction in abundance of these microbes in obese and NASH subjects ( Zhu et al . , 2013 ) . Little is known about the biochemical activities of these microbes; however , it has been shown that dietary composition could affect their abundance in humans ( Walker et al . , 2011 ) . Studies of diet-induced obesity in mice suggest that abundance of Oscillospira may affect gut barrier function in the proximal colon ( Lam et al . , 2012 ) . Further analysis of the genomic sequences to identify the specific strains involved could provide new insights into the role of these microbes in the pathogenesis of NAFLD . Our study has generated a rich resource for future analyses of NAFLD . The primary data are available from the authors and summary tables are posted at http://systems . genetics . ucla . edu .
The HMDP strains have previously been described in detail ( Bennett et al . , 2010 ) . The strain names and number of mice used in this study are listed in Table 11 . All mice were obtained from the Jackson Laboratory and bred at University of California , Los Angeles . The experimental design for the diet studies was previously described ( Parks et al . , 2013 ) . Briefly , male mice were maintained on a chow diet ( Ralston Purina Company ) until 8 weeks of age when they were given a HF/HS diet ( Research Diets-D12266B , New Brunswick , NJ ) with the following composition: 16 . 8% kcal protein , 51 . 4% kcal carbohydrate , 31 . 8% kcal fat . After 8 weeks on the HF/HS diet , mice were sacrificed after a 4-hr fast . Blood was collected by retro-orbital bleeding under isoflurane anesthesia . Plasma and livers were frozen immediately in liquid nitrogen and stored at −80°C until analysis . All assays and phenotypes were from the same mice . Mice were housed in rooms with a 14-hr light/10-hr dark cycle ( light is on between 6 a . m . and 8 p . m . ) . On the day of experiment , mice were fasted at 6:30 a . m . and sacrificed between 10:30 a . m . and noon . All animal procedures were approved by the Institutional Care and Use Committee ( IACUC ) at University of California , Los Angeles . 10 . 7554/eLife . 05607 . 023Table 11 . Inbred and recombinant inbred strains used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 05607 . 023Inbred strainsRecombinant inbred strainsStrainnStrainnStrainn129X1/SvJ4AXB10/PgnJ3BXD1/TyJ3A/J4AXB12/PgnJ5BXD11/TyJ2AKR/J5AXB13/PgnJ5BXD12/TyJ5BALB/cJ5AXB15/PgnJ4BXD13/TyJ2BTBR T<+> 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 Liver lipids were extracted by the method of Folch et al . ( 1957 ) . About 100 mg of livers was used for lipid extraction and the dried organic extract was dissolved in 1 . 8% ( wt/vol ) Triton X-100 . The amount of lipids in each extract was determined by colorimetric assay from Sigma ( St . Louis , MO ) ( triglyceride , TC and unesterified cholesterol ) and Wako ( Richmond , VA ) ( phospholipids ) according to the manufacturer's instructions . Plasma ALT activity was assayed using a kinetic colorimetric assay kit from Pointe Scientific , Inc ( Canton , MI , USA ) according to the manufacturer's protocol . Activity of ALT was determined by the rate of decrease in NADH as measured by the change in absorbance at 340 nm . Flash-frozen liver and epididymal adipose samples were weighed and homogenized in Qiazol ( Qiagen , Valencia , CA ) , and RNA was isolated according to the manufacturer's protocol using RNeasy columns ( Qiagen ) . Isolated RNA was analyzed for global gene expression using Affymetrix HT_MG430A arrays , and data from the microarray analysis were filtered as described ( Bennett et al . , 2010 ) . Gene enrichment analysis was performed using the Database for Annotation , Visualization and Integrated Discovery ( DAVID , v6 . 7 ) program ( Huang da et al . , 2009; Huang da et al . , 2009 ) . Genotypes for 113 strains of mice ( 29 classical inbred , 84 recombinant inbred ) were obtained from Jackson Laboratories using the Mouse Diversity Array ( Yang et al . , 2009 ) . After removing SNPs that were flagged as poor quality , 459911 SNPs remained . We further filtered these to about 200 , 000 SNPs by removing SNPs that did not have a minor allele frequency of >5% and a missing genotype rate of <10% . Genome-wide association mapping of the hepatic TG content was performed using FaST-LMM ( Factored Spectrally Transformed Linear Mixed Models ) , which uses a linear mixed model to correct for population structure ( Listgarten et al . , 2012 ) . Cut-off values for genome-wide significance were determined by computing the FDR estimated by the q values ( Storey and& Tibshirani , 2003 ) . Recombinant adenovirus was generated using the AdEasy system as previously described ( Bennett et al . , 2013 ) . Briefly , a shuttle vector containing the full-length mouse Gde1 cDNA and a C-terminal c-myc tag or bacterial LacZ sequence was cotransformed with the adenoviral backbone plasmid pAdEasy-1 for homologous recombination in Escherichia coli BJ5183 cells . Positive recombinants were linearized and transfected into 293 cells for virus packaging and propagation . Adenoviruses were purified by CsCl banding and stored at −80°C until use . Adenovirus expressing shRNA for Gde1 ( 5′-CCGGGACATCGAGTTTACTTCTGATCTCGAGATCAGAAGTAAACTCGATGTCTTTTTG-3′ ) driven by a U6 promoter was generated in a similar fashion . For adenoviral infection , 8-week-old chow-fed male C57BL/6 mice ( 8 per group ) were injected with adenoviral construct ( 1 × 109 pfu diluted in 0 . 2 ml saline , i . v . ) via the tail vein . After injection , the mice were switched to a HF/HS diet . The control group consisted of mice injected with adenoviral construct expressing the LacZ gene in the overexpression studies or empty virus in shRNA knockdown experiments . Mice were sacrificed 7 days post injection after a 4-hr fast . Expression of Gde1 in the liver was assessed by qPCR and Western blotting . Data from mice showing the absence of overexpression or knockdown were excluded from the analysis . Metabolic profiling by LC-MS of amino acids , biogenic amines , and other polar metabolites in plasma was performed as previously described ( Wang et al . , 2011; Roberts et al . , 2012 ) . Metabolite concentrations were determined using the standard addition method ( Ito and Tsukada , 2002 ) . Microbial DNA was isolated from the cecum using the PowerSoil DNA Isolation Kit according to the manufacturer's instructions ( MO BIO Laboratories , Carlsbad , CA ) . Region-specific primers including the Illumina flowcell adapter sequences were used for amplifying the V4 region of the 16S rRNA gene . The reverse amplicon primer contains a 12-base barcode sequence that allows sample pooling for sequencing . The barcoded primers and sample preparation were performed as previously described ( Hamady et al . , 2008; Costello et al . , 2009; Caporaso et al . , 2012 ) . Each sample was amplified in triplicate , combined and cleaned using the PCR clean-up kit ( MO BIO Laboratories , Carlsbad , CA ) . Cleaned and quantified amplicons were sequenced with the Illumina MiSeq machine at the GenoSeq Core Facility at the University of California , Los Angeles using 500-cycle PE kit . The MiSeq run contained a control library , which was made from phiX174 as described ( Caporaso et al . , 2012 ) . The raw data from the MiSeq run were first processed through a quality filter using established guidelines ( Bokulich et al . , 2013 ) . The remaining sequences were analyzed using the open source software package Quantitative Insights Into Microbial Ecology ( QIIME ) version 1 . 7 . 0 ( Caporaso et al . , 2010; Kuczynski et al . , 2011 ) . Demultiplexed sequences from all of the samples were clustered into operational taxonomic units ( OTUs ) based on their sequence similarity ( 97% identity ) using a reference based OTU picking protocol in QIIME . The taxonomic composition was assigned to the representative sequence of each OTU using Ribosomal Database Project ( RDP ) Classifier 2 . 0 . 1 ( greengenes 13-08 ) ( Wang et al . , 2007 ) . The relative abundance of bacteria at each taxonomic level ( e . g . , phylum , class , order , family , and genus ) was computed for each mouse . Biweight midcorrelation ( bicor ) values between microbiota and measured traits were determined using R statistic program . | Non-alcoholic fatty liver disease is a major health problem worldwide and is caused by an abnormal build-up of fat molecules in liver cells that disrupts how the cells work . Although many people with the disease show only mild or no symptoms , if the disease progresses the consequences—such as organ damage and an increased risk of liver cancer—can be severe . Although non-alcoholic fatty liver disease has been linked with obesity and diabetes , how it develops is poorly understood . The most widely supported explanation suggests that the disease begins with an imbalance in the process that normally maintains the correct amount of fat molecules called triglycerides inside cells . As a result , triglycerides accumulate in the liver cells in a process known as steatosis , which is then thought to make the liver vulnerable to further problems . However , this theory has been questioned by genetic experiments that suggest triglyceride build-up actually protects cells from other kinds of damage . Hui et al . studied mice that had been fed a diet that was high in fat and sugar . The extent of liver steatosis varied considerably between the mice , with some mice accumulating 30 times more triglyceride in their liver than others . The underlying variation in the genes of the mice was then examined to investigate whether this can explain the differences in liver condition . This revealed at least three DNA stretches that appear to be linked to triglyceride accumulation in the liver , including several genes that appear to be active during steatosis . One of these genes , known as Gde1 , had not previously been shown to have a role in controlling how cells make and use triglycerides . To confirm the role of Gde1 , Hui et al . artificially turned the gene on in some mice and prevented it from turning on in others . Turning on Gde1 significantly increased the amount of triglyceride in the liver and keeping it turned off decreased triglyceride levels . Hui et al . suggest that this is because Gde1 helps to make a precursor molecule that is needed to build triglycerides . Certain gut bacteria also appear to be linked to steatosis . This study used a population-based approach in mice to examine genetic factors in the development of fatty liver disease . The challenge now is to find out how the genes work and to understand their interactions with each other and with the environment . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"genetics",
"and",
"genomics"
] | 2015 | The genetic architecture of NAFLD among inbred strains of mice |
Interdisciplinary syntheses are needed to scale up discovery of the environmental drivers and molecular basis of adaptation in nature . Here we integrated novel approaches using whole genome sequences , satellite remote sensing , and transgenic experiments to study natural loss-of-function alleles associated with drought histories in wild Arabidopsis thaliana . The genes we identified exhibit population genetic signatures of parallel molecular evolution , selection for loss-of-function , and shared associations with flowering time phenotypes in directions consistent with longstanding adaptive hypotheses seven times more often than expected by chance . We then confirmed predicted phenotypes experimentally in transgenic knockout lines . These findings reveal the importance of drought timing to explain the evolution of alternative drought tolerance strategies and further challenge popular assumptions about the adaptive value of genetic loss-of-function in nature . These results also motivate improved species-wide sequencing efforts to better identify loss-of-function variants and inspire new opportunities for engineering climate resilience in crops .
Discovering the environmental drivers and functional genetics of adaptation in nature is a key goal of evolutionary biology and valuable to advance applied genetics in agriculture . Understanding the genetics of drought adaptation in plants is particularly important as crop losses resulting from droughts affect billions of people each year , posing the greatest threat to global food stability . Because droughts also impose strong selection on natural plant populations , investigating drought adaptation in wild species is both useful for addressing fundamental questions of evolutionary biology , such as determining whether adaptation proceeds by few or many alleles , and informative for efforts to reverse engineer drought tolerance in crops ( Mickelbart et al . , 2015 ) . Such an evolutionary research program is motivated by the need to understand adaptive drought tolerance strategies for different types of drought conditions , which can vary in severity and timing ( Tardieu , 2012 ) . Furthermore , previous limitations of single gene approaches have reinforced the necessity of developing methods to identify beneficial alleles at genomic scales and functional molecular resolutions ( Dean and Thornton , 2007; Passioura , 2010 ) . Drought stress can occur throughout the year and drought timing is forecast to change over the next century ( Trenberth et al . , 2014 ) . While dramatic evolutionary responses to drought events have been documented , ( e . g . Franks et al . , 2007 ) , little is known about the relationship between drought timing and adaptation . However , the observation both in nature and agriculture that plants are particularly susceptible to drought while flowering ( Nam et al . , 2001; Dietrich and Smith , 2016 ) has contributed to the longstanding hypothesis that adaptive flowering time should reflect patterns in the seasonal timing of drought events ( Passioura , 1996 ) . Detailed studies of life history also reveal that locally adapted Arabidopsis thaliana ( Arabidopsis hereafter ) populations begin flowering in their home environments just prior to and after periods of increased historical drought frequency ( Mojica et al . , 2016 ) . Flowering time in Arabidopsis is correlated with other drought tolerance traits such as water use efficiency and can serve as a proxy for alternative drought tolerance strategies , with early flowering genotypes being associated with low water use efficiency ( drought escape strategy ) and late flowering genotypes with high water use efficiency ( dehydration avoidance strategy ) ( McKay et al . , 2003; Lovell et al . , 2013; Kenney et al . , 2014 ) . Thus , the historical timing of drought experienced by locally adapted populations may explain the evolution of these strategies and the distribution of alleles responsible for natural flowering time variation . This hypothesis motivated our investigation to identify alleles associated with drought timing and test the prediction that they contribute to adaptive flowering time evolution . Identifying functionally relevant genetic variation contributing to adaptation is needed to understand fundamental evolutionary processes . In contrast to early theoretical predictions and popular assumptions , loss-of-function ( LoF ) alleles , those that eliminate or ‘knockout’ a gene’s molecular function , are overrepresented among alleles reported as responsible for crop improvement and often produce adaptive phenotypes in wild species ( Hoekstra et al . , 2006; Rausher , 2008; Olsen and Wendel , 2013; Alonso-Blanco and Méndez-Vigo , 2014; Weigel and Nordborg , 2015b; Torkamaneh et al . , 2018 ) . Indeed , a number of individual genes exhibiting evidence of locally adaptive loss-of-function have been documented in Arabidopsis ( Grant et al . , 1998; Johanson et al . , 2000; Kliebenstein , 2001; Kroymann et al . , 2003; Mouchel et al . , 2004; Aukerman , 1997; Hauser et al . , 2001; Mauricio et al . , 2003; Alonso-Blanco et al . , 2005; Werner et al . , 2005; Barboza et al . , 2013; Xiang et al . , 2014 ) . Discovering adaptive LoF alleles is particularly valuable for inspiring targeted molecular breeding because functionally similar mutations can be mined from the breeding pool or generated directly by non-transgenic native gene editing . Unfortunately , traditional genome-wide association scans based on the one-locus two-allele model perform poorly at detecting adaptive LoF alleles , which because of the large number of mutations that can create them , are likely to arise through parallel molecular evolution ( Pennings and Hermisson , 2006; Barboza et al . , 2013; Kerdaffrec et al . , 2016 ) . Species-wide whole genome sequences however , present the opportunity to advance beyond previous mapping and scanning methods that relied on linked polymorphisms by instead characterizing and contrasting functionally defined alleles . Here , we combined long-term satellite-detected drought histories , whole genome sequence scans based on allele function , and transgenic knockout experiments in Arabidopsis to test historical predictions about how drought timing shapes the evolution of flowering time and outline a broadly scalable approach for discovering loss-of-function gene variants contributing to plant climate adaptation .
To study global seasonal drought timing , satellite-detected measurements offer a valuable historical record . One such measurement , the Vegetative Health Index ( VHI ) has been used for decades to monitor drought , including in many places across the natural range of Arabidopsis ( Kogan , 1997 ) . Though primarily used as a tool to predict crop productivity , by quantifying drought induced vegetative stress this index also provides a resource for evolutionary ecologists to study seasonal patterns in drought-related episodes of natural selection . We analyzed 34 years of VHI data to characterize drought regimens at the home environments of Arabidopsis ecotypes ( Figure 1 , Supplementary file 1 ) . We found that drought frequency during the spring ( ß = 50 . 016 , p < 2×10−16 ) and summer ( ß = −28 . 035 , p = 4 . 4×10−7 ) significantly predict flowering time among Arabidopsis ecotypes ( Supplementary file 2A ) . We then generated a drought-timing index that quantifies the relative frequency of drought between spring and summer over the typical reproductive growing season and observed substantial differences in drought timing experienced by ecotypes ( Figure 1—figure supplement 1 ) . This environmental variation presented a useful cline to address classical hypotheses about the evolution of flowering time in relation to drought timing and identify LoF alleles potentially contributing to this evolution . To identify candidate LoF alleles underlying drought adaptation and flowering time evolution , we analyzed whole genome sequences in Arabidopsis . We first surveyed the genomes of 1135 ecotypes ( 1001 Genomes Consortium , 2016 ) for LoF alleles in protein coding genes predicted to encode truncated amino acid sequences ( Supplementary file 3A ) . To overcome the likely parallel evolutionary origins of LoF alleles that would have challenged previous methods , we classified alleles based functional allele state rather than individual polymorphisms for association testing . After filtering to reduce the likelihood of false positives ( see materials and methods ) , we thus tested 2088 genes for LoF allele associations with drought timing ( Figure 2A ) and flowering time ( Figure 2B ) . These analyses identified 247 genes in which LoF alleles are significantly associated with drought timing and/or flowering time after accounting for population structure and multiple testing ( Supplementary file 3B ) . In contrast , when we performed these analyses on a permuted LoF genotype matrix , we found no genes that were significantly associated with drought timing or flowering time ( Figure 1—figure supplement 1 ) . It should be noted that the 2088 genes tested for associations to flowering time and drought timing are not a complete representation of LoF alleles in Arabidopsis . In some cases , previously studied LoF alleles did not pass filtering steps ( Supplementary file 3D , E ) . This was primarily because the frequency or quality of LoF allele calls in these genes fell below our filtering requirements ( see materials and methods ) . In other cases , the Col-0 reference genome already has a documented LOF allele . Finally , we expect LoF alleles to be undetectable if they are the product of large insertions or deletions which cannot be properly identified with currently available resequencing data . Thus , while the methods used here are designed to minimize false positives ( alleles classified as LoF , but which are actually functional ) , the likely occurrence of false negatives ( undetected LoF alleles ) in available data motivates the need for more sophisticated species wide genome sequencing efforts including a greater diversity of de-novo quality genomes for comprehensive detection of functionally relevant genetic variation across the species . Associations to drought timing predicted associations of LoF alleles to flowering time directly . Together , summer drought and earlier flowering associated genes ( Figure 2C ) , and spring drought and later flowering associated genes ( Figure 2D ) overlapped seven times more often than expected by chance ( χ2=492 , p < 2 × 10−16 ) and no shared associations were observed in the opposite direction . The strengths of the associations between LoF alleles and drought timing ( P values ) was also strongly correlated with the strengths of the associations to flowering time ( r2 = 0 . 48 . Figure 2—figure supplement 1E , Figure 2C , D ) . This result is comparable to overlapping peaks in a ‘Manhattan plot’ generated from a traditional genome wide association scan ( e . g . Bosse et al . , 2017 ) . In contrast , these associations were weakly correlated when genotypes were permuted ( r2 = 0 . 01 Figure 2—figure supplement 1F ) , indicating that the result is not simply explained as an artifact of allele frequencies or by the relationship between drought timing and flowering time ( i . e . Supplementary file 1A ) . Thus , satellite-detected drought histories and a functional genome-wide scanning approach prove useful for predicting the direction and molecular targets of phenotypic evolution . Similar investigations with ecologically meaningful environmental variation could be valuable for discovering candidates underlying other important traits that are especially difficult to measure . These results further support the classical hypothesis that the relationship between phenology and drought timing is the most important feature of plant drought tolerance ( Passioura , 1996 ) , indicating the evolution of ‘drought escape’ through earlier flowering in summer drought environments , and ‘dehydration avoidance’ by later flowering genotypes in spring drought environments . Because most Arabidopsis populations appear to exhibit a winter annual life habit , germinating in the fall and overwintering as a rosette ( Ratcliffe , 1961; Thompson , 1994; Burghardt et al . , 2015 ) , late flowering genotypes in spring drought environments are expected to still encounter drought conditions . However , delayed flowering may ensure that droughts co-occur with vegetative growth rather than during the drought sensitive reproductive phase . This pattern is also consistent with hypotheses explaining the more water conservative water use and stomatal traits observed in late flowering genotypes ( McKay et al . , 2003; Lovell et al . , 2013; Kenney et al . , 2014; Kooyers , 2015 ) and those from spring drought environments ( Dittberner et al . , 2018 ) . Future experimental work will be valuable to identify other plant physiological traits affected by the LoF alleles associated with drought timing . These results provide new insight into the ecology and genetics of Arabidopsis life history evolution , but the complex ecological reality of these processes is undoubtedly beyond the scope of this study . We found that drought timing remains a significant predictor of allele associations to flowering time when controlling for allele associations with latitude and minimum temperature ( slope estimate in multiple linear regression , p < 2×10−16 , Supplementary file 2B ) . However , other unknown climatic variables or environmental interactions and non-linearities likely contribute to the flowering time adaptation as well . Flowering time is only one component of phenology and other adaptive life history transitions such a germination timing ( Donohue , 2002 ) may also be influenced by drought timing and could change how drought timing affects the evolution of flowering time , a hypothesis that warrants further investigation . Furthermore , measuring flowering time in other environments , such alternate light regimes , may yield a different set of candidate genes using similar approaches . Signatures of selection in the genes identified differ from the genome average and neutral expectations . As expected for genes harboring LoF alleles , these show parallel evolution of LoF and accelerated amino acid sequence evolution among Arabidopsis ecotypes ( Figure 2—figure supplement 2A , B , Supplementary file 2C ) . We also found evidence of positive selection for LoF alleles in genes associated with drought timing and/or flowering time . While these genes have similar global frequencies of LoF alleles compared to genes not showing associations with drought timing and/or flowering time ( Figure 2—figure supplement 2C ) , they tend to have significantly fewer unique LoF alleles ( Figure 2—figure supplement 2D ) and greater frequencies of each independent LoF allele ( Figure 2E ) . This pattern is consistent with theoretical predictions and results from simulations of adaptation by parallel molecular evolution involving recurrent mutation combined with more rapid local fixation of alleles experiencing positive selection ( Pennings and Hermisson , 2006 ) . In cases where adaptation proceeds through the fixation of a single adaptive allele , traditional genome scanning approaches may be sufficient to detect causal loci . However , when genetic variation consists of multiple independent alleles , as is often the case for the genes examined here ( Figure 2—figure supplement 2D ) , classifying alleles functionally before testing for associations is likely necessary . The extent of LoF responsible for adaptive phenotypic evolution is much greater than once assumed ( Smith , 1970; Albalat and Cañestro , 2016 ) . LoF alleles identified were overwhelmingly associated with spring drought or later flowering rather than summer drought or earlier flowering ( χ2 = 132 , p < 2 × 10−16 , Figure 2 ) . Because the reference genome and gene models are from an early flowering Arabidopsis line , Col-0 , this is consistent with the hypothesis that LoF alleles are particularly important in the evolution of phenotypic divergence ( Rausher , 2008 ) . This result also highlights the need to develop functional genomics resources informed by multiple de-novo quality reference genomes . We found that flowering time is strongly predicted by the accumulation of LoF alleles across the 214 candidate genes associated to spring drought and/or later flowering time ( Figure 3A–E ) , estimating a 1 day increase for every three additional LoF alleles across these candidate genes ( Figure 3F ) . This relationship is best represented as a simple linear regression; the addition of a non-linear quadratic predictor variable did not significantly improve the fit of the model ( F = 0 . 7005 , p = 0 . 4028 ) . Importantly , we did not find a broader overabundance of LoF alleles in later flowering ecotypes or those from spring drought environments that would explain this relationship ( e . g . Figure 2—figure supplement 3 ) . Rather , these findings support a model of climate-associated evolution in complex traits that includes a substantial contribution from widespread genetic LoF and give promise to targeted LoF for directed phenotypic engineering . Experimental knockout lines confirmed the later flowering times predicted from natural allele associations . To test phenotypic effects , we screened a panel of confirmed T-DNA insertion mutants representing a sample of candidate LoF alleles associated with spring drought and/or later flowering . As predicted by variation among Arabidopsis ecotypes ( Figure 2D ) , the vast majority of knockout lines in these candidate genes ( 57 of 59 , χ2 = 51 , p = 8 . 045e-13 ) flowered later on average than the wild type genotype ( Figure 3G , Supplementary file SF ) . LoF alleles identified through these analyses and experiments include those previously linked to flowering time ( Cui et al . , 2007 ) and drought responses ( Aghdasi et al . , 2012; Qin et al . , 2017 ) . Implementing a functional genome-wide association scan , we find that allele associations with ecologically meaningful environmental variation ( drought timing ) accurately predict associations with adaptive phenotypes directly ( flowering time ) . Together with validation in transgenic lines , these findings outline a scalable model for gaining deeper insights into the functional genomics of climate adaptation in nature . Combining large scale knockout experiments with functional genome wide association scans may be a valuable approach for future research to quantify the power to predict LoF allele effects . These results also further challenge historical assumptions about molecular adaptation that have implications for influencing evolutionary theory and public attitudes toward emerging molecular breeding approaches . Groundbreaking yield increases during the green revolution of the 1960 s were largely attributable to semi-dwarf phenotypes caused by LoF alleles in both rice and barley ( Spielmeyer et al . , 2002; Jia et al . , 2009 ) . Later it was found that natural LoF alleles of the same gene in wild Arabidopsis produce similar phenotypes ( Barboza et al . , 2013 ) , suggesting the potential to mine ecological species for information directly useful for crop improvement . Visions of a second green revolution powered and informed by such natural variation call for discoveries in evolutionary functional genomics at scales that have now become possible . The genes identified here could inspire future molecular breeding of climate resilient crops and this work more broadly highlights the value of integrating diverse disciplines to scale up the discovery of the climatic drivers of adaptation and functionally significant genetic variation at molecular resolutions .
To study patterns in historical drought , the remotely sensed Vegetative Health Index ( VHI ) was used , a satellite-detected drought measurement tool whose advantage is that it includes information about vegetative impacts of drought ( Passioura , 1996; AghaKouchak et al . , 2015 ) . This index is based on multiple data sources from NOAA satellites , combining deviations from historic climatic ( Temperature Condition Index derived from AVHRR-based observations in thermal bands ) and vegetative conditions ( Vegetative Condition Index derived from NDVI ) to detect periods of ecological drought conditions and distinguish between other sources of vegetative stress such as cold ( Kogan , 1997; Kogan et al . , 2005; Rojas et al . , 2011 ) . VHI was collected weekly since 1981 at 16 km2 resolution on a scale from 0 to 100 , where values below 40 reflect drought conditions ( Kogan , 1997 ) ( Figure 1A ) . The frequencies of observing drought conditions during photoperiodic spring ( quarter surrounding spring equinox ) , summer ( quarter surrounding summer solstice ) , fall ( quarter surrounding fall equinox ) , and winter ( quarter surrounding winter solstice ) were calculated globally from 1981 to 2015 ( Figure 1B ) in R ( R Core Development Team , 2017 ) using the raster package ( Hijmans , 2016 ) . After removing ecotypes with missing location data or locations falling within pixels classified as water , seasonal drought frequencies and drought timing were calculated at the location of origin for 1 , 097 Arabidopsis ecotypes that were included as part of the 1001 Genomes Project ( 1001 Genomes Consortium , 2016 ) ( Figure 1C , Supplementary file 1 ) . Up to date global map files of seasonal drought frequency and the drought-timing index used here are available on Dryad and greymonroe . github . io/data alongside a brief tutorial showing how to extract data for points of interest in R . We tested whether seasonal drought frequencies significantly predicted with flowering time ( flowering time described in subsequent section regarding LoF associations ) by multiple linear regression ( Supplementary file 2A ) To characterize the seasonal timing of droughts during an important period of Arabidopsis’ life history , a univariate drought-timing index was generated that quantifies whether the historical frequency of drought increases or decreases over the course of the typical Arabidopsis reproductive growing season ( Ratcliffe , 1961; Thompson , 1994; Burghardt et al . , 2015 ) . Specifically , this index is equal to the natural log transformed ratio between spring and summer drought frequency . More negative values reflect environments where drought frequency increases from spring to summer and are referred to here as ‘summer drought environments , ’ ( e . g . Figure 1B left ) . Conversely , more positive values reflect environments where drought frequency decreases from spring to summer and are referred to here as ‘spring drought environments , ’ ( e . g . Figure 1B right ) . To identify functionally definitive gene variants ( Hoekstra and Coyne , 2007; Weigel and Nordborg , 2015a; Byers et al . , 2017 ) , LoF alleles ( Albalat and Cañestro , 2016 ) were identified from whole genome sequence data of 1 , 135 Arabidopsis accessions ( Olson , 1999; Cutter and Jovelin , 2015; 1001 Genomes Consortium , 2016 ) using R scripts . First , genes were filtered to those containing at least 5% frequency of predicted frameshift or premature stop mutations and less than 5% missing allele calls from results generated by the 1 , 001 Genomes Consortium ( 1001 Genomes Consortium , 2016 ) using ‘SnpEff’ ( Cingolani et al . , 2012 ) . To reduce instances where exon skipping might ameliorate LoF mutations ( Gan et al . , 2011 ) , genes were filtered to those with a single predicted gene model ( Lamesch et al . , 2012 ) . Additionally , to preclude false LoF calls for cases where compensatory mutations restore gene function or in which an insignificant portion of the final protein product is affected by putative LoF mutations ( MacArthur et al . , 2012 ) , coding regions were translated into predicted amino acid sequences from which lengths from start to stop codon were calculated in R . LoF alleles were defined as those producing protein products with at least 10% lost because of late start codons and/or prematurely truncated translation . Allelic heterogeneity expected to mask these genes from traditional GWAS ( Remington , 2015; Monroe et al . , 2016; Flood and Hancock , 2017 ) was corrected for by classifying all alleles as either functional ( 0 ) or non-functional ( 1 ) . A final frequency filter was re-applied ( 5% global LoF allele frequency ) , resulting in 2088 genes for downstream association analyses ( Supplementary file 3B ) . Finally , to compare the results of this pipeline to genes known to harbor natural LoF alleles ( Mouchel et al . , 2004; Shindo et al . , 2008; Gujas et al . , 2012; Kliebenstein , 2001; Kroymann et al . , 2003; Grant et al . , 1998; Tian et al . , 2003; Mauricio et al . , 2003; Werner et al . , 2005; Aukerman , 1997; Flowers et al . , 2009; Xiang et al . , 2014; Xiang et al . , 2016; Amiguet-Vercher et al . , 2015; Johanson et al . , 2000; Le Corre et al . , 2002; McKay et al . , 2003; Stinchcombe et al . , 2004; Shindo et al . , 2005; Flowers et al . , 2009; Méndez-Vigo et al . , 2011; Lovell et al . , 2013; Hauser et al . , 2001; Bloomer et al . , 2012; Alonso-Blanco et al . , 2005; Zhen and Ungerer , 2008; Kang et al . , 2013; Monroe et al . , 2016; Zhu et al . , 2015; Barboza et al . , 2013 ) , we manually performed this functional allele calling approach on a set of 16 genes ( Supplementary file D , E ) To identify candidate LoF alleles responsible for climate adaptation and phenotypic evolution , the relationships between functional allele state and drought timing and between functional allele state and flowering time were evaluated for each of the 2088 genes that passed preceding filtering steps . Specifically , the association between functional allele state among Arabidopsis ecotypes and historical drought timing at their locations of origin was tested by logistic regression in a generalized linear model in R ( R Core Development Team , 2017 ) . This association study differs from traditional GWAS in several respects . First , because the alleles studied here are functionally defined , they are expected to be more likely to have a phenotypic impact than random SNPs . Second , the scope of our analyses were restricted to a subset of the genome - 2088 genes with high confidence LoF allele calls that passed previous filtering steps , rather than tens of thousands to millions of SNPs . Finally , in contrast to traditional GWAS , which is designed to identify associated chromosomal regions rather than functionally definitive genetic variations , our approach is motivated by the ability to identify alleles at molecular resolutions whose functional relevance can be tested empirically . Thus , the balance of opportunity costs related to trade-offs between false positive and false negative associations that generally challenge GWAS are shifted to reduce false negatives rather than minimizing false positives . For these reasons , we implemented analyses based on ( Price et al . , 2006 ) to balance false positives and false negatives . Population structure was accounted for by performing a principal component analysis on the kinship matrix among all ecotypes and including in each model the first three resulting principal components , which explain >75% of variance in relatedness between ecotypes ( Price et al . , 2006 ) . The P-values ( Pdrought timing ) of the slope estimates ( βdrought timing ) for drought timing in these models were adjusted to account for multiple tests by a Bonferroni correction to identify those significantly associated ( Supplementary file 3C ) . Summer drought genes were identified as those in which LoF alleles are found in ecotypes that experience a significantly ( βdrought timing <0 and Pdrought timing <0 . 05 ) more negative drought-timing index ( summer drought environments where drought frequency increases over the course of the reproductive growing season , Figure 1B left and Figure 2A top ) . Conversely , spring drought genes were identified as those in which LoF alleles are found in ecotypes that experience a significantly ( βdrought timing >0 and Pdrought timing <0 . 05 ) more positive drought-timing index ( spring drought environments where drought frequency decreases over the course of the reproductive growing season , Figure 1B right and Figure 2A bottom ) . The above analytical approach was repeated to test whether functional allele state is associated with the reported common garden flowering times of Arabidopsis ecotypes ( Alonso-Blanco and Méndez-Vigo , 2014 ) ( Supplementary file 1 ) . See Alonso-Blanco et al . ( Alonso-Blanco and Méndez-Vigo , 2014 ) for details , but in brief , flowering time was measured in growth chambers at 10°C ( considerably less missing data than experiment at 16°C ) under 16 hour days . Earlier flowering genes were identified as those in which LoF alleles are found in ecotypes that flower significantly ( βflowering time <0 and Pflowering time <0 . 05 ) earlier than ecotypes with a functional allele ( Figure 2B top ) . Later flowering genes were identified as those in which LoF alleles are found in ecotypes that flower significantly ( βflowering time >0 and Pflowering time <0 . 05 ) later than ecotypes with a functional allele ( Figure 2B bottom ) . The preceding analyses revealed considerable overlap between genes associated with both drought timing and flowering time . To assess whether this result was an artifact of the binary LoF allele calls , we randomly permuted the genotype matrix and repeated the analyses described above , testing for significant associations between allele states and drought timing and/or flowering time . Quantile-quantile plots of P values were visualized using qqPlot in the GWASTools package in R ( Gogarten et al . , 2012 ) ( Figure 2—figure supplement 1A–D ) To address the longstanding hypothesis that flowering time reflects adaptation to drought timing ( Fox , 1990; Passioura , 1996; Kooyers , 2015 ) , and to test the corresponding prediction that alleles associated with drought timing are also associated with flowering time , the groups of genes identified with significant associations to drought timing or flowering time were compared ( Figure 2C and D ) . Deviation from the null hypothesis of independent associations to drought timing and flowering time was evaluated by a chi-squared test ( Expected number of co-associated genes = 12 , Observed = 83 , χ2 = 492 , p = 2×10−16 ) . The magnitude of P-values have historically served as the basis of selecting candidate loci for further examination toward their contribution to environmental adaptation or phenotypic evolution in quantitative trait locus mapping and genome wide association scans [e . g . ( Bosse et al . , 2017 ) . To test whether associations to environment ( drought timing ) can be used to identify loci associated with phenotypes ( flowering time ) directly , the correlation between log transformed P-values describing allele associations with drought timing ( Pdrought timing ) and with flowering time ( Pflowering time ) was calculated ( Figure 2—figure supplement 1E , r2 = 0 . 48 , ) and visualized separately for genes associated to summer drought/earlier flowering ( Figure 2C ) and to spring drought/later flowering ( Figure 2D ) . To control for the possibility that allele frequencies or the relationship between drought timing and flowering time explained these observations , we also tested whether allele associations were correlated when generated from association analyses using a matrix of randomly permuted genotypes with the same allele frequencies ( Figure 2—figure supplement 1F , r2 = 0 . 01 ) . Finally , to control for the possibility that correlated LoF allele associations were explained by confounding environmental variables we tested whether the LoF allele associations to drought timing remained predictive while accounting for LoF allele associations with latitude and minimum temperature of the coldest month ( Hijmans et al . , 2005 ) using a multiple linear regression in R ( Supplementary file 3B ) . To do so , we repeated the association analyses described in the previous section but instead tested for LoF allele associations with latitude and minimum temperatures . We then included these P values ( Supplementary file 2B ) in a multiple linear regression where the strength of the association to flowering time was predicted by the associations to drought timing , latitude , and minimum temperature simultaneously . To assess whether histories of selection for genes identified differ from the genome wide expectation , measures of amino acid sequence evolution were evaluated for 122 genes in which loss-of-function is associated with drought timing or flowering time and for which there are orthologs identified between A . lyrata and A . thaliana ( Goodstein et al . , 2012 ) . For each gene , sequences were aligned using MAFFT ( Katoh and Standley , 2013 ) , codons with gaps removed , and the number of non-synonymous and synonymous polymorphisms among A . thaliana accessions ( PN and PS ) as well as synonymous and non-synonymous divergence ( DN and DS ) from A . lyrata were measured using mkTest . rb ( https://github . com/kern-lab/ ) . The ratios PN/PS and DN/DS were then calculated to measure the proportion of variants predicted to affect amino acid sequences that are segregating among ecotypes and diverged from A . lyrata , respectively . These calculations were also performed for genes not associated to drought timing or flowering time ( n = 912 ) and the remaining genes across the A . thaliana genome ( n = 20373 ) with orthologs between A . lyrata and A . thaliana . To test whether genes identified show evidence of accelerated protein sequence evolution , comparisons were made to genes associated with drought timing or flowering time for both PN/PS ( Figure 2—figure supplement 2A ) and DN/DS ( ( Figure 2—figure supplement 2A , B ) by two-sided students t-tests ( α = 0 . 05 ) in R ( R Core Development Team , 2017 ) . Because theory predicts adaptation by loss-of-function to proceed through multiple independent alleles , but to exhibit a fewer number of different alleles than in neutral loci at similar LoF allele frequencies ( Pennings and Hermisson , 2006; Ralph and Coop , 2010; Ralph and Coop , 2015 ) , the number of unique LoF alleles was estimated by protein length in the genes that passed preceding filtering steps . To address the hypothesis that genes in which LoF alleles are associated to drought history or flowering time are likely to reflect positive selection compared to genes in which LoF are random with respect to drought history or flowering time , the total number of unique LoF alleles between these groups was compared using a two-sided students t-test ( log10 transformed , p = 5 . 8×10−7 , ( Figure 2—figure supplement 2D ) . To control for the possibility that this result in an artifact of reduced frequency of LoF alleles in genes identified , the global frequency of LoF was also compared between these groups ( log10 transformed , two-sided students t-test , p = 0 . 11 , ( Figure 2—figure supplement 2C ) . Finally , to further test the prediction that LoF alleles in genes identified have increased in frequency because of more positive selection , the frequency per specific LoF allele was compared between groups ( log10 transformed , two-sided students t-test , p = 3 . 4×10−7 , Figure 2E ) . The significance of the tendency for LoF associations to spring drought/later flowering time ( Figure 2D ) was tested by chi-squared tests ( spring drought vs . summer drought , p < 2×10−16; later vs . earlier flowering , p < 2×10−16 , spring drought/later flowering vs . summer drought/earlier flowering , p < 2×10−16 ) . The chromosomal locations of candidate genes ( those associated to spring drought/later flowering time ) were mapped onto the Arabidopsis genome ( Lamesch et al . , 2012 ) ( Figure 3A ) . To address the hypothesis that widespread LoF contributes to later flowering time phenotypes , the total number of LoF in candidate genes for each ecotype was calculated and the correlation between this value and flowering time evaluated ( Figure 3F , r2 = 0 . 39 , p < 2×10−16 ) . We also tested whether a model which included a non-linear predictor ( squared value of the total number of LoF in candidate genes ) was a better fit than the simple linear model by an analysis of variance ( F = 0 . 7005 , p = 0 . 4028 ) . The preceding analyses provided compelling evidence of LoF in candidate genes as important in the evolution of later flowering time phenotypes . To test the prediction that non-functionalization of these genes causes increased flowering time , phenotypes were measured in transgenic lines in a subsample of candidate genes showing a significant association between loss-of-function and spring drought environments and/or later flowering time . Motivated by the general need to develop a high throughput approach of studying naturally adaptive LoF , knockout lines from the Arabidopsis Biological Resource Center were chosen from a collection created by the SALK Institute in which a T-DNA insertion in an exon of candidate genes has already been identified and confirmed to be homozygous ( O'Malley and Ecker , 2010; Rutter et al . , 2017 ) . These T-DNA knockout lines were generated by the SALK institute ( Supplementary file 3F ) and exist in a common genetic background ( Columbia ) ( Alonso et al . , 2003 ) . Seeds were planted in 2’ pots containing wet potting soil and stratified for 5 days at 4°C . Seedlings were thinned to a single plant per pot one week after stratification . Plants were grown ( 59 T-DNA knockout lines , 10 reps of each line and 30 reps Columbia ) in a stratified ( by shelf ) , randomized design in growth chambers ( Conviron ATC60 , Controlled Environments , Winnipeg , MB ) under 16 hr of light at 20°C . Flowering time was measured as days after planting to the emergence of the first open flower , based on the definition of flowering time used by the 1 , 001 Genomes Consortium ( 1001 Genomes Consortium , 2016 ) . We calculated the least squares mean ( lsmean from ‘lsmeans’ package in R ) flowering time for each line from a mixed model where shelf and tray were included as random effects ( Supplementary file 3F ) . We tested the prediction that knockout lines would flower later ( have higher lsmean flowering time estimates ) than the wild type Columbia genotype by a chi-squared test ( p = 8 . 1×10−13 ) . | Water shortages caused by droughts lead to crop losses that affect billions of people around the world each year . By discovering how wild plants adapt to drought , it may be possible to identify traits and genes that help to improve the growth of crop plants when water is scarce . It has been suggested that plants have adapted to droughts by flowering at times of the year when droughts are less likely to occur . For example , if droughts are more likely to happen in spring , the plants may delay flowering until the summer . Arabidopsis thaliana is a small plant that is found across Eurasia , Africa and North America , including in areas that are prone to drought at different times of the year . Individual plants of the same species may carry different versions of the same gene ( known as alleles ) . Some of these alleles may not work properly and are referred to as loss-of-function alleles . Monroe et al . investigated whether A . thaliana plants carry any loss-of-function alleles that are associated with droughts happening in the spring or summer , and whether they are linked to when those plants will flower . Monroe et al . analyzed satellite images collected over the last 30 years to measure when droughts have occurred . Next , they searched genome sequences of Arabidopsis thaliana for alleles that might help the plants to adapt to droughts in the spring or summer . Combining the two approaches revealed that loss-of-function alleles associated with spring droughts were strongly predicted to be associated with the plants flowering later in the year . Similarly , loss-of-function alleles associated with summer droughts were predicted to be associated with the plants flowering earlier in the year . These findings support the idea that plants can adapt to drought by changing when they produce flowers , and suggest that loss-of-function alleles play a major role in this process . New techniques for editing genes mean it is easier than ever to generate new loss-of-function alleles in specific genes . Therefore , the results presented by Monroe et al . may help researchers to develop new varieties of crop plants that are better adapted to droughts . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"genetics",
"and",
"genomics"
] | 2018 | Drought adaptation in Arabidopsis thaliana by extensive genetic loss-of-function |
Neural networks are typically defined by their synaptic connectivity , yet synaptic wiring diagrams often provide limited insight into network function . This is due partly to the importance of non-synaptic communication by neuromodulators , which can dynamically reconfigure circuit activity to alter its output . Here , we systematically map the patterns of neuromodulatory connectivity in a network that governs a developmentally critical behavioral sequence in Drosophila . This sequence , which mediates pupal ecdysis , is governed by the serial release of several key factors , which act both somatically as hormones and within the brain as neuromodulators . By identifying and characterizing the functions of the neuronal targets of these factors , we find that they define hierarchically organized layers of the network controlling the pupal ecdysis sequence: a modular input layer , an intermediate central pattern generating layer , and a motor output layer . Mapping neuromodulatory connections in this system thus defines the functional architecture of the network .
Neuromodulators constitute a major channel of communication in the nervous system and act at virtually all levels of sensorimotor processing to tune the intrinsic and synaptic properties of neurons ( Marder , 2012; Nadim and Bucher , 2014; van den Pol , 2012 ) . How these properties are tuned can profoundly influence the function of not only the individual neurons , but also the circuits in which they participate . Dynamic and coordinated regulation of multiple brain circuits is required for behavior , and an attractive idea with deep historical roots is that neuromodulators , which typically act over timescales and distances that are long compared with synaptic neurotransmission , may serve to coordinate activity in broadly distributed circuits to bias the performance of behaviors appropriate for a given set of circumstances ( Bargmann , 2012; Bicker and Menzel , 1989; Harris-Warrick and Marder , 1991 ) . Although the generality and scope of this viewpoint remain unclear , it implies that identifying sites of neuromodulator action may represent a productive strategy for mapping critical circuits involved in generating a behavior of interest . The strategy of mapping sites of neuromodulator action emerged from early observations on the activating effects of various brain-derived hormones and biogenic amines . When introduced into the nervous system , these compounds were found to induce complex motor programs ( Harris-Warrick , 1988 ) , such as emergence of moths from their cocoons ( Truman and Riddiford , 1970 ) , egg-laying in sea hares ( Kupfermann , 1967 ) , and postural changes in lobsters ( Livingstone et al . , 1980 ) . Attempts by Sombati and Hoyle ( 1984 ) to map the sites of action of the insect locomotor activator , octopamine ( OA ) , in the metathoracic ganglion of locusts led to Hoyle’s ‘orchestration hypothesis , ’ according to which flight and other motor programs are encoded in the activities of subpopulations of OA neurons ( Hoyle , 1985 ) . Although the explicit statement of this influential hypothesis remains to be proven , a generalized role for OA in coordinating insect flight was supported by subsequent discoveries that OA neurons modulate muscle metabolism and visual motion processing during flight ( Libersat and Pflueger , 2004; Suver et al . , 2012 ) . Mechanistic insight into how neuromodulators regulate and coordinate circuit function came from the intensive functional and anatomical investigation of small circuits ( Bargmann and Marder , 2013; Selverston , 2010 ) . This work included fine-mapping sites of neuromodulator action by painstaking physiological characterization of single neuron responses in , among other systems , the crustacean stomatogastric ganglion ( Flamm and Harris-Warrick , 1986; Hooper and Marder , 1987; Swensen and Marder , 2001 ) . The STG houses two principal central pattern generators ( CPGs ) that drive digestive rhythms and their activity patterns , the pyloric and gastric mill rhythms , are both dependent upon , and can be variously reconfigured by , the actions of neuromodulators ( Marder and Bucher , 2007 ) . Two of these , the neuropeptides proctolin and C . borealis Tachykinin-Related Peptide Ia ( CabTRP ) , offer a simple example of how neuromodulators acting at different sites can coordinate changes in two ( overlapping ) circuits ( Nusbaum , 2002; Nusbaum et al . , 2001 ) . Both proctolin and CabTRP are released from a neuromodulatory projection neuron ( MCN1 ) into the STG ( Blitz et al . , 1999 ) , and although both peptides activate the same inward current ( Swensen and Marder , 2000 ) , their effects on the pyloric and gastric mill rhythms differs because they target distinct cells within the respective CPGs ( Blitz et al . , 1999; Wood et al . , 2000 ) . Proctolin principally excites the pyloric circuit and can activate it from quiescence , while CabTRP is required for the gastric mill rhythm and acts on a key neuron in its CPG in addition to neurons of the pyloric circuit . Activation of MCN1 by mechanosensory inputs from the stomach , induces a gastric mill rhythm via the action of CabTRP and alters the pyloric rhythm in response to the actions of both peptides ( Beenhakker and Nusbaum , 2004; Blitz et al . , 2004 ) . Sensory information , conveyed by two neuromodulators , thus produces coordinated changes in two functionally related circuits . The significance and adaptive value of many neuromodulatory effects characterized in the STG remains unknown , and , in general , the sheer abundance of circuit neuromodulation revealed by studies of this and other small systems challenges the simple idea of ‘chemical coding’ of behavior by neuromodulators . This complexity is also underscored by analyses of neuromodulator receptor-distributions , first undertaken by ligand autoradiography in the 1980s . On the one hand , these studies supplied strong evidence that neuromodulators could act at many sites and over long distances , but they also highlighted the difficulty of establishing which sites were relevant for performance of specific behaviors without knowledge of where and under what circumstances each neuromodulator was released ( Herkenham , 1987 ) . For neuromodulators already implicated in specific behaviors , however , the receptor distributions sometimes spectacularly confirmed the idea that neuromodulators target ethologically significant circuits ( Insel and Young , 2000 ) . For example , cross-species differences in striatal expression of vasopressin receptors in two closely related vole species were shown to correlate with , and in fact cause , monogamous and polygamous predispositions in mating ( Hammock and Young , 2005; Young et al . , 1997 ) . Based on these and other examples , variations in neuromodulator receptor expression during speciation have been proposed to be a major driver of behavioral evolution ( Katz and Lillvis , 2014 ) . The recent development of genetic techniques for targeting and functionally manipulating neurons in genetic model animals has facilitated the functional characterization of neuronal populations on which neuromodulators act ( Spangler and Bruchas , 2017 ) . This work again provides examples of neuromodulators that coordinate activity in broadly distributed circuits . The evidence is particularly compelling for conserved neuromodulators , such oxytocin ( Mitre et al . , 2016; Stoop , 2012; 2014 ) , which in mice regulates distinct circuits that promote social behaviors , including conspecific recognition ( Ferguson et al . , 2001 ) , pup retrieval ( Marlin et al . , 2015 ) , and social learning ( Choe et al . , 2015 ) . Oxytocin’s homologs likewise act on circuits that facilitate behaviors related to affiliation and reproduction in species as diverse as worms ( Garrison et al . , 2012 ) , leeches ( Wagenaar et al . , 2010 ) , fish ( Reddon et al . , 2015 ) , and birds ( Kelly and Goodson , 2014a ) . Similarly , members of the Neuropeptide Y ( NPY ) signaling pathway have been shown to act on circuits that promote feeding in multiple species ( Taghert and Nitabach , 2012 ) . Like many neuromodulatory signaling systems , however , oxytocin , NPY , and their receptors are widely distributed in nervous systems and are likely to function in multiple contexts ( Chronwall and Zukowska , 2004; Kelly and Goodson , 2014b ) . This added complexity in neuromodulator action , together with the observation that neuromodulators rarely , if ever , act in isolation , has made it difficult to simply generalize the conclusion that neuromodulators organize activity in broadly distributed circuits to produce adaptive changes in behavioral expression . Given the evidence that favors such an organizational role , however , it remains a potentially useful strategy to map sites of neuromodulator action to identify key circuits involved in the generation of behaviors of interest . Particularly suitable behaviors for this approach are those for which both the neuromodulators important for behavioral performance and the circumstances under which they are released are known . The hormonally governed behaviors that underlie insect molting meet these criteria ( Truman , 2005; Zitnan and Adams , 2012 ) . Molting is accomplished by the serial execution of several motor programs in what is called an ecdysis sequence ( White and Ewer , 2014 ) . Ecdysis sequences are initiated by the peripheral release of Ecdysis Triggering Hormone ( ETH ) , which facilitates the secretion of multiple other peptide hormones , including Eclosion Hormone ( EH ) , Crustacean Cardioactive Peptide ( CCAP ) , and Bursicon . These factors function as neuromodulators within the nervous system to orchestrate the progression of the ecdysis sequence , and their action has been extensively studied in Drosophila at the pupal stage ( Diao et al . , 2016; Kim et al . , 2015; Kim et al . , 2006; Mena et al . , 2016 ) . The pupal ecdysis sequence of the fly consists of three distinct behavioral phases , the second of which is governed by CCAP and Bursicon released from neurons that are targets of ETH ( Diao et al . , 2015; Kim et al . , 2015; Kim et al . , 2006; Lahr et al . , 2012 ) . Here , we use the recently developed Trojan exon method ( Diao et al . , 2015 ) , which permits the high-fidelity targeting of neurons that express specific neuromodulator receptors , to investigate the downstream effectors of ETH , CCAP , and Bursicon and show that the sites of action of these factors expose the structure and operational logic of the fly pupal ecdysis circuit .
Each of the three phases of the Drosophila pupal ecdysis sequence is characterized by a dominant abdominal motor rhythm ( Video 1 , Kim et al . , 2006 ) . The entire sequence can be induced by injection of ETH1 ( one of two ETH peptides encoded by the ETH gene in Drosophila ) , and all three phases have been proposed to be under the control of distinct peptidergic neurons ( Kim et al . , 2006 ) . This is clearly true of the second behavioral phase ( Phase II , sometimes referred to as ‘ecdysis’ ) , which is specifically dependent on CCAP and the heterodimeric hormone Bursicon ( Kim et al . , 2006; Lahr et al . , 2012 ) . We have previously shown that a subset of CCAP-expressing neurons ( ETHRA/CCAP neurons ) also expresses the A-isoform of the ETH receptor ( ETHRA ) and is required for head eversion , a signature event of pupal ecdysis that occurs during Phase II ( Diao et al . , 2016 ) . The ETHRA/CCAP neurons also include the subset expressing Bursicon ( Figure 1A ) , and we began our investigation by confirming that chronic suppression of these neurons using the inward rectifying channel Kir2 . 1 blocks execution of Phase II . Consistent with previous observations , this manipulation also inhibits execution of the third motor program ( Phase III , or ‘post-ecdysis’ ) and extends the duration of the first ( Phase I or ‘pre-ecdysis , '; Figure 1B ) , suggesting that Phase I may be terminated by the onset of Phase II ( Kim et al . , 2015; Mena et al . , 2016 ) . To determine whether this is the case , we used the temperature-sensitive dTrpA1 channel to activate the ETHRA/CCAP neurons immediately after the onset of abdominal lifting , which initiates Phase I and is accompanied by rolling waves of anteriorly directed contractions of the lateral body wall that alternate from one side of the animal to the other . We observed rapid termination of Phase I and initiation of Phase II , which was then followed by execution of Phase III ( Figure 1C ) . This was the case whether dTrpA1-mediated activation was sustained throughout the observation period or was transient ( i . e . 1 min; Video 2 ) . The rapid termination of Phase I upon activation of ETHRA/CCAP neurons , together with the extended duration of this phase when these same neurons are suppressed , demonstrates that ETHRA/CCAP neuron activity is necessary and sufficient for normal Phase I termination in addition to Phase II initiation . Furthermore , because Phase III behaviors follow , or fail with , those of Phase II when ETHRA/CCAP neurons are activated or suppressed , respectively , we conclude that these neurons are also important determinants of Phase III . To establish how the ETHRA/CCAP neurons regulate the three phases of the pupal ecdysis sequence , we sought to identify and characterize their downstream signaling partners . Genetic data demonstrate that CCAP and Bursicon jointly mediate signaling by the ETHRA/CCAP neurons: Most animals bearing null mutations in both the CCAP gene and the gene encoding Pburs—one of the two subunits of Bursicon—execute Phase I , but not Phase II behaviors , and 70–90% do not evert their heads during the vigorous side-to-side swinging that characterizes Phase II ( Lahr et al . , 2012 ) . The severe head eversion deficits seen in CCAP/Pburs double mutants must result from synergistic actions of the two hormones because animals with null mutations only in CCAP display relatively normal pupal ecdysis behavior , and over half of animals with null mutations in Pburs are able to complete pupal ecdysis , although most show delays in head eversion . To characterize the downstream neurons that mediate the effects of CCAP and Bursicon , we used the Trojan exon method to generate transgenic fly lines that express Gal4 or Split Gal4 components specifically in cells that express either the Bursicon receptor ( encoded by the rickets ( rk ) gene ) or the CCAP receptor ( CCAP-R ) . In addition , we used a previously generated and strongly expressing Rk-Gal4 driver line ( Rkpan-Gal4 Diao and White , 2012 ) . The expression patterns of these lines reveal that both Rk and CCAP-R are broadly expressed in the CNS at the time of pupal ecdysis , but that few neurons express both receptors , and those only very weakly ( Figure 2A ) . Rk and CCAP-R are thus expressed in almost completely distinct populations of neurons indicating that the synergistic effects of the two hormones released from the ETHRA/CCAP neurons is not due to both signals converging on a common set of targets . While the effects of individually eliminating Bursicon and CCAP function by mutation differ significantly , blocking activity in the two receptor-expressing populations of neurons targeted by these factors is similar . Suppression of either population starting at the third larval instar results in severe pupal ecdysis deficits and lethality in 100% of animals ( Figure 2B ) . Behavioral analysis , however , reveals small differences at the level of motor function . Animals in which the Rk-expressing neurons are suppressed are most impaired , lacking all movement and therefore all phases of the pupal ecdysis sequence ( Figure 2C , top; Video 3 ) . Interestingly , pupal development in these animals is otherwise normal: a bubble appears in the abdomen ( Figure 2B , left , arrow ) and the fat body degrades on schedule , but the overt gut movements that herald the onset of Phase I behaviors ( Robertson , 1936 ) fail to appear . The presence of a sustained heartbeat throughout indicates that the animals remain viable for many hours , despite not initiating pupal ecdysis . Animals in which the CCAP-R neurons are suppressed also lack normal ecdysis behavior , executing only a mixture of weak and irregular contractions during a period of abdominal lifting with some resemblance to Phase I ( Figure 2C , bottom; Video 3 ) . We conclude that the two distinct groups of neurons targeted by Bursicon and CCAP are both essential for pupal ecdysis . We focused first on characterizing the function of the Rk-expressing neurons . The results of neuronal suppression demonstrate that some or all the Rk-expressing neurons are essential for initiating and/or generating all phases of the pupal ecdysis sequence . To determine whether the Rk-expressing neurons include motor neurons in the ecdysis circuit essential to its output , we performed intersectional labeling using a Rk hemidriver together with a vesicular glutamate ( VGlut ) hemidriver , which expresses in glutamatergic neurons , including all Drosophila motor neurons ( Diao et al . , 2015 ) . We observed expression in only a handful of neurons of the thoracic ganglia , none of which extend axons to muscles , indicating that Rk is not expressed in motor neurons ( Figure 3A ) . We likewise find that Rk is not highly expressed in neurons that receive the hormonal input that initiates the ecdysis sequence as determined using a Rk hemidriver in conjunction with ETHRA and ETHRB hemidrivers to identify neurons that co-express Rk and either the A- or B-isoform of the ETHR ( Figure 3A’ , A” , Diao et al . , 2016 ) . For both isoforms , only a small number of neurons was identified that co-expressed Rk , and the suppression of these neurons with 2X UAS-Kir2 . 1 failed to block pupal ecdysis ( data not shown ) . The Rk-expressing neurons essential for pupal ecdysis thus do not belong to either the input or the output layers of the ecdysis network and must therefore occupy an intermediate position in the circuit hierarchy . To gain insight into the function of the Rk-expressing neurons , we investigated their response to upstream input from the Bursicon-expressing neurons , using the physiogenetic ATP/P2X2 system ( Yao et al . , 2012 ) . To selectively activate the Bursicon-expressing neurons in excised pupal nervous systems by exposure to ATP , we expressed the purinergic P2X2 channel under the control of a Burs-LexA::GADfl driver and monitored the response of Rk-expressing neurons using the calcium sensor UAS-GCaMP6s driven by Rk-Gal4 . Ca++ activity was measured in the large population of neurons located in the ventral nerve cord ( VNC-Rk neurons; Figure 2A , left , box ) by laser scanning confocal microscopy , sampling at approximately 1 Hz . We found that ATP-induced phasic Ca++ activity in the VNC-Rk neurons , characterized by distinct , alternating , left-right oscillations across the ventral midline ( Figure 3B , C ) , which were absent in control preparations lacking P2X2 expression . Quantifying the alternating oscillations , we found that the midline oscillations were , on average , sustained for 9 min . and consisted of 15 cycles of oscillation . This striking pattern of activity was reminiscent of the Phase II swinging motor program , which is likewise induced by activation of neurons that express Bursicon and , as shown by muscle Ca++ imaging below ( Figure 4C ) , consists of approximately 18 bouts of alternating left-right abdominal swinging and lasts approximately 17 min . The induced activity of the Rk-expressing neurons in the isolated nervous system thus appears to be correlated with the motor output induced by a similar stimulus in the intact animal . These results suggest that the Rk-expressing neurons may compose part of the central pattern generator governing pupal ecdysis , and because they are essential for all phases of pupal ecdysis we sought to investigate their activity throughout the ecdysis sequence . We took advantage of the fact that exposure of excised pupal nervous systems to the ETH peptide , ETH1 , stimulates a fictive pupal ecdysis sequence ( Diao et al . , 2016; Kim et al . , 2006; Mena et al . , 2016 ) . We reasoned that Rk-expressing neurons should be activated by ETH1 and that their temporal dynamics might reveal phasic activity corresponding to the phases of ecdysis behavior . We monitored UAS-GCaMP6s activity in the VNC-Rk neurons as before and found that in excised CNS preparations treated with ETH1 these neurons clearly showed enhanced Ca++ activity relative to preparations that did not receive ETH1 ( Figure 4A ) . Compared to time traces of activity in the latter preparations , which were distinguished by relatively flat baselines and slow , low amplitude Ca++ oscillations , the ETH1-induced traces exhibited considerable complexity . The traces could be divided into three principal phases , denoted with Arabic numbers to distinguish them from the behavioral phases , which we denote with Roman numerals . Typically , baseline Ca++ activity rose over approximately the first 10–15 min . after ETH1 addition ( Figure 4 , Phase 1 ) , reached a peak during the next 20 min . ( Phase 2 ) , and then slowly declined ( Phase 3 ) . Superimposed on this baseline activity were Ca++ oscillations , which initially exhibited relatively low amplitude and high frequency , but which increased suddenly in amplitude during peak baseline activity in Phase 2 . After a transition period of mixed amplitude and frequency during the baseline decline in Phase 3 , the oscillations slowed and became more uniformly large and regular . These features were sufficiently stereotyped across preparations that detection of the different Ca++ activity phases could be automated ( see Materials and methods ) , and using custom Matlab code ( i . e . ‘PhaseFinder; https://github . com/BenjaminHWhite/PhaseFinder’ [White , 2016; copy archived at https://github . com/elifesciences-publications/PhaseFinder]; Figure 4—figure supplement 1 ) to analyze the traces , we were able to define their average times of onset ( Figure 4A ) . Consistent with the hypothesis that activity of the VNC-Rk neurons is correlated with the phases of the ecdysis motor programs , the phases of ETH1-induced Ca++ activity have durations similar to the ecdysis behavioral phases observed in live animals . To permit a more direct comparison of ecdysis motor program activity with VNC-Rk neuron activity , we developed an imaging strategy that allowed us to quantify behavior by directly monitoring the Ca++-mediated muscle contractions that drive the body wall movements ( see Materials and methods ) . In this way , ecdysis behavior could be analyzed from Ca++ activity signals using the same methods used to analyze the neuronal activity . To implement this strategy , we used the 24B-Gal4 driver to express UAS-GCaMP6s in muscles and monitored Ca++ signals in animals during pupal ecdysis ( Video 4; Figure 4B ) . The integrated Ca++ signal over the abdominal musculature in such preparations ( dotted box ) typically exhibited a profile similar to that of the VNC-Rk neurons and divided into three principal phases when processed using the PhaseFinder program . The average onset times of these phases closely matched those calculated for the three phases of VNC-Rk activity by the same program ( Figure 4C ) , again indicating a close correspondence between the neuronal activity and the three behavioral phases . Analysis of the spatiotemporal patterns of muscle contraction visualized by Ca++ imaging also revealed details not easily seen by observation of body wall movements alone . Some details previously described only from observations of animals removed from their puparia ( Kim et al . , 2006 ) were clearly evident from the muscle Ca++ imaging , such as the mixed intervals of abdominal swinging and peristalsis that occur at the transition between Phases II and III . As noted above , the transition of the VNC-Rk Ca++ activity from Phase 2 to Phase 3 is also characterized by a variable interval with oscillations of mixed amplitude and frequency . To assess the similarity of this transition period to the observed interval of mixed behavior in the muscle Ca++ traces , we modified the PhaseFinder program to identify this transition period in the Phase III Ca++ data and found an analogous transition in the muscle Ca++ activity that corresponded to the interval of mixed behavior . The calculated durations of the transition periods in the two experiments were not significantly different , although both exhibited a high degree of variation across preparations ( Figure 4C ) . We interpret this additional , and initially unexpected , correspondence in the VNC-Rk and muscle Ca++ data as further evidence that VNC-Rk neuronal activity is correlated with the generation of the ecdysis motor programs . Further support for such a correlation comes from analyzing the spatiotemporal patterns of Ca++ signaling associated with each phase of VNC-Rk neuron activity , the characteristics of which differ from each other in ways similar to those of the ecdysis motor patterns ( Figure 5 ) . The similarity between the Ca++ activity of Phase 2 and the abdominal swinging of Phase II—evident in the activation experiments described above—was also seen in ETH1-induced activity data . Conspicuous left-right oscillations in the Phase 2 Ca++ signal occur both in the images ( Video 5 , Figure 5A ) and in Ca++ traces representing the signals derived from neurons on either side of the ventral midline ( Figure 5B ) . These signals consistently oscillate in antiphase during Phase 2 ( Pearson’s correlation coefficient , R = −0 . 30 ± 0 . 13 , p=0 . 0004 , n = 8 ) , but not , for example , during Phase 1 , or the latter half of Phase 3 ( Figure 5C , bottom ) when the oscillations are coincident . The corresponding analysis of muscle-generated Ca++ signals shows that Phase II is similarly characterized by strong rhythmic activity alternating across the midline with robust , temporally anti-correlated peaks ( R = −0 . 27 ± 0 . 213 , p=0 . 0016 , n = 11; Figure 5D , E ) . To further compare the properties of Phase II behavior and Phase 2 Ca++ dynamics , we analyzed the frequency and number of mid-line oscillations , which for Phase II activity conforms to swings of the body wall . We found that both frequencies ( 0 . 02 ± 0 . 05 Hz for VNC-Rk neurons vs . 0 . 03 ± 0 . 09 Hz for 24B muscles ) and oscillation numbers ( 23 ± 7 for VNC-Rk neurons vs . 18 ± 2 for 24B muscles ) were similar . Taken together , these data are consistent with the hypothesis that Phase 2 activity of the VNC-Rk neurons drives the execution of the Phase II motor program . There is also evidence that the VNC-Rk neuron activity of Phases 1 and 3 similarly generates the motor patterns of Phases I and III of pupal ecdysis . We have already noted that the Transition Period of Phase 3 has an apparent behavioral correlate in the mixed Phase II and III behaviors observed in the muscular activity . In addition , the muscular activity responsible for the stretch compressions of Phase III exhibit bilaterally coincident peaks of Ca++ activity ( R = 0 . 57 ± 0 . 14 , p<0 . 0001 , n = 11; Figure 5F , bottom traces ) , similar to those of VNC-Rk neuron activity during late Phase 3 ( Figure 5C , bottom ) . In some preparations , symmetric and rhythmic anterior-to-posterior waves of VNC-Rk neuron Ca++ activity are also evident , as might be expected for the neuronal activity that drives Phase III abdominal peristalsis ( Video 6 ) . Similarly , the distinct spatiotemporal patterns of VNC-Rk neuron activity associated with Phase 1 are frequently reminiscent of the lateralized , alternating posterior-to-anterior peristaltic waves of contraction that traverse the body wall during Phase I ( Video 7 ) . These patterns , however , differ from those of Phase 2 in that they are not generated by anatomically isolated neuronal populations that dominate the global Ca++ signal , but instead derive from multiple , anatomically intermingled signals . Resolving the individual components of these intermingled signals and reproducibly identifying them across preparations will require more refined methods , but to perform a preliminary decomposition of the Ca++ signal produced by the Rk-expressing neurons , we analyzed the activity in a representative preparation , examining 95 small regions of interest ( ROIs; Figure 5—figure supplement 1A , B ) . Although we cannot be certain that these ROIs correspond to individual cells due to the limited resolution of the Ca++ signal in the z-dimension , their Ca++ activity traces fell into two broad categories: those with large , slow changes in baseline amplitude ( i . e . #28 , 84 , 16 ) , and those with oscillatory activity , but relatively constant baseline ( i . e . #89 , 2 , 74 ) . This suggests that the VNC-Rk population contains at least two types of neurons: one that may represent the oscillatory output of the network and one that may be involved in sustaining phasic activity within it . Records for both types of ROI , contained examples in which activity was predominantly restricted to one ( #28 , 89 ) , two ( #84 , 2 ) , or three ( #16 , 74 ) phases and some records with oscillatory activity exhibited changes in oscillation frequency with phase ( e . g . #2 , 74 ) . Analysis of the average frequency of oscillations as a function of phase for all 95 ROIs indicates that approximately 10% of the ROIs had frequency profiles similar to that of the global Ca++ trace ( Figure 5—figure supplement 1C ) . This observation is consistent with the conclusion that multifunctional neurons participate in the generation of all three phases of behavior , but more refined observation and manipulation of individual VNC-Rk neurons will be required to definitively determine their role ( s ) in central pattern generation . Overall , however , our analysis of the spatiotemporal structure of ETH1-induced Ca++ activity of the Rk-expressing neurons suggests that they comprise a multifunctional central pattern generator ( CPG ) for the pupal ecdysis sequence . The hypothesis that Rk-expressing neurons act as central pattern generators for the ecdysis rhythms predicts that these neurons communicate their output to downstream motor neurons . To directly determine whether motor neurons receive input from Rk-expressing neurons , we used the ATP/P2X2 system to stimulate Rk-expressing neurons while monitoring the response of glutamatergic neurons in the VNC—95% of which are motor neurons ( Daniels et al . , 2008 ) . To do so , we used the VGlut-LexA::QFAD driver to express LexAop-GCaMP6s , while expressing UAS-P2X2 under the control of Rk-Gal4 . We found that ATP induced a substantial increase in Ca++ signal in a large number of neurons ( Figure 5G ) , indicating that glutamatergic motor neurons of the VNC are downstream targets of the Rk-expressing neurons . The identity of some or all the motor neurons activated by the stimulation of Rk-expressing neurons was revealed by our parallel investigation of the targets of CCAP signaling . As shown above , CCAP-R-expressing neurons are essential for pupal ecdysis , and as for the Rk-expressing neurons , we used intersectional methods to ask whether they include ETHR-expressing neurons of the input layer or glutamatergic motor neurons of the output layer . We find that very few CCAP-R-expressing neurons in the pupal CNS co-express either of the ETHR isoforms , but that a significant complement are glutamatergic , as identified by the intersectional driver for CCAP-R and VGlut . A fillet preparation of the pupal body wall in which the muscles are labeled with phalloidin ( Figure 6A , magenta ) , reveals that many of the glutamatergic neurons labeled by UAS-6XGFP ( Figure 6A , green ) , are motor neurons , sending their axons out the abdominal nerves and forming synapses on muscles in the body wall ( Figure 6A’ , arrowheads ) . Using the same intersectional driver to express UAS-GCaMP6s in these neurons ( i . e . CCAP-R/VGlut neurons ) , we selectively monitored their response to ETH1 in excised pupal CNS preparations . Interestingly , the profile of the induced Ca++ activity is multiphasic and decomposes into three principal phases as distinguished by the PhaseFinder program ( Figure 6B ) . As for the Ca++ activity of the VNC-Rk neurons , the last phase contains an identifiable transition period with Ca++ oscillations of mixed amplitude and frequency , and the activity of Phase 2 , when analyzed bilaterally across the midline , shows characteristic left-right alternations ( Figure 6C ) . The durations ( Figure 6D ) and frequencies ( Figure 6E ) of Ca++ activity for the CCAP-R/VGlut and VNC-Rk neurons were comparable , with no statistically significant differences found for any Phase . These observations are consistent with a tight coupling between the activities of the VNC-Rk and CCAP-R/VGlut motor neurons . To directly determine whether activity in CCAP-R neurons can be driven by the Rk-expressing neurons , we used the ATP/P2X2 system and found that selective activation of the Rk-expressing neurons induced a large Ca++ response in the CCAP-R VNC neurons ( Figure 6F ) . We conclude that motor neurons expressing the CCAP-R act downstream of CPG neurons that express the Bursicon receptor , Rickets . The above data are consistent with a model in which the hormones Bursicon and CCAP , released from neurons in the input layer of the ecdysis network , act on two subsequent layers in the network hierarchy responsible for central pattern generation and motor output . The latter two layers , defined by their expression of Bursicon- and CCAP-receptors , respectively , appear to be broadly involved in generating all three phases of the ecdysis sequence . Bursicon and CCAP , however , are unlikely to be released from the ETHRA/CCAP neurons until Phase II , the first motor program for which they are required , and their regulation of Phase I behavior is inhibitory . The initiation and maintenance of Phase I is thus likely to depend on other factors . To ask which neurons within the input layer might regulate Phase I , we used two drivers that together allowed us to interrogate the function of the two subsets of ETHR-expressing neurons outside of those that secrete CCAP and Bursicon . These drivers target neurons that express the B-isoform of the ETHR ( i . e . ETHRB-Gal4; Figure 7A ) , a population almost completely distinct from the ETHRA-expressing subset ( Diao et al . , 2016 ) , and neurons that express ETHRA , but do not co-express CCAP ( i . e . non-CCAP/ETHRA neurons; Figure 7B ) . Silencing the activity of ETHRB-expressing neurons with UAS-Kir2 . 1 , caused a variety of behavioral defects , including rapid , tremulous movements of the body wall , but the most salient feature was the lack of an overt Phase I ( Figure 7C; Video 8 ) . Phase II and Phase III motor programs were readily discernible in affected animals , but the abdominal lifting which initiates Phase I was absent , and although some abdominal movements preceded Phase II , they appeared to be exaggerated versions of the body wall movements that normally prefigure the onset of Phase I , and they never displaced the abdominal air bubble as they do in normal animals . Animals in which the non-CCAP/ETHRA neurons were silenced also exhibited obvious deficits in Phase I , which was executed by most animals ( n = 10/15 ) , but was consistently shorter than it was in controls ( 3 . 82 ± 2 . 8 vs . 7 . 42 ± 3 . 5 , p=0 . 0137; Figure 7D; Video 9 ) . This behavioral difference was not attributable to the bloating commonly observed in these animals due to suppression of ETHRA-expressing neurons that co-express the diuretic hormone Leucokinin ( Diao et al . , 2016 ) . Bloating did , however , make it difficult to consistently distinguish Phases II and III , which sometimes appeared persistently intermingled . These results indicate specific requirements for ETHRB and non-CCAP/ETHRA neurons in the initiation and maintenance of Phase I . Interestingly , Phase I was not induced by activating either of these populations of neurons prior to the onset of ecdysis using UAS-TrpA1 , and further experiments will be required to identify the sufficient causes for initiation of this phase .
The schematic shown in Figure 7D broadly augments existing models of the pupal ecdysis network ( Kim et al . , 2015; Mena et al . , 2016; Zitnan and Adams , 2012 ) , both by providing a more comprehensive description of the input layer than has previously been possible and by identifying the motor circuits on which this layer acts . A principal finding reported here is that the downstream targets of Bursicon and CCAP are shared components of the pupal ecdysis network and are used to generate all three motor rhythms . Our results draw particular attention to the centrality of neurons that express the Bursicon receptor ( Rk ) , which are absolutely required for all pupal ecdysis behavior . A role in central pattern generation is indicated both from the effects of their suppression , which eliminates all motor activity , and from their pattern of ETH1-induced Ca++ activity , which matches the phases of ecdysis behavior . The fact that ETH1-induced Ca++ activity is observed in the excised nervous system and thus in the absence of sensory feedback , demonstrates that it is centrally generated and further supports the identification of the VNC-Rk neurons as central pattern generators . Conclusive evidence that some or all VNC-Rk neurons participate in central pattern generation will require more precise observations and perturbations than those performed here , as will determining the functional roles of individual neurons . However , our preliminary observation that regions containing at most small numbers of VNC-Rk neurons exhibit activity that is phasically coupled to two or more motor patterns argues that the ecdysis circuitry includes multifunctional CPG neurons that express Rk and are subject to modulation by distinct input layer modules , as indicated in Figure 7D . Similar architectures have been described in other motor networks where two CPGs formed from overlapping pools of neurons can switch between activity states to generate distinct behaviors ( Kristan and Gillette , 2007 ) . How input layer neurons modulate the pupal ecdysis CPG is exemplified by the control of Phase II by ETHRA/CCAP neurons . Direct activation of these neurons induces Phase II-like rhythmic activity in the VNC-Rk neurons , an observation that is easily explained if Bursicon secreted from ETHRA/CCAP neurons shifts the mode of activity of the VNC-Rk CPG . This mechanism is consistent with the neuromodulatory control of CPGs described in numerous other systems ( Briggman and Kristan , 2008; Dickinson , 2006; Marder and Bucher , 2007 ) and accounts for the long-standing observation that CCAP- and Bursicon-expressing neurons are important for pupal ecdysis ( Kim et al . , 2006; Lahr et al . , 2012; Park et al . , 2003a; 2008 ) , including Phase II initiation and Phase I termination ( Kim et al . , 2015 ) . The CCAP- and Bursicon-expressing neurons are known to express additional neuropeptides , including Myoinhibitory Peptides and Allatostatin C , and it is likely that these neuromodulators also play a role in regulating these phases . The mixed activity patterns that define the transition from Phase II to Phase III , first described by Kim et al . ( 2006 ) and further characterized here , are also readily interpreted as a period of bistability in which CPG modes transiently alternate , perhaps as Bursicon and/or other co-released neuromodulator concentrations fall . In addition to neurons that switch CPG activity from Phase I to Phase II , the input layer must also contain neurons that initiate pupal ecdysis by inducing Phase I . The search for such neurons has focused primarily on those that express ETHRA ( Kim et al . , 2006; Krüger et al . , 2015; Lahr et al . , 2012; Mena et al . , 2016 ) , but no components of this group have yet been identified that are required for ecdysis initiation . To identify the ETH targets responsible for Phase I , we systematically parsed ETHR-expressing neurons into three , nearly mutually exclusive subsets that together cover the entire input layer . Our results indicate that the largely uncharacterized neurons that express the B-isoform of ETHR are required to initiate Phase I , and that the non-CCAP/ETHRA neurons are important for maintaining that phase . The essential role of ETHRB-expressing neurons in Phase I initiation is consistent with the significantly higher affinity for ETH peptides of ETHRB compared with ETHRA ( Iversen et al . , 2002; Park et al . , 2003b ) . ETHRB-expressing neurons may thus initiate Phase I by responding to rising titers of ETH earlier than neurons expressing ETHRA . How they regulate the VNC-Rk CPG neurons remains to be determined , but their mechanism of action appears to be different from that of the ETHRA/CCAP neurons insofar as the Phase I motor program cannot be evoked by TrpA1-mediated activation . It could be that this manipulation fails to induce the correct pattern of activity in ETHRB-expressing neurons . Preliminary imaging results show that ETHRB-expressing neurons respond to ETH1 with oscillatory activity ( data not shown ) , and it is possible that these neurons directly couple to the Rk-expressing neurons through synaptic or electrical contacts and participate in generating Phase I behavior . However , further characterization of the activity of both the ETHRB- and non-CCAP/ETHRA neurons will be required to determine how they modulate VNC-Rk CPG activity . Two input layer neurons that are common to the ETHRB- and non-CCAP/ETHRA groups express the major ecdysis neuromodulator , EH ( Diao et al . , 2015 ) . The EH-expressing neurons , which are among the few cells to express both ETHRA and ETHRB , respond to ETH1 application at the onset of Phase II ( Kim et al . , 2006 ) , and evidence from other insects indicates that EH targets CCAP-expressing neurons ( Ewer and Truman , 1996 ) . EH is thus thought to be responsible for the release of CCAP and Bursicon , but this has not yet been verified in Drosophila where the EH receptor has yet to be identified . We were thus not able to target EH receptor-expressing neurons in this study , but the identity and function of such neurons is likely to be critical to understanding the progression of the ecdysis sequence . In general , it is worth noting that the neuromodulators regulating the ecdysis sequence are of the type called ‘extrinsic , ’ because they are released from neurons that do not function in the circuits upon which they act ( Katz and Frost , 1996 ) . Extrinsic neuromodulatory neurons , however , must be components of the broader neural networks that generate behaviors , and the mechanisms that organize their activities are only beginning to be understood ( Brezina , 2010 ) . In some cases , these mechanisms are surprising . For example , the neuromodulatory connections between neurons that govern two foraging states in C . elegans are orthogonal to the sensory-to-motor synaptic connections between these neurons , which are not involved in the state decision ( Flavell et al . , 2013 ) . There are currently few studies that jointly examine patterns of neuromodulatory and synaptic connectivity ( but see Schlegel et al . , 2016 ) , and to understand how extrinsic neuromodulatory neurons integrate into the broader networks in which they function more examples of such networks are required . Elucidating the interactions of neurons in the input layer of the ecdysis network—in addition to interactions of the input layer with neurons in other layers—should provide insight into this general problem . Our finding that the motor output of the pupal ecdysis network is mediated by neurons that express the CCAP-R provides insight into the hitherto poorly understood mechanism of action of CCAP . This neuropeptide plays critical roles in the ecdysis of other insects ( Arakane et al . , 2008; Gammie and Truman , 1997 ) , but genetic data demonstrate that in Drosophila it plays a subsidiary role to Bursicon , acting synergistically with that hormone to render pupal ecdysis more robust ( Lahr et al . , 2012 ) . Our results indicate that it does so by acting on motor neurons , and because CCAP is co-released with Bursicon from the ETHRA/CCAP neurons to govern the CPG transition at Phase II , this suggests a role for feed-forward signaling in the pupal ecdysis circuit . Neuromodulatory feedforward pathways have been previously described ( Wu et al . , 2010 ) and appear to be a common motif in motor network architectures ( Taghert and Nitabach , 2012 ) . Feedforward loops of the type posited here for Bursicon and CCAP may be important in adjusting the coupling between Rk-expressing CPG neurons and their downstream motor neuron targets during Phase II . Compensatory adjustments in CPG , motor neuron and muscle activity by a single neuropeptide released from two different nodes in a feedforward loop have been described in the Aplysia feeding network where they guarantee stability of network output ( Jing et al . , 2010 ) . Coordinating CPG activity with motor neuron activity may be particularly important for multifunctional CPGs , in which individual neurons participate in multiple motor patterns , as for example , in the leech swim/crawl network in which multifunctional neurons fire in phase with the contraction of one muscle group during swimming , but not necessarily during crawling ( Briggman and Kristan , 2006 ) . The architecture of the pupal ecdysis network revealed here is similar to that of other motor circuits , such as those governing locomotion , feeding , and breathing in which higher order neurons modulate the activity of core CPGs to generate varied motor patterns ( Feldman et al . , 2013; Mullins et al . , 2011; Nusbaum et al . , 2001 ) . What is striking about neuromodulator action in the ecdysis circuit is its broad scope . ETH acts throughout the input layer to control different phases of pupal ecdysis behavior; Bursicon similarly regulates a large and essential set of neurons constituting the ecdysis CPG; and CCAP acts on many motor neurons necessary for generating the rhythmic ecdysis movements . The observation that Bursicon and CCAP signal from the input layer speaks to an organizational logic in which the ecdysial neuromodulators function together to provide coherence to the operation of the pupal ecdysis network by acting both within each hierarchical layer and by acting coordinately across layers . This organization is consistent with a generalized role for neuromodulatory systems in organizing neural activity to generate behavior ( Marder , 2012 ) . Our results also support the rationale of mapping neuromodulatory pathways as a strategy for identifying essential network circuits and their functional organization . It is worth noting that our mapping of the pupal ecdysis network was done without reference to patterns of synaptic connectivity . Synaptic connectomes have proved difficult to interpret , in part due to their dense interconnectivity . If , as has been previously emphasized ( Bargmann and Marder , 2013; Brezina , 2010; Marder , 2012 ) , this interconnectivity reflects the multifunctionality of the underlying networks , and if the functional configuration of a network at any given time is determined by where and how neuromodulators are acting on its components , then patterns of neuromodulatory connectivity may provide a necessary complement to synaptic maps to render them interpretable . A key challenge will lie in identifying which neuromodulator systems play critical roles in establishing a network’s output , but as the work here demonstrates , when these are known , the neuromodulatory connections can deliver substantial insight into how a neural network is organized .
ETH1 was synthetized by GenScript ( Piscataway , NJ ) ; all oligonucleotides were synthesized by Integrated DNA Technologies , Inc ( Coralville , IA ) ; and all gene synthesis were carried out by Epoch Life Science , Inc ( Sugar Land , TX ) . All restriction enzymes were from New England Biolabs ( Ipswich , MA ) . Vinegar flies of the species Drosophila melanogaster were used in this study . Flies were raised on cornmeal-molasses-yeast medium and housed at 25°C and 65% humidity . Both male and female progeny of the genotypes indicated in Supplementary file 1 were used in this study and all experiments analyzed animals at the time of pupal ecdysis , approximately 12 hr after puparium formation . Fly stocks described in previous publications include: ETHRA-Gal4 ( i . e . ETHRAMI00949-Gal4 ) , ETHRB-Gal4 ( i . e . ETHRBMI00949-Gal4 ) , ETHRA-p65AD ( i . e . ETHRAMI00949-p65AD ) , all from Diao et al . ( 2016 ) ; CCAP-R-Gal4 ( CCAP-RMI05804-Gal4 ) , VGlut-Gal4DBD ( i . e . VGlutMI04979-Gal4DBD ) and VGlut-LexA::QFAD ( VGlutMI04979-LexA::QFAD ) , all from Diao et al . ( 2015 ) ; Rkpan-Gal4 ( Diao and White , 2012 ) ; and CCAP-Gal4DBD ( Luan et al . , 2006b ) . The ETHRB-p65AD line was made by ΦC31-mediated cassette exchange into MiMIC insertion MI00949 in the ETHR gene using a strategy previously used to make the ETHRA-p65AD line ( Diao et al . , 2016 ) . Briefly , we created an ‘ETHRBMI00949-p65AD in 4b’ construct ( Supplementary file 2 ) by combining two fragments: one was a PCR-generated fragment encoding T2A-p65AD , amplified from the pBS-KS-attB-SA-SD-0-T2A-P65AD vector ( Diao et al . , 2015 ) using the primers listed in the Key Resources Table , and the other was a synthesized gene fragment corresponding to the ETHR genomic region from the MI00949 insertion point to the 3’ end of exon 4b . The latter fragment included an extension containing an Hsp70 polyadenylation signal and was flanked by Sal I restriction sites , which were used to subclone the fragment into the pBS-KS-attB1-2 vector ( Venken et al . , 2011 ) . The T2A-p65AD fragment was inserted in frame into a unique Bgl II site just prior to the stop codon of Exon 4b in the synthesized fragment using the In-Fusion Cloning Kit from Takara Bio USA , Inc ( Mountain View , CA ) . The resulting vector ( pBS-KS-ETHRMI00949-T2A-p65AD in 4B ) was used for ΦC31-mediated transgenesis . All other lines created for use in this paper were generated using the Trojan exon technology by plasmid injection as described in Diao et al . ( 2015 ) . All injections were made by Rainbow Transgenic Flies , Inc ( Camarillo , CA ) . CCAP-R-specific Split Gal4 lines ( CCAP-R-Gal4DBD and CCAP-R-p65AD ) , were generated by inserting the indicated Trojan exons into the MI05804 MiMIC site in intron 4 of the CCAP-R gene . New Rk-specific lines were generated by first inserting the Trojan Gal4 Expression Module ( T-GEM ) into the Rk locus using Crispr/Cas technology as described previously ( Diao et al . , 2015 ) . The guide RNA ( sgRNA ) used to target T-GEM insertion was specific for a PAM site in intron 13 of the Rk gene and was made by annealing the two oligos listed in the Key Resources Table . The sgRNA was then inserted into the U6b-sgRNA-short plasmid of Ren et al . ( 2013 ) after digestion with Bbs I . The T-GEM construct was flanked by left ( HAL ) and right ( HAR ) homology arms of approximately 1 kb in length amplified by PCR using the primers indicated in the Key Resources Table ( where upper case indicates sequences homologous to Rk and lower case indicates sequence homologous to the pT-GEM ( 1 ) plasmid . ) The PCR products were cloned into the linearized pT-GEM ( 1 ) plasmid using the In-Fusion Cloning Kit from Takara Bio USA , Inc ( Mountain View , CA ) . The RkTGEM-Gal4 transgenic flies were made by microinjecting embryos of the {nosCas9} attP2 line ( Ren et al . , 2013 ) with the sgRNA and pT-GEM plasmid DNA . The microinjection was made by Rainbow Transgenic Flies , Inc ( CA ) . The G0 adults were crossed with yw;Sp/Cyo;Dr/TM3 , Sb flies and the progeny were screened by fluorescence for those with eye-specific expression of the selection marker , RFP . Rk-specific Split Gal4 hemidriver lines were subsequently generated from this RkTGEM-Gal4 line by substituting Gal4DBD and p65AD Trojan exons into the site of T-GEM insertion . The Burs-LexA::VP16AD line was generated from a drosophilized LexA::VP16AD DNA construct ( Supplementary file 2 ) in pBlueScript synthesized by Epoch Life Science , Inc ( Sugar Land , TX ) . Flanking NotI and AscI restriction sites in the construct were used to subclone this construct into the pCAST-BursGal4DBD plasmid ( Luan et al . , 2012 ) after excision of the Gal4DBD sequence . Burs-LexA::VP16AD flies were made by standard P-element transgenesis by injecting the resulting pCAST-Burs-LexA::VP16AD plasmid into the embryos of w1118 flies and a line was established with the transgene inserted on Chromosome II . The UAS-dTrpA1 and UAS-P2X2 lines were the kind gifts of Paul Garrity and Orie Shafer , respectively . As indicated in the Key Resources Table , all other fly lines were obtained from the Bloomington Drosophila Stock Center at Indiana University including the MI05804 line , which was generated by the Drosophila Gene Disruption Project , http://flypush . imgen . bcm . tmc . edu/pscreen/mimic . html ( Nagarkar-Jaiswal et al . , 2015 ) . All neuronal suppression experiments were conducted using two copies of UAS-Kir2 . 1 , by combining insertions on Chromosomes II and III . To restrict suppression to the pupal stage , we used the temperature-sensitive Gal4 inhibitor , tsGal80 , expressed under the control of the ubiquitous tubulin promoter ( i . e . tub-Gal80ts ) ( McGuire et al . , 2003 ) . Animals were shifted to 31°C at the wandering L3 stage . Neuronal activation of ETHRA/CCAP neurons was accomplished using UAS-dTrpA1 using temperature shifts from 18°C to 29°C . Transient temperatures shifts were accomplished by heating to 29°C for 1 min before returning to 18°C . The method used for videorecording of pupal ecdysis behaviors was described in Diao et al . ( 2016 ) . Briefly , cryptocephalic pupae were selected for videorecording just prior to pupal ecdysis , after the abdominal bubble had appeared and vigorous movement of the gut had commenced . Puparia were coated with a mixture of halocarbon oil and water ( ~2:1 ) to increase their transparency and placed ventral side down on a cover glass , which was attached to a glass slide with doublestick tape to form a small chamber . Behavior was recorded from the ventral side for 1–2 hr at 20X magnification using a Sony NEX VG20 camcorder mounted on an Olympus SZX12 stereomicroscope . Videorecords were imported into iMovie ( Apple Inc . , Cupertino , CA ) software for behavioral analysis . Scoring of the pupal ecdysis phases largely followed the criteria described by Kim et al . ( 2006 ) : Phase I ( pre-ecdysis ) begins when the tip of the abdomen is lifted , creating an air pocket at the posterior end of the puparium , and continues with posterior-to-anterior ‘rolling’ contractions of the lateral body wall; Phase II ( ecdysis ) is characterized by persistent swinging , resulting from alternating lateral contractions of the abdomen; and Phase III ( post-ecdysis ) consists predominantly of anterior-to-posterior ‘stretch-compressions’ of the abdomen . The transition from Phase II to Phase III is not well-defined and consists of mixed abdominal movements that include intermingled swings and stretch-compressions . These various movements are often poorly resolved when viewed through the puparium , even when the latter are clarified by treatment with an oil/water mixture . To unambiguously define the end of Phase II for behavioral experiments , and to simplify the analysis , we therefore defined the last occurrence of any swinging ( scored by playing the videos backwards and marking the time of the ‘first’ swinging bout ) as the end of Phase II . For the muscle Ca++ imaging experiments , where the transition from consolidated swinging to mixed abdominal movements could be resolved , we defined the end of consolidated swinging as the end of Phase II , and divided Phase III into a ‘Transition Period’ consisting of the mixed behavioral phase followed by a ‘late Phase III’ period of consolidated stretch-compressions . This division conformed well with the patterns of neural Ca++ activity recorded in VNC-Rk and glutamatergic CCAP-R-expressing neurons . Nervous system whole mounts were excised from stage four pupae with an air bubble ( Bainbridge and Bownes , 1981 ) and prepared for immunolabeling as described previously , using normal donkey or goat serum in the blocking solution ( Luan et al . , 2006a ) . Rabbit anti-pBurs ( kind gift of Aaron Hsueh and Willi Honegger ) was used at 1:1000 dilution . Guinea pig secondary antibodies were conjugated to Alexa Fluor 555 ( Invitrogen , Carlsbad , CA ) . Expression patterns of Gal4 and Split Gal4 lines were visualized using either UAS-6XEGFP or UAS-6XmCherry . To visualize labeling of motor axons by CCAP-R-p65AD ∩ VGlut-Gal4DBD , wandering third instar larvae were briefly anesthetized under CO2 , immersed in 100% ETOH , and then pinned out and filleted from the dorsal side in PBS . The head and internal organs were removed before fixation and staining . Muscle was visualized using Alexa Fluor 594-conjugated phalloidin ( Invitrogen , Carlsbad , CA ) . Preparations were incubated with phalloidin at 1:1000 dilution for 2 hr , followed by three 10 min washes . Confocal imaging was done using a Nikon C2 personal confocal microscope with a 20X/0 . 75 NA air objective . Unless otherwise noted , the images presented are maximum intensity projection images of a Z-stack collected through the entire preparation . Ca++ imaging of excised pupal nervous systems was carried out using conditions similar to those originally described by Kim et al . ( 2006 ) . Cryptocephalic pupae expressing two copies of UAS-GCaMP6s ( on Chromosomes II and III ) under the control of Rk-Gal4 , ETHRB-Gal4 , or the Split Gal4 driver CCAP-R-P65AD ∩ vGlut-Gal4DBD were dissected under cold phosphate-buffered saline ( 137 mM NaCl , 2 . 7 mM KCl , 10 mM Na2HPO4 and 2 mM KH2PO4 , pH 7 . 3 ) approximately 2 hr prior to pupal ecdysis as determined by the appearance of the abdominal gas bubble and the onset of gut movement . CNSs were excised and placed on poly-lysine ( Sigma-Aldrich , St . Louis , MO ) coated cover glass , dorsal side up , and then covered with Schneider's Insect Medium . GCaMP6s fluorescence in the VNC was imaged at a frequency of approximately 1 Hz for 90 min using a Nikon C2 confocal microscope with a 20X/0 . 75 NA air objective . Imaging was carried out using the largest pinhole ( optical section thickness of 150 μm ) and focusing on a plane approximately 20–35 μm below the dorsal surface , to maximize the number of neurons in the field . Preparations were imaged for 5 min to measure baseline Ca++ activity prior to addition of 600 nM synthetic ETH1 . Muscle Ca++ activity was imaged in pupal animals that expressed two copies of UAS-GCaMP6s under the control of the muscle driver 24B-Gal4 . These animals were selected prior to pupal ecdysis as described above , and their puparia were rinsed with 50% bleach for 3 min to permit optimal clarity . Puparia were then mounted as described for behavioral videorecording in a halocarbon oil/water mixture , and GCaMP6s florescence was imaged from the ventral side at a frequency of approximately 1 Hz for 90 min using a Nikon C2 confocal microscope with a 4X/0 . 13 NA air objective . By using the largest pinhole and a single image plane , Ca++ activity of most bodywall muscles could be visualized . The Nikon C2 imaging files collected during GCaMP imaging were saved as ND2 files for quantification . All ND2 format images were background-subtracted and mean intensities over the regions of interest ( ROI ) measured using ImageJ software ( Schneider et al . , 2012 ) . ROI are as indicated in the figures and Ca++ traces were then normalized to the average signal ( F ) measured during the initial 5-min period prior to addition of ETH and are presented as ΔF/F in the figures . Muscle Ca++ traces were normalized to the average signal measured over the first 250 frames . For the experiment shown in Figure 5—figure supplement 1 , the frequencies of Ca2+ oscillations for single ROIs were determined using the PhaseFinder program ( described below ) and average frequencies were calculated for each Phase , the duration of which was calculated by PhaseFinder using the global Ca++ signal for the Rk-VNC neurons . Heatmaps showing the average Phase frequencies for each individual ROI ( and the whole population ) were created with MatLab using the built-in heatmap function . Movies prepared for presentation were exported as . avi files and then converted into iMovie format for editing and display . Custom MatLab code ( available at GitHub: https://github . com/BenjaminHWhite/PhaseFinder ) was written to objectively identify the onset of each phase of Ca++ activity . This was achieved as shown schematically in Figure 4—figure supplement 1 . Using ImageJ , the mean Ca++ activity within an ROI drawn over the whole VNC was calculated for each image in a timeseries to generate a Ca++ trace . The PhaseFinder code operated on such a trace by first identifying peaks , and then a sliding window was used to compare the frequency or average amplitude of the peaks in one time-window to those in the next . The onset of Phase 1 was defined as the time of the first peak of activity following the addition of ETH . To find the onset of Phase 2 ( and offset of Phase 1 ) , the difference in average peak amplitude was calculated for consecutive windows and the first window for which this difference was more than one standard deviation above the average peak amplitude across all windows was defined as the beginning of Phase 2 . Phase 3 was the most difficult to define because of the complexity of its Ca++ activity patterns , but it was generally distinguished from the two preceding phases by a change in peak frequency . As described in the text , we divided this phase into a ‘Transition Period , ’ of mixed frequencies , and a period of more uniform frequency ( ‘late Phase 3’ ) . The onset of Phase 3 ( which also marked the offset of Phase 2 ) was defined by the decrement in frequency that occurred at the time of phase transition . This was most easily identified by analyzing the time series data in reverse , using time windows starting at the end of the Ca++ trace and looking for the first window for which the difference in frequency was positive ( indicating a decrement in frequency moving forward in time ) . The uniformity of activity in late Phase 3 allowed its onset ( and the offset of the Transition Period of Phase 3 ) to be defined simply as the first window for which the difference in peak frequency exceeded the average peak frequency for all windows by more than one standard deviation . Phase durations were calculated by subtracting the onset of each phase from the onset of the next phase . Further details about each parameter used in the PhaseFinder program can be found in the documentation available at Github ( see below ) . Importantly , the parameter settings used to analyze all neuronal datasets ( i . e . for both Rk- and CCAP-R-expressing neurons ) were the same . Muscle Ca++ traces required somewhat different parameters because of the significantly larger signal amplitudes in integrated mean fluorescence . For all experiments , the number of biological replicates ( i . e . the number of animals or CNS preparations of a given genotype ) analyzed was at least five , with the actual numbers for each experiment given in the figures or figure legends . Estimation of required sample sizes were made using the procedure of Campbell et al . ( 1995 ) for binary categorical variables , since most experiments involved determining whether manipulation of a specific neuronal population ( e . g . one expressing the receptors for Bursicon or CCAP ) resulted in pupal ecdysis deficits or not . Similar manipulations previously applied to neurons expressing Bursicon ( Peabody et al . , 2008 ) or CCAP ( Park et al . , 2003a ) report effect sizes on pupal ecdysis ranging from approximately 0 . 45 to 0 . 9 , indicating the use of samples sizes of 4 to 10 . The correlation analyses of Ca++ signals measured on the left- and right-hand sides of the midline for VNC-Rk neurons and muscles were performed in MatLab using the built-in Corrcoeff function . For the statistical analysis of Ca++ oscillation frequencies in Figure 6E , GraphPad Prism was used to conduct a one-way ANOVA with multiple comparisons . Brown-Forsythe and Bartlett’s tests determined that the variance between groups was not significant and a Sidak’s multiple comparison’s test showed no significant differences between Rk and CCAPR-Vglut frequencies at any of the phases . | Why do animals behave the way they do ? Behavior occurs in response to signals from the environment , such as those indicating food or danger , or signals from the body , such as those indicating hunger or thirst . The nervous system detects these signals and triggers an appropriate response , such as seeking food or fleeing a threat . But because much of the nervous system takes part in generating these responses , it can make it difficult to understand how even simple behaviors come about . One behavior that has been studied extensively is molting in insects . Molting enables insects to grow and develop , and involves casting off the outer skeleton of the previous developmental stage . To do this , the insect performs a series of repetitive movements , known as an ecdysis sequence . In the fruit fly , the pupal ecdysis sequence consists of three distinct patterns rhythmic abdominal movement . A hormone called ecdysis triggering hormone , or ETH for short , initiates this sequence by triggering the release of two further hormones , Bursicon and CCAP . All three hormones act on the nervous system to coordinate molting behavior , but exactly how they do so is unclear . Diao et al . have now used genetic tools called Trojan exons to identify the neurons of fruit flies on which these hormones act . Trojan exons are short sequences of DNA that can be inserted into non-coding regions of a target gene to mark or manipulate the cells that express it . When a cell uses its copy of the target gene to make a protein , it also makes the product encoded by the Trojan exon . Using this technique , Diao et al . identified three sets of neurons that produce receptor proteins that recognize the molting hormones . Neurons with ETH receptors start the molting process by activating neurons that make Bursicon and CCAP . Neurons with Bursicon receptors then generate motor rhythms within the nervous system . Finally , neurons with CCAP receptors respond to these rhythms and produce the abdominal movements of the ecdysis sequence . Many other animal behaviors depend on substances like ETH , Bursicon and CCAP , which act within the brain to change the activity of neurons and circuits . The work of Diao et al . suggests that identifying the sites at which such substances act can help reveal the circuits that govern complex behaviors . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Neuromodulatory connectivity defines the structure of a behavioral neural network |
Hematopoietic stem cells ( HSCs ) are maintained by a perivascular niche in bone marrow but it is unclear whether the niche is reciprocally regulated by HSCs . Here , we systematically assessed the expression and function of Angiopoietin-1 ( Angpt1 ) in bone marrow . Angpt1 was not expressed by osteoblasts . Angpt1 was most highly expressed by HSCs , and at lower levels by c-kit+ hematopoietic progenitors , megakaryocytes , and Leptin Receptor+ ( LepR+ ) stromal cells . Global conditional deletion of Angpt1 , or deletion from osteoblasts , LepR+ cells , Nes-cre-expressing cells , megakaryocytes , endothelial cells or hematopoietic cells in normal mice did not affect hematopoiesis , HSC maintenance , or HSC quiescence . Deletion of Angpt1 from hematopoietic cells and LepR+ cells had little effect on vasculature or HSC frequency under steady-state conditions but accelerated vascular and hematopoietic recovery after irradiation while increasing vascular leakiness . Hematopoietic stem/progenitor cells and LepR+ stromal cells regulate niche regeneration by secreting Angpt1 , reducing vascular leakiness but slowing niche recovery .
Hematopoietic stem cells ( HSCs ) reside in a specialized bone marrow niche in which Leptin Receptor+ ( LepR+ ) perivascular stromal cells and endothelial cells secrete factors that promote their maintenance ( Kobayashi et al . , 2010; Ding et al . , 2012; Ding and Morrison , 2013; Greenbaum et al . , 2013; Poulos et al . , 2013; Morrison and Scadden , 2014 ) . Nearly all the cells that express high levels of Scf ( Kitl ) or Cxcl12 in the bone marrow are LepR+ ( Zhou et al . , 2014 ) . Conditional deletion of Scf from LepR+ cells and endothelial cells leads to loss of all quiescent and serially-transplantable HSCs from adult bone marrow ( Oguro et al . , 2013 ) . These LepR+ niche cells have also been identified based on their expression of high levels of Cxcl12 ( Sugiyama et al . , 2006; Ding and Morrison , 2013; Omatsu et al . , 2014 ) , low levels of the Nestin-GFP transgene ( Mendez-Ferrer et al . , 2010; Kunisaki et al . , 2013 ) , PDGFRα ( Morikawa et al . , 2009; Zhou et al . , 2014 ) , and Prx1-Cre ( also known as Prrx1-Cre ) ( Greenbaum et al . , 2013 ) . Consistent with the conclusion that HSC niche cells include mesenchymal stem/stromal cells ( Sacchetti et al . , 2007; Mendez-Ferrer et al . , 2010 ) , the LepR+ cells are highly enriched for CFU-F and give rise to most of the osteoblasts and fat cells that form in adult bone marrow ( Zhou et al . , 2014 ) . Angpt1 has been proposed to be expressed by osteoblasts in the bone marrow and to promote the maintenance of quiescent HSCs in an osteoblastic niche ( Arai et al . , 2004 ) . However , HSCs and perivascular stromal cells also express Angpt1 ( Takakura et al . , 2000; Ivanova et al . , 2002; Forsberg et al . , 2005; Kiel et al . , 2005; Sacchetti et al . , 2007; Ding et al . , 2012 ) . Moreover , it has not been tested whether Angpt1 deficiency affects HSC function in vivo . Thus , the physiological function and sources of Angpt1 in the bone marrow remain uncertain . Angpt1 ( Suri et al . , 1996 ) , and its receptor Tie2 ( Dumont et al . , 1994; Puri et al . , 1995; Sato et al . , 1995; Davis et al . , 1996 ) , are necessary for embryonic vascular development . Tie2 is mainly expressed by endothelial cells ( Schnurch and Risau , 1993; Kopp et al . , 2005 ) but also by HSCs ( Iwama et al . , 1993; Arai et al . , 2004 ) . Angpt1 over-expression promotes the development of larger , more numerous , more highly branched , and less leaky blood vessels ( Suri et al . , 1998; Thurston et al . , 1999; Cho et al . , 2005 ) . Angpt1 expression by primitive hematopoietic progenitors ( HPCs ) promotes angiogenesis during embryonic development ( Takakura et al . , 2000 ) . Global conditional deletion of Angpt1 between embryonic day ( E ) 10 . 5 and E12 . 5 increases the size and number of blood vessels in fetal tissues but later deletion has little effect on vascular development ( Jeansson et al . , 2011 ) . Nonetheless , Angpt1 does regulate angiogenesis in response to a variety of injuries in adult tissues ( Kopp et al . , 2005; Jeansson et al . , 2011; Lee et al . , 2013 ) , promoting angiogenesis in some contexts ( Thurston et al . , 1999 ) while negatively regulating angiogenesis in other contexts ( Visconti et al . , 2002; Augustin et al . , 2009; Jeansson et al . , 2011; Lee et al . , 2014 ) . A key function of Angpt1 is to reduce the leakiness of blood vessels , perhaps by tightening junctions between endothelial cells ( Thurston et al . , 1999; Brindle et al . , 2006; Lee et al . , 2013 , 2014 ) . Irradiation and chemotherapy not only deplete HSCs but also disrupt their niche in the bone marrow , particularly the sinusoids ( Knospe et al . , 1966; Kopp et al . , 2005; Li et al . , 2008; Hooper et al . , 2009 ) around which most HSCs ( Kiel et al . , 2005 ) as well as Scf- , Cxcl12- , and LepR-expressing stromal cells reside ( Ding et al . , 2012; Ding and Morrison , 2013; Omatsu et al . , 2014; Zhou et al . , 2014 ) . Regeneration of this perivascular niche after injury , including endothelial and stromal components , is necessary for regeneration of HSCs and hematopoiesis ( Kopp et al . , 2005; Hooper et al . , 2009 ) . After 5-fluorouracil treatment , Tie2 signaling ( which is regulated by its ligands Angpt1 , Angpt2 , and possibly Angpt3 [Augustin et al . , 2009; Eklund and Saharinen , 2013; Fagiani and Christofori , 2013; Thomson et al . , 2014] ) regulates the remodeling of blood vessels in the bone marrow and adenoviral over-expression of Angpt1 accelerates the recovery of hematopoiesis ( Kopp et al . , 2005 ) . This raises the question of whether endogenous Angpt1 is necessary for niche recovery and whether it acts by promoting HSC function in an osteoblastic niche or by regulating vascular regeneration .
We first assessed the Angpt1 expression using a commercially available antibody to stain bone marrow sections . Most bone marrow cells did not stain positively and we were unable to detect any staining among bone-lining cells where osteoblasts localize ( Figure 1A–C ) . The most prominent staining was in large CD41+ megakaryocytes ( Figure 1D–F ) and in c-kit+ HPCs ( Figure 1G–I ) . 10 . 7554/eLife . 05521 . 003Figure 1 . Angpt1 was expressed by megakaryocytes and hematopoietic stem/progenitor cells in the bone marrow . ( A–C ) Immunostaining of femur sections from Col1a1*2 . 3-GFP mice with anti-Angpt1 antibody showed that Angpt1 was not detectably expressed by bone lining Col1a1*2 . 3-GFP+ osteoblasts . Nuclei were stained with DAPI ( blue ) . ( n = 3 mice from 3 independent experiments ) . ( D–I ) Representative femur sections from wild-type mice showed that anti-Angpt1 antibody stained CD41+ megakaryocytes ( arrows , D–F ) and c-kit+ hematopoietic progenitors ( HPCs ) ( arrows , G–I ) throughout the bone marrow . * in F indicates trabecular bone—note the lack of Angpt1 staining in bone-lining cells ( n = 3 mice from 3 independent experiments ) . ( J–O ) Images of femur sections from Angpt1GFP mice showed that GFP was expressed by CD41+ megakaryocytes ( arrows , J–L ) and c-kit+ HPCs ( arrows , M–O ) ( n = 3 mice from 3 independent experiments ) . ( P–Y ) Flow cytometric analysis of non-enzymatically dissociated Angpt1GFP bone marrow cells ( which contains hematopoietic but few stromal cells ) showed that GFP was rarely expressed by whole bone marrow ( WBM ) cells ( P ) or c-kit− cells ( Q ) but was expressed by most c-kit+ cells ( R ) , CD150+CD48−LSK hematopoietic stem cells ( HSCs ) ( S ) , CD150−CD48−LSK multipotent progenitor cells ( MPPs ) ( T ) , CD48+LSK HPC cells ( U ) , Flt3+IL7Rα+Lineage−Sca1lowc-kitlow common lymphoid progenitors ( CLPs ) ( V ) , CD34+FcγR−Lineage−Sca1−c-kit+ common myeloid progenitor cells ( CMPs ) ( W ) and CD34+FcγR+Lineage−Sca1−c-kit+ granulocyte/macrophage progenitors ( GMPs ) ( X ) . CD34−FcγR−Lineage−Sca1−c-kit+ megakaryocytic/erythroid progenitors ( MEPs ) expressed little GFP ( Y ) . Data represent mean ± s . d . from 4 mice from 4 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 00310 . 7554/eLife . 05521 . 004Figure 1—figure supplement 1 . Generation of Angpt1GFP knock-in mice . ( A ) Targeting strategy to generate the Angpt1GFP knock-in allele . A BAC clone containing the Angpt1 genomic region was used to generate the targeting vector by recombineering . The knock-in allele resulted in the replacement of the first exon of Angpt1 by GFP , in-frame with the ATG start codon of Angpt1 . ( B ) The targeting vector was electroporated into Bruce4 ES cells . Correctly targeted clones were identified by Southern blotting with probes indicated in panel A . These ES cells were used to generate chimeric mice by blastomere injection . Chimeric mice were then bred with C57BL/6-Tyrc-2J to obtain germline transmission . These mice were bred with Flpe mice ( Rodriguez et al . , 2000 ) to remove the Neo cassette . ( C ) PCR genotyping demonstrated germline transmission of the Angpt1GFP-Neo allele . ( D ) No live Angpt1GFP/GFP pups were born from Angpt1GFP × Angpt1GFP matings , as would be expected for Angpt1 deficient mice , indicating that Angpt1GFP is a strong loss-of-function allele . ( E ) In mechanically dissociated bone marrow from Angpt1GFP mice ( which contains hematopoietic but not stromal cells ) 85% of GFP+ cells were c-kit+ and 76% of c-kit+ cells were GFP+ ( n = 4 mice from 4 independent experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 004 To analyze Angpt1 expression by flow cytometry , we generated Angpt1GFP knock-in mice by recombining GFP into the endogenous Angpt1 locus ( Figure 1—figure supplement 1A–D ) . Consistent with the antibody staining pattern , GFP was expressed by CD41+ megakaryocytes ( Figure 1J–L ) and c-kit+ HPCs throughout bone marrow ( Figure 1M–O ) . By flow cytometry , only 1 . 5 ± 0 . 8% of mechanically dissociated bone marrow cells ( which include few stromal cells ) were GFP+ ( Figure 1P ) . Overall , 85% of GFP+ hematopoietic cells were c-kit+ ( Figure 1—figure supplement 1E ) : 72 ± 13% of c-kit+ cells were GFP+ and only 1 . 3 ± 0 . 7% of c-kit− cells were GFP+ ( Figure 1Q , R ) . All CD150+CD48−LSK HSCs expressed high levels of GFP ( Figure 1S ) . All CD150−CD48−LSK multipotent progenitors ( MPPs ) were also positive for GFP , though at somewhat lower levels per cell than HSCs ( Figure 1T ) . Virtually all CD48+LSK HPCs , Lineage−Sca1lowc-kitlowFlt3+IL7Rα+ common lymphoid progenitors ( CLPs; Kondo et al . , 1997 ) , CD34+FcγR−Lineage−Sca1−c-kit+ common myeloid progenitors ( CMPs; Akashi et al . , 2000 ) , and CD34+FcγR+Lineage−Sca1−c-kit+ granulocyte-monocyte progenitors ( GMPs; Akashi et al . , 2000 ) were GFP+ but with successively lower expression levels per cell relative to HSCs ( Figure 1U–X ) . Few CD34−FcγR−Lineage−Sca1−c-kit+ megakaryocyte-erythroid progenitors ( MEPs; Akashi et al . , 2000 ) were positive for GFP ( Figure 1Y ) . Angpt1 was thus broadly expressed by early HPCs , at levels that declined as progenitors matured . Virtually all of the Angpt1 expression by stromal cells in the bone marrow was by LepR+ cells . GFP+LepR+ stromal cells localized mainly around sinusoids throughout the bone marrow ( Figure 2A–C ) but were also present near arterioles ( data not shown ) . GFP was expressed by 94 ± 3 . 2% of LepR+ stromal cells and 94 ± 2 . 5% of GFP+ stromal cells ( CD45−Ter119− ) were LepR+ ( Figure 2F ) . Consistent with this , nearly all GFP+ stromal cells were PDGFRα+ , a marker of mesenchymal stem/stromal cells ( Morikawa et al . , 2009 ) expressed by LepR+ bone marrow cells ( Zhou et al . , 2014 ) . Nearly all PDGFRα+ cells were GFP+ ( Figure 2G ) . Consistent with an earlier report ( Sacchetti et al . , 2007 ) , this suggests that Angpt1 is widely expressed by mesenchymal stem/stromal cells in the bone marrow as LepR+ cells exhibit most of the CFU-F and osteogenic activity in adult mouse bone marrow ( Zhou et al . , 2014 ) . We were unable to detect GFP expression by endothelial cells ( Figure 2H ) , or by Osteopontin+ bone lining cells in the diaphysis ( Figure 2D ) or metaphysis ( Figure 2E ) . Quantitative RT-PCR ( qRT-PCR ) analysis found that the highest levels of Angpt1 in the bone marrow were in HSCs ( 350-fold higher than whole bone marrow cells [WBM] ) , followed by LepR+ stromal cells ( 200-fold higher than WBM ) , LSK primitive HPCs ( 120-fold ) , c-kit+ HPCs ( 60-fold ) , and megakaryocytes ( 70-fold; Figure 2I ) . 10 . 7554/eLife . 05521 . 005Figure 2 . Angpt1 was expressed by Leptin Receptor+ ( LepR+ ) perivascular stromal cells but not endothelial cells or osteoblasts in bone marrow . ( A–C ) Representative femur sections showed that LepR+ perivascular stromal cells ( Tomato+ ) expressed GFP in Leprcre; tdTomato; Angpt1GFP mice . Endothelial cells were stained with anti-VE-cadherin antibody ( blue ) ( n = 3 mice from 3 independent experiments ) . Note that Angpt1 expression in LepR+ cells is much easier to see in sections from GFP mice than in antibody stained sections . ( D and E ) Representative femur sections from Angpt1GFP mice showed that GFP was not detectably expressed by Osteopontin+ ( red ) osteoblasts in the diaphyseal ( D ) or metaphyseal ( E ) regions ( n = 3 mice from 3 independent experiments ) . ( F ) In the bone marrow stroma from Leprcre; tdTomato; Angpt1GFP mice , nearly all LepR+ cells were positive for GFP , and vice versa . Data represent mean ± s . d . from 4 mice from 3 independent experiments . ( G ) In the bone marrow stroma from Angpt1GFP mice , nearly all PDGFRα+ cells were positive for GFP , and vice versa . Data represent mean ± s . d . from 4 mice in 3 experiments . ( H ) Bone marrow CD45−Ter119−VE-cadherin+ endothelial cells did not express detectable GFP ( n = 3 mice from 3 independent experiments ) . ( I and J ) Angpt1 ( I ) and Tie2 ( J ) transcript expression levels by qRT-PCR of unfractionated bone marrow cells , HSCs , LSK cells , c-kit+ cells , EYFP+ cells from Leprcre; loxp-EYFP mice , CD41+ megakaryocytes , Col1a1*2 . 3-GFP+ osteoblasts , VE-cadherin+ bone marrow endothelial cells . All data represent mean ± s . d . from 3–8 mice/genotype in 3 independent experiments . Two-tailed Student's t-tests were used to assess statistical significance relative to unfractionated bone marrow cells ( *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 005 Consistent with prior studies ( Iwama et al . , 1993; Schnurch and Risau , 1993; Arai et al . , 2004; Kopp et al . , 2005 ) , qRT-PCR showed that the Angpt1 receptor , Tie2 , was expressed most prominently by endothelial cells ( 167-fold higher than WBM ) and HSCs ( 21-fold higher; Figure 2J ) . Tie2 protein was expressed by most c-kit+ HPCs and endothelial cells in normal adult bone marrow and after irradiation ( Figure 6—figure supplement 2 ) . To study Angpt1 function under physiological conditions in adult bone marrow , we generated a floxed allele of Angpt1 ( Angpt1fl ) ( Figure 3—figure supplement 1A–C ) and an Angpt1 deficient allele ( Angpt1− ) by recombining Angpt1fl in the germline using CMV-cre . Consistent with the embryonic lethal phenotype of two independent Angpt1 null alleles that were previously described ( Suri et al . , 1996; Jeansson et al . , 2011 ) , the mating of Angpt1+/− heterozygous mice did not lead to the birth of any Angpt1−/− pups ( Figure 3—figure supplement 1D ) . Angpt1−/− embryos were found dead when timed pregnancies were examined at E12 . 5 ( data not shown ) . Angpt1 transcripts could not be detected in the fetal livers of Angpt1−/− mice ( Figure 3—figure supplement 1E ) . Thus , germline recombination of the Angpt1fl allele leads to a severe loss of Angpt1 function . We conditionally deleted Angpt1 from osteoblasts using Col1a1*2 . 3-cre; Angpt1fl/fl mice . Col1a1*2 . 3-cre recombines efficiently in fetal and postnatal osteoblasts ( Liu et al . , 2004; Ding et al . , 2012 ) . Col1a1*2 . 3-cre deleted 94 ± 3 . 0% of Angpt1fl alleles in Col1a1*2 . 3-GFP+ osteoblasts from Col1a1*2 . 3-cre; Angpt1fl/fl; Col1a1*2 . 3-GFP mice ( Figure 3—figure supplement 2A ) . Ten to 13 week-old adult Col1a1*2 . 3-cre; Angpt1fl/fl mice had normal blood cell counts ( Figure 3—figure supplement 2B ) , normal lineage composition in bone marrow , spleen and thymus ( Figure 3—figure supplement 2C ) , and normal cellularity in the bone marrow , spleen and thymus ( Figure 3A and Figure 3—figure supplement 2D ) . CD150+CD48−LSK HSC frequency was normal in the bone marrow and spleens of Col1a1*2 . 3-cre; Angpt1fl/fl mice relative to littermate controls ( Figure 3B ) . Col1a1*2 . 3-cre; Angpt1fl/fl bone marrow cells also had normal frequencies of CLPs ( Figure 3—figure supplement 2E ) , colony-forming progenitors in culture ( Figure 3C ) and dividing HSCs that incorporated BrdU ( 5'-bromo-2'-deoxyuridine ) over 10 days in vivo ( Figure 3D ) . Col1a1*2 . 3-cre; Angpt1fl/fl and control bone marrow cells gave rise to similar levels of long-term multilineage reconstitution upon transplantation into irradiated mice ( Figure 3E ) . 10 . 7554/eLife . 05521 . 006Figure 3 . Angpt1 was dispensable for HSC maintenance and hematopoiesis . ( A–E ) Deletion of Angpt1 from osteoblasts using Col1a1*2 . 3-cre did not significantly affect bone marrow or spleen cellularity ( A , n = 3 mice/genotype from 3 independent experiments ) , HSC frequency ( B , n = 3 mice/genotype from 3 independent experiments ) , colony-forming progenitor frequency in bone marrow ( C , n = 3 mice/genotype 3 independent experiments ) , incorporation of a 10-day pulse of BrdU by HSCs ( D , n = 3 pairs of male mice and 3 pairs of female mice/genotype ) , or reconstituting capacity of bone marrow cells in a competitive reconstitution assay ( E , n = 14–15 recipient mice/genotype from 3 independent experiments ) . ( F–J ) Leprcre; Angpt1GFP/fl mice had normal bone marrow and spleen cellularity ( F , n = 4 mice/genotype from 4 independent experiments ) , HSC frequency in bone marrow and spleen ( G , n = 5–6 mice/genotype from 5 independent experiments ) , colony-forming cell frequency in bone marrow ( H , n = 3 mice/genotype from 3 independent experiments ) , BrdU incorporation into HSCs ( I , n = 3 pairs of male mice and 3 pairs of female mice/genotype ) , and reconstituting capacity upon transplantation into irradiated mice ( J , n = 13 recipient mice/genotype from 3 independent experiments ) . Angpt1fl/fl and Angpt1GFP/fl mice ( lacking Cre ) were indistinguishable and were therefore pooled together as controls . ( K–O ) Mx1-cre; Angpt1fl/fl mice had normal bone marrow and spleen cellularity ( A , n = 3 mice/genotype ) , HSC frequency in bone marrow and spleen ( K , n = 3 mice/genotype ) , colony-forming cell frequency in bone marrow ( L , n = 6 mice/genotype from 4 independent experiments ) , BrdU incorporation into HSCs ( M , n = 3 pairs of male mice and 3 pairs of female mice/genotype ) , and reconstituting capacity upon transplantation into irradiated mice ( N , n = 10–14 recipient mice/genotype from 3 independent experiments ) . ( P–S ) Global deletion of Angpt1 in adult mice using UBC-Cre/ER ( 2–5 months after tamoxifen treatment ) did not significantly affect cellularity in the bone marrow or spleen ( P , n = 9–11 mice/genotype from 7 independent experiments ) , HSC frequency in the bone marrow ( Q , n = 9–11 mice/genotype from 7 independent experiments ) , colony-forming progenitor frequency in bone marrow ( R , n = 5 mice/genotype from 3 independent experiments ) , or reconstituting capacity of bone marrow cells upon transplantation into irradiated mice ( S , n = 13–14 recipient mice/genotype from 3 independent experiments ) . Two-tailed Student's t-tests were used to assess statistical significance . See Figure 3—figure supplement 2 for data on recombination efficiency . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 00610 . 7554/eLife . 05521 . 007Figure 3—figure supplement 1 . Generation of Angpt1fl mice . ( A ) Targeting strategy for the generation of the Angpt1fl allele . A BAC clone containing the Angpt1 genomic region was used to generate the targeting vector by recombineering . An Frt-Neo-Frt-loxp cassette was inserted 5′ of exon1 and a loxp site was inserted 3′ of exon1 . Sequence conservation among species was examined to avoid disrupting conserved intronic regulatory elements . Upon Cre mediated recombination , exon1 containing the translational start codon was excised , leading to the loss of the 5′ UTR and the first 99 amino acids of Angpt1 . Linearized targeting vector was electroporated into Bruce4 ES cells . ( B ) Southern blotting identified correctly targeted ES clones using the probes indicated in panel E . These ES cells were injected into blastomeres to generate chimeric mice . Chimeric mice were bred with C57BL/6-Tyrc-2J mice to obtain germline transmission . These mice were then bred with Flpe mice to remove the Neo cassette . ( C ) PCR genotyping confirmed germline transmission of the Angpt1fl allele . ( D ) A predicted null allele of Angpt1 ( Angpt1− ) was generated by mating Angpt1fl/+ mice with CMV-cre mice . No live Angpt1−/− mice were born from Angpt1−/+ × Angpt1−/+ matings . Consistent with the reported E12 . 5 lethal phenotype of Angpt1 null mice ( Suri et al . , 1996; Jeansson et al . , 2011 ) , Angpt1−/− mice were found dead at E12 . 5 . ( E ) qRT-PCR analysis confirmed the absence of Angpt1 mRNA in cells from Angpt1−/− fetal livers ( n = 3–5 mice/genotype ) . Data represent mean ± s . d . Two-tailed Student's t-tests were used to assess statistical significance . *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 00710 . 7554/eLife . 05521 . 008Figure 3—figure supplement 2 . Deletion of Angpt1 did not significantly affect blood cell counts . ( A , F , K ) Col1a1*2 . 3-cre , Leprcre and UBC-Cre/ER efficiently recombined Angpt1fl alleles in Col2 . 3-GFP+ osteoblasts from Col1a1*2 . 3-cre; Ang1fl/fl; Col2 . 3-GFP mice ( A ) , LepR+ cells from Leprcre; Angpt1fl/GFP mice ( F ) and LSK cells and LepR+ cells from UBC-cre/ER; Ang1fl/fl mice ( K ) . The recombination efficiency of Angpt1fl was measured by real-time PCR analysis of genomic DNA from flow cytometrically isolated cells . The amplification of the recombined allele in Col1a1*2 . 3-cre; Ang1fl/fl; Col2 . 3-GFP cells or UBC-cre/ER; Ang1fl/fl cells was compared to the amplification of the same product from Angpt1fl/fl cells . An unrelated genomic locus was amplified in parallel to normalize DNA content . The amplification of the recombined allele in Leprcre; Angpt1fl/GFP cells was compared to the amplification of the same product from Angpt1−/+ cells ( germline heterozygous for the recombined allele ) ( F; n = 3 mice/genotype from 3 independent experiments ) . ( H ) Genotyping of hematopoietic colonies formed by individual HSCs from Mx1-cre; Angpt1fl/fl mice showed efficient recombination of the Angpt1fl allele . Overall , 65 of 66 colonies examined ( >98% ) exhibited complete recombination of the Angpt1fl allele ( * , a single clone that was not recombined ) . ( B , G , I , L ) Normal white blood cell , red blood cell , and platelet counts in young adult Col1a1*2 . 3-cre; Angpt1fl/fl mice ( A; n = 3 mice/genotype from 3 independent experiments ) , Leprcre; Angpt1fl/fl or Leprcre; Angpt1fl/GFP mice ( C , n = 6–7 from 5 independent experiments ) , Mx1-cre; Angpt1fl/fl mice ( E , n = 3 mice/genotype from 3 independent experiments ) and UBC-cre/ER; Angpt1fl/fl mice 2–5 months after tamoxifen treatment ( F , n = 7–8 mice/genotype from 6 independent experiments ) . ( C–E ) Col1a1*2 . 3-cre; Ang1fl/fl mice had normal frequencies of CD4+ and/or CD8+ T cells in the thymus ( C ) , thymus cellularity ( D ) and CLP frequency in the bone marrow ( E ) ( n = 3 mice from 3 independent experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 008 To further test whether osteolineage progenitors are a source of Angpt1 for HSC maintenance , we deleted Angpt1 using Osx-Cre ( Sp7-Cre ) ( Rodda and McMahon , 2006 ) . Osx-Cre recombined 93% of Angpt1fl alleles from CD105+PDGFRα+CD45−Ter119−CD31− osteoprogenitors ( Park et al . , 2012 ) in Osx-cre; Angpt1fl/fl mice ( Figure 4A ) . Osx-cre; Angpt1fl/fl mice also had normal blood cell counts ( Figure 4B ) , normal cellularity , and HSC frequency in the bone marrow and spleen ( Figure 4C , D ) , normal frequencies of colony-forming progenitors in culture ( Figure 4E ) and normal levels of long-term multilineage reconstitution upon transplanting bone marrow cells into irradiated mice ( Figure 4F ) . Angpt1 from osteoblasts and their restricted progenitors are thus not required for hematopoiesis or HSC maintenance in normal adult mice . 10 . 7554/eLife . 05521 . 009Figure 4 . Angpt1 from osteoblast progenitors or Nestin-Cre-expressing cells is dispensable for HSC maintenance and hematopoiesis . ( A ) Osx-Cre recombined 93 ± 3 . 0% of Angpt1fl alleles in CD105+PDGFRα+CD45−Ter119−CD31− osteoprogenitors from Osx-cre; Angpt1fl/fl mice . Recombination efficiency was measured as described in Figure 3—figure supplement 2A ( n = 3 mice from 3 independent experiments ) . ( B–F ) Osx-cre; Angpt1fl/fl mice had normal blood cell counts ( B , n = 6 mice/genotype from 3 independent experiments ) , bone marrow and spleen cellularity ( C , n = 6–7 mice/genotype from 6 independent experiments ) , HSC frequency in bone marrow and spleen ( D , n = 4 mice/genotype from 4 independent experiments ) , colony-forming cell frequency in bone marrow ( E , n = 5 mice/genotype from 5 independent experiments ) , and reconstituting capacity upon transplantation into irradiated mice ( F , n = 23–24 recipient mice/genotype from 5 independent experiments ) . ( G–K ) Young adult Nestin-cre; Angpt1fl/fl mice had normal white blood cell counts , red blood cell counts , and platelet counts ( G , n = 4 mice/genotype from 3 independent experiments ) , bone marrow and spleen cellularity ( G , n = 4 mice/genotype from 3 independent experiments ) , HSC frequency ( I , n = 4 mice/genotype from 3 independent experiments ) , colony-forming progenitor frequency in the bone marrow ( J , n = 3 mice/genotype ) , and competitive reconstituting capacity upon transplantation into irradiated mice ( K , n = 9–10 recipient mice/genotype from 2 independent experiments ) . Two-tailed Student's t-tests were used to assess statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 009 To test whether Angpt1 from LepR+ stromal cells is required for HSC maintenance we generated Leprcre; Angpt1fl/fl or Leprcre; Angpt1fl/GFP mice . Lepr-Cre deleted 91% of Angpt1fl alleles in LepR+ bone marrow stromal cells from Leprcre; Angpt1fl/GFP mice ( Figure 3—figure supplement 2F ) . At 8 to 13 weeks of age , Leprcre; Angpt1fl/GFP mice had normal blood cell counts ( Figure 3—figure supplement 2G ) , normal lineage composition in bone marrow and spleen ( data not shown ) , and normal cellularity in the bone marrow and spleen ( Figure 3F ) . They also had normal frequencies of CD150+CD48−LSK HSCs in the bone marrow and spleen ( Figure 3G ) , colony-forming progenitors in bone marrow ( Figure 3H ) , and dividing HSCs ( Figure 3I ) . Leprcre; Angpt1fl/GFP and control bone marrow cells gave rise to similar levels of long-term multilineage reconstitution upon transplantation into irradiated mice ( Figure 3J ) . Angpt1 from LepR+ stromal cells is thus not required for hematopoiesis , HSC maintenance , or HSC quiescence in normal adult mice . Similar results were obtained from adult Nestin-cre; Angpt1fl/fl mice ( Figure 4G–K ) . To test if Angpt1 expressed by hematopoietic cells regulates HSC function we generated Mx1-cre; Angpt1fl/fl mice . pIpC ( polyinosinic-polycytidylic acid ) was administered to mice at 2 months of age then the mice were examined 3 months later . We observed complete recombination in 98% of colonies formed by HSCs in culture ( Figure 3—figure supplement 2H ) . Mx1-cre; Angpt1fl/fl mice had normal blood cell counts ( Figure 3—figure supplement 2I ) , normal lineage composition in bone marrow and spleen ( data not shown ) , and normal bone marrow and spleen cellularity ( Figure 3K ) . Mx1-cre; Angpt1fl/fl mice also had normal frequencies of CD150+CD48−LSK HSCs in the bone marrow and spleen ( Figure 3L ) , colony-forming progenitors in bone marrow ( Figure 3M ) , and dividing HSCs that incorporated BrdU during a 10-day pulse ( Figure 3N ) . Mx1-cre; Angpt1fl/fl and control bone marrow cells gave rise to similar levels of long-term multilineage reconstitution upon transplantation into irradiated mice ( Figure 3O ) . Similar results were obtained when we conditionally deleted Angpt1 from fetal hematopoietic and endothelial cells by generating Tie2-cre; Angpt1fl/fl mice ( Figure 5A–G ) and when we conditionally deleted Angpt1 from megakaryocytes by generating Pf4-cre; Angpt1fl/fl mice ( Figure 5H–K ) . Angpt1 from endothelial cells and hematopoietic cells , including megakaryocytes , is thus not required for hematopoiesis , HSC maintenance , or HSC quiescence in normal adult mice . 10 . 7554/eLife . 05521 . 010Figure 5 . Angpt1 from endothelial cells or megakaryocytes is dispensable for HSC maintenance and hematopoiesis . ( A ) Tie2-Cre recombined 97 ± 0 . 4% of Angpt1fl alleles in CD45+/Ter119+ hematopoietic cells and 97 ± 0 . 6% in VE-Cadherin+ endothelial cells from Tie2-cre; Angpt1fl/fl mice ( measured as described in Figure 3—figure supplement 2A; n = 3 mice from 3 independent experiments ) . ( B–G ) Tie2-cre; Angpt1fl/fl mice had normal blood counts ( B , n = 3–6 from 3 independent experiments ) , bone marrow and spleen cellularity ( C , n = 5–10 mice/genotype from 4 independent experiments ) , HSC frequency in bone marrow and spleen ( D , n = 5–10 mice/genotype from 4 independent experiments ) , colony-forming cell frequency in bone marrow ( E , n = 3 mice/genotype from 3 independent experiments ) , vascular density and morphology ( F , n = 3 mice/genotype from 3 independent experiments ) and reconstituting capacity upon transplantation into irradiated mice ( F , n = 8 recipient mice/genotype from 2 independent experiments ) . All data represent mean ± s . d . Two-tailed Student's t-tests were used to assess statistical significance . ( H–K ) Pf4-cre; Angpt1fl/fl mice had normal bone marrow and spleen cellularity ( H , n = 5 mice/genotype from 4 independent experiments ) , HSC frequency in bone marrow and spleen ( I , n = 5 mice/genotype from 4 independent experiments ) , colony-forming cell frequency in bone marrow ( J , n = 5 mice/genotype from 5 independent experiments ) and reconstituting capacity upon transplantation into irradiated mice ( K , n = 14–15 recipient mice/genotype from 3 independent experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 010 To globally delete Angpt1 we generated UBC-cre/ER; Angpt1fl/fl mice . UBC-Cre/ER ubiquitously recombines in adult mice upon tamoxifen administration ( Ruzankina et al . , 2007 ) . We administrated tamoxifen-containing chow to 8-week old UBC-cre/ER; Angpt1fl/fl mice as well as littermate controls for 2–5 months then sacrificed them for analysis . UBC-cre/ER recombined 95% of Angpt1fl alleles in LSK cells and 96% of Angpt1fl alleles in LepR+ cells in the bone marrow of UBC-cre/ER; Angpt1fl/fl mice ( Figure 3—figure supplement 2K ) . UBC-cre/ER; Angpt1fl/fl mice had normal blood cell counts ( Figure 3—figure supplement 2L ) , normal bone marrow and spleen hematopoietic lineage composition ( data not shown ) and normal bone marrow and spleen cellularity relative to controls ( Figure 3P ) . UBC-cre/ER; Angpt1fl/fl mice also had normal frequencies of CD150+CD48−LSK HSCs in the bone marrow and spleen ( Figure 3Q ) , colony-forming progenitors in bone marrow ( Figure 3R ) , and long-term multilineage reconstituting bone marrow cells upon transplantation into irradiated mice ( Figure 3S ) . Angpt1 is thus dispensable for hematopoiesis and for the maintenance and function of HSCs in normal adult mice . We next tested whether Angpt1 regulates the recovery of hematopoiesis after irradiation . Since hematopoietic cells ( HSCs , c-kit+ HPCs , and megakaryocytes ) and LepR+ stromal cells were the major sources of Angpt1 in the bone marrow ( Figures 1 , 2 ) we reconstituted irradiated Leprcre; Angpt1fl/GFP or control recipients by transplanting 1 × 106 mechanically dissociated bone marrow cells from Mx1-cre; Angpt1fl/fl or control donors 1 month after pIpC treatment . This allowed us to test whether Angpt1 from LepR+ stromal cells and/or hematopoietic cells influenced the regeneration of hematopoiesis after irradiation . We did not detect significant changes in the patterns of Angpt1 , Tie2 , or Angpt2 expression in the bone marrow after irradiation and bone marrow transplantation ( Figure 6—figure supplement 2 ) . In both normal adult bone marrow , and after irradiation and transplantation , Tie2 was expressed primarily by endothelial cells and c-kit+ HPCs ( Figure 6—figure supplement 2G ) while Angpt2 was expressed primarily by endothelial cells . As expected , non-irradiated adult Leprcre; Angpt1fl/GFP mice , Mx1-cre; Angpt1fl/fl mice , and Leprcre; Mx1-cre; Angpt1fl/GFP mice all had normal bone marrow cellularity ( Figure 6A ) , normal numbers of LSK cells in the bone marrow ( Figure 6B ) , and normal CD150+CD48−LSK HSC frequency ( Figure 6C ) . However , at 8 and 12 days after irradiation , Leprcre; Angpt1fl/GFP mice that had been transplanted with Mx1-cre; Angpt1fl/fl bone marrow cells , and to a significantly lesser extent Leprcre; Angpt1fl/GFP mice that had been transplanted with wild-type bone marrow cells , exhibited significantly higher bone marrow cellularity ( Figure 6A ) and significantly higher numbers of LSK cells in the bone marrow ( Figure 6B ) as compared to wild-type mice transplanted with wild-type bone marrow cells . At 16 days after irradiation , most of these differences persisted but by 28 days after irradiation mice in all treatments had similar bone marrow cellularities and LSK ( Lineage−Sca1+c-kit+ cells ) numbers ( Figure 6A , B ) . The accelerated recovery of HPCs and hematopoiesis in the absence of Angpt1 was also evident in white blood cell counts and in the numbers of myeloid , lymphoid , and erythroid cells in the bone marrow ( Figure 6—figure supplement 1 ) . Angpt1 expression by hematopoietic cells and LepR+ stromal cells therefore negatively regulates the recovery of hematopoiesis after irradiation . 10 . 7554/eLife . 05521 . 011Figure 6 . Angpt1 deficiency in hematopoietic stem/progenitor cells and LepR+ stromal cells accelerated the recovery of HSCs and hematopoiesis after irradiation . One million bone marrow cells from Angpt1fl/fl ( Cre− ) or Mx1-cre; Angpt1fl/fl ( Mx1 ) mice were transplanted into irradiated Angpt1fl/GFP or Angpt1GFP ( Cre− ) or Leprcre; Angpt1fl/GFP ( Lepr ) mice ( all panels reflect mean ± s . d . from 6–11 mice/genotype/time point from 5 independent experiments ) . Bone marrow cellularity ( A ) and LSK cell numbers ( B ) were analyzed at the indicated time points after irradiation and transplantation , always in two femurs and two tibias per mouse . ( C ) Mx1-cre; Leprcre; Angpt1fl/GFP mice had a normal frequency of CD150+CD48−LSK HSCs in the bone marrow as compared to control ( Angpt1fl/fl or Angpt1fl/GFP ) , Mx1-cre; Angpt1fl/fl , or Leprcre; Angpt1fl/GFP mice . ( n = 6 mice/genotype from 5 independent experiments ) . ( D ) Competitive long-term multilineage reconstitution assay in which 1 . 5 × 106 donor bone marrow cells from the indicated primary recipient mice 12 days after irradiation were transplanted along with 3 × 105 recipient bone marrow cells into irradiated secondary recipient mice . The recipient cells were previously-transplanted compromised bone marrow cells . ( n = 11–15 recipient mice/genotype from 3 independent experiments ) Two-tailed Student's t-tests were used to assess statistical significance ( * or #p < 0 . 05 , ** or ##p < 0 . 01 , *** or ###p < 0 . 001 ) . * indicates statistical significance relative to Cre− control cells . # indicates statistical significance relative to Mx1 cells . ( E and F ) 4000 LSK cells from Angpt1fl/fl ( Cre− ) or Mx1-cre; Angpt1fl/fl ( Mx1 ) mice were transplanted into irradiated Leprcre; Angpt1fl/GFP ( Lepr ) mice . Bone marrow cellularity ( E ) and LSK cell number in the bone marrow ( F ) were analyzed 14 and 28 days after irradiation and bone marrow transplantation . Data represent mean ± s . d . from 4 mice/genotype/time point from 3 independent experiments . Two-tailed Student's t-tests were used to assess statistical significance ( *p < 0 . 05 ) . ( G and H ) One million bone marrow cells from Angpt1fl/fl ( Cre− ) or Pf4-cre; Angpt1fl/fl ( Pf4 ) mice were transplanted into irradiated Leprcre; Angpt1fl/GFP ( Lepr ) mice . Bone marrow cellularity ( G ) and LSK cell number in the bone marrow ( H ) were analyzed at 8 , 12 , 16 , and 28 days after irradiation and transplantation . Data represent mean ± s . d . from 4 mice/genotype/time point from 3 independent experiments . Two-tailed Student's t-tests were used to assess statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 01110 . 7554/eLife . 05521 . 012Figure 6—figure supplement 1 . Angpt1 deletion accelerated hematopoietic recovery after irradiation . One million of bone marrow cells from control ( Cre- ) or Mx1-cre; Angpt1fl/fl ( Mx1 ) mice were transplanted into lethally irradiated control ( Cre- ) or Leprcre; Angpt1fl/GFP ( Lepr ) mice . White blood count ( A ) , red blood count ( B ) , platelet count ( C ) , Gr-1+Mac1+ myeloid cell number ( D ) , B220+ B cell number ( E ) , CD3+ T cell number ( F ) and Ter119+ erythroid cell number ( G ) from two femurs and two tibias . Data represent mean ± s . d . from 6–11 mice/genotype/time point from 5 independent experiments . Two-tailed Student's t-tests were used to assess statistical significance ( * or #p < 0 . 05 , ** or ##p < 0 . 01 , *** or ###p < 0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 01210 . 7554/eLife . 05521 . 013Figure 6—figure supplement 2 . Angpt1 , Tie2 , and Angpt2 expression patterns were similar in adult bone marrow before and after irradiation . One million of WBMs from Angpt1GFP mice were transplanted into lethally irradiated Angpt1GFP mice , which were then analyzed at 8 , 12 , 16 and 28 days after irradiation and transplantation ( n = 3 donor and 3 recipient mice from 3 independent experiments ) . ( A ) Flow cytometric analysis of mechanically dissociated bone marrow cells ( containing hematopoietic cells but not stromal cells ) showed that most GFP+ bone marrow cells were c-kit+ before and after irradiation . ( B and C ) Flow cytometric analysis of enzymatically dissociated bone marrow showed that nearly all LepR+ stromal cells were GFP+ , and vice versa , before and after irradiation . ( D ) VE-Cadherin+ endothelial cells from enzymatically dissociated bone marrow did not express GFP before or after irradiation . ( E and F ) VE-Cadherin+ endothelial cells ( E ) and LSK cells ( F ) uniformly expressed Tie2 before and after irradiation . Day 8 after irradiation was not included in ( F ) because of few LSK cells . ( G ) Percentage of Tie2+ cells among WBMs , c-kit− hematopoietic cells , c-kit+ hematopoietic cells , LSK cells , VE-Cadherin+CD45−Ter119− endothelial cells ( Endo ) and LepR+ perivascular stromal cells ( Lepr+ ) before and after irradiation . ( H ) Representative femur sections showing that GFP was not detectably expressed by NG2+ osteoblasts before or after irradiation . ( I ) Angpt2 transcript expression levels by qRT-PCR of unfractionated bone marrow cells and VE-cadherin+ bone marrow endothelial cells before and after irradiation . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 013 We tested whether Angpt1 also influenced the expansion of HSC numbers after irradiation by transplanting WBMs from mice in each of the treatments described above at 12 days after transplantation . Bone marrow cells from Leprcre; Angpt1fl/GFP mice transplanted with Mx1-cre; Angpt1fl/fl bone marrow gave significantly higher levels of donor cell reconstitution in all lineages as compared to bone marrow cells from wild-type mice transplanted with wild-type bone marrow ( Figure 6D ) . To a lesser extent , bone marrow cells from Leprcre; Angpt1fl/GFP mice that had been transplanted with wild-type bone marrow also gave significantly higher levels of donor cell reconstitution as compared to bone marrow cells from wild-type mice transplanted with wild-type bone marrow ( Figure 6D ) . The expansion in HSC numbers during reconstitution is thus negatively regulated by Angpt1 expressed by hematopoietic cells and LepR+ stromal cells . Since WBMs could potentially contain Angpt1-expressing stromal cells in addition to hematopoietic stem/progenitor cells and megakaryocytes we undertook a series of additional experiments to test whether hematopoietic stem/progenitor cells are a functionally important source of Angpt1 for hematopoietic regeneration . First , we transplanted 4000 LSK ( Lineage−Sca-1+c-kit+ ) cells from control or Mx1-cre; Angpt1fl/fl mice into Leprcre; Angpt1GFP/fl mice to test the effects of HPCs uncontaminated by stromal cells on hematopoietic regeneration after irradiation . We found that the mice transplanted with Angpt1 deficient LSK cells had significantly higher bone marrow cellularity ( Figure 6E ) and LSK cell numbers ( Figure 6F ) than mice transplanted with control LSK cells at 14 days after irradiation . These data prove that Angpt1 expression by hematopoietic cells regulates hematopoietic recovery after irradiation . The only hematopoietic cells other than c-kit+ hematopoietic stem and progenitor cells that express Angpt1 are megakaryocytes ( Figure 1D–F , J–L ) . To test whether Angpt1 expression by megakaryocytes contributes to the regulation of hematopoietic recovery we conditionally deleted Angpt1 from megakaryocyte lineage cells using Pf4-Cre and transplanted WBMs from control and Pf4-cre; Angpt1fl/fl mice into Leprcre; Angpt1GFP/fl recipients . We did not detect any significant differences in hematopoietic recovery between mice transplanted with control vs Pf4-cre; Ang1fl/fl bone marrow ( Figure 6G , H ) . These data indicate that Angpt1 expression by megakaryocyte lineage cells has little effect on hematopoietic recovery after irradiation . Together , our data demonstrate that Angpt1 expression by hematopoietic stem and progenitor cells and LepR+ stromal cells regulate hematopoietic recovery after irradiation . Consistent with a prior study ( Jeansson et al . , 2011 ) , loss of Angpt1 expression in the bone marrow had no detectable effect on the bone marrow vasculature in normal young adult mice . Deletion of Angpt1 from hematopoietic cells in Mx1-cre; Angpt1fl/fl mice , or perivascular stromal cells in Leprcre; Angpt1fl/GFP mice , or both in Mx1-cre; Leprcre; Angpt1fl/GFP mice , did not affect the numbers of VE-cadherin+ endothelial cells or LepR+ perivascular cells in the bone marrow ( Figure 7A , B ) , or the density or morphology of the vasculature in bone marrow relative to control mice ( Figure 7—figure supplement 1A ) . 10 . 7554/eLife . 05521 . 014Figure 7 . Angpt1 deficiency in hematopoietic stem/progenitor cells and LepR+ stromal cells increased endothelial cell proliferation and accelerated the recovery of vascular morphology after irradiation . ( A and B ) Deletion of Angpt1 from hematopoietic cells ( Mx1 ) , LepR+ stromal cells ( Lepr ) , or both ( Lepr and Mx1 ) did not significantly affect the number of VE-cadherin+ endothelial cells ( A ) or LepR+ perivascular stromal cells ( B ) in the bone marrow of normal young adult mice . Cell number in enzymatically dissociated bone marrow cells was determined in 2 pairs of femurs and tibias per mouse ( n = 3 mice/genotype from 3 independent experiments ) . ( C and D ) Representative images showing normal ( C ) and regressed ( D ) sinusoids in transverse femur sections . Regressed sinusoids were distinguished from non-regressed sinusoids by being dilated and having few hematopoietic cells around them . ( E–K ) One million bone marrow cells from Angpt1fl/fl ( Cre− ) or Mx1-cre; Angpt1fl/fl ( Mx1 ) mice were transplanted into irradiated Angpt1fl/GFP or Angpt1GFP ( Cre− ) or Leprcre; Angpt1fl/GFP ( Lepr ) mice . Three-dimensional reconstructions of 50 μm thick sections of femoral bone marrow stained with anti-VE-cadherin antibody revealed the regression and regeneration of blood vessels after irradiation . Representative images for control ( Cre− ) mice were taken at steady state ( E ) , 8 days ( F ) , 12 days ( G ) , 16 days ( H ) and 28 days ( I ) after irradiation and transplantation . Representative images for Mx1 → Cre− , Cre− → Lepr and Mx1 → Lepr mice were taken 12 days after irradiation ( G ) . ( J ) The percentage of regressed sinusoids in sections through the bone marrow . Data represent mean ± s . d . from 5–6 mice/genotype/time point from 4 independent experiments . ( K ) Incorporation of a 24-hr pulse of BrdU into VE-cadherin+ endothelial cells ( mean ± s . d . from 3–4 mice/genotype/time from 3 experiments ) . Two-tailed Student's t-tests were used to assess statistical significance ( * or #p < 0 . 05; ** or ##p < 0 . 01; *** or ###p < 0 . 001 ) . ( L and M ) 4000 LSK cells from Angpt1fl/fl ( Cre− ) or Mx1-cre; Angpt1fl/fl ( Mx1 ) mice were transplanted into irradiated Leprcre; Angpt1fl/GFP ( Lepr ) mice . Vascular morphology ( M ) and the percentage of regressed sinusoids ( L ) were analyzed at the indicated time points . ( N and O ) One million bone marrow cells from Angpt1fl/fl ( Cre− ) or Pf4-cre; Angpt1fl/fl ( Pf4 ) mice were transplanted into irradiated Leprcre; Angpt1fl/GFP ( Lepr ) mice . Vascular morphology ( O ) and the percentage of regressed sinusoids ( N ) were analyzed at the indicated time points . Two-tailed Student's t-tests were used to assess statistical significance ( *p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 01410 . 7554/eLife . 05521 . 015Figure 7—figure supplement 1 . Angpt1 deficiency accelerated the recovery of vascular morphology after irradiation . Representative confocal images of VE-cadherin ( red ) and DAPI ( blue ) stained bone marrow sections from control ( Cre- ) or Mx1-cre; Angpt1fl/fl ( Mx1 ) mice under normal conditions ( A ) , or at 8 ( B ) , 16 ( C ) , or 28 ( D ) day after transplantation of bone marrow from control ( Cre- ) or Leprcre; Angpt1fl/GFP ( Lepr ) mice ( n = 3–4 mice/genotype/time point in 4 independent experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 015 Irradiation induces vascular regression followed by regeneration from surviving endothelial cells ( Heissig et al . , 2005; Kopp et al . , 2005; Li et al . , 2008; Hooper et al . , 2009 ) . Consistent with this we observed dilated regressed sinusoids throughout the bone marrow 8 days after irradiation and bone marrow transplantation ( compare Figure 7C–F , J ) . Few hematopoietic cells clustered around these regressed sinusoids relative to normal bone marrow ( compare Figure 7C–F ) . Morphologically normal , hematopoietic cell-invested sinusoids were evident in some areas of the bone marrow 12 days after transplantation ( Figure 7G , J ) , and their frequency increased 16 days after transplantation ( Figure 7H , J ) . By day 28 , regressed vessels were no longer observed in the bone marrow in any treatment ( Figure 7I , J ) . When Mx1-cre; Angpt1fl/fl bone marrow cells were transplanted into wild-type recipients ( Mx1 → Cre− ) , vascular recovery was indistinguishable from control mice ( Cre− → Cre− ) ( Figure 7G , J and Figure 7—figure supplement 1 ) . However , when wild-type bone marrow cells were transplanted into Leprcre; Angpt1fl/GFP recipients ( Cre− → Lepr ) we observed accelerated morphological recovery of the vasculature , with significantly fewer dilated regressed vessels at 12 and 16 days after transplantation relative to control mice ( Figure 7G , J and Figure 7—figure supplement 1C ) . The accelerated recovery was significantly more pronounced when we transplanted Mx1-cre; Angpt1fl/fl bone marrow cells into Leprcre; Angpt1fl/GFP recipients ( Mx1 → Lepr ) . These mice exhibited significantly fewer regressed vessels at 12 and 16 days after transplantation relative to control mice and Cre− → Lepr mice ( Figure 7G , J and Figure 7—figure supplement 1C ) . By 28 days after irradiation mice in all treatments had reacquired morphologically normal bone marrow vasculature ( Figure 7I , J and Figure 7—figure supplement 1D ) . These data indicate that Angpt1 produced by hematopoietic cells and LepR+ stromal cells slows the morphological recovery of blood vessels after irradiation . When combined with the observation that Angpt1 also slows the regeneration of HSCs ( Figure 6B , D ) and bone marrow hematopoiesis ( Figure 6A ) after irradiation , the data indicate that Angpt1 negatively regulates the regeneration of the HSC niche in bone marrow after irradiation . We transplanted 4000 LSK cells from control and Mx1-cre; Angpt1fl/fl mice into Leprcre; Angpt1GFP/fl mice to test the effects of HPCs uncontaminated by stromal cells on vascular regeneration after irradiation . We found that the mice transplanted with Angpt1 deficient LSK cells had significantly better vascular morphology in the bone marrow ( Figure 7L , M ) than mice transplanted with control LSK cells at 14 days after irradiation . To test whether Angpt1 expression by megakaryocytes contributed to the vascular recovery we transplanted WBMs from control and Pf4-cre; Angpt1fl/fl mice into Leprcre; Angpt1GFP/fl recipients . We did not detect any significant differences in vascular recovery between mice transplanted with control vs Pf4-cre; Ang1fl/fl bone marrow ( Figure 7N , O ) . Angpt1 expression by hematopoietic stem and progenitor cells and LepR+ stromal cells therefore regulate both hematopoietic and vascular recovery after irradiation . To investigate the cellular mechanism by which Angpt1 influences vascular recovery after irradiation we assessed the proliferation of bone marrow endothelial cells . In normal adult bone marrow few endothelial cells incorporated a 24-hr pulse of BrdU and deletion of Angpt1 from hematopoietic cells , LepR+ cells , or both did not influence this frequency ( Figure 7K ) . After irradiation and bone marrow transplantation , endothelial cells were recruited into cycle ( Figure 7K ) . Deletion of Angpt1 from hematopoietic cells and LepR+ cells significantly increased the frequency of dividing endothelial cells 12 days after transplantation ( Figure 7K ) . These data suggest that Angpt1 slows the recovery of the vasculature and the HSC niche partly by negatively regulating the proliferation of endothelial cells after irradiation . To test whether Angpt1 regulates vascular leakiness in the bone marrow we assessed Evans blue extravasation . Evans blue binds to serum albumin and can be used to trace macromolecule flux across blood vessels ( Radu and Chernoff , 2013 ) . In normal bone marrow we observed little Evans blue extravasation , irrespective of whether Angpt1 was deleted from hematopoietic cells , LepR+ cells , or both ( Figure 8A ) , suggesting that Angpt1 is dispensable for maintaining vascular integrity in normal adult bone marrow . In contrast , 12 days after irradiation we observed uniformly high levels of Evans blue extravasation in bone marrow from mice in all treatments ( Figure 8A and Figure 8—figure supplement 1A ) , consistent with the leakiness that would be expected from regenerating blood vessels ( Hooper et al . , 2009 ) . When morphological recovery of bone marrow vessels was complete 28 days after irradiation ( Figure 7J and Figure 7—figure supplement 1D ) , control mice ( Cre− → Cre− ) and wild-type mice transplanted with Angpt1 deficient bone marrow cells ( Mx1 → Cre− ) had largely re-established vascular integrity , with little Evans blue extravasation ( Figure 8A ) . Leprcre; Angpt1fl/GFP mice transplanted with wild-type bone marrow cells ( Cre− → Lepr ) showed a trend toward increased Evans blue extravasation but the effect was not statistically significant relative to control mice ( Figure 8A ) . In contrast , Leprcre; Angpt1fl/GFP mice transplanted with Angpt1 deficient bone marrow cells ( Mx1 → Lepr ) exhibited significantly higher levels of Evans blue extravasation at 28 days after transplantation ( Figure 8A ) . Leaky vasculature was not observed in the spleen ( Figure 8—figure supplement 1B ) . Angpt1 from hematopoietic cells and LepR+ stromal cells is thus required to promote vascular integrity in the bone marrow after regeneration at the expense of slowing endothelial cell proliferation and the morphological recovery of blood vessels , slowing the regeneration of the HSC niche . 10 . 7554/eLife . 05521 . 016Figure 8 . Angpt1 deficiency in hematopoietic stem/progenitor cells and LepR+ stromal cells increases the leakiness of regenerated blood vessels . One million bone marrow cells from Angpt1fl/fl ( Cre− ) or Mx1-cre; Angpt1fl/fl ( Mx1 ) mice were transplanted into lethally irradiated Angpt1GFP/fl or Angpt1GFP ( Cre− ) or Leprcre; Angpt1GFP/fl ( Lepr ) mice ( A , D , H ) ( n = 4 mice/genotype/time point from 3 independent experiments ) . 4000 LSK cells Angpt1fl/fl ( Cre− ) or Mx1-cre; Angpt1fl/fl ( Mx1 ) mice were transplanted into Leprcre; Angpt1GFP/fl mice ( Lepr ) ( B and E ) ( n = 4 mice/genotype/time point from 3 independent experiments ) . One million bone marrow cells from Angpt1fl/fl ( Cre− ) or Pf4-cre; Angpt1fl/fl ( Pf4 ) were transplanted into Leprcre; Angpt1GFP/fl ( Lepr ) mice ( C and F ) ( n = 4 mice/genotype/time point from 3 independent experiments ) . ( A–C ) Extravasation of intravenously injected Evans blue into femoral bone marrow at the indicated time points after irradiation and bone marrow transplantation . ( D–F ) Live imaging of calvarial bone marrow at 28 days after irradiation and bone marrow transplantation to assess dextran-FITC extravasation ( arrows ) . The mice were injected with dextran-FITC and anti-VE-cadherin antibody before microscopy . ( G ) Quantification of the number of pores larger than 100 nm in diameter per 50 µm2 of sinusoidal endothelium ( n = 3–5 mice/genotype from 3 independent experiments ) . Two-tailed Student's t-tests were used to assess statistical significance ( * or # , p < 0 . 05; ** or ## , p < 0 . 01; *** or ### , p < 0 . 001 ) . ( H ) Scanning electron microscopy of bone marrow sinusoids from Cre− → Cre− , Mx1 → Cre− , Cre− → Lepr , and Mx1 → Lepr mice at 28 days after irradiation . Arrows indicate pores greater than 100 nm in diameter in sinusoidal endothelium 28 days after irradiation . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 01610 . 7554/eLife . 05521 . 017Figure 8—figure supplement 1 . Angpt1 deficiency led to the persistence of pores and leakiness in blood vessels after irradiation . ( A ) Femoral bone marrow was heavily infiltrated by intravenously injected Evans blue that extravasated 12 days after irradiation . ( B ) Lack of extravasation of intravenously injected Evans blue into spleen from Cre− → Cre− , Mx1 → Cre− , Cre− → Lepr , and Mx1 → Lepr mice ( n = 4 mice/genotype/time point from 3 independent experiments ) . ( C and D ) Dextran-FITC was confined within VE-cadherin+ blood vessels in the calvarial bone marrow in normal adult mice ( B ) but extravasated out of vessels 12 days after irradiation ( C ) ( n = 3–4 mice from 3 independent experiments ) . ( E–H ) Scanning electron microscopy of sinusoidal ( D and E ) and arterial ( F and G ) blood vessels in normal adult bone marrow ( E and G ) or 15 days after irradiation ( F and H ) . We only detected pores in sinusoidal endothelium ( n = 3–6 mice from 3 independent experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 017 To independently assess vascular integrity , we performed live imaging of the vasculature in the calvarium bone marrow of mice intravenously administered anti-VE-cadherin antibody and dextran-FITC ( 70 kDa ) . In normal mice , dextran-FITC fluorescence was tightly restricted within VE-cadherin+ vessels ( Figure 8—figure supplement 1C ) . 12 days after irradiation , calvarium blood vessels became dilated and dextran-FITC leaked throughout the medullary cavity ( Figure 8—figure supplement 1D ) . At 28 days after transplantation , control mice ( Cre− → Cre− ) , wild-type mice transplanted with Angpt1 deficient hematopoietic cells ( Mx1 → Cre− ) , and Leprcre; Angpt1fl/GFP mice transplanted with wild-type bone marrow cells ( Cre− → Lepr ) all exhibited vascular integrity , with little discernible leakage of dextran-FITC ( Figure 8D ) . In contrast , Leprcre; Angpt1fl/GFP mice transplanted with Angpt1 deficient bone marrow cells ( Mx1 → Lepr ) exhibited leaky vessels in the calvarium with dextran-FITC infiltrating the bone marrow ( Figure 8D , see arrows ) . Thus , consistent with the Evans blue assay , Angpt1 from hematopoietic cells and LepR+ cells promotes bone marrow vascular integrity during regeneration . We transplanted 4000 LSK cells from control and Mx1-cre; Angpt1fl/fl mice into Leprcre; Angpt1GFP/fl mice to test the effects of HPCs uncontaminated by stromal cells on vascular integrity after irradiation . The mice transplanted with Angpt1 deficient , but not control , LSK cells exhibited vascular leakage in the bone marrow 28 days after irradiation , as evidenced by high level of Evans blue and Dextran-FITC extravasation ( Figure 8B , E ) . To test whether Angpt1 expression by megakaryocytes contributed to the vascular integrity we transplanted WBMs from control and Pf4-cre; Angpt1fl/fl mice into Leprcre; Angpt1GFP/fl recipients . We did not detect any significant differences in vascular leakiness between mice transplanted with control vs Pf4-cre; Ang1fl/fl bone marrow ( Figure 8C , F ) . Angpt1 expression by hematopoietic stem and progenitor cells and LepR+ stromal cells therefore promotes vascular integrity during regeneration after irradiation . We performed scanning electron microscopy to better understand the loss of blood vessel integrity after Angpt1 deletion . Normal bone marrow sinusoids were 5–20 µm in luminal diameter ( Figure 8—figure supplement 1E ) . They could be readily distinguished from bone marrow arterioles , which had thicker walls and a different morphology ( Figure 8—figure supplement 1G ) . At 8–16 days after irradiation , sinusoid diameter in the bone marrow increased ( compare Figure 8—figure supplement 1E , F ) and the endothelial lining was marked by small pores ( Figure 8—figure supplement 1F ) , consistent with the finding that irradiation causes discontinuities in bone marrow blood vessels ( Daldrup et al . , 1999 ) . At 28 days after irradiation , pores were rare in sinusoids from wild-type mice transplanted with wild-type marrow ( Cre− → Cre− ) , wild-type mice transplanted with Angpt1 deficient bone marrow ( Mx1 → Cre− ) , and Leprcre; Angpt1fl/GFP mice transplanted with wild-type bone marrow ( Cre− → Lepr ) ( Figure 8G , H ) , consistent with the integrity of vessels in these mice . In contrast , pores remained common in blood vessels in bone marrow from Leprcre; Angpt1fl/GFP mice transplanted with Angpt1 deficient bone marrow cells ( Mx1 → Lepr ) ( Figure 8H , see arrow; Figure 8G ) . These data suggest that in the absence of Angpt1 , vascular integrity is reduced in regenerating bone marrow blood vessels because of the persistence of pores or discontinuities among endothelial cells . To test whether the accelerated hematopoietic recovery in the absence of Angpt1 is caused by the increase in vascular leakiness we treated mice with cavtratin , an anti-permeability agent unrelated to Angpt1 function ( Gratton et al . , 2003 ) . We injected cavtratin ( 2 . 5 mg/kg/day i . p . ) into control ( control bone marrow transplanted into control mice ) and Mx1 → Lepr recipients from 7 to 13 days after irradiation then analyzed the mice 14 days after irradiation and bone marrow transplantation . Based on both Dextran-FITC live-imaging and Evans blue extravasation , cavtratin significantly reduced vascular leakiness in both control and Mx1 → Lepr recipient mice ( Figure 9A , B ) . However , cavtratin administration did not significantly affect the recovery of bone marrow cellularity or LSK cell numbers in the bone marrow of control or Mx1 → Lepr recipients ( Figure 9C , D ) . Mx1 → Lepr recipients continued to regenerate bone marrow cellularity and LSK cell numbers significantly faster than control mice , irrespective of cavtratin treatment . The accelerated recovery of hematopoietic stem/progenitor cells and hematopoiesis in the absence of Angpt1 is therefore not caused by increased vascular leakiness . These appear to reflect independent effects of Angpt1 . 10 . 7554/eLife . 05521 . 018Figure 9 . Vascular leakage does not promote hematopoietic regeneration in Angpt1 mutant mice . One million whole WBMs from control and Mx1-cre; Angpt1fl/fl mice were transplanted into irradiated control and Leprcre; Angpt1GFP/fl mice , respectively . Cavtratin was administered into control ( Cre− → Cre− ) , and mutant ( Mx1 → Lepr ) recipients from 7 to 13 days after irradiation and transplantation . 14 days after irradiation mice were analyzed for Dextran-FITC extravasation in calvarial bone marrow ( A , n = 3 mice/genotype from 3 independent experiments ) , Evans blue extravasation in femoral bone marrow ( B , n = 3 mice/genotype from 3 independent experiments ) , bone marrow cellularity , and LSK cell number in the bone marrow ( C and D , n = 4 mice/genotype from 3 independent experiments ) . Cell numbers reflect two femurs and two tibias . Two-tailed Student's t-tests were used to assess statistical significance ( *p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05521 . 018
Several gene-expression profiling studies are consistent with our finding that HSCs are a major source of Angpt1 ( Ivanova et al . , 2002; Akashi et al . , 2003; Forsberg et al . , 2010; Cabezas-Wallscheid et al . , 2014 ) . In addition , an Angpt1-LacZ knockin allele is expressed in an HSC-enriched population ( Takakura et al . , 2000 ) . Several prior studies have also demonstrated that perivascular stromal cells are a significant source of Angpt1 in the bone marrow ( Sacchetti et al . , 2007; Mendez-Ferrer et al . , 2010; Ding et al . , 2012 ) . In contrast , we have been unable to detect Angpt1 expression in osteoblasts in Angpt1GFP knock-in mice ( Figure 2D , E ) , by anti-Angpt1 antibody staining ( Figure 1A–C ) , by qRT-PCR on sorted cells ( Figure 2I ) , or by gene-expression profiling of osteoblasts ( data not shown ) . Megakaryocytes also express Angpt1 ( Figure 1J–L ) but unlike deletion in hematopoietic stem/progenitor cells and LepR+ stromal cells , conditional deletion in megakaryocyte lineage cells did not affect hematopoietic or vascular regeneration after irradiation ( Figures 6–9 ) . Thus , hematopoietic stem/progenitor cells and LepR+ stromal cells are the functionally important sources of Angpt1 in the bone marrow . Tie1 and Tie2 are the receptors for Angpt1 , Angpt2 , and possibly Angpt3 ( Augustin et al . , 2009; Eklund and Saharinen , 2013; Fagiani and Christofori , 2013; D'Amico et al . , 2014; Thomson et al . , 2014 ) . Tie1 is not required for the development or maintenance of fetal liver or adult bone marrow HSCs ( Corash et al . , 1989; Rodewald and Sato , 1996 ) . Tie1/Tie2 double knockout ES cells contribute to fetal but not adult hematopoiesis ( Puri and Bernstein , 2003 ) . Given that Angpt1 is not required for the maintenance of adult hematopoiesis , some combination of Angpt2 and Angpt3 may be required for adult hematopoiesis . Global Angpt1 deletion ( Figure 3P–S ) , deletion from osteoblasts and their progenitors ( Figure 3A–E and Figure 4A–F ) , or deletion from hematopoietic and/or LepR+ stromal cells ( Figure 3F–O ) , did not affect HSC frequency or HSC function in normal adult mice . Angpt1 deletion from these cell populations had little effect on the bone marrow vasculature in normal adult mice ( Figures 7 , 8 ) . We observed only rare pores in sinusoidal epithelium from Leprcre; Mx1-cre; Angpt1fl/GFP mice ( Figure 8G , H ) and virtually no Evans blue leakage ( Figure 8A ) . In contrast , Angpt1 deletion from LepR+ stromal cells and hematopoietic cells had much larger effects on the regeneration of the vasculature and the HSC niche after irradiation . Angpt1 deficiency from these cell populations accelerated the recovery of hematopoiesis ( Figure 6A ) , the regeneration of LSK cells ( Figure 6B ) , the regeneration of long-term multilineage reconstituting HSCs ( Figure 6D ) , the proliferation of endothelial cells ( Figure 7K ) , and the morphological recovery of bone marrow blood vessels ( Figure 7E–J and Figure 7—figure supplement 1 ) . However , Angpt1 deficiency also increased the leakiness of the regenerated blood vessels ( Figure 8A , D ) by allowing pores to persist in sinusoidal endothelium ( Figure 8G , H ) . Together , the data indicate that Angpt1 produced by LepR+ stromal cells and hematopoietic cells promotes vascular integrity in regenerating blood vessels in the bone marrow at the cost of slowing the regeneration of HSC niches and hematopoiesis .
Targeting vectors for generating Angpt1GFP and Angpt1fl/+ mice were constructed by recombineering ( Liu et al . , 2003 ) . Linearized targeting vectors were electroporated into Bruce4 ES cells . Corrected targeted ES cell clones were identified by Southern blotting and injected into C57BL/6-Tyrc-2J blastocysts . The resulting chimeric mice were bred with C57BL/6-Tyrc-2J mice to obtain germline transmission . Then the Frt-Neo-Frt cassette introduced by the targeting vector was removed by mating with Flpe mice ( Rodriguez et al . , 2000 ) . These mice were backcrossed onto a C57BL/Ka background . Other mice used in this study were: Col1a1*2 . 3-cre ( Liu et al . , 2004 ) , Leprcre ( DeFalco et al . , 2001 ) , Osx-cre ( Rodda and McMahon , 2006 ) , Nestin-cre ( Tronche et al . , 1999 ) , Pf4-cre ( Tiedt et al . , 2007 ) , Tie2-cre ( Koni et al . , 2001 ) , Mx1-cre ( Kühn et al . , 1995 ) , UBC-cre/ER ( Ruzankina et al . , 2007 ) , Col1a1*2 . 3-GFP ( Kalajzic et al . , 2002 ) , Loxp-EYFP ( Srinivas et al . , 2001 ) , and Loxp-tdTomato ( Madisen et al . , 2010 ) . For induction of UBC-Cre/ER , Tamoxifen chow ( Harlan , Indianapolis , IN ) containing tamoxifen citrate at 40 mg/kg , with 5% sucrose , was administrated to mice for 2–5 months before analysis . C57BL/6-SJL ( CD45 . 1 ) mice were used as recipients in transplantation experiments unless otherwise indicated . All mice were housed at the Unit for Laboratory Animal Medicine at the University of Michigan or in the Animal Resource Center at the University of Texas Southwestern Medical Center . All protocols were approved by the University of Michigan Committee on the Use and Care Animals and by the University of Texas Southwestern Institutional Animal Care and Use Committee . The following primers were used for genotyping . For Angpt1GFP , OLD308: 5′-gggaaagagtcaaacaagcag-3′ OLD309: 5′-aaccgaaagcgatcacttac-3′ and OLD292: 5′-cggacacgctgaacttgtgg-3′ . For Angpt1fl , OLD335: 5′-ggactcaacttcctgggtaagc-3′ and OLD336: 5′-ggctttgacagtcaaaatgcc-3′ . For Angpt1− , OLD3111: 5′-cagggttggcataaaatttgc-3′ and OLD350: 5′-tcctggtctttgcacttgactg-3′ . Cells were directly sorted into Trizol . Total RNA was extracted per the manufacture's instructions . SuperScript III ( Lifetech , Grand Island , NY ) was used to generate cDNA . Quantitative real-time PCR was performed using SYBR green on a LightCycler 480 or Stepone Plus . β-actin amplification was used to normalize the transcript content of samples . Primers used in this study were: Angpt1: OLD7: 5′-gggggaggttggacagtaat-3′ and OLD8: 5′-catcagctcaatcctcagca-3′ . Tie2: forward: 5′-gattttggattgtcccgaggtcaag-3′ and reverse: 5′-caccaatatctgggcaaatgatgg-3′ . β-actin: OLD27: 5′-gctcttttccagccttcctt-3′ OLD28: 5′-cttctgcatcctgtcagcaa-3′ . Cells were sorted or directly pipetted into methylcellulose culture medium ( 3434 , Stemcell Technologies , Vancouver , BC , Canada ) and incubated at 37°C for 14 days in a humidified chamber . Bone marrow cells were isolated by flushing or by crushing the long bones with a mortar and pestle in Ca2+ and Mg2+ free HBSS with 2% heat-inactivated bovine serum . Spleen cells were obtained by crushing the spleen between two glass slides . The cells were dissociated to a single cell suspension by gently passing through a 25G needle then filtering through a 70 µm nylon mesh . The following antibodies were used to isolate HSCs: anti-CD150 ( TC15-12F12 . 2 ) , anti-CD48 ( HM48-1 ) , anti-Sca1 ( E13-161 . 7 ) , anti-c-kit ( 2B8 ) , and the following antibodies against lineage markers ( anti-Ter119 , anti-B220 [6B2] , anti-Gr1 [8C5] , anti-CD2 [RM2-5] , anti-CD3 [17A2] , anti-CD5 [53-7 . 3] , and anti-CD8 [53-6 . 7] ) . HPCs were identified by flow cytometry using the following antibodies: anti-Sca1 ( E13-161 . 7 ) , anti-c-kit ( 2B8 ) , and the following antibodies against lineage markers ( anti-Ter119 , anti-B220 [6B2] , anti-Gr1 [8C5] , anti-CD2 [RM2-5] , anti-CD3 [17A2] , anti-CD5 [53-7 . 3] and anti-CD8 [53-6 . 7] ) , anti-CD34 ( RAM34 ) , anti-CD135 ( Flt3 ) ( A2F10 ) , anti-CD16/32 ( FcγR ) ( 93 ) , anti-CD127 ( IL7Rα ) ( A7R34 ) , anti-CD24 ( M1/69 ) , anti-CD43 ( 1B11 ) , anti-B220 ( 6B2 ) , anti-IgM ( II/41 ) , anti-CD3 ( 17A2 ) , anti-Gr1 ( 8C5 ) , anti-Mac1 ( M1/70 ) , anti-CD41 ( MWReg30 ) , anti-CD71 ( C2 ) , anti-Ter119 , anti-CD44 ( IM7 ) and anti-CD25 ( PC61 ) . DAPI was used to exclude dead cells . Unless otherwise indicated , antibodies were obtained from eBioscience ( San Diego , CA ) or BD Bioscience ( San Jose , CA ) . For flow cytometric analysis of bone marrow stromal cells , bone marrow was flushed using HBSS with 2% bovine serum . Then , whole bone marrow was digested with DNase I ( 200 U/ml ) and Collagenase IV ( 200 U/ml ) or liberase ( Roche , San Francisco , CA ) at 37°C for 15 min . Samples were then stained with antibodies and analyzed by flow cytometry . Anti- PDGFRα ( APA5 ) , anti-CD45 ( 30F-11 ) , anti-CD31 ( 390 ) , and anti-Ter119 antibodies were used to isolate perivascular stromal cells . For analysis of bone marrow endothelial cells , mice were i . v . injected with 10 µg/mouse Alexa Fluor 647 conjugated anti-VE-cadherin antibody ( BV13 , eBiosciences ) ( Butler et al . , 2010 ) . 10 min later , the long bones were removed and bone marrow was flushed , digested , and stained as above . Samples were analyzed using a FACSAria or FACSCanto II flow cytometer ( BD Biosciences ) . Data were analyzed by FACSDiva ( BD Biosciences ) or FlowJo ( Tree Star ) software . Adult recipient mice were lethally irradiated by a Cesium 137 GammaCell40 Irradiator ( MDS Nordia ) or an XRAD 320 x-ray irradiator ( Precision X-Ray Inc . , North Branford , CT ) with two doses of 540 rad ( total 1080 rad ) delivered at least 2 hr apart . Cells were transplanted intravenously into the retro-orbital venous sinus of anesthetized mice . 3 × 105 bone marrow cells were transplanted together with 3 × 105 recipient type competitor cells unless otherwise noted . Mice were maintained on antibiotic water ( neomycin sulfate 1 . 11 g/l and polymixinB 0 . 121 g/l ) for 14 days then switched to regular water . Recipient mice were periodically bled to assess the level of donor-derived blood cells , including myeloid , B and T cells for at least 16 weeks . Blood was subjected to ammonium chloride/potassium red cell lysis before antibody staining . Antibodies including anti-CD45 . 2 ( 104 ) , anti-CD45 . 1 ( A20 ) , anti-Gr1 ( 8C5 ) , anti-Mac-1 ( M1/70 ) , anti-B220 ( 6B2 ) , and anti-CD3 ( KT31 . 1 ) were used to stain cells for analysis by flow cytometry . For BrdU incorporation assays , mice were given an intraperitoneal injection of 1 mg BrdU ( Sigma , St . Louis , MO ) per 6 g of body mass in PBS ( Phosphate Buffered Saline ) and maintained on 1 mg/ml of BrdU in the drinking water for 24 hr ( endothelial cells ) or 10 days ( HSCs ) . Bone marrow endothelial cells were pre-stained by i . v . injection of Alexa Fluor 555 conjugated anti-VE-cadherin antibody ( BV13 , eBiosciences ) . The frequency of BrdU+ cells was then analyzed by flow cytometry using the APC BrdU Flow Kit ( BD Biosciences ) . Freshly dissected bones were fixed in 4% paraformaldehyde overnight followed by 3-day decalcification in 10% EDTA . Bones were sectioned using the CryoJane tape-transfer system ( Instrumedics , St . Louis , MO ) . Sections were blocked in PBS with 10% horse serum for 1 hr and then stained overnight with goat-anti-Angpt1 ( Santa Cruz , Dallas , TX , 1:200 ) , chicken-anti-GFP ( Aves , Tigard , OR , 1:1000 ) , anti-CD41-PE ( eBioscience , clone eBioMWReg30 , 1:200 ) and/or goat-anti-Osteopontin ( R&D , Minneapolis , MN , 1:400 ) antibodies . Donkey-anti-goat Alexa Fluor 647 , donkey-anti-chicken Alexa Fluor 488 , and/or Donkey-anti-goat Alexa Fluor 555 were used as secondary antibodies ( Invitrogen , Grand Island , NY , 1:400 ) . Slides were mounted with anti-fade prolong gold ( Invitrogen ) and images were acquired with a LSM780 confocal microscope ( Zeiss , San Diego , CA ) . For thick sections , the specimens were cleared overnight with Benzyl Alcohol/Benzyl Benzoate ( 1:2 ) solution ( Sigma ) . 3D reconstruction of bone marrow was achieved by Z stack of tiled images of femoral bone marrow with a Zeiss LSM780 confocal microscope . We defined regressed sinusoids according to the criteria used in a previous publication ( Hooper et al . , 2009 ) . We analyzed sinusoid morphology in thin optical sections through a segment of the femurs of mice . The sections were transverse sections through the longitudinal axis of the femurs , such that we observed cross-sections through most sinusoids . We first counted the total number of sinusoids in the section . Sinusoids were identified in these sections based on vessel morphology and bright VE-cadherin staining ( VE-cadherin staining was dimmer in arterioles and capillaries ) . We then counted the number of regressed sinusoids in the same sections to arrive at the percentage of regressed sinusoids . Regressed sinusoids ( Figure 7D ) were distinguished from non-regressed sinusoids ( Figure 7C ) by being larger in diameter and having few hematopoietic cells around them . This assay was modified from a published method ( Radu and Chernoff , 2013 ) . Mice were retro-orbitally injected with 200 μl of 0 . 5% Evans blue in PBS and sacrificed 15 min later . Femurs and spleens were collected , crushed , and then Evans blue in these tissues was eluted in a set volume of PBS . After a brief centrifugation , the concentration of Evans blue in the supernatant was measured on a Nanodrop spectrophotometer ( Thermo Scientific , Waltham , MA ) at a wavelength of 610 nm . Femurs and spleens from mice without Evans blue injection were used as negative controls and blanks . Mice were anaesthetized by i . p . injection of ketamine/xylazine . Before imaging , the mice received a retro-orbital injection of 100 μl PBS solution containing 10 μg Alexa Fluor 660 conjugated anti-VE-cadherin antibody ( BV13 , eBiosciences ) and 100 μg Dextran-FITC ( 70-kDa , Sigma ) . Then , the mouse was placed on a heated stage with the skull positioned under the objective using a stereotaxic device . Dextran-FITC fluorescence and autofluorescence from bone collagen were captured using two-photon imaging while Alexa Fluor 660-anti-VE-cadherin fluorescence was captured using confocal imaging on the same LSM780 microscope ( Zeiss ) . An approximately 4 × 6 mm area of the calvarium encompassing most of the parasagittal bone marrow cavities within the left and right frontal bones was scanned in each imaging section . Euthanized mice were pre-fixed by vascular perfusion via the left ventricle for 10 min with a solution containing 2% glutaraldehyde , 2% paraformaldehyde , and 0 . 1 M cacodylate buffer at pH 7 . 3 . A 1 ml syringe fitted with a 23-gauge needle ( BD Biosciences ) containing ice-cold PBS was inserted into the growth plate and the then entire marrow plug was gently flushed from the marrow cavity . The marrow plugs were post-fixed in 2 . 5% glutaraldehyde overnight . They were partially dehydrated in ethanol , fractured in liquid nitrogen , rehydrated , and then fixed in 1% osmium tetroxide for another 2 hr . After full dehydration using a graded series of ethanol concentrations followed by hexamethyldisilazane , the specimens were coated with sliver . Two to three specimens per mouse were randomly chosen and examined on a Zeiss Sigma VP FE-SEM at 5–10 mkV at the UT Southwestern Electron Microscopy Core Facility . Cavstratin , a fusion peptide of the putative scaffolding domain of caveolin-1 ( amino acids 82–101: DGIWKASFTTFTVTKYWFYR ) and the antennapedia internalization sequence ( RQIKIWFQNRRMKWKK ) , was synthesized as previously described ( Gratton et al . , 2003 ) at the UT Southwestern Protein Chemistry Technology Center . Peptides were dissolved initially in DMSO and diluted 1000-fold in sterile PBS before in vivo administration ( 2 . 5 mg/kg per mouse ) . | In adults , blood cells develop from a set of stem cells that are found in bone marrow . There are also specialized blood vessels and cells called ‘stromal cells’ within the bone marrow that provide these stem cells with oxygen , nutrients , and other molecules . This local environment , or ‘niche’ , plays an important role in regulating the maintenance of these stem cells . But it has not been known whether stem cells can reciprocally regulate their niches . Unfortunately , radiation used to treat cancer obliterates the stem cells and their niche; both must recover after such a treatment before the patient can produce blood cells normally again . A protein called Angpt1 is thought to play a role in this post-treatment recovery . Angpt1 is known to regulate blood vessels in the bone marrow , and one influential study had previously suggested that bone cells produce Angpt1 , which promotes and regulates the maintenance of the stem cells within the niche . However , this previous study did not directly test this . Thus , it was not clear whether Angpt1 promotes the regeneration of the stem cells themselves or if it regulates the rebuilding of the niche . Now , Zhou , Ding and Morrison have genetically engineered mice to make a ‘reporter’ molecule—which glows green when viewed under a microscope—wherever and whenever the gene for Angpt1 is active . These experiments showed where the protein is produced , and unexpectedly revealed that the bone cells do not make Angpt1 . Instead , it is the stem cells and the stromal cells in the niche that made the protein . Further experiments showed that deleting the gene for Angpt1 from mice , or just from their bone cells , did not affect blood cell production; nor did it affect the maintenance or regulation of the stem cells . Next , Zhou , Ding and Morrison looked at whether Angpt1 might be involved in rebuilding the niche after being exposed to radiation . Some of these irradiated mice had been genetically engineered to lack Angpt1; and , in these mice , blood stem cells and blood cell production recovered more quickly than in mice with Angpt1 . The blood vessels in the niche also grew back more quickly in the irradiated mice that lacked Angpt1 . However , these regenerated blood vessels were leaky . This suggests that blood stem cells produce Angpt1 to slow the recovery of the niche and reduce leakage from the blood vessels . Thus , blood stem cells can regulate the regeneration of the niches that maintain them . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine"
] | 2015 | Hematopoietic stem and progenitor cells regulate the regeneration of their niche by secreting Angiopoietin-1 |
How tissue shape emerges from the collective mechanical properties and behavior of individual cells is not understood . We combine experiment and theory to study this problem in the developing wing epithelium of Drosophila . At pupal stages , the wing-hinge contraction contributes to anisotropic tissue flows that reshape the wing blade . Here , we quantitatively account for this wing-blade shape change on the basis of cell divisions , cell rearrangements and cell shape changes . We show that cells both generate and respond to epithelial stresses during this process , and that the nature of this interplay specifies the pattern of junctional network remodeling that changes wing shape . We show that patterned constraints exerted on the tissue by the extracellular matrix are key to force the tissue into the right shape . We present a continuum mechanical model that quantitatively describes the relationship between epithelial stresses and cell dynamics , and how their interplay reshapes the wing .
The dynamic choreography of tissue shape changes that occur during development dramatically illustrates the fact that morphogenesis depends on organized cellular force generation . The mechanisms that control the orientation and patterning of these active processes and the corresponding tissue stresses are beginning to be explored in a variety of developmental systems , for review ( Lecuit and Lenne , 2007; Keller , 2012; Heisenberg and Bellaiche , 2013 ) . However , a complete understanding of the mechanical basis of morphogenesis will require not only a description of cell autonomously generated forces , but also quantitative insights into how cells respond to tissue stresses . Cells exert forces on extracellular matrices , but also on each other—this is particularly true of epithelial cells , which are tightly connected by specialized adhesive junctions . Thus , stresses generated by one epithelial cell can be transmitted to others throughout the tissue . In vitro experiments have shown that tissues respond to stress elastically over short time scales but that they can plastically remodel when subjected to stress over longer times ( Beloussov et al . , 2000; Harris et al . , 2012 ) . This can occur as a result of cell shape changes , cell rearrangements or both , and appears to vary with the cell types examined . Furthermore , experiments with cultured epithelial cells suggest that tissue compression can limit cell proliferation in vitro ( Puliafito et al . , 2012; Streichan et al . , 2014 ) . How these cellular responses might influence tissue size and shape in vivo is not clear . Nevertheless , these in vitro observations suggest that a complete and quantitative understanding of tissue morphogenesis will require new insights into tissue viscoelasticity in vivo and the cellular mechanisms that give rise to it . Drosophila pupal wing morphogenesis is an ideal system in which to study the interplay of cellular force generation and tissue material properties in vivo . During pupal stages , anisotropic stresses along the proximal-distal ( PD ) axis of the wing blade epithelium help guide anisotropic tissue flows that reshape the blade—elongating it in the PD axis and narrowing it in the anterior-posterior ( AP ) axis , for review ( Eaton and Julicher , 2011 ) . The mechanisms that produce PD-oriented stresses in the wing blade are not fully understood . They are generated in part by contraction of cells in the wing hinge , which connects to the wing blade on its proximal side . However , we do not understand the origin of counterforces that restrain movement of the wing blade at the margin . Analyzing cells in a subregion of the wing blade showed that tissue flows are associated with cell shape changes , cell divisions and cell rearrangements that are oriented along the PD axis ( Aigouy et al . , 2010 ) . To quantitatively understand the cellular basis of this tissue shape change , we must determine the global patterns of these cellular events throughout the wing blade . Furthermore , while hinge contraction contributes to PD tissue stresses in the blade , cells in the wing blade might also contribute autonomously to tissue flows and stresses . Thus , to understand the mechanical basis of pupal wing morphogenesis , we must understand the emergence of PD-oriented stresses in the wing blade , and distinguish stresses autonomously generated by wing epithelial cells from the response of epithelial cells to these stresses . Here , we combine several quantitative methods to investigate how cell flows and global tissue shape changes emerge from the collective behavior and mechanical properties of many wing epithelial cells . We develop image analysis methods to track the majority of cells in the wing throughout morphogenesis , and analyze cell shapes and rearrangements of the junctional network . Furthermore , we develop theoretical methods to quantify the cellular contributions to tissue shear and area homeostasis in the wing blade . We show that localized apical extracellular matrix connections to the cuticle at the wing margin provide the counterforce to hinge contraction , and are required for the development of normal stresses in the wing blade . These stresses are essential to reshape the pupal wing while maintaining wing area homeostasis . We distinguish autonomously controlled from stress-driven cellular events , and present a continuum mechanical model that quantitatively explains wing shape changes on the basis of the relationship between tissue stress , cell elongation and cell rearrangements .
The emergence of two-dimensional stresses in the plane of the wing blade suggests that there are physical constraints on the movement of wing epithelial cells near the margin . We wondered whether there might be a matrix connecting the wing epithelium to the overlying pupal cuticle in this region . To investigate this , we used a laser to destroy the region between the margin of the E-Cadherin:GFP expressing wing epithelium and the cuticle after the two had separated as a consequence of molting . Although this treatment does not apparently damage either the wing or the cuticle , it causes the wing epithelium to rapidly retract away from the cuticle within seconds ( Figure 1A–B′′ , Video 1 ) . Laser ablation causes epithelial retraction when performed at any region along the wing blade margin—anteriorly , posteriorly or distally . During tissue flows , the now disconnected margin moves even further away from the cuticle , producing abnormal wing shapes ( Figure 1C–F ) . This shows that the wing is physically restrained by apical extracellular matrix connections to the overlying cuticle , and that these connections are required to shape the wing during tissue flows . 10 . 7554/eLife . 07090 . 003Figure 1 . Physical constraints at the margin maintain epithelial tension in the wing . ( A ) Cartoon depicting a pupal wing at 32 hAPF . Dashed double-sided arrows depict the proximal-distal ( PD ) and anterior-posterior ( AP ) axes . The PD axis is defined by a regression line passing through selected sensory organs ( red dots ) that are easily identifiable in Ecad::GFP expressing wings . The x axis is defined to correspond to the PD axis pointing distally , and the y axis is defined to correspond to the AP axis pointing anteriorly . L2–L5 indicate longitudinal veins 2–5 . Brown dashed line outlines the cuticular sac surrounding the wing epithelium . Scale bar 20 µm . ( B , B′ ) Show the distal end of a wild-type ( WT ) Ecad::GFP-expressing wing at 24 hAPF ( greyscale in B , B′ ) and the same wing 3 . 5 min after laser ablation in the space between wing margin and cuticle ( magenta in B′ ) . The blue dashed line indicates the site of laser ablation . ( B′′ ) Shows wing margin displacement measured with respect to the cuticle ( brown dashed line in B′ ) along the white dotted line in ( B′ ) . Experimental points ( magenta ) were interpolated by a polynomial ( blue line ) . ( C–F ) Show 32 hAPF wings that were unperturbed ( C ) or subjected to laser ablation at 22 hAPF ( D–F ) . Ablation of the connections between the wing margin and the cuticle were performed in different regions , indicated by blue dashed lines in ( D–F ) , and lead to altered wing shapes at 32 hAPF compared to the unperturbed control ( C ) . Scale bar 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 00310 . 7554/eLife . 07090 . 004Video 1 . Laser ablations of the apical extracellular matrix present in the space between the tissue and the cuticle . Green lines indicate the site of ablation right before ablation . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 004 We wondered whether the large apical extracellular matrix protein Dumpy might contribute to these connections . Dumpy is a 2 . 5 MDa protein that is predicted to form filaments at least 1 µm long ( Wilkin et al . , 2000 ) . It forms an elastic matrix in the embryonic tracheal lumen , and provides mechanical resilience of tendon cell attachments to the overlying cuticle ( Dong et al . , 2014 ) . While dumpy null mutations are lethal , some hypomorphs produce wings that are short and misshapen—a defect that arises during pupal development ( Waddington , 1939 , 1940 ) . To ask whether shape defects in dumpy wings might arise during pupal tissue flows , we imaged dumpyov1 pupal wings that expressed E-Cadherin:GFP . The shape of dumpyov1 wings is normal at 16 hr after puparium formation ( APF ) , before molting occurs ( Figure 2A , B ) . Shortly afterwards , when hinge contraction begins , the shape of the dumpyov1 mutant wing blade begins to differ from wild type ( WT ) . The wing blade epithelium retracts abnormally far from the distal cuticle and fails to elongate in the PD axis . By the time tissue flows have ended , the characteristic abnormal shape of the dumpyov1 wing is apparent ( Video 2 and Figure 2A–B′′ ) . 10 . 7554/eLife . 07090 . 005Figure 2 . Dumpy-dependent apical attachments of wing tissue to the cuticle act as a counter-force to hinge contraction . ( A–B′′ ) Show individual frames from a time-lapse video of dumpyov1 mutant and control WT wings expressing Ecad::GFP , and depict wings at 16 hAPF ( A , B ) , 22 hAPF ( A′ , B′ ) , and 32 hAPF ( A′′ , B′′ ) . The position of the cuticle is indicated by a brown dashed line . Scale bar 100 µm . ( C ) Cartoon depicting the pupal wing and surrounding cuticular sac , labeled to indicate the optical sections shown in panels ( D–F ) . ( D–F′′′ ) Show optical sections through a 22 hAPF wing from a pupa harboring a Dumpy::YFP protein trap at the endogenous locus . ( D–E ) Show sections in the plane of the wing epithelium near the anterior ( D ) and posterior ( E ) margins . Dumpy::YFP is present in the space between the wing margin and the cuticle . Scale bar 20 µm . ( F–F′′′ ) Show optical sections orthogonal to the wing epithelium at different proximal-distal positions ( indicated in C ) . Dumpy::YFP connects the dorsal wing surface and the cuticle in specific positions ( arrowheads ) . Scale bars: 20 µm . ( G ) Summarizes the pattern of Dumpy::YFP connections between the dorsal wing surface and overlying cuticle—these lie over the hinge and vein regions . ( H ) Shows the pattern of Dumpy::YFP connections between the wing margin and cuticle—these extend around the wing margin except in the posterior/proximal regions . ( I–I′ ) Summarize quantifications of circular laser ablation experiments performed in nine specific regions of five WT and five dumpyov1 mutant wings . The size of the red circles indicates the initial rate of area expansion of the perimeter of the circular cut , which reflects the isotropic tissue stress . Green bars represent the direction and magnitude of the elliptical deformation of the initially circular cut , reflecting the anisotropy of tissue stress ( see ‘Materials and methods’ , Analysis of circular laser ablations ) . Magenta bars depict the orientation and magnitude of local cell elongation . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 00510 . 7554/eLife . 07090 . 006Figure 2—figure supplement 1 . Dumpy::YFP cuticle imprint and Dumpy apical connections . ( A ) Shows Dumpy::YFP fluorescence in the cuticle ( >10 µm away from the apical wing surface ) of a 22 hAPF wing ( red square indicates enlarged region ) . Dumpy::YFP highlights imprints of cell boundaries in the cuticle . ( B ) Shows total Dumpy::YFP present between the dorsal wing surface and the cuticle . Dumpy::YFP signal was manually segmented and colored in green , then overlaid over a 22 hAPF WT Ecad::GFP expressing wing to highlight its position relative to wing veins ( see also Video 3 ) . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 00610 . 7554/eLife . 07090 . 007Figure 2—figure supplement 2 . dumpyov1 weakens distal attachments between wing margin and cuticle . ( A ) Cartoon of a dumpyov1 mutant wing indicating sites of laser cuts ( blue lines ) performed in wings shown in ( B–F′′ ) . ( B–F′′ ) Show 22 hAPF dumpyov1 mutant wings before ( greyscale ) and shortly after ( magenta ) laser severing in different regions between the margin and the cuticle . Blue dashed lines indicate cut site , and green arrows the amount and direction of retraction after the cut . Time after ablation is indicated ( minutes , seconds ) . Tension-bearing connections are present anteriorly and posteriorly , but not distally in dumpyov1 . Scale bar 20 µm . ( G ) dumpyov1 mutant wing at 37 hAPF , 15 hr after ablation of margin-cuticle connections in the whole region comprising panels B–F′′ . Scale bar 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 00710 . 7554/eLife . 07090 . 008Figure 2—figure supplement 3 . Method to determine stresses in WT and dumpyov1 mutant . ( A , B ) Enlarged regions of the wing epithelium at 22 . 5 hAPF , before ( A ) and 50 s after ( B ) circular laser ablation in the epithelium . The green circle depicts the 14 µm circular cut in diameter . The red ellipse is a fit to the manually segmented perimeter of the cut region at 50 s . Minor ( blue ) and major ( magenta ) axes of this ellipse are used to define orthogonal kymographs . Scale bar 20 µm . ( C , D ) Kymographs defined in ( B ) . Arrowheads depict the lines that were segmented using Fiji . Δx shows the relative increase in wing tissue displacement along the major and minor axes after the cut . ( E ) Graph showing an example of the relative tissue displacement along the major and minor axes . These displacements are used to estimate the initial velocity gradient of recoil after laser ablation ( ‘Materials and methods’ , Analysis of circular laser ablations ) . The initial velocity gradient reflects the isotropic and and anisotropic stresses in the tissue . ( F , G ) Comparison of stresses between WT control and dumpyov1 mutant wings . Circular cuts were performed in nine different locations as depicted in Figure 2I , I′ . Error bars show standard deviation over five replicates for each location and genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 00810 . 7554/eLife . 07090 . 009Video 2 . Synchronized time-lapses of wild-type ( WT ) and dumpyov1 wings . The synchronization is based on the time when histoblast nests merge at ∼26 . 5 hAPF . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 009 To examine Dumpy distribution , we imaged wings from flies harboring a protein trap construct that expresses YFP:Dumpy from the endogenous chromosomal locus . YFP:Dumpy is present on the apical surface of epithelial cells throughout the wing , and within the overlying cuticle ( Figure 2—figure supplement 1A , Video 3 ) . Interestingly , Dumpy is also present in a fibrous-appearing matrix that connects the wing to the overlying cuticle in specific places . This matrix lies between the cuticle and the margin of the wing ( Figure 2C–E , H ) , as well as in stripes that run on the dorsal surface of the wing between longitudinal veins L3 and L4 , and over veins L2 and L5 ( Figure 2F , G , Figure 2—figure supplement 1B ) . Dumpy-containing matrix also connects a subregion of the wing hinge to the overlying cuticle ( Figure 2F , G , Figure 2—figure supplement 1B , Video 3 ) . 10 . 7554/eLife . 07090 . 010Video 3 . Dumpy::YFP distribution in a 40 µm deep z-stack that was manually annotated to identify the regions where the protein is present . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 010 To investigate the extent to which wing margin constraints had been relieved by the dumpyov1 mutation , we performed laser-severing experiments in the dumpyov1 mutant background . Cutting a dumpyov1 wing between the cuticle and the distal wing blade revealed almost undetectable retraction , suggesting that distal attachments of the wing blade to the cuticle are severely compromised . However retraction was still observed when severing was performed between the cuticle and the anterior or posterior margins ( Figure 2—figure supplement 2 ) . Furthermore , severing the matrix around the entire margin causes dumpyov1 mutant wings to develop even more dramatic wing shape abnormalities during tissue flows ( compare Figure 2—figure supplement 2G to Figure 2A′′ ) . This suggests that apical matrix connections to the cuticle are not completely abrogated in dumpyov1 wings . To ask how the dumpyov1 mutation influenced PD tension in the wing blade , we performed circular laser cuts covering about 5–10 cells in different regions of WT and dumpyov1 wings ( Figure 2I , I′ , Figure 2—figure supplement 3 ) . We observed a recoil of the ablated region , indicating that the blade epithelium is under tension . From the recoil , we can compare both the isotropic and the anisotropic components of epithelial stress in WT and dumpyov1 mutant wings ( Figure 2I , I′ and Figure 2—figure supplement 3 ) . These stress patterns differ between WT and dumpyov1 wings . The orientation of anisotropic stress in dumpyov1 is somewhat splayed and not as well aligned with the PD axis . Furthermore , anisotropic tension in dumpyov1 wings tends to be reduced in the central region and increased anteriorly and posteriorly . Overall , Dumpy-dependent elastic connections are key to the emergence of the stress pattern during morphogenesis . This suggests that these stresses play an important role in guiding tissue flows . What are the cellular events that shape the wing blade during tissue flows ? To quantitatively address this question , we developed methods to quantify cell shape changes , cell divisions , cell rearrangements and cell extrusions during wing morphogenesis . We imaged three E-Cadherin:GFP-expressing wings at cellular resolution every 5 min between 16 and 32 hr APF . We then extracted and projected the planes containing the apical adherens junctions , automatically detected cell edges , and tracked each cell in the wing over the course of the videos ( Figure 3A , A′ and Videos 4 , 5 ) . We designed a relational database ( DB ) to store information pertaining to all cells in a given video ( ‘Materials and methods’ , Long-terms time-lapse imaging and Data handling and image processing ) . Querying these DBs provides information about individual cellular properties such as shape , area , and associated cell boundaries . It also provides information about neighbor and lineage relationships , identifying neighbor exchanges ( T1 transitions ) , cell divisions and cell extrusions ( T2 transitions ) . 10 . 7554/eLife . 07090 . 011Figure 3 . Cellular contributions to wing blade area changes . ( A , A′ ) WT Ecad::GFP expressing wing at 16 hAPF ( A ) and ( A′ ) at 32 hAPF . The hinge is colored purple . Green shaded region indicates the region of the blade in which segmented cells could be tracked from the beginning to the end of the video . This region was used for further analysis ( D , E ) . ( B ) Cartoon illustrating cellular contributions to wing blade area change: cell area change , cell division and cell extrusion . The equation decomposes the relative area change of the entire wing blade ( v ) into the relative area changes due to each cellular contribution throughout the wing blade . ( C ) dumpyov1 mutant wing at 32 hAPF . Green shaded region indicates tracked region used for analysis in ( F , G ) . ( D , E ) Relative cellular contributions to wing blade area change over time , averaged over three WT wings . The rates of relative area change are shown in ( D ) and their cumulative sums are shown in ( E ) . Lighter shaded regions in indicate standard deviations between wings . ( F , G ) Cellular contributions to wing blade area change in a dumpyov1 mutant wing . Cumulative plots ( G ) were generated starting at 16 . 5 hAPF , the earliest time common to all compared videos . Scale bar 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 01110 . 7554/eLife . 07090 . 012Figure 3—figure supplement 1 . Reproducibility of cell number and cell area in three WT wings . ( A ) A WT Ecad::GFP-expressing wing at 32 hAPF . The tracked and analyzed region is highlighted in green . Scale bar 100 µm . ( B ) Cell number change in 3 WT wing blades . ( C ) Average cell area ( µm2 ) in the blade of 3 WT wings . ( D ) Cell division rates in 3 WT wing blades ( divisions per cell per hour ) . The time was divided into 1 hr intervals in which the cell division rate was averaged . Error bars represent the standard error to the mean . ( E ) Cumulative increase in cell number due to cell division in 3 WT wing blades . ( F ) Cell extrusion rates ( extrusions per cell per hour ) in 3 WT wing blades . These were obtained similarly to panel D . ( G ) Cumulative numbers of cells lost due to cell extrusion throughout 3 WT wing blades . The same 3 WT wings are indicated in red , blue and green in panels ( B–G ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 01210 . 7554/eLife . 07090 . 013Figure 3—figure supplement 2 . Epithelial tension is required to maintain wing blade area . ( A–F ) Cumulative cellular contributions to changes in wing blade area in three mechanically perturbed WT wings . Green highlighting indicates tracked regions subjected to further analysis . Wings were severed either distally within the wing blade ( A , D ) and ( B , E ) , or close to the hinge-blade interface ( C , F ) starting the ablation at 15 . 5 hAPF . ( D , E , F ) Show cumulative area change of the tracked region ( blue curves ) along with cumulative cellular contributions due to cell divisions ( orange ) , cell area changes ( green ) and cell extrusions ( turquoise ) . To calculate these changes starting at the time of laser ablation , we must first estimate the outline of the tracked region at the time of laser severing—after laser severing , the wing changes its area rapidly before time-lapse imaging can begin . Since molting has not yet occurred at 15 . 5 hAPF , the outline of the cuticle serves as a marker for the initial anterior and posterior edges of the wing . We estimate the initial hinge-blade interface position in laser severed wings by overlaying the cuticle outline of the laser severed wing with that of a non-perturbed WT wing at 15 . 5 hAPF . We determine the initial position of the cut from the laser-burned imprint in the cuticle . By using these landmarks to calculate the initial area , we estimate the area change that occurred before the onset of recording , and offset each cumulative area change ( blue ) curve by this amount . The offsets are negative for dist-sev#1 and #2 , that is , wing blade area shrinks rapidly in response to laser ablation . In contrast , the offset is zero for the proximally severed wing blade , which does not shrink immediately after ablation . ( G ) Final ( 32 hAPF ) measured area of unperturbed and the indicated laser severed wing blades . The red error bar indicates the standard deviation amongst the 3 WT wing blades . The absolute blade area of dist-sev#1 reflects the sum of the areas of tissue proximal and distal to the cut . ( H ) Average cell area ( µm2 ) over time in WT wing blades and wing blades subject to the indicated perturbations ( dumpyov1 , dist-sev#1 , dist-sev#2 , prox-sev ) . When PD stresses are weakened , the average cell area shrinks to ∼10 µm2 as compared with ∼15 µm2 in unperturbed WT . ( I ) Cell division rates ( per cell per hour ) in WT and perturbed wing blades . The time was divided into 1 hr intervals in which the cell division rate was averaged . Error bars represent the standard error to the mean . Colors refer to the same wings as in panel H . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 01310 . 7554/eLife . 07090 . 014Video 4 . Cell outline obtained from the segmentation procedure . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 01410 . 7554/eLife . 07090 . 015Video 5 . High resolution video of a WT wing expressing Ecad::GFP . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 015 We showed previously that the area of the wing blade remains fairly constant during pupal morphogenesis , despite the fact that cells are dividing ( Aigouy et al . , 2010 ) . To understand how the wing blade maintains area homeostasis , we quantified wing blade area , cell divisions , cell area changes and cell extrusions between 16 and 32 hAPF in three different WT wings ( Figure 3A–E ) . We analyzed a large region of the wing blade in which all cells could be tracked during the whole video ( green in Figure 3A , C ) . The tissue area expansion rate is v=1AdAdt , where A is the tissue area and d/dt denotes the total time derivative . The tissue expansion rate can be decomposed into its cellular contributions as ( 1 ) v=1adadt+kd−ke , where a is average cell area , kd and ke are cell division and extrusion rates , respectively . The cumulative area expansion rate of the whole wing blade is ln ( A/A0 ) , where A0 is the tissue area when recording starts ( ∼16 hAPF ) , can be obtained by integrating the area expansion rate v over time . Figure 3D shows both the tissue area expansion rate ( dark blue ) and the contributions to this expansion rate from cell area changes , cell divisions and extrusions . These are averages over three WT wings . Figure 3E shows the corresponding cumulative quantities . The dynamics of wing area changes in the 3 WT wing blades are extremely similar—after contracting slightly during the first half of morphogenesis , blade area gradually returns to very close to its original value ( dark blue lines in Figure 3D , E ) . This almost constant area reflects a balance between cell divisions ( orange ) on the one hand , and cell area changes ( green ) and extrusions ( light blue ) on the other . Interestingly , blade area in three analyzed wings is more reproducible than would be expected from the variation in each cellular contribution , if they were independent of each other ( see shaded regions depicting standard deviations in Figure 3D , E ) . To quantify this observation we compared the variance of overall relative area change with the sum of the variances of the cumulative cellular contributions . We find that sum of variances is about 20 times larger than variance of the sum . This shows that cellular contributions are not independent and that normal variations in the rate of cell division can be compensated by changes in cell area and/or extrusion to maintain wing blade area ( Figure 3—figure supplement 1 ) . To ask whether connections to the cuticle were required to maintain wing blade area , we examined blade area changes and the underlying cellular contributions in dumpyov1 mutant wings ( Figure 3F , G ) . As a complementary approach , we quantified the cellular contributions to area changes in laser-severed wings . We severed the wing either between the hinge and the blade , or at the very distal tip before hinge contraction occurred ( Figure 3—figure supplement 2 ) . We also severed connections between the wing margin and the cuticle ( Figure 1D , F ) at about 22 hAPF . In contrast to unperturbed WT wings , total wing area decreases dramatically when connections at the margin are weakened by dumpyov1 mutation or by laser severing ( Figure 3G and Figure 3—figure supplement 2 , see dark blue curve ) . Thus , connections to the cuticle are required to maintain wing blade area during morphogenesis . These connections provide mechanical linkages that permit the buildup of tensile stresses while maintaining wing blade area . How does epithelial stress influence the cellular events contributing to area homeostasis ? To answer this question , we first compared cellular contributions to area change during morphogenesis of WT and dumpyov1 mutant wings . Wing blade area decrease in dumpyov1 mutant wings is not a consequence of fewer cell divisions—cells actually divide more than in WT ( Figure 3—figure supplement 2D–F , I ) . Cells in dumpyov1 mutant wings have a similar maximum division rate but divide over a longer period of time , resulting in more cells at the end of morphogenesis ( Figure 3G and Figure 3—figure supplement 2D–F , yellow curve ) . The reduced wing blade area in dumpyov1 is quantitatively explained by reductions in cell area and by cell extrusions , which more than compensate the increased proliferation . Thus , reduced epithelial stresses in dumpyov1 wings perturb the balance between cell divisions , area changes and extrusions seen in WT . All laser-severing perturbations decrease the final wing area , similar to dumpyov1 mutant wings ( Figure 3—figure supplement 2 ) . In these wings , the analysis of cellular contributions to wing area changes is complicated by the delay between cutting and the initiation of time-lapse imaging ( about 45 min ) . During this intervening time , the wing responds acutely to reduced tension , and both wing and cell area decrease to values below those expected for WT wings of the same stage ( Figure 3—figure supplement 2D–F , H ) . While we can estimate changes to cell area during this time , we cannot know the rates of cell division and extrusion . Nevertheless , several interesting conclusions can be drawn by analyzing final cell area , and the rates of division , area change and extrusion after time-lapse imaging begins . Wings that have been severed before hinge contraction ( whether at the hinge-blade interface or at the distal tip ) behave similarly to dumpyov1 mutant wings . After an initial delay , the rate of cell division increases and cells divide more than in unperturbed wings . However , cell extrusions and decreasing cell area more than compensate for increased cell division to produce smaller wings . When cuticle connections are severed later at 22 hAPF , most cell divisions have already occurred , and this treatment does not increase proliferation . In this case , the number of cell extrusions increases , and the final cell area is smaller than that of unperturbed wings . Taken together , analyzing dumpyov1 and laser-severed wings shows that epithelial stresses are required to balance cell divisions with cell extrusions and cell area changes to maintain area homeostasis during morphogenesis . This is consistent with observations in the thorax , where overcrowding drives delamination ( Marinari et al . , 2012 ) . In the preceding section , we discussed how cellular processes contributed to wing blade area changes . These area changes correspond to the isotropic component of a tensor characterizing the tissue strain ( Figure 4A ) . Now , we discuss shape changes of the wing blade , which correspond to the anisotropic part of this tensor and characterize the process of elongation along an axis ( i . e . , pure shear ) . The rate of change of pure shear is described by the pure shear rate tensor v∼ ( Figure 4B ) with v∼xx characterizing the rate of elongation along the PD axis ( Figure 1A ) . Note that pure shear , that is , convergence-extension flow , is different from so-called simple shear , which results from a superposition of pure shear and a rotation ( Figure 4C ) . In the following , we use the term shear to refer to pure shear . We now discuss how tissue shear can be decomposed into contributions from cell shape changes and topological changes . These include cell divisions , cell neighbor exchanges ( T1 transitions ) , and cell extrusions ( T2 transitions ) ( Figure 4D–G ) . 10 . 7554/eLife . 07090 . 016Figure 4 . A method to quantify cellular contributions to wing blade deformation . ( A , B ) Isotropic part of tissue deformation , that is , tissue area growth ( A ) , and anisotropic part of tissue deformation , that is , tissue pure shear , along the x axis ( B ) . The deformation occurs during a time interval δt . ( B ) The sign of the pure shear rate component v∼xx indicates the shear direction . Positive means that the shear deformation occurs along the PD axis ( also referred to as x axis ) . Negative corresponds to shear along the AP axis ( or y axis ) . ( C ) A simple shear deformation corresponds to a superposition of a pure shear deformation and a rotation . ( D–H ) Cartoons depicting how changes in cell shape , cell rearrangements , cell divisions , cell extrusions and correlation effects could produce tissue shear . ( I ) The equation decomposes the tissue shear rate into shear contributed by each of these cellular processes ( color code as in D–H ) . For simplicity , the tensorial equation was projected onto the PD axis as most of the deformation occurs along the PD axis . ( J , K ) Triangulation method . ( J ) The cellular network is tiled with triangles: each vertex ( red dot ) of the cellular network that touches three cells gives rise to a single triangle ( red ) , whose corners are defined by the centers of the three cells ( green dots ) . ( K ) The resulting set of triangles tiles the cellular network without gaps or overlaps . ( L–O ) Triangle network modifications upon topological changes due to cell rearrangements ( L ) , cell divisions ( N ) , and extrusions ( O ) . ( M ) The discontinuous change in average triangle elongation during a given topological change is used to calculate the shear induced by the topological change . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 016 To understand the cellular contributions to the overall shear of the wing blade , we developed a method to distinguish and quantify shear caused by cell shape changes and shear caused by topological changes . In a piece of tissue where no topological changes occur within the cellular network , the deformation of the individual cells defines the deformation of the whole piece of tissue . However , when topological changes occur , deformation of individual cells no longer completely accounts for the overall shear ( Figure 4D–H ) . The triangle method we outline below represents an exact geometrical formalism to decompose large-scale deformations of the wing blade into contributions by cell deformation and by each kind of topological change . First , we tile the cellular network with triangles as follows . Each vertex of the cellular network that touches three cells ( red dot in Figure 4J ) gives rise to a single triangle ( red ) , whose corners are defined by the centers of the three cells ( green dots ) . Vertices that touch more than three cells are treated as described in Appendix 1 , ‘Triangulation procedure’ . The resulting set of triangles tiles the cellular network without gaps or overlaps ( Figure 4K ) . We choose a tiling into triangles , because the deformation of a single triangle between two frames of the video can be uniquely characterized by a single 2 × 2 tensor describing a linear transformation ( see Appendix 1 , ‘Triangle deformation’ ) . Note that such a characterization by a 2 × 2 tensor is in general not possible for polygons with more than three sides . For each triangle and time point , this tensor describes relative area changes , rotation , and shear of the triangle . The average shear rate of all triangles in the tissue corresponds to the overall tissue shear rate . To connect the tissue shear rate to cell elongation changes , we define a nematic tensor Q characterizing the state of triangle elongation ( see Appendix 1 , ‘Triangle elongation’ ) . Then , the change of triangle elongation corresponds exactly to triangle shear . Cell elongation is obtained as the average of the elongation tensors Q of those triangles that belong to a given cell . Hence , in the absence of topological changes , we find an exact relation between cell elongation change and overall tissue shear ( Appendix 1 , ‘Large-scale shear in the absence of topological transitions’ ) . If topological changes occur , then average cell elongation change is not equal to tissue shear . To include the effect of topological transitions , we write ( Appendix 1 , ‘Contributions to shear by topological changes’ ) : ( 2 ) v∼=DQDt+R , where v∼ is the overall tissue shear rate tensor , DQ/Dt is a corotational rate of change in average triangle elongation , and R is the shear rate tensor due to topological changes . Contributions to R include T1 transitions ( T ) , cell divisions ( C ) , T2 transitions ( E ) , as well as correlated cell shape changes and cell rotations ( D ) : R = T + C + E + D ( see Figure 4I ) . How can we define the contributions by topological changes ( T , C , and E ) to tissue shear ? During a topological transition , the triangulation changes and thus the average triangle elongation changes ( Figure 4L–O ) . However , at the moment the topological change occurs there is no tissue shear . Therefore , tissue shear and triangle elongation are no longer the same . This can be compensated by introducing a contribution to tissue shear by topological transitions . This contribution corresponds to the negative change in the average triangle elongation caused by the change in the triangulation ( Figure 4M and Appendix 1 , ‘Intermediate network states’ and Appendix 1 , ‘Contributions to shear by topological changes’ ) . In the definition of R in Equation 2 , we have also introduced the contribution D to tissue shear , which accounts for collective cellular events that combine to increase average triangle elongation in the absence of tissue shear and topological transitions . This occurs when several triangles have fluctuating shapes , such that the instantaneous elongation and the rotation rate or area expansion rate of triangles are correlated . Note that this effect does not occur when several triangles undergo equal deformations and rotations . One example of a cellular network deformation that produces the contribution D to tissue shear is shown in Figure 4H . Here , cells in neighboring rows slide relative to each other in alternating directions , such that no net pure shear occurs . However , there are alternating rows of simple shear and a net change in triangle elongation . We introduce the contribution D in the definition of R in order to compensate for the increase in the average triangle elongation DQ/Dt stemming from such correlations , should they exist in the wing blade . We first used the triangle method to calculate the patterns of tissue shear and cellular contributions to this tissue shear in WT ( Figure 5 , Video 6 , and Appendix 1 , ‘Spatially resolved shear patterns’ ) and dumpyov1 mutant wings ( Figure 5—figure supplement 1 and Video 7 ) . To visualize these patterns we averaged all quantities within squares of about 26 × 26 µm2 . Figure 5 shows shear patterns in WT at early , intermediate , and late time points during pupal wing morphogenesis . Shear is indicated by a line whose orientation represents the shear axis and whose magnitude corresponds to the shear rate . 10 . 7554/eLife . 07090 . 017Figure 5 . Patterns of cellular contributions to tissue shear in unperturbed WT wings . ( A–E′′ ) Patterns of local tissue shear rates ( A–A′′ ) , local shear rates contributed by cell rearrangements ( T1 transitions , B–B′′ ) , cell shape changes ( C–C′′ ) , cell divisions ( D–D′′ ) , and correlation effects ( E–E′′ ) , in a WT wing at 17 . 5 hAPF ( phase I ) , 21 hAPF ( intermediate phase ) and 26 hAPF ( phase II ) . The shear rate and shear rate contribution tensors were locally averaged within 26 × 26 µm2 square elements ( 25–50 cells ) of a fixed grid . A 45 min time window was used to smooth the shear values within each grid element . The resulting nematic tensors are represented by line segments whose length and direction correspond to the norm and orientation of the tensor , respectively . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 01710 . 7554/eLife . 07090 . 018Figure 5—figure supplement 1 . Patterns of cellular contributions to tissue shear in dumpyov1 mutant . ( A–E′′ ) Patterns of local tissue shear rates ( A–A′′ ) and local shear rates contributed by cell rearrangements ( B–B′′ ) , cell shape changes ( C′ , C′′ ) , cell divisions ( D–D′′ ) , and correlation effects ( E–E′′ ) , in a dumpyov1 mutant wing at 17 . 5 hAPF , 21 hAPF and 26 hAPF . The shear calculations were performed as described in Figure 5 . The resulting nematic tensors are represented by line segments whose length corresponds to the amplitude of the shear and whose orientation to the shear axis . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 01810 . 7554/eLife . 07090 . 019Video 6 . Dynamic patterns of tissue shear and of its cellular contributions in a WT wing . Line segments are nematic representations of shear . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 01910 . 7554/eLife . 07090 . 020Video 7 . Dynamic patterns of tissue shear and of its cellular contributions in a dumpyov1 mutant . Line segments are nematic representations of shear . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 020 Tissue shear in the WT wing blade is oriented in a fan-shaped pattern with a strong PD component ( Figure 5A–A′′ ) . At about 21 hAPF , the shear pattern develops a sharp reorientation between veins L3 and L4 , where shear is oriented along the AP axis . This region corresponds to a stripe of Dumpy-containing matrix that attaches the blade to the cuticle ( Figure 2—figure supplement 1B ) . Shear decreases over time , beginning distally and proceeding proximally , and finishes by about 32 hAPF . What are the patterns of cellular contributions to the tissue shear ? These patterns reveal a surprising complexity that changes with time . Shear caused by cell elongation and cell rearrangements ( T1 transitions ) display significant contributions that are antagonistic . Unexpectedly , T1 transitions cause shear along the AP axis early in the process ( Figure 5B–B′′ ) . This shear is opposed by increasing cell elongation along the PD axis ( Figure 5C–C′′ ) . At intermediate and late stages , T1 transitions shift their average orientation to cause PD shear—at the same time , cells reduce their elongation along the PD axis , causing AP shear . The contribution of cell extrusions to tissue shear is negligible ( not shown ) and cell divisions result in significant shear only during the early stages of wing morphogenesis ( Figure 5D–D′′ ) . Finally , plotting the contribution D to shear due to correlation effects reveals that these effects do exist in the wing ( Figure 5E–E′′ ) . Thus , patterns of cell elongation , cell rearrangement , cell division and correlation effects make dynamically changing contributions to tissue shear that are sometimes antagonistic . To investigate the cause of altered wing shape in dumpyov1 , we performed a similar analysis . The patterns of cellular contributions to tissue shear in a dumpyov1 mutant wing display subtle abnormalities ( Figure 5—figure supplement 1 ) . However , a more quantitative analysis is required to understand the origin of the altered dumpyov1 mutant wing shape . To better understand the quantitative relationships between the cellular processes contributing to tissue shear , we studied spatially averaged shear in the wing blade projected on the PD axis . We quantified average tissue shear , and shear caused by each cellular contribution over time in three different WT videos . These averages were taken over the tracked region shown in Figure 3A and the resulting quantities were further averaged over the three videos . These averages and the standard deviations between the videos are shown in Figure 6A , B . Positive values indicate shear along the PD axis and negative values indicate shear orthogonal to the PD axis ( i . e . , shear in the AP axis ) . Adding the contributions of shear caused by cell divisions , cell rearrangements , cell shape changes , and correlation effects ( light-pink line in Figure 6—figure supplement 1A ) reproduces the independently calculated total shear curve ( blue line in Figure 6A and Figure 6—figure supplement 1A ) . Small differences between these two curves ( about 3% ) stem from small inaccuracies ( see Appendix 1 , ‘Decomposition of the large-scale tissue shear rate’ ) . Thus , we can decompose tissue shear into its individual cellular contributions . 10 . 7554/eLife . 07090 . 021Figure 6 . Total cellular contributions to tissue shear throughout the WT wing blade . ( A ) Shows the tissue shear rate ( blue ) over time , and shear rates contributed by cell rearrangements ( red ) , cell shape changes ( green ) , cell divisions ( orange ) , and correlation effects ( magenta ) , averaged throughout the wing blade . These averages were taken over the tracked region shown in Figure 3A by averaging nematic tensors throughout the wing blade . The resulting quantities were further projected onto the PD axis and averaged over the three WT videos . Ribbons indicate the standard deviation between wings . The sign of the shear rate defines its orientation ( >0 is PD-oriented and <0 is AP-oriented ) . ( B ) Shows the accumulated tissue shear over time throughout the blade , and the accumulated contributions of each cellular process ( color code as in A ) . ( C ) Pattern of local tissue rotation rate at 21 hr after puparium formation ( APF ) . The local tissue rotation rate ωm is plotted separately for each triangle m . Red circles correspond to a counter-clockwise rotation and blue circles correspond to a clockwise rotation . The area of each circle scales with the absolute rotation rate . ( D ) The spatial power spectrum of the local tissue rotation rate corresponding to the pattern in panel C ( see Appendix 1 , ‘Power spectrum of local tissue rotation’ ) . The power spectrum is a function of a wave vector q = ( qx , qy ) , which is measured in units of a typical cell diameter d0 = 4 μm . The two peaks in the power spectrum at qpeak ≈ ( 0 , ±0 . 3d0/2π ) correspond to the existence of horizontal bands of alternating tissue rotation that are separated by about 1 . 5 cell diameters ( compare panels F , G ) . ( E ) Correlation effects contributing to shear along the PD axis , Dxx ( magenta curve ) . Dxx can be decomposed into an area expansion part Dxxe ( green curve ) , which corresponds to a correlation between the local area expansion rate vm and local triangle elongation Qxxm: Dxxe=− ( 〈vmQxxm〉−〈vm〉〈Qxxm〉 ) , and into a rotational part Dxxr ( blue curve ) , which corresponds to a correlation between the local tissue rotation rate ωm and local triangle elongation: Dxxr≃2 ( 〈ωmQxym〉−〈ωm〉〈Qxym〉 ) ( see Appendix 1 , ‘Large-scale shear in the absence of topological transitions’ ) . The rotational part dominates the shear by correlation effects . ( F ) Enlargement of the rotation pattern in panel C with an additional indication of the pattern of the local shear rate tensor by green bars . Length and orientation of a bar correspond to magnitude and axis of the local shear rate , respectively . The axis of local shear is correlated with the sign of local rotation ( indicated by red and blue circles ) . ( G ) The same region of the wing in the subsequent frame ( about 5 min later ) . Three corresponding triangles in panels F and G are colored in cyan , yellow and orange , respectively . The patterns of local shear and rotation change on time scales of minutes . ( H ) A correlation of local rotation and local shear within bands as shown in panels F , G corresponds to bands of alternating simple shear . ( I ) Contribution to the shear due to correlation effects of the group of triangles that are going to disappear due to a T1 transition within nine video frames ( <45 min ) ( Appendix 1 , ‘Role of T1 transitions in the correlation-induced shear’ ) . Inset: the area of this group is small compared to the total blade area , although it accounts for a significant amount of shear due to correlation effects in the blade . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 02110 . 7554/eLife . 07090 . 022Figure 6—figure supplement 1 . The shear decomposition method effectively describes tissue deformation and cellular contributions to tissue shear . ( A ) Verification of the shear decomposition method . The total tissue shear rate can be obtained by averaging changes in triangle elongation , without redefining triangles after topological changes ( blue curve , see Appendix 1 , ‘Decomposition of the large-scale tissue shear rate’ ) . The same values are obtained independently by decomposing shear into the cellular contributions due to cell shape changes , cell divisions , cell extrusions , cell rearrangements and correlation effects to shear ( pink curve ) . ( B ) Independent methods quantitatively account for the shape change of the wing blade . The blade shape was first characterized by a nematic Qxxt determined by the outline of the tracked region . The change of this nematic with respect to its initial value Q0t over time is shown ( dotted yellow line ) together with the cumulative tissue shear v˜xx ( blue ) obtained from the triangle method . ( C ) Here , we illustrate a scenario without overall tissue shear and where cell elongation does not change in the long run . However , the shear contribution by T1 transitions T is positive along the horizontal axis . This shear contribution is exactly compensated by the correlation effects D . ( D ) As a proxy for cell boundary orientation , we use the orientation of the line ( yellow ) connecting the centers ( blue points ) of the two corresponding triangles ( red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 022 Does the cumulative tissue shear we calculate account for the shape change of the wing blade ? To verify this , we characterize the blade shape by a nematic determined by the outline of the tracked region ( see Appendix 1 , ‘Characterization of wing blade anisotropy’ ) . The change of this nematic with respect to its initial value over time is shown together with the cumulative tissue shear obtained by the triangle method ( Figure 6—figure supplement 1B ) . These quantities agree well , indicating that the shear projected on the PD axis accounts for the main features of tissue shape changes . Thus , we can now use the shear decomposition method to discuss how different cellular events contribute to shape change of the wing over time . On average , the wing shears smoothly along its PD axis between 17 and 32 hAPF as the hinge contracts . In contrast , the cellular processes that combine to produce tissue shear change over time . During the first 6 hr of our videos , shear caused by cell elongation in the PD axis ( green curves in Figure 6A , B ) is even larger than PD tissue shear ( blue curves ) . This shows that cell elongation increases more than the tissue elongates suggesting that active cellular processes also contribute to PD cell elongation . Subsequently , starting at about 22 hAPF , cell elongation decreases although the blade continues to elongate . These discrepancies between cell shape changes and tissue shape changes require topological changes in the cell network . To more clearly discuss these events , we define two distinct phases of wing morphogenesis ( phases I and II ) that are separated by the peak of cell elongation occurring at about 22 . 5 hAPF . Quantifying shear caused by T1 transitions and by cell elongation reveals that they change dynamically throughout morphogenesis with a striking reciprocal relationship ( green and red curves in Figure 6A , B ) . This reciprocal relationship accounts to a large extent for the discrepancies between cell elongation and tissue shear . It further suggests that the active contribution to cell elongation ( i . e . , the amount of cell elongation that exceeds tissue shear ) may be linked to AP-oriented T1 transitions—the orientation of these T1s , which work against the observed tissue shear , suggests that they are autonomously controlled . Active AP-oriented T1 transitions could produce PD cell elongation if mechanical constraints prevent the wing from shearing . In principle , it is also possible that active PD-oriented cell shape changes could produce AP-oriented T1 transitions under the same constraints . Cell divisions also contribute to PD shear in the wing blade . Although cell divisions initially cause a small amount of AP shear , their direction changes during phase I such that their net contribution shears the wing in the PD axis . In addition , correlation effects produce significant shear in the AP direction and contribute most strongly at the time that T1 transitions are changing their orientation . In summary , the continuous large-scale deformation of the wing blade emerges from complex patterns of cell dynamics on small scales . During phase I , cells undergo AP-oriented T1 transitions while elongating in the PD axis . Cell divisions during phase I contribute shear along the PD axis . During phase II , the orientation of T1 transitions shifts to the PD axis and cells relax their shape . Correlation effects contribute AP shear , and peak roughly at the time that T1's shift their orientation . To ask whether the cellular contributions to tissue shear were independent of each other , we compared the sum of the variances of the final cellular contributions to tissue shear with the variance of final tissue shear itself . The ratio of these values is about 25 , indicating that the cellular contributions to tissue shear—like the contributions to area change—are not independent of each other . Thus , the overall tissue shear is more reproducible than would be expected from the variations of the cellular contributions . What exactly are cells doing that results in correlation effects ? We found that shear due to correlation effects was mainly generated by correlations between local elongation and rotations ( Figure 6C–I ) . To investigate this further , we determined the magnitude and orientation of shear and the rotation rate associated with each triangle for each frame of the video . We observed that triangles rotated and sheared in striking spatial patterns that rapidly fluctuate in time ( Figure 6C , F , G ) . These patterns correspond to rows of correlated shear and rotation that are distributed throughout the wing blade . To characterize these correlated patterns , we calculated the spatial power spectrum of the local tissue rotation rate ( Figure 6D , Appendix 1 , ‘Power spectrum of local tissue rotation’ ) . This revealed that shear and rotation are correlated in regions corresponding to PD-oriented rows that were about 3–7 triangles long . These rows consist of triangles that all rotate and shear in the same direction . The rows are interspersed with other rows of similar length with mirrored patterns of shear and rotation ( note blue and red rows in Figure 6F , G ) . Such a pattern of rotation and pure shear is characteristic of neighboring rows of triangles and cells undergoing simple shear in alternating directions ( Figure 6H ) . This would occur if PD-oriented rows of cells slide with respect to each other . As discussed above , such rearrangements can indeed contribute to correlation terms ( see cartoon in Figure 4H ) . If rows of cells slide past each other , cells typically engage in T1 transitions . Since the peak of AP correlation effects coincided with a shift in the net orientation of T1 transitions ( i . e . , when the red curve in Figure 6A crosses 0 ) , we wondered whether correlation effects could be associated with T1 transitions at this time ( see also Figure 6—figure supplement 1C ) . We therefore examined whether correlation effects were associated with a particular type of topological change . Indeed , correlation effects are mostly accounted for by those cells that are about to undergo a T1 transition within the next 9 frames , although they cover only a small fraction of the total area ( Figure 6I , Appendix 1 , ‘Role of T1 transitions in the correlation-induced shear’ ) . For rows of cells to slide past each other , cells would have to undergo a peculiar type of T1 transition in which the orientation of the boundaries gained and lost are similar . Boundary orientations around T1 transitions are difficult to measure , because boundary length is small . The triangulation method provides us with a more robust measure of a bond orientation . As a proxy for cell boundary orientation , we use the orientation of the line connecting the centers of the two corresponding triangles ( Figure 6—figure supplement 1D ) . We calculated the proportion of connection losses or acquisitions occurring in different directions over the course of morphogenesis , and compared them with the distribution of connection angles in general . We observe that at early times , when T1-induced shear is mainly along the AP axis , connections are typically lost along the PD axis and gained along the AP axis ( Figure 7A–D ) . At intermediate times , when the correlation term is maximal , both lost and gained connections were oriented along the PD axis ( Figure 7B ) . This is consistent with the unusual T1 transitions associated with sliding cell rows . Finally , as T1 transitions begin to cause net PD shear , AP connections are preferentially lost and PD connections are gained ( Figure 7D ) . Thus , as T1 transitions shift from an AP- to a PD-oriented shear , they pass through an intermediate state where connection gains and losses are still oriented but do not cause shear . Interestingly , at the same time , the correlation term has maximal magnitude . This suggests that the correlation effects are related to these unusual patterns of T1 transitions at intermediate times . 10 . 7554/eLife . 07090 . 023Figure 7 . Changes in angular distribution of lost and newly formed cell–cell junctions . ( A–D ) Effective proportions of cell–cell connections that are lost ( blue ) or gained ( red ) in different directions as a consequence of cell rearrangements in a WT wing . To calculate these effective proportions , we subtracted the angular distribution of all cell boundaries from the angular distribution of cell–cell connections that were lost or gained by cell rearrangements , revealing the orientation of cell boundaries with a disproportionate tendency to be lost or gained . Rose diagrams show angles of cell boundaries that are more likely to be gained ( red ) or lost ( blue ) at specific times during the video corresponding to important changes in cell dynamics: ( A ) 18 . 9 hAPF ( peak of negative shear rate by cell rearrangements ) , ( B ) 21 hAPF ( peak of correlation effects ) , ( C ) 21 . 5 hAPF ( shear rate by cell rearrangements crosses zero ) and ( D ) 24 . 5 hAPF ( peak of positive shear rate by cell rearrangements ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 023 To investigate which cellular events depended on epithelial stresses , we quantified shape changes and cellular contributions to tissue shear in the dumpyov1 mutant wing blade ( Figure 8A , B ) . As a complementary approach , we also studied these events in wings that had been subjected to laser severing ( Figure 8C–F and Figure 8—figure supplement 1 ) . In dumpyov1 wings , cells experience hinge contraction but the counterforces exerted by cuticle connections seem to be reduced ( Figure 2A–A′′ ) . Tissue shear is dramatically altered in dumpyov1 mutant wings as compared to WT—instead of shearing in the PD axis , these wings shear on average along the AP axis ( Figure 8A , B ) . Examining the different contributions to tissue shear in dumpyov1 wings shows that the rates of AP shear caused by T1 transitions and by correlation effects are similar to WT and persist for longer times . Thus , these processes are likely to be autonomously driven . By the end of the video , they cause more accumulated AP-oriented shear than in WT . Analogously , cell divisions cause more cumulative PD shear than in WT—consistent with the increased number of cell divisions in the dumpyov1 mutant wing . In contrast , cell elongation during phase I causes less PD shear than in WT . Thus , PD-oriented epithelial stresses must contribute to PD cell elongation . Interestingly , the increase of cell elongation in the PD axis still exceeds the increase of elongation of the blade in the dumpyov1 mutant wing . This suggests that autonomous cellular processes cause the residual PD cell elongation in dumpyov1 mutant wings . Finally , PD shear by T1 transitions in phase II is smaller than in WT . This is not due to a premature cessation of T1's—indeed quantifying the rate of T1 transitions ( regardless of orientation ) shows that they occur at a higher rate and for a longer time than in WT wings ( Figure 8—figure supplement 2 ) . Rather , T1 transitions fail to orient as effectively with the PD axis in the dumpyov1 mutant wing ( see shear patterns in Figure 5—figure supplement 1B–B′′ ) . 10 . 7554/eLife . 07090 . 024Figure 8 . Total cellular contributions to tissue shear throughout perturbed wing blades . ( A–F ) Show total shear rates ( A , C , E ) and total accumulated shear ( B , D , F ) , along with their cellular contributions , in the dumpyov1 wing blade ( A , B ) , and wing blades severed either distally ( C , D ) or proximally ( E , F ) before hinge contraction ( ∼16 hAPF ) . Blue = total tissue shear , Red = shear due to T1 transitions , Green = shear due to cell elongation change , Orange = shear due to cell division , Magenta = shear due to correlation effects . Corresponding plots for WT wings ( identical to those in Figure 6 ) are inset in the upper right corners of ( A , B ) for the purposes of comparison . Insets in left corners of ( A , C , E ) show the tracked regions of each wing in green at 32 hAPF . To plot accumulated tissue shear in laser ablated wings , we offset the calculated accumulated tissue shear ( blue ) by a value corresponding to the difference in blade elongation before ablation and at start of recording ( see Figure 3—figure supplement 2 ) . All videos were aligned in time by taking the histoblast nests fusion time as a reference at about 26 . 5 hAPF . All cumulated shear curves start at 16 . 2 hAPF , which is the earliest common time point registered in all compared videos , including the dist-sev#2 video shown in Figure 8—figure supplement 1A , B . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 02410 . 7554/eLife . 07090 . 025Figure 8—figure supplement 1 . Total cellular contributions to tissue shear throughout perturbed wing blades . ( A-C ) Show total shear rate ( A ) and total accumulated shear ( B ) , along with their cellular contributions , in a distally severed wing blade ( dist-sev#2 ) . A small piece of WT tissue ( about 500 cells ) was ablated in the blade at ∼16 hAPF , leaving a small distal piece of tissue . Left corner inset in ( A ) shows the severed wing at 32 hAPF , with the tracked region highlighted in green . Blue = total tissue shear , Red = shear due to T1 transitions , Green = shear due to cell elongation change , Orange = shear due to cell division , Magenta = shear due to correlation effects . The offset for accumulated tissue shear ( blue in panel B ) was calculated as in Figure 8 ( see also Figure 3—figure supplement 2 ) . ( C ) Shows the wing blade aspect ratio ( ar ) in WT unperturbed wings at 16 hAPF and 32 hAPF , and in laser ablated wings at 32 hAPF . The aspect ratio is obtained from the measured blade elongation by the relation ar = exp ( 2Qxx ) , ( see ‘Materials and methods’ , Measurements of wing dimensions ) . Red error bars: standard deviation over the 3 WT wings . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 02510 . 7554/eLife . 07090 . 026Figure 8—figure supplement 2 . Effect of mechanical perturbations on T1 transitions . ( A , B ) Show the rate of T1 transitions over time in unperturbed WT wing blades and under different genetic and mechanical perturbations . ( A ) Shows WT ( blue ) , dumpyov1 ( green ) , and the three wings that were mechanically severed before hinge contraction: dist-sev#1 ( dark red ) , dist-sev#2 ( purple ) and prox-sev ( yellow ) . ( B ) Shows T1 transition rates in the two wings suffering laser ablation in the extra-cellular matrix ( ECM ) just prior to phase II: ECM_AntCut ( cyan ) , ECM_DistCut ( black ) , along with WT ( blue ) for comparison . ( C–F ) Effective proportions of cell–cell connections that are lost ( blue ) or gained ( red ) in different directions in the severed hinge-blade video ( prox-sev ) . These effectives proportions were calculated for all time points as described in Figure 7 , and are displayed at 17 . 5 hAPF ( peak of negative shear rate by cell rearrangements ) , 21 hAPF ( peak of correlation effects ) , 24 hAPF ( shear rate by cell rearrangements crosses zero ) and 25 . 5 hAPF ( peaks of positive shear rate by cell rearrangements ) . Note that the timing of these events differs slightly from the corresponding times in unperturbed wings . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 02610 . 7554/eLife . 07090 . 027Figure 8—figure supplement 3 . Total cellular contributions to tissue shear in wing blades after laser-severing of the extracellular matrix . ( A ) Blade anisotropic deformation rate and its cellular contributions in a WT wing in which the extracellular matrix was distally ablated shortly before phase II ( ECM_DistCut ) . ( B ) Blade anisotropic deformation rate and its cellular contributions in a WT wing in which the extracellular matrix was anteriorly ablated shortly before phase II ( ECM_AntCut ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 027 As a complementary approach , we asked how tissue shear and cell dynamics were altered in wings that had been subjected to laser severing . We first examined wings that had undergone laser severing before hinge contraction at the distal tip of the wing blade ( Figure 8C , D and Figure 8—figure supplement 1A , B ) . These wings experience mechanical conditions similar to those in dumpyov1 wings in that they undergo hinge contraction but cannot attach properly to the distal cuticle . Consistent with this , the final shape of distally severed wings resembles that of dumpyov1 wings . Furthermore , distally severed wings show similar tissue shear and cellular contributions to shear as dumpyov1 mutant wings . These observations suggest that weakened connections to the distal cuticle are key to the altered wing shape of dumpyov1 mutants . To perturb PD stresses even more strongly , we severed the region between the hinge and the blade before hinge contraction occurred ( Figure 8E , F ) . These wings are not subjected to externally generated PD stresses at all , although they may still experience autonomously-induced stresses . We expected that cellular events that depended on externally generated stresses would be even more strongly perturbed in proximally severed wings than in dumpyov1 or distally-severed wings . The proximally severed wing undergoes significant but short-lived PD shear during the molting process until 18 hAPF . Subsequently , PD shear stops and becomes even negative by 20 hAPF . Cells do not elongate in the PD axis as much as in WT wings , although PD cell elongation still exceeds PD tissue shear . Again , this suggests that only a fraction of PD cell elongation is normally caused by external stresses , and that autonomous cellular events must produce the residual PD cell elongation in proximally severed wings . PD shear due to cell division is larger than in WT ( as it is in dumpyov1 and distally severed wings ) confirming that these divisions do not depend on external stresses . Furthermore , like dumpyov1 and distally severed wings , proximally severed wings undergo greater AP shear resulting from correlation effects . Thus the cellular events underlying correlation effects produce even more shear when stresses are reduced . However later , T1 transitions fail to generate significant PD shear in proximally severed wings . The reduction in T1-dependent PD shear is much stronger than in either dumpyov1 or distally severed wings , confirming that reorientation of T1 transitions in phase II is dependent on externally generated PD stress . At the very beginning of the video , PD-oriented connections are preferentially lost and AP-oriented connections are preferentially gained , consistent with the AP shear caused by T1 transitions at this time ( Figure 8—figure supplement 2C–F ) . However , unlike WT , the preferential loss of connections never shifts towards the AP axis . Loss of connections remains biased towards the PD axis throughout the video—despite the fact that net shear caused by T1 transitions becomes very small . Shear caused by T1 transitions becomes small because the preferred orientation of gained connections gradually shifts from the AP to the PD axis . By the end of the video , both the assembly and disassembly of cell boundaries are preferentially oriented along the PD axis . These observations suggest that PD-oriented cell boundaries have a greater tendency to disassemble than those oriented at other angles , and that this is an autonomous , planar polarized feature of wing epithelial cells . To disturb connections to the overlying cuticle without damaging the wing epithelium we disrupted apical extracellular matrix between the wing margin and the cuticle shortly before the onset of phase II , when this region becomes accessible ( see Videos 8 , 9 ) . When anterior connections are severed , the wing blade shears even more in the PD axis while the area decreases slightly ( Figure 8—figure supplement 3 ) . This suggests that these connections restrain the narrowing of the wing blade in the AP axis . Increased PD tissue shear is mainly a consequence of T1 transitions . Since severing anterior connections reduces AP stress while PD stress does not change , this supports the idea that T1 events during phase II are oriented by anisotropic stress . 10 . 7554/eLife . 07090 . 028Video 8 . Video of a WT wing in which the extracellular matrix was laser-ablated anteriorly . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 02810 . 7554/eLife . 07090 . 029Video 9 . Video of a WT wing in which the extracellular matrix was laser-ablated distally . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 029 When distal connections are severed , cells rapidly reduce their elongation and area ( Figure 8—figure supplement 3 ) . Surprisingly , the rate of PD shear caused by T1 transitions continues to increase as it does in WT at this stage . However , after about 4 hr the PD shear rate due to T1 transitions decreases prematurely . Since the analysis of dumpyov1 and proximally severed wing shows that PD T1 transitions are stress dependent , this suggests that there is a time delay between the change in tissue stress and the resulting T1 transitions . Taken together , the analyses of dumpyov1 wing blades and laser-severed wing blades distinguish autonomously driven cellular processes from those induced by tissue stresses . PD cell elongation and PD-oriented T1 transitions clearly depend on tissue stresses , whereas PD-oriented cell divisions are autonomously driven . AP-oriented T1 transitions and a corresponding fraction of PD-oriented cell elongation are also driven autonomously , as are the cellular events underlying correlation effects . The overall picture that emerges is that changes of wing blade shape arise due to force balances that involve stresses exerted at the boundary of the tissue , and internal tissue stresses . Boundary stresses are due to hinge contraction and to the resistance of extracellular matrix attachments to the cuticle . Internal tissue stresses are generated by cell autonomous processes , like T1 transitions , and by elastic cell deformations . Tissue mechanics depends strongly on elastic connections of the wing to the cuticle . We must now understand how these isotropic and anisotropic mechanical stresses in the tissue , combined with boundary stresses , lead to cell and tissue remodeling . The interplay between boundary stresses and forces generated in the tissue is complex and requires a physical approach . We now present a continuum mechanical theory to understand these force balances and to calculate both tissue and cell shape changes . We first define tissue material properties , starting with elastic properties , adding cell autonomous stresses and tissue shear due to topological changes . We then introduce elastic linkers to the surrounding cuticle . Finally , we compare predictions of this theory to the experimentally measured cell and tissue shape changes and determine key biophysical parameters characterizing tissue material properties . We first investigated stresses present in the tissue . In an elastic tissue , tissue deformations stem from cell shape changes and cell elasticity is responsible for tissue elasticity . For small deformations , Hooke's law states that the mechanical stress in the material is proportional to its deformation . We write the isotropic part of the stress ( 3 ) P=−K¯ lnaa0 , where P denotes two-dimensional tissue pressure , a and a0 are cell area and preferred cell area respectively , and K¯ is the area compressibility . The preferred cell area a0 changes when cells divide . For small ( a − a0 ) /a0 , Equation 3 corresponds to Hooke's law . Equation 3 implicitly contains the active contribution to pressure , which influences the preferred cell area a0 ( see Appendix 2 , ‘Constitutive equation for the tissue stress’ ) . We now focus on the anisotropic part of the stress σ∼ , also called the shear stress . For simplicity , we write the elastic anisotropic stress in the form of Hooke's law σ∼e=2KQ , where K is a shear elastic modulus and Q is the cell elongation . In this expression , the cell shape is isotropic in the absence of anisotropic stresses . However , in tissues , planar polarized cells may spontaneously elongate . Therefore , we postulate the following constitutive equation for the anisotropic tissue stress ( 4 ) σ∼=2KQ+ζ , where ζ is a tensor that can be interpreted as a cell autonomous active stress related to spontaneous cell elongation . To test whether Equation 4 accounts for anisotropic stresses present in the tissue , we further analyzed the circular laser ablations performed in different regions of WT and dumpyov1 wings at 22 hAPF ( Figure 2I , I′ ) . We defined a shear rate v∼cut that characterizes the anisotropic recoil of the circular cut boundary into an elliptic shape ( see ‘Materials and methods’ , Analysis of circular laser ablations ) . We plotted the projection of this shear rate on the PD axis v∼xxcut as a function of the projected average cell elongation Qxx ( Figure 9A ) . Cell elongation was determined as the average cell elongation within the corresponding region in unperturbed wings . We found that the anisotropic part of the shear rate varied linearly with cell elongation ( Figure 9A ) . The positive slope of this linear relation indicates that the shear modulus K is positive . In this argument , we use the recoil shear rate v∼cut as a measure proportional to tissue stress . A linear fit to this data also has a positive intercept , corresponding to a positive ζxx in Equation 4 . This implies that wing blade cells would spontaneously elongate along the AP axis in the absence of stress . Equivalently , in the absence of cell deformation ( Q = 0 ) wing blade cells are exerting higher contractile stress in the PD axis than in the AP axis . The relative contributions of elastic stress and stress associated with spontaneous cell elongation can be quantified from this linear fit . The ratio of the intercept and the slope of the linear fit equals ζxx/2K , which leads to the estimate ζxx/K = 0 . 333 ± 0 . 003 in WT and ζxx/K = 0 . 316 ± 0 . 004 in dumpyov1 wing . Experimental data obtained in WT and dumpyov1 wings fall on similar lines ( Figure 9A ) , suggesting that internal mechanical properties of the tissue are not perturbed in dumpyov1 mutant wings . Since dumpy mutant cells are less elongated than those of WT ( Figure 9—figure supplement 1A ) , their similar mechanical properties imply that they are under less anisotropic stress consistent with the loosened connections to the overlying cuticle . 10 . 7554/eLife . 07090 . 030Figure 9 . Dependency of stresses and topological changes on cell deformation . ( A ) Anisotropic recoil of circular cut boundaries after laser ablation of the blade at 22 hAPF , as a function of the projected average cell elongation in the region of the cut , for WT and dumpyov1 wings . ( B ) Isotropic recoil of circular cut boundaries after laser ablation of the blade at 22 hAPF , as a function of the average cell area in the region of the cut , for WT and dumpyov1 wings . ( C ) Topological changes are driven either by cell elongation or by polarity-dependent processes , with a delay τd . ( D ) Shear due to topological changes as a function of cell elongation in the blade . Experimental points are color-coded according to time . The black line is a fit of Equation 5 shown in ( C ) to the experimental data . Due to a delay τd in the response of topological changes to cell elongation , data points follow a spiraling curve during wing morphogenesis . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 03010 . 7554/eLife . 07090 . 031Figure 9—figure supplement 1 . Cell elongation state . ( A ) Average cell elongation state in the wing blade as a function of time . This quantity is obtained from the triangulated network of cells in WT and dumpyov1 wings ( blue and green curves , respectively ) . ( B ) Average final cell elongation as a function of the final cell area in the blade , for WT , dumpyov1 and cdc2E1-E24 wings , as well as WT wings in which the extracellular matrix was ablated . Note that in cdc2E1-E24 mutant wings , cell divisions were inhibited from about 16 hAPF on . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 03110 . 7554/eLife . 07090 . 032Figure 9—figure supplement 2 . Shear due to topological changes as a function of cell elongation in the blade for 6 analyzed wings . Experimental points are color-coded according to time . Black lines are the results of a joint fit of Equation 5 to the six wings , with a single choice of the parameters τr and τd , and different values of λxx ( see Table 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 032 If we perform the same analysis for the isotropic part of the tissue stress , we use the area expansion rate of the cut circle during recoil as a proxy for negative tissue pressure −P . Plotting this rate as a function of average cell area in a given region , we find that a0 is smaller than cell area a , which reveals that the tissue is under contractile tension or negative P ( Figure 9B ) . This is consistent with the observed tissue area contraction in dumpyov1 mutant and severed wings . However , parameter values a0 and K¯/K cannot be reliably estimated by this method because the average cell area a varies too little , and because cell divisions may introduce heterogeneity in the preferred cell area a0 . Overall , laser ablation experiments indicate that there are two contributions to anisotropic stress in the PD axis . First , blade cells are elastically deformed along the PD axis , in response to hinge contraction and margin attachments . Second , these cells would tend to spontaneously shorten in the PD axis even in the absence of external stresses . So far , we have characterized the elastic properties of the tissue . On long time scales , elastic stresses can be relaxed by topological changes of the cellular network generating tissue shear viscosity . We therefore now develop a model describing the average rate and orientation of tissue shear R due to such topological changes . Although R includes the shear caused by T1 and T2 transitions , cell divisions and correlation effects , it is dominated by T1 transitions ( Figure 6A , B ) . We asked how the shear rate R due to topological changes is regulated in the wing blade . If cells in the tissue are elongated , this elongation could drive topological changes that give rise to the shear rate R = Q/τr . Here , τr can be interpreted as the time during which cell elongation relaxes . Indeed , τr is also the time beyond which the tissue behaves as a viscous fluid with viscosity Kτr , due to elastic stress relaxation by topological changes including cell rearrangements and cell divisions . In addition , our observations of laser-severed wings suggest that the response of R to cell elongation is not instantaneous , but follows with a time delay τd of a few hours ( see previous discussion and Figure 8—figure supplement 3 ) , which also must be incorporated in the theory . In addition to elongation-induced rearrangements , planar polarized tissues may undergo oriented topological changes even in the absence of cell elongation . For instance , topological changes could be driven autonomously by increased contractility of certain bonds . Taking into account all these contributions , we postulate the following relation between the shear rate R and cell elongation ( Figure 9C ) : ( 5 ) ( 1+τdddt ) R=1τrQ+λ . Here , the delayed response to cell elongation is introduced via the time derivative of the shear R . This way of implementing a delay implies that the current value of cell elongation has the strongest impact on shear due to topological changes and the effect of recent values of cell elongation fades exponentially over time , disappearing beyond the time τd . Shear driven by polarity dependent processes is characterized by the tensor λ . Thus , even if cells are not elongated ( Q = 0 ) , λ drives shear due to topological changes . For shear along the AP axis λxx < 0 , while for shear along the PD axis λxx > 0 . In order to verify whether Equation 5 captures the dynamics of the shear created by topological changes during pupal wing morphogenesis , we plotted Rxx vs the cell deformation component Qxx for different times ( Figure 9D ) . For an instantaneous response of R ( no delay , τd = 0 ) these points would fall on a straight line . For non-zero delay , the history of the process matters and these points follow a curve spiraling towards a fixed point ( see Appendix 2 , ‘Effect of delay in topological changes’ ) . We find indeed that experimental data for each WT wing follow a spiral , confirming the existence of a delay . A fit of the theory to the data allows us to estimate the coefficients in Equation 5 ( τd = ( 3 . 7 ± 0 . 9 ) h , τr = ( 1 . 8 ± 0 . 6 ) h and λxx = ( −0 . 10 ± 0 . 04 ) h−1 ) . Interestingly , λxx is negative , indicating that polarity driven topological changes create AP shear—consistent with conclusions from laser ablation experiments . Polarity driven AP-oriented topological changes may be related to increased tension of PD-oriented as compared to AP-oriented cell bonds . This anisotropy of cell bond tension could account for the positive value of ζ in Equation 4 because it would tend to elongate cells in the AP-axis . The fact that a constant negative value of λxx accounts for the experimental data suggests that the tendency to undergo polarity-driven topological changes exists during both phase I and II . The transition from phase I to phase II occurs when cell elongation-driven topological changes begin to exceed the effect of polarity-driven topological changes . During phase I topological changes are largely polarity driven along the AP axis . Equation 2 then implies that these topological changes , together with hinge contraction , contribute to the buildup of cell elongation in the PD axis . Once flows have stopped , the value of λxx determines the final value of cell elongation . Indeed , Equations 2 , 5 predict that in a steady state without shear flows ( R = 0 , dQ/dt = 0 ) , the cell elongation is given by Q = −τr λ . We can therefore use the cell elongation at the end of the videos , where the tissue is almost stationary , to estimate τr λxx . In other words , the final cell elongation should be independent of external stresses . Instead final cell elongation is governed only by the internal dynamics of the tissue , that is , the tendency to undergo AP-oriented topological changes λxx as well as the stress relaxation time-scale τr . Interestingly , examining different perturbed conditions , we find that the final cell elongation and hence λxx depend on cell area ( Figure 9—figure supplement 1B and Appendix 4 , ‘Area dependence of λxx’ ) . Overall , we find that the shear created by topological changes is driven in part by cell elongation and in part by cell polarity-dependent processes . Furthermore , topological changes respond to cell elongation with a delay of several hours . Equations 1–5 constitute a full theory for tissue mechanics taking into account cell shape changes . We now want to test whether this tissue description quantitatively accounts for cell and tissue shape changes during wing morphogenesis . We ask here whether the tissue and cell properties described by our equations can quantitatively account for the observed changes in the shapes of hinge and blade as well as the elongation and topological changes of their constituent cells . We use a simplified description of the wing where the hinge and blade are represented by two rectangles that are attached to each other . These rectangles have the same areas as the hinge and blade and they undergo pure shear that is the average shear in the respective parts of the wing . To describe tissue flow , Equations 1–5 , which characterize local tissue properties , have to be complemented by the condition of force balance . At the boundary of the tissue , elastic linkers connected to the cuticle impose external forces that have to balance tissue stresses . In our model the rectangles are therefore connected to an external frame by elastic elements ( Figure 10A and Figure 10—figure supplement 1 ) . These external elements correspond to the extracellular matrix and the frame corresponds to the cuticle . The elastic elements provide resistance to extension of blade and hinge along the PD and AP axes ( see Appendix 3 , ‘Boundary stresses’ ) . 10 . 7554/eLife . 07090 . 033Figure 10 . Continuum mechanical model of wing morphogenesis . ( A ) Schematics of the wing model: the hinge and blade are represented by rectangles . Within each rectangle , the tissue is subjected to cell-autonomous anisotropic and isotropic stresses ζ and ζ¯ , and to topological changes driven by cell polarity-dependent processes λ . The complex elastic material connecting the wing to the cuticle is represented by AP-oriented elastic links ( green and red springs on the cartoon ) and PD-oriented springs ( see Figure 10—figure supplement 1 ) . In WT wings , the blade distal end is fixed , while it is free to move in the dumpyov1 mutant . ( B ) ( Left ) Experimental ( solid line ) and theoretical ( dashed line ) time courses of tissue shear rate ( blue curves ) , cell elongation change ( green curves ) and shear due to topological changes ( red curves ) , in the blade and along the PD axis . ( Middle ) Experimental and theoretical time courses of cumulative tissue shear ( blue curve ) , cell elongation ( green curve ) and cumulative shear due to topological changes ( red curves ) , in the blade and along the PD axis . ( Right ) Experimental and theoretical cumulative relative blade area change . Model parameters were obtained by a fitting procedure to experimental data ( Tables 1 , 2 ) . The continuum mechanical model recapitulates cell shape changes and tissue flow during wing morphogenesis . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 03310 . 7554/eLife . 07090 . 034Figure 10—figure supplement 1 . Schematics of the rectangle model . ( A ) The hinge and blade are represented by two rectangles connected to an external frame representing the cuticle ( blue rectangle ) . AP-oriented elastic springs at the anterior and posterior sides , with elastic moduli kH and k , resist change of heights of the rectangle . In addition , two PD-oriented elastic springs constrain the length of the hinge and blade with elastic moduli kPD and kPDH . In WT wings , the blade is firmly attached distally to the cuticle . To represent the wing whose ECM is ablated distally , a free boundary is introduced on the distal side of the blade rectangle . Similary , anterior ablation of the ECM is represented by removing anterior springs , k = 0 . To represent the dumpyov1 mutant , distal links are removed while anterior and posterior links are weakened . ( B ) Geometry of rectangle deformation . The position of the hinge-blade interface is labeled xBH , the total length of the wing L , the heights of the hinge and blade hH and h . Lower plots: representation of the velocity profiles in the x and y direction in the hinge and blade , in the rectangle model , for WT wings . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 03410 . 7554/eLife . 07090 . 035Figure 10—figure supplement 2 . Continuum mechanical model in the hinge . ( Left ) Experimental and theoretical time courses of tissue shear rate ( blue curves ) , cell elongation change ( green curves ) and shear due to topological changes ( red curves ) , in the hinge and along the PD axis . ( Right ) Experimental and theoretical time courses of cumulative tissue shear ( blue curve ) , cell elongation ( green curve ) and cumulative shear due to topological changes ( red curves ) , in the hinge and along the PD axis . Model parameters were obtained by a fitting procedure to experimental data ( Tables 1 , 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 035 In each rectangle , anisotropic stresses and shear are described by Equations 4 , 5 , with different tissue parameters in hinge and blade . Isotropic internal stresses in the blade are described by Equation 3 . Preferred cell area a0 changes rapidly when cells divide and is modulated by the cell contractility ζ¯ ( see Appendix 2 , ‘Constitutive equation for the tissue stress’ ) . In the hinge , we do not solve the full mechanical problem but rather impose the observed hinge area contraction . We do so by adjusting the pressure in the hinge to match the observed hinge area . We start with rectangles whose areas and aspect ratio are consistent with those of hinge and blade at 16 hAPF ( see Appendix 3 ) . The initial conditions for Q and R are the observed average cell elongation and the initial shear rate due to topological changes . To trigger cell flows in the model , we turn on active stresses ζ , ζ¯ and polarity dependent topological changes λ . We solved Equations 1–5 in two rectangles that correspond to hinge and blade . Note that the pressure in the hinge is not determined by Equation 3 , but by imposing hinge area . In the blade , we used the measured rates of cell division and T2 transitions to calculate the blade area changes ( Equation 1 ) . We normalized all elastic moduli and friction coefficients to the blade shear modulus K . The solutions Q ( t ) , a ( t ) , v ( t ) , and v∼ ( t ) of Equations 1–5 depend on a set of parameters characterizing the tissue both in hinge and blade ( see Tables 1 , 2 ) . In addition , elastic coefficients describe constraints imposed by linkers at the boundary . Friction due to motion with respect to the cuticle is captured by a friction coefficient γ . Furthermore , to fully account for the observed flows , we found that we needed to add a contribution of tissue area viscosity η¯ to Equation 3 ( see Appendix 2 , ‘Constitutive equation for the tissue stress’ , Equation 52 ) . We include this term in all subsequent calculations . Laser ablation of the extracellular matrix is introduced in our model by removing the elastic linkers at the tissue boundary at the side of ablation . 10 . 7554/eLife . 07090 . 036Table 1 . Coefficients describing tissue properties in Equation 5 are fitted with single values of time-scales τr and τd in hinge and blade while λxx was allowed to vary among different wings ( see Figure 9—figure supplement 2 ) DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 036WT #1WT #2WT #3ECMAntCutECMDistCutDPBladeτr [h]1 . 7 ± 0 . 1τd [h]4 . 2 ± 0 . 3λxx [h−1]−0 . 11 ± 0 . 01−0 . 11 ± 0 . 01−0 . 10 ± 0 . 01−0 . 10 ± 0 . 01−0 . 068 ± 0 . 007−0 . 094 ± 0 . 008HingeτrH [h]4 . 6 ± 2τdH [h]2 . 4 ± 1λxxH [h−1]−0 . 05 ± 0 . 01−0 . 05 ± 0 . 01−0 . 04 ± 0 . 01−0 . 03 ± 0 . 01−0 . 01 ± 0 . 01−0 . 04 ± 0 . 01WT; wild type . 10 . 7554/eLife . 07090 . 037Table 2 . Parameters of the rectangle model are divided in three groups describing blade tissue properties , hinge tissue properties and external linksDOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 037WTECMDistCutECMAntCutDpTissuebladecell autonomous shear stressζxx/K0 . 333 ± 0 . 0030 . 316 ± 0 . 004shear elastic modulusK/K1cell area contractilityζ¯/K0 . 05 ± 0 . 03area elastic modulusK¯/K2 . 07 ± 0 . 09area viscosity coefficientη¯/K[h]49 ± 2hingecell autonomous shear stressζxxH/K0shear elastic modulusKH/K0External linksbladeeffective AP elastic constantkL0/K0 . 5 ± 0 . 100 . 005 ± 0 . 007effective PD elastic constantkPD/K4 . 91 ± 0 . 045 . 3 ± 0 . 2friction coefficientγ/K[h]21 . 3 ± 0 . 822 . 1 ± 0 . 6distal connections–YesNoYesNohingeeffective AP elastic constantkPDH/K67 . 8 ± 0 . 478 ± 2effective PD elastic constantkPDH/K9 . 50 ± 0 . 0716 . 8 ± 0 . 6friction coefficientγ/K[h]21 . 3 ± 0 . 822 . 1 ± 0 . 6Cell autonomous shear stress in wing blade of WT and dumpyov1 are determined from circular laser cut experiments . Unperturbed and mechanically perturbed WT wings are first simultaneously fitted using results listed in Table 1 . Then , the dumpyov1 wing is fitted keeping the values of hinge and blade tissue parameters the same as in WT . The effective anterior-posterior ( AP ) and PD elastic constants describe effects of external elastic elements providing resistance to changes in size of blade and hinge along the AP and PD direction . All quantities are normalized by the elastic shear modulus of the blade tissue K . Quantities containing spatial dimensions are also normalized by the initial length L0 of the WT wing . Uncertainties reported for the parameters in this table ( expect for the cell autonomous shear stress ζxx ) were determined by the fit . Note that they do not reflect uncertainties arising from approximations made in the rectangle model ( supplement section 4 ) and from pre-processing of experimental data ( supplement section 1 . 6 ) . We first analyzed WT wings in both unperturbed and mechanically perturbed conditions where only the extracellular matrix was ablated . Because blade tissue was not damaged by matrix ablation we expected that most parameters characterizing the hinge and blade would be identical to those in unperturbed wings . However , λxx can differ in wings characterized by different cell areas , as noted above . We therefore used the values of τr , τd and λxx determined by jointly fitting Equation 5 to the data ( see Table 1 and Figure 9—figure supplement 2 ) . In addition , we used experimentally determined values of ζ/K estimated from circular laser ablation experiments in WT wing blades ( Figure 9A ) . A similar fit is performed for the hinge and corresponding parameters are superscripted with an index H . This leaves 10 parameters unspecified . These characterize the isotropic stresses in the blade , the anisotropic stresses in the hinge , the stiffness of the elastic linkers at the boundary , and the coefficient describing external friction ( Table 2 ) . We estimated these parameters by performing joint fits of Equations 1–5 to the quantified time dependence of tissue area changes and shear in unperturbed and mechanically perturbed WT wings . We found that anisotropic stresses in the hinge are small compared to isotropic stresses and therefore we could set ζH and KH to zero . A single set of the remaining 8 parameters accounted well for all time dependent data ( see Figure 10B and Figure 10—figure supplement 2 and Table 2 ) . It is remarkable that this very simplified model can account for the main features of the complex shape changes that occur during pupal wing morphogenesis . The choice of rectangles indeed captures the average shear and area change that determine tissue shape change . Displaying the time evolution of the rectangles within a time-lapse video of the pupal wing reveals the close agreement between the calculated and observed shape changes of hinge and blade ( see Videos 10–13 ) . 10 . 7554/eLife . 07090 . 038Video 10 . Time evolution of the rectangles obtained from the rectangle model in a WT wing . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 03810 . 7554/eLife . 07090 . 039Video 11 . Time evolution of the rectangles obtained from the rectangle model in a dumpyov1 wing . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 03910 . 7554/eLife . 07090 . 040Video 12 . Time evolution of the rectangles obtained from the rectangle model in a WT wing where the extracellular matrix was ablated distally at the onset of phase II . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 04010 . 7554/eLife . 07090 . 041Video 13 . Time evolution of the rectangles obtained from the rectangle model in a WT wing where the extracellular matrix was ablated anteriorly at the onset of phase II . Note that length of the hinge is increased by a correction term described in Appendix 4 , ‘Fitting of the rectangle model to cell and tissue shape in the hinge and blade’ . DOI: http://dx . doi . org/10 . 7554/eLife . 07090 . 041 Our experimental data suggest that distal elastic connections to the cuticle are weakened in the dumpyov1 mutant wing . We therefore asked whether we could fit the rectangle model described above using WT tissue parameters but allowing parameters describing connections to the cuticle to change . We removed distal boundary connections to the blade ( Figure 10—figure supplement 1 ) as suggested by laser ablation experiments and allowed other parameters characterizing external linkers to change . We used these values and the measured value of ζ/K ( which is almost the same as in WT , see Figure 9A ) in fits of the full dynamics of the rectangle model to the dumpyov1 mutant data ( see Figure 10B and Figure 10—figure supplement 2 ) . These calculations show that distal and lateral connections to the wing blade margin are weakened , as expected ( see Table 2 ) . Overall , the continuum theory of epithelia outlined in Equations 1–5 together with appropriate boundary conditions recapitulates the dynamics of cell and tissue shape over 16 hr of morphogenesis . By fitting the theory to experimental measurements , we determine the values of tissue parameters characterizing intrinsic tissue time-scales , cell autonomous stresses , tissue elastic moduli and elastic moduli of external linkers . Using this theory to investigate mechanical parameters in a dumpyov1 mutant wing , we confirm the existence of Dumpy-dependent elastic connections between the wing and cuticle . Thus , this theoretical approach is a powerful tool for studying how mutations influence specific aspects of cell and tissue mechanics .
Developing tissues are active viscoelastic materials that generate and respond to mechanical stresses to change their shapes . How tissue material properties emerge from the properties and behavior of their many constituent cells , and how these properties quantitatively account for tissue shape change is a major question in developmental biology . These questions have been difficult to answer because of both experimental and theoretical limitations . Methodologies for large-scale imaging and image analysis of entire tissues at the cellular level were insufficient , and direct measurement of forces and stresses in vivo have been rare ( Hutson et al . , 2003; Farhadifar et al . , 2007; Rauzi et al . , 2008; Bosveld et al . , 2012; Bambardekar et al . , 2015 ) . Furthermore , theoretical approaches are required to understand how tissue shear emerges from different kinds of cellular processes . By developing new experimental and theoretical approaches , we have been able for the first time to account for the shape change of a developing tissue on the basis of its active and its viscoelastic properties . Our approach allows us to quantitatively understand how such tissue properties emerge from the interactions of its constituent cells . Our work is stimulated by previous approaches to tissue dynamics . For example , previous work on tissue tectonics highlighted the decomposition of tissue shear in cell shape changes and cell intercalation ( Brodland et al . , 2006; Graner et al . , 2008; Marmottant et al . , 2008; Blanchard et al . , 2009; Butler et al . , 2009 ) . Our triangle method refines these ideas using a geometrical scheme that defines and distinguishes exact cellular contributions to tissue shear . These are contributions due to cell deformation , cell division , T1 and T2 transitions . This approach has also led to the discovery of a new correlation effect that contributes significantly to tissue shear . This collective effect can result from local correlations between cell elongation and tissue rotation , or between cell elongation and cell area changes . For instance , this correlation effect can be produced by relative sliding of neighboring cell rows ( Figure 4H ) . In our work we combine this approach with a novel continuum mechanical model based on patterns of cell elongation . Our model combines visco-elastic material properties with additional cell autonomous so-called active stresses . These active stresses contribute to force balances and also can drive topological rearrangements . This model allows us to quantitatively discuss tissue deformations that emerge from the dynamic interplay of externally applied forces , internally generated stresses and the resulting collective cell rearrangements . Our model is related to continuum models of active gels and tissues ( Kruse et al . , 2005; Brodland et al . , 2006; Ranft et al . , 2010 ) that can be obtained by coarse-graining cell-based models ( Honda et al . , 1984; Graner and Glazier , 1992; Drasdo et al . , 1995; Chen and Brodland , 2000; Brodland et al . , 2007; Farhadifar et al . , 2007; Basan et al . , 2011 ) . However , our model differs from existing continuum models in several ways: it is directly based on measurable cellular contributions to tissue shear , it includes a new time scale corresponding to a delay for the generation of T1 transitions , and it introduces a cell-autonomous contribution to T1 transitions . Such cell-autonomous T1 transitions ( Irvine and Wieschaus , 1994 ) may for example be generated by orientation-dependent contractility of cell boundaries ( Bertet et al . , 2004; Skoglund et al . , 2008; Rauzi et al . , 2010; Sawyer et al . , 2011 ) . Overall , our approach can bridge scales from individual cellular events to cell flows and large-scale tissue shape changes . We find that the wing tissue shapes itself through patterned contractions and shear that occur in the context of patterned extra-cellular matrix ( ECM ) connections to a surrounding cuticular scaffold . Thus , by pulling on these connections , the tissue forces itself into the right shape . In dumpy mutants , these connections are weakened and the final tissue shape is dramatically altered as compared to WT ( Waddington , 1940; Carmon et al . , 2010 ) . Dumpy plays a similar role in shaping the Drosophila trachea in embryonic development ( Dong et al . , 2014 ) . The mechanical function of such ECM connections may be generally important during epithelial morphogenesis in vivo . Our quantitative analysis has provided new insights into the biology of pupal wing morphogenesis . We have discovered that cell-autonomous planar polarized T1 transitions occur early , during the first phase of the process . Surprisingly , they occur along the AP axis and actively increase cell elongation in the PD axis . Why were these AP-oriented T1 transitions not detected in our previous analysis ( Aigouy et al . , 2010 ) ? Comparing the rate of T1 transitions with the amount of shear they generate shows that the orientation of T1 transitions at this time is less focused than it is later when T1 transitions cause shear in the PD axis . However , because their overall rate is higher in phase I , a slight bias in their orientation causes significant AP shear . The fact that there is only a slight AP bias in these T1 transitions likely explains why they have not been detected before , and highlights the importance of the large-scale quantitative analysis that we employ here . What mechanisms might underlie AP-oriented T1 transitions that occur in phase I ? Quantitative analysis of laser ablation data indicates that the preferred shape of wing epithelial cells in the absence of external forces is not isotropic , but rather elongated in the AP axis . Increased contractility of PD-oriented cell boundaries could explain this preference , and would also tend to favor AP-oriented T1 transitions . We note that components of both the Fat and Core planar cell polarity pathways are enriched on PD-oriented cell boundaries during phase I ( Merkel et al . , 2014 ) . These systems are known to regulate both cell boundary tension and Cadherin turnover ( Classen et al . , 2005; Rogulja et al . , 2008; Mao et al . , 2011; Bosveld et al . , 2012; Warrington et al . , 2013; Nagaoka et al . , 2014 ) , and it will be interesting to investigate their involvement in orienting T1 transitions at this stage . Our theoretical analysis suggests that AP-oriented T1 transitions are unlikely to strongly influence the final shape of the wing . However , they have a strong influence on peak cell elongation and final cell shape . What function could this serve during morphogenesis ? One possibility is that AP-oriented T1 transitions influence the morphology of cuticular ridges on the adult wing surface . The cuticle of the adult wing is shaped in a reproducible pattern of ridges that form between cell rows ( Doyle et al . , 2008; Merkel et al . , 2014 ) . These ridges likely influence anisotropic mechanical properties of the wing during flight . Orderly packing of the wing epithelium into hexagons is necessary for the long-range order of these ridges ( Merkel et al . , 2014 ) , and their spacing presumably depends on the spacing of cell rows at the time the adult cuticle is secreted . AP-oriented T1 transitions could influence both these features . First , they are predicted to increase the final magnitude of PD cell elongation and may therefore influence ridge spacing . Second , increased cell elongation contributed by AP-oriented T1 transitions may help focus the direction of the subsequent PD-oriented T1 transitions and increase hexagonal order . Interestingly , PCP mutant wings are characterized by irregular cell packing geometry ( Classen et al . , 2005 ) , and less long-range order in the pattern of cuticular ridges ( Merkel et al . , 2014 ) . Our analysis has revealed unexpected properties of the cell elongation-dependent shift of T1 transitions towards the PD axis . Separately quantifying the orientations of cell boundary loss and cell boundary expansions during this shift shows that these two components of a T1 transition respond differently to cell elongation . The average orientation of cell boundary expansion shifts from the AP to the PD axis at 21 hAPF before the complementary change in orientation of cell boundary loss ( Figure 7 ) . This suggests that PD-oriented epithelial stresses promote expansion of cell boundaries along the PD axis at a lower threshold than is required to block shortening of PD-oriented boundaries . Thus , the mechanisms that operate to contract cell boundaries and to expand new ones respond differently to epithelial stresses . This difference means that cells tend to both gain and lose boundaries along the PD axis during the shift from AP to PD-oriented T1 transitions . Interestingly , this does not always simply represent the contraction and re-expansion of the same cell boundaries . Rather , contractions and expansions of different cell boundaries in the same direction occur as rows of cells move with respect to each other—a process that is associated with the maximal contribution of measured correlation effects ( compare Figure 6A magenta line and Figure 7 ) . PD-oriented cell divisions also contribute to the shape change of the wing blade during pupal morphogenesis . Interestingly , while epithelial stresses promote cell division in cultured epithelial cells ( Puliafito et al . , 2012; Streichan et al . , 2014 ) , this does not appear to be the case in the wing . In fact cell divisions occur at a similar rate and for longer times in laser ablated and dumpyov1 mutant wings . Thus , there are more cell divisions overall in situations where stresses are reduced . This suggests that the cell divisions that occur normally are not initiated by the stresses associated with hinge contraction but are controlled by other factors . Examining their pattern suggests that signals from veins may play a role ( data not shown , [Garcia-Bellido et al . , 1994] ) . It is also possible that hormonal signals such as ecdysone could control their timing . We do not yet understand why perturbations that reduce stress result in slightly more cell divisions . These extra divisions may occur by different mechanisms that respond to wounding . Furthermore , the magnitude of PD shear caused by cell divisions is proportional to the rate of cell divisions , suggesting that cell division orientation is also unaffected by reducing PD stresses . Thus , it may be that cell divisions in the pupal wing are autonomously controlled by planar polarized cortical cues . Alternatively , because the AP-oriented T1 transitions still cause some PD cell elongation in laser severed and dumpyov1 mutant wings , it may be that residual cell elongation is still sufficient to orient cell divisions . During development , tissues are shaped with extreme reproducibility . It has been estimated that the shape of the Drosophila wing is precise to within 1 cell diameter ( Abouchar et al . , 2014 ) . How is such reproducibility achieved ? The cell lineage in the wing is indeterminate , and both wing shape and wing area can be preserved in the face of a variety of developmental insults—including cell death and differential growth rates . This requires that all cells within the wing be able to sense its overall size and shape . How do cells acquire this information ? Based on our findings , we would propose that stresses in the wing epithelium could be one cue . Interestingly , the total area and shear of WT wings is more reproducible than would be expected from the variation in individual cellular contributions ( Figures 3 , 6 ) . Thus , these cellular events influence and can compensate for each other . The fact that epithelial tension is required to maintain area homeostasis and to control tissue shear suggests a mechanism for this compensation . Variations in cell division , cell shape change , cell rearrangements and cell extrusions could be detected by cells as changes in epithelial tension . The ability of cells to sense and respond to tension could underlie this compensation and produce wings of reproducible sizes and shapes .
Flies were raised at 25°C under standard conditions unless stated otherwise . Pupae were collected for imaging as described previously ( Classen et al . , 2008 ) . Ecad::GFP flies ( Huang et al . , 2009 ) were used as control for all live imaging experiments . The dumpyov1 mutation ( Sturtevant et al . , 1929 ) was recombined with Ecad::GFP for imaging , and for quantifying tissue flows and cell behaviors in a dumpyov1 mutant ( Bloomington , reference number 276 ) . The Dumpy::YFP protein trap line ( DGRC ) was used to describe Dumpy distribution in the pupal wing . To inhibit cell divisions in the pupal wing while imaging it , we used the thermosensitive allele cdc2E1-E24 of cdk1 that we recombined with Ecad::GFP ( Stern et al . , 1993 ) . Cells expressing two copies of cdc2E1-E24 and shifted to 30°C arrest in G2 phase just prior to entering mitosis ( Weigmann et al . , 1997 ) . Pupae were prepared for live imaging as previously described ( Classen et al . , 2008 ) . Long-term time-lapses were acquired with a Zeiss spinning disk microscope driven by the Axiovision software and equipped with an inverted Axio Observer stand , a motorized xyz stage , and a Zeiss LCI Plan-Neofluar 63× 1 . 3 NA Imm Corr lens associated to an objective heater set to 25°C . The fly pupa was placed in a temperature-controlled chamber set to 25°C and equipped with a humidifier to prevent desiccation . Images were recorded with an AxioCamMR3 camera ( 2 × 2 binning ) . Laser power was measured through a 10×/0 . 45 NA lens using a power-meter ( PT9610 , Gigahertz-Optik ) . A power of 0 . 980 mW was found to be optimal to prevent noticeable bleaching during 24 hr of continuous acquisition with an exposure time of 265 ms . Briefly , the dorsal cell layer of the pupal wing was scanned within 5 min in the ( x , y , z ) dimensions , over 24 overlapping positions of about 30 z-sections each . This scan procedure was continuously repeated about 260 times to cover more than 20 hr of development . Custom Fiji macros helped keeping the tissue in focus in a semi-automated way . We benefited from the flexible architecture provided by our computer department to handle several TB of data using custom unix bash scripts to archive , compress and store data on tape and retrieve them easily . Image pre-processing steps were mostly performed using custom Fiji macros called from a master bash script to enable parallelization using GNU parallel ( Tange , 2011 ) . A low-pass filter was first applied to images to remove high frequency noise . For each z-stack , the signal of highest contrast was projected using a C++ algorithm as previously described ( Merkel et al . , 2014 ) . Tiles were stitched using Fiji ( Preibisch , 2009; Schindelin et al . , 2012 ) . Stitched images were then loaded in Packing Analyzer v8 . 5 ( PA8 . 5 ) for cell edge detection by using a seeded-watershed algorithm . To facilitate cell tracking , cross-correlation of subsequent images was performed in PA8 . 5 to calculate local tissue displacement beforehand . Resulting cell-tracking masks were parsed using a custom C++ parser that extracted cell shape properties , cell lineages and cell neighbor relationships for further storage in a SQLite relational DB . Topology was kept in the DB by ordering cell neighbors counter-clockwisely around each cell , namely by following a directed path of cell–cell junctions around each cell . One DB file was created per video . All DB queries were carried out using R ( R Development Core Team , 2014 ) or Python ( www . python . org; Hunter , 2007; Pérez and Granger , 2007; van der Walt et al . , 2011 ) . We restricted our analyses to a subset of cells that were trackable throughout the entire course of the video , excluding cells that became visible only at later stages due to the apposition of the two cell layers . At 32 hAPF , most cells of the dorsal cell layer are visible . Therefore , we manually drew regions of interest ( ROIs ) on 32 hAPF wings using the veins and hinge–blade interface as landmarks , where the Ecad::GFP signal is more intense . These ROIs include the blade region delimited by the hinge–blade interface and the wing margin , but also the hinge–blade interface itself . In addition , we defined a triangular portion of the hinge delimited by the hinge–blade interface and the most anterior sensory organ located in the bulk of the hinge . In other hinge regions , cells were too small and elongated to be consistently tracked . Starting from these ROIs at 32 hAPF , we developed a backward-tracking algorithm that uses information about cell divisions and cell extrusions to reconstitute the corresponding group of cells at start of recording ( usually ∼16 hAPF ) , discarding margin cells that were not present through the course of the video . In the resulting set of cells , cells in contact with the wing margin at start of recording ( about 16 hAPF ) were not perfectly segmented and therefore further discarded . This gave rise to a group of cells that is fully consistent in time . Since part of the hinge could not be analyzed at cellular resolution , we complemented our analysis of the small portion of the hinge by using particle image velocimetry ( PIV ) ( Adrian and Westerweel , 2011 ) to extract information about the entire hinge deformation as described thereafter . PIV was implemented by Benoit Aigouy and described elsewhere ( Merkel et al . , 2014 ) . Laser ablations were performed using an ultraviolet laser microdissection apparatus as described elsewhere ( Grill et al . , 2001 ) . For tissue severing , a C-apochromat-40×/1 . 2 water immersion lens was used to focus the beam along AP-oriented line segments to ablate both dorsal and ventral cell layers . Subsequent long-term imaging of the whole wing was then carried out on our dedicated Zeiss spinning disk . Extracellular-matrix ablations were done using a 63×/1 . 4 oil lens to focus the beam between the tissue and the cuticle and to cut over 10 μm in depth . Circular cuts were also conducted with the 63×/1 . 4 oil lens but over a depth of 5 μm . Each circular cut experiment was repeated on five distinct pupae . Fly wings expressing Ecad::GFP were used in all circular cut experiments . Circular cuts were performed to disconnect a small subset of cells present inside the circle from the rest of the tissue . The interface of the tissue with the ablated circular region was always visible and well spatially defined after laser ablation . Therefore we quantified its displacement to calculate the initial velocity gradient that describes the immediate tissue deformation in response to the ablation . To do so , we first fitted an ellipse to this manually segmented interface , at 50 s after ablation . The major and minor axes of the ellipse define the orientation of two orthogonal kymographs , each intersecting the ellipse in two points corresponding to two opposite sides of the interface between the tissue and the ablated region . Each kymograph depicts the displacement as a function of time of the two opposite sides of this interface , which were manually segmented using sub-pixel resolution ROI . Thus , we quantified the displacement of four points corresponding to the intersections of the minor and major axes with the ellipse ( see Figure 2—figure supplement 3 ) . This procedure was semi-automated in a custom Fiji macro . The initial response of the tissue reflects the orientation and amplitude of tissue stresses . This can be captured by the initial velocity gradient within a few seconds after the 4 s ablation . Indeed , the initial displacement normal to the cut boundary was approximately linear as a function of time within the first 5 s after ablation . A linear fit to the data provided the initial normal velocities V|| and V⊥ along the major and minor axis of the ellipse , respectively . We then define the velocity gradient tensor in the coordinate system of the ellipse and rotate it into the image coordinate system: ( 6 ) vijcut= ( cosθ−sinθsinθcosθ ) ( V‖/r‖00V⊥/r⊥ ) ( cosθsinθ−sinθcosθ ) , where θ is the angle between the major axis of the ellipse and the PD axis ( x axis ) . The radii r|| and r⊥ are the half lengths of major and minor axes of the ellipsoidal shape at the time the velocities are determined . This velocity gradient can be decomposed into trace and a traceless part as: ( 7 ) vijcut= ( C00C ) + ( v∼xxcutv∼xycutv∼xycut−v∼xxcut ) , where vkkcut=2C is the isotropic expansion rate , v∼xxcut is the shear rate projected onto the PD-axis . While in the main text , tensors are denoted as bold face symbols , in the supplement we often use an explicit index or matrix notation for tensors . Latin indices i , j denote the x and y coordinates of a cartesian coordinate system of the tissue . Circular cuts were performed in nine different regions of the wing where cell elongation differed ( Figure 2I , I′ ) . Due to the fast imaging settings optimized to catch the initial velocity gradient , the image quality was not sufficient for cell segmentation . Therefore we used wings of different animals at the same stage ( 22 . 5 hAPF ) to estimate cell elongation Qxx for each of the nine ablation regions that were easily located using morphological landmarks such sensory organs and veins ( see Figure 9A ) . Average cell area acell was estimated using a similar approach ( see Figure 9B ) . We find that the shear rate projected onto the PD axis v∼xxcut increases with cell elongation . We find that data of v∼xxcut as a function of v∼xxcut can be accurately fitted by a linear function both in WT and dumpyov1 wing . Fit parameters are given byWT: v∼xx[h−1]= ( 0 . 018±0 . 008 ) Qxx+ ( 0 . 003±0 . 001 ) Dp: v∼xx[h−1]= ( 0 . 019±0 . 005 ) Qxx+ ( 0 . 003±0 . 002 ) The correlation between the isotropic expansion rate vkkcut and the logarithm of cell area is less apparent . However , following Equation 8 , we performed a linear fit to the data and found the following best fit parameters:WT: 12vkk[h−1]= ( 0 . 008±0 . 007 ) ln ( acell/aref ) + ( 0 . 017±0 . 002 ) Dp: 12vkk[h−1]= ( 0 . 005±0 . 008 ) ln ( acell/aref ) + ( 0 . 015±0 . 002 ) Parameters obtained in isotropic expansion rate fits have high uncertainties and we do not use them in the rest of the paper . In order to compare theory to the experimental data we need to approximate the observed hinge and blade shapes by rectangles . We first show how a characteristic height and length can be associated to an arbitrary two-dimensional shape . We then specify how hinge and blade regions were selected to obtain the associated height and length . Note that for the purpose of estimating the wing dimensions , we do not used tracked regions of a subpart of the wing as in ‘Data handling and image processing’ , but we use all available segmented data . Finally , we introduce corrections that account for changes of the visible part of the tissue in the field of view . | The individual cells in a developing animal embryo organize themselves into tissues with specific and reproducible shapes , which requires the cells to communicate with one another . Cells in tissues exert forces on their neighbors , and respond to being pushed and pulled by the cells around them . In the fruit fly Drosophila melanogaster , each wing consists mainly of a framework of proteins and other molecules that is built by epithelial cells . These epithelial cells divide and grow during the life of a fly larva , and then reorganize themselves into the shape of the wing after it forms into a pupa . During this reshaping , epithelial cells in some regions of the wing experience powerful contractions . Previous work had suggested that and these forces produced tension in the rest of the wing to pull it into its final elongated shape . But it wasn't clear what exactly these contractions were pulling against to produce the tension . Nor was it understood exactly how wing epithelial cells responded to tension to reorganize themselves into a different wing shape . Now , Etournay , Popović , Merkel , Nandi et al . have analyzed the forces acting across the entire wing blade and how these forces shape the wing . All cell divisions , cell neighbor exchanges and changes in cell shape in the developing wing blade were tracked under a microscope; this revealed how each one of them contributed to the change in wing shape . Further experiments revealed that localized contractile forces produce tension in the wing because it is connected around its edge to surrounding structures via an extracellular protein called Dumpy . Releasing these contacts , by severing them with a laser or by mutating Dumpy , caused the wing to develop into abnormal shapes , showing that the tension in the wing blade has an important role in determining wing shape . Furthermore , by tracking cells in wings that had been severed by a laser , or mutated for Dumpy Etournay , Popović , Merkel , Nandi et al . could figure out exactly which cellular processes were guided by epithelial tension . Etournay Popović , Merkel , Nandi et al . also present a theoretical model that describes how the interplay between active force generation and the response of cells to the resulting tension shapes the wings of fruit flies . They propose that epithelial tension provides a mechanism through which cells can communicate with each other to ensure that together the combined behavior of these cells generates reproducible shapes . Further studies are required to analyze how active force generation is patterned and cells sense and respond to external forces during development . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"physics",
"of",
"living",
"systems"
] | 2015 | Interplay of cell dynamics and epithelial tension during morphogenesis of the Drosophila pupal wing |
The cholinergic interneurons ( CINs ) of the striatum are crucial for normal motor and behavioral functions of the basal ganglia . Striatal CINs exhibit tonic firing punctuated by distinct pauses . Pauses occur in response to motivationally significant events , but their function is unknown . Here we investigated the effects of pauses in CIN firing on spiny projection neurons ( SPNs ) – the output neurons of the striatum – using in vivo whole cell and juxtacellular recordings in mice . We found that optogenetically-induced pauses in CIN firing inhibited subthreshold membrane potential activity and decreased firing of SPNs . During pauses , SPN membrane potential fluctuations became more hyperpolarized and UP state durations became shorter . In addition , short-term plasticity of corticostriatal inputs was decreased during pauses . Our results indicate that , in vivo , the net effect of the pause in CIN firing on SPNs activity is inhibition and provide a novel mechanism for cholinergic control of striatal output .
Acetylcholine – the first neurotransmitter to be identified – is present in high concentrations in the striatum of the basal ganglia ( Zhou et al . , 2001 ) where it is released by a population of intrinsic cholinergic interneurons ( CINs ) that are distinct from the cholinergic projection neurons that innervate the cortex ( Woolf and Butcher , 1981 ) . In addition , recent studies indicate an external cholinergic innervation from the pedunculopontine and laterodorsal tegmental nuclei to the dorsal striatum and nucleus accumbens ( Dautan et al . , 2016 , 2014 ) . CINs play an important role in behavior by modulating the activity of the principal output neurons of the striatum , the spiny projection neurons ( SPNs ) ( Witten et al . , 2010 ) . Previously there has been great interest in the role of CINs in disorders of the basal ganglia ( Maurice et al . , 2015; Pisani et al . , 2007; Shen et al . , 2015 ) and more recently in behavioral flexibility ( Aoki et al . , 2015; Bradfield et al . , 2013; Okada et al . , 2014 ) . The neuronal mechanisms for regulation of SPN activity by CINs are therefore of intense current interest . CINs are tonically active ( Wilson et al . , 1990 ) and although they represent only 1% of striatal neurons their extensive axonal fields allow them to control large striatal regions ( Bolam et al . , 1984 ) . These features combine with a diverse array of receptors for acetylcholine both at pre- and post-synaptic locations ( Hernández-Echeagaray et al . , 1998; Hersch et al . , 1994; Sugita et al . , 1991 ) . Thus , CINs are crucial modulators of striatal function . In particular , CINs control SPN excitability and response to glutamatergic inputs via activation of postsynaptic muscarinic type 1 receptors ( M1 ) . At the presynaptic level CINs exert a tonic inhibitory influence over incoming information via the activation of muscarinic type 2 ( M2/M4 ) receptors located on cortical and thalamic glutamatergic terminals . The firing pattern of CINs in vivo varies according to behavioural context . During slow wave activity or activated brain states , spontaneous CIN activity ranges from tonic firing , which may be regular or irregular , to bursts of firing separated by pauses of variable duration ( Sharott et al . , 2012 ) . In awake animals exposed to repeated , motivationally significant stimuli , putative CINs respond by a pause in firing ( Graybiel et al . , 1994; Kimura et al . , 1984; Morris et al . , 2004 ) , sometimes preceded and often followed by a burst . Previous work ( Crittenden et al . , 2017; Ding et al . , 2010; Doig et al . , 2014; English et al . , 2011; Nelson et al . , 2014 ) has focused on the effects of the burst component . In awake animals , the pause has been reported to have either no reliable effects with pauses of short duration ( English et al . , 2011 ) , or a mixture of excitation and inhibition during pauses of long duration ( Witten et al . , 2010 ) on firing activity of SPNs recorded extracellularly . Here we focus on the effects of the pause on subthreshold membrane potential fluctuations of SPNs recorded intracellularly , as well as supratheshold firing activity of SPNs identified by juxtacellular recording in vivo . We used optogenetic silencing of CINs – optimized to produce a pause in nearby CINs – during whole-cell and juxtacellular electrophysiological recording from CINs and SPNs in vivo . We found that a pause in CIN firing caused an inhibitory effect on SPN activity in vivo , reversing expectations based on in vitro studies and computational theories ( Ashby and Crossley , 2011; Franklin and Frank , 2015; Pakhotin and Bracci , 2007 ) and complementing previous in vivo findings ( English et al . , 2011; Witten et al . , 2010 ) . In particular , a pause in CIN firing hyperpolarizes the membrane potential of SPNs and reduces the short-term plasticity at cortical inputs , leading to decreased firing of SPNs in vivo .
We injected an adeno-associated virus ( AAV ) encoding the neuronal silencer Halorhodopsin ( NpHR ) into the striatum of ChAT-cre mice to induce pauses in CINs that mimic naturally occurring pauses ( Figure 1A ) . Histological analysis confirmed that the expression of NpHR was restricted to CINs in the dorsal striatum ( Figure 1B and Figure 1C ) . To assess the specificity of viral transduction we counted cells that express NpHR-eYFP and stained for anti-ChAT . We found that the 99 . 0 ± 0 . 48% of the cells that expressed NpHR-eYFP also stained for anti-ChAT leaving only 1 . 0 ± 0 . 48% that were non-specifically expressing NpHR-eYFP . Our injections targeting the dorsal striatum infected 57 . 35 ± 4 . 33% of ChAT positive neurons ( Figure 1D ) . Whole cell recordings obtained both in acute slices and in vivo from anaesthetized mice showed that a brief pulse of light produced a hyperpolarization of the membrane potential of CINs , causing a pause of their tonic firing ( Figure 1—figure supplement 1 ) . The average peristimulus time histogram ( PSTH ) shows that in all CINs tested , NpHR-induced inhibition caused complete silencing of firing during light ( Figure 1—figure supplement 1A and B ) . To measure the effect of the pause in CIN firing on SPN activity in vivo , we first obtained juxtacellular recordings from adjacent SPNs . Optically identified CINs were separated from putative SPNs based on their distinct firing properties and spike waveforms ( Figure 1—figure supplement 2 and Figure 1—figure supplements 2—source data 1 ) . To differentiate optogenetically induced pauses from those occurring during the resetting of autonomous activity ( interspike interval ranging from 138 to 902 ms ) ( Figure 1—figure supplement 2 and Figure 1—figure supplements 2—source data 1 . ) , we delivered a light pulse of 0 . 5 or 1 s which induced a complete pause of ongoing firing in all CINs ( Figure 1—figure supplement 3A and Figure 1G ) . The pause was followed by rebound firing with a variable onset ( 0 . 5 s: 105 . 6 ± 11 . 2 ms , n = 16; 1 s: 69 . 37 ± 8 . 8 ms , n = 16 ) and stayed two standard deviations ( S . D . ) above the mean for 79 . 3 ± 12 . 6 ms ( 0 . 5 s pause ) and 138 . 12 ± 14 . 6 ms ( 1 s pause ) ( Figure 1—figure supplement 3A and Figure 1G ) . The light-induced pause in CIN firing caused inhibition of firing activity in SPNs ( Figure 1I and Figure 1J ) which dropped below 2 s . D . from the mean firing frequency ( Figure 1K ) . To calculate the significance of changes in firing rate in individual neurons , taking account of successive bins , a change in firing rate was considered significant if the cumulative sum of the observed deviation of the firing rate from the mean was less than a critical value predicted from a Poisson distribution ( Ellaway , 1978; Imamura and Onoda , 1983 ) , ( see Materials and methods ) . In response to a 1 s exposure to light , the group average ( n = 22 cells ) PSTH showed a significant decrease in firing rate commencing at the 400–500 ms bin ( red bins , Figure 1K ) at the p<0 . 01 level ( see Materials and methods ) . When individual cells were analyzed , within the group of cells exposed to 1 s light , a total of 14/22 cells showed a significant ( p<0 . 05 ) decrease in firing , at a median latency of 700 ms from the light pulse . Three cells showed a significant ( p<0 . 05 ) increase in firing after the light pulse , one cell showed a significant increase ( p<0 . 05 ) followed by a significant decrease ( p<0 . 05 ) , and four cells showed no significant change in firing rate . However , these changes occurred at much longer latency after the light offset . To determine which responses were due to the pause ( and not the rebound firing activity ) , we tested cells for inhibition during the light on period . A total of 9/14 of the cells showing an inhibitory response ( 64% ) showed a significant ( p<0 . 05 ) decrease within the 1 s light on period , median latency of 400 msec . Light-induced pauses of CINs resulted in a similar inhibition of SPNs firing both in anesthetized mice ( 1 s: 43 . 7 ± 6 . 32 % n=22 ) and in awake , head restrained mice ( 51 . 52 ± 13 . 9 % n=7 ) ( Figure 1L ) . These results indicate that a pause in CIN firing of 1 s duration significantly reduces the firing of SPNs in vivo . We next tested the effect of a shorter light pulse ( 0 . 5 s ) on the SPNs firing . In the cells exposed to a 0 . 5 s light on period ( n = 17 cells ) , the group average PSTH showed a significant decrease in firing rate at the 400–500 ms bin ( red bins , Figure 1—figure supplement 3B ) . When individual cells were analyzed , however , none of the cells ( 0/17 ) showed a significant decrease during the 0 . 5 s period of light on . These results indicate that a pause in CIN firing of 0 . 5 s duration was not of sufficient duration for the effect that was observed reliably with 1 s duration . To determine the effect of the transition from light on to light offset , we analyzed the activity of SPNs in the period after the light offset , relative to the last 500 ms of the light on period . There was no significant increase from this new baseline until 700–800 ms after the light off ( p<0 . 01 , see Materials and methods ) . Consistent with this , the firing rate remained more than 2 s . D . below the initial baseline mean ( magenta bins , Figure 1K ) . To understand the cellular mechanisms underlying the reduction of firing in SPNs , we then obtained in vivo whole cell recordings of SPNs in the dorsal striatum of mice anesthetized with a ketamine/xylazine mix ( Figure 2A and Figure 2B ) . Under these conditions SPNs showed a characteristic bimodal distribution of the membrane potential , fluctuating between a DOWN state , characterized by hyperpolarized potentials , and an UP state , where the cell is close to threshold for firing ( Figure 2B ) . The best-fit of the sum of two Gaussian distributions to the all-amplitudes distribution of the membrane potential revealed that , during a pause ( 5 s ) in CIN firing , the membrane potential of SPNs shifts to more hyperpolarized values in the DOWN state ( before: −71 . 59 ± 1 . 82 mV; pause: −72 . 77 ± 1 . 86 mV; after: −71 . 66 ± 1 . 75 mV; n = 13; Repeated Measure One-way ANOVA , F ( 2 , 24 ) =7 . 692; p=0 . 0026; Figure 2C and Figure 2D ) . The UP state potential was also significantly more hyperpolarized during a pause ( before: −56 . 35 ± 1 . 85 mV; pause: −57 . 79 ± 2 . 01 mV; after: −55 . 89 ± 1 . 83 mV; n = 13; Repeated Measure One-way ANOVA , F ( 2 , 24 ) =6 . 139; p = 0 . 0070 Figure 2C and Figure 2D ) . We also observed a decrease in the number of points during the UP state , thus suggesting a shortened duration ( Figure 2C ) . We then used a crossover of moving averages to detect and quantify UP and DOWN state durations . We observed significantly shorter UP states during a pause ( before: 0 . 42 ± 0 . 02 s; pause: 0 . 39 ± 0 . 01 s; after: 0 . 41 ± 0 . 02 s; n = 13; Repeated Measure One-way ANOVA , F ( 2 , 24 ) =11 . 56 , p=0 . 0003; Figure 2E ) while DOWN state durations were significantly increased ( before: 0 . 51 ± 0 . 03 s; pause: 0 . 55 ± 0 . 04 s; after: 0 . 5 ± 0 . 04 s; n = 13; Repeated Measure One-way ANOVA , F ( 2 , 24 ) = 5 . 874 , p=0 . 0084; Figure 2E ) . The number of action potentials fired in the UP states during a pause was significantly decreased ( before: 1 . 94 ± 0 . 11 Hz; pause: 0 . 86 ± 0 . 08 Hz; after: 1 . 97 ± 0 . 13 Hz; n = 334 , 319 , 325 UP states respectively; One-way ANOVA , F ( 2 , 975 ) =33 . 16 , p<0 . 0001; Figure 2F ) UP states did not exhibit significant changes in the mean amplitude ( before: 17 . 23 ± 1 . 02 mV; pause: 16 . 34 ± 1 . 27 mV; after: 16 . 83 ± 1 . 12 mV; n = 13; Repeated Measure One-way ANOVA , F ( 2 , 24 ) =1 . 88 p=0 . 1748; data not shown ) , nor in their frequency ( before: 1 . 03 ± 0 . 04 Hz; pause: 0 . 98 ± 0 . 05 Hz; after: 1 . 0 ± 0 . 05 Hz; n = 13; Repeated Measure One-way ANOVA , F ( 2 , 24 ) =2 . 039 , p=0 . 1521; data not shown ) . However , the transition from DOWN to UP state was significantly slower during a pause ( before: 145 . 9 ± 17 . 52 mV/s; pause: 127 . 7 ± 16 . 87 mV/s; after: 139 ± 17 . 76 mV/s; n = 13; Repeated Measure One-way ANOVA , F ( 2 , 24 ) =5 . 09 , p=0 . 0144; Figure 2G and Figure 2H ) . Altogether , these data indicate that a pause in CIN firing decreases the excitability of SPNs in vivo . The altered DOWN and UP state dynamics we observed could not be due a change in the activity of extrinsic afferents because the optogenetic stimulation and expression of NpHR is virtually specific to the striatal CINs . However , cholinergic effects on presynaptic or postsynaptic receptors are possibilities . Because UP states in the anesthetized animal represent the synchronous activation of cortical inputs we investigated if a pause in CIN firing affected the responsiveness of SPNs to repetitive synaptic inputs . We measured excitatory post-synaptic potentials ( EPSPs ) in response to stimulation of glutamatergic inputs from the motor cortex in vivo ( Figure 3A ) . In control conditions ( before a pause in CIN firing ) , a train of 10 stimuli at 25 Hz , induced a gradual facilitation and summation of EPSPs over the course of the train ( Figure 3B and Figure 3C ) . During the pause in CIN firing , we also observed a significant increase in the amplitude of the first EPSP in the stimulus train ( before: 4 . 3 ± 0 . 65 mV; pause: 5 . 16 ± 0 . 67 mV; n = 9; paired t-test , p=0 . 0115; Figure 3B and Figure 3D ) . However , during a pause ( 1 s ) , the ratio of the last to the first EPSP , was significantly decreased ( before: 1 . 59 ± 0 . 12; pause: 1 . 29 ± 0 . 06; n = 9; paired t-test , p=0 . 0146; Figure 3B and Figure 3E ) . These data suggest that a pause in CIN firing gates cortical information transfer at striatal synapses in vivo . Our finding that a pause in CIN firing modulates short term plasticity at excitatory synaptic inputs and decreases firing of SPNs in vivo could be attributed to a decrease in the intrinsic excitability of SPNs . We measured the effect of a pause on the intrinsic membrane responses of SPNs in acute slices ( Figure 4 ) . In agreement with previous observations ( Maurice et al . , 2015 ) the light-induced pause produced a significant increase in rheobase current compared to control or after light ( control: 310 ± 22 . 09; light: 333 . 3 ± 20 . 94; after: 313 . 3 ± 21 . 08; n = 12; Repeated Measure One-way ANOVA , F ( 2 , 22 ) =33 . 79 , p<0 . 0001; Figure 4B ) , and a significant decrease in the number of action potentials fired in response to a current pulse ( n = 12; two-way ANOVA , current x light interaction , F ( 16 , 176 ) =1 . 77 , p=0 . 039; Figure 4C ) . Thus , we confirmed that a pause in CIN firing decreased intrinsic excitability of SPNs . Interruption of tonic M1 receptors activation is a possible mechanism for the decreased excitability observed during a pause . Under physiological conditions , cholinergic activation of M1 receptors increases SPN excitability , their responsiveness to cortical and thalamic inputs , and their firing by modulating multiple intracellular conductances ( Akins et al . , 1990; Gabel and Nisenbaum , 1999; Galarraga et al . , 1999; Hsu et al . , 1996; Shen et al . , 2005 , 2007 ) . We therefore tested the hypothesis that a pause could decrease intrinsic excitability of SPNs via modulation of M1 receptors in acute slices . We found that , similar to the effect of a pause , bath application of low concentrations of M1 receptor antagonist pirenzepine ( 100 nM ) , increased the rheobase current ( control: 280 ± 32 . 66; light: 308 . 6 ± 32 . 32; pirenzepine: 302 . 9 ± 30 . 99; pirenzepine and light: 305 . 7 ± 34 . 01; n = 7; Repeated Measure One-way ANOVA , F ( 3 , 18 ) =16 . 37 , p<0 . 0001; Figure 4E ) and significantly shifted the relationship between current injection and number of action potential to the right ( n = 7; two-way ANOVA , current x light interaction , F ( 24 , 144 ) =2 . 519 , p=0 . 0004; Figure 4F ) . Moreover , the M1 receptor antagonist occluded the effect of the light , in that the combined effect of pirenzepine and light on the intrinsic excitability of SPNs was not significantly different from the individual effects ( Figure 4E ) . These results suggest that interruption of M1 receptor activation during a pause is sufficient to decrease SPN excitability ( Figure 4G ) .
Based on in vivo whole cell recordings we show that optogenetically-induced pauses in CIN firing have direct and reproducible inhibitory effects on SPN activity . Pauses in CIN firing caused hyperpolarization of subthreshold membrane potential fluctuations measured by in vivo whole-cell recording and caused a significant decrease in the firing rate of SPNs measured using in vivo juxtacellular recordings . Short-term facilitation of corticostriatal synaptic transmission was also modulated during a pause . The main novel feature of the present study is a focus on the effect of the pause uncontaminated by preceding or following bursts , which required measurement during pauses of sufficient duration to separate pause effects from effects of rebound firing . The use of in vivo whole-cell recording configuration furthermore , enabled the effects of the pause on the subthreshold membrane potential to be measured intracellularly , which revealed subthreshold changes associated with the significantly decreased firing of SPNs during the pause . In addition , in vivo whole cell recordings permitted measurement of EPSPs in response to stimulation of the motor cortex , which showed that short term plasticity of cortical input was modulated by the pause . We believe these findings constitute the first direct evidence that the pause in CIN firing carries a meaningful signal to postsynaptic SPNs , complementing previous studies that have focused on the effects of burst-pause or pause-burst sequences . The subthreshold membrane potential fluctuations measured by in vivo whole-cell recording revealed that during a pause there was a hyperpolarizing shift of subthreshold membrane potential fluctuations , a slower transition between hyperpolarized and depolarized membrane potentials , and reduced duration of spontaneously occurring synchronous depolarizations . The underlying membrane potential fluctuations are seen during slow-wave sleep and anesthesia , but during wakefulness the membrane potential fluctuations are less stereotyped , with depolarizing synaptic events of variable amplitude occurring in temporally unstructured sequences ( Mahon et al . , 2006; Sippy et al . , 2015 ) . Understanding exactly how a pause would modulate these less stereotyped membrane potential fluctuations requires further whole cell recording experiments in awake animals . Our finding that pauses in CIN firing cause inhibition of SPN firing complements an earlier study ( English et al . , 2011 ) that showed a decrease in firing of SPNs after an optogenetically induced pause-burst sequence . The inhibitory effect reported in this earlier study was synchronous with the rebound increase in firing that often occurs after an optogenetically induced pause . Taken together with our findings , there thus appear to be two successive inhibitory effects related to the pause-burst sequence . First , our data show a pause mediated inhibition caused by interruption of cholinergic excitation of SPNs . Second , English et al . ( 2011 ) have shown a burst-mediated inhibition caused by activation of GABAergic inputs to SPNs from other neurons excited by acetylcholine . The onset of the inhibitory effect on SPNs of the pause in CINs was statistically significant 400 ms after the onset of the pause , both in the group where the pause lasted 500 ms and in the group where the pause duration was 1 s . Consistent with this delayed effect , English et al . ( 2011 ) reported no inhibition of SPNs during a pause of 200 ms duration . The delayed onset of inhibition that we observed at 400 ms could not have been detected in this 200 ms timeframe . In an additional group of 3 cells , English et al . ( 2011 ) also did not detect an inhibition of MSNs during a 1 s exposure to light . The detection of inhibitory responses in cells with low firing rates and a high coefficient of variation in firing times is challenging , and in a small sample it is not possible to rule out an inhibitory effect . The absence of pause-related inhibition in the cells exposed to 1 s light in the earlier study might also reflect differences in the power output of the optical fibre ( 4–10 mW versus 10–30 mW output ) and the diameter – 200 µm versus 125 µm etched down to 50 µm used in English et al . ( 2011 ) . The use of pauses of longer duration enabled the effect of the pause to be measured in isolation from the effects of any prior or subsequent rebound firing of CINs . In the natural firing patterns of CINS , the duration of the pauses varies considerably in different behavioural situations . Pauses occurring in response to sensory cues paired with motivationally significant events acquire a pause duration in the range of 200–300 ms ( Aosaki et al . , 1995; Morris et al . , 2004 ) . In anaesthetized animals pauses of duration 380–800 ms have been reported in response to a light stimulus ( Schulz et al . , 2011 ) . In vivo , spontaneous firing activity is broken up by pauses that vary in duration up to one second ( Sharott et al . , 2012 ) . Thus , pauses of similar duration to those used in the current experiments occur naturally as well as in controlled behavioural paradigms . The use of longer duration pauses made it possible to distinguish effects of pauses from rebound firing . In our juxtacellular recordings both 0 . 5 s and 1 s light pulses showed a significant reduction of SPNs firing in the group average starting at 400 ms . Single cell analysis showed that during a light pulse of 1 s , 64% of the SPNs displayed a significant inhibition in their firing rate . We did not detect , however , any significant effect on the firing of individual SPNs in response to 0 . 5 s light . These results thus leave an open question about the modulatory role of short pauses of CINs ( <400 ms ) on the activity of SPNs and the striatal processing that occurs in response to sensory cues paired with motivationally significant events . We speculate that short pauses may exert their effects mainly by rebound increases in firing , acting on nicotinic receptors on GABA interneurons as shown by English et al . ( 2011 ) , while longer pauses may engage a dual mechanism involving pause-related removal of muscarinic excitation via M1 receptors , in addition to GABA inhibition activated on the rebound . We used NpHR to optogenetically silence CINs during 0 . 5 and 1 s pauses for recording responses of SPNs in juxtacellular configuration , and changes in membrane potential fluctuations in cells recorded in whole-cell mode during 5 s pauses . Light activation of NpHR caused effective silencing of CINs , as shown by measurements of action potential firing at the soma in our recordings . One caveat for the use of the inward Cl- pump - especially during longer pulses - is the increased rebound firing , evident at the light offset , which may intensify the naturally occurring rebound firing that often follow a pause in CIN firing ( Mahn et al . , 2016; Raimondo et al . , 2012 ) . Although in the current experiments the main focus is on the effects during the period of the silencing rather than at the time of the rebound , it may be helpful to use other tools in parallel in future experiments to avoid this possibility . Cell by cell analysis of the effects of the CINs pause on SPNs showed that while the group average showed a significant decrease in firing rate during the 1 s pause , a fraction of the neurons in the group did not show an individually significant decrease in firing rate . There are several possible explanations for why some SPNs did not respond to optogenetic manipulation of CINs . It is theoretically possible that not all SPN neurons in the sample express M1 receptors . However , this seems unlikely given that nearly all medium spiny neurons have M1 receptor messenger RNA ( Yan et al . , 2001 ) and express M1 receptor protein ( Hersch et al . , 1994 ) . The remaining possibilities include that the non-responding SPNs were beyond the reach of the axonal arborization of a pausing CIN; or in a location where CINs did not express the neuronal silencer NpHR . The present results add new knowledge – the inhibitory effects on SPNs of the pause in CIN firing – to existing knowledge concerning the effects of burst-pause or pause-burst sequences . We observed inhibition caused by the pause , occurring before the onset of the burst , after a delay of approximately 400 ms . We consider the most consistent interpretation to be that the clearance of acetylcholine levels from the synaptic cleft by acetylcholinesterase is the crucial factor for the onset of the inhibition we observed . It has been an open question whether the dynamics of acetylcholine clearance are fast enough for acetylcholine concentration to track momentary pauses in firing . If clearance were not extremely fast , the extracellular space would act as a capacitor and smooth out the effects of pauses . Clearance by hydrolysis of acetylcholine is catalyzed by acetylcholinesterase ( AChE ) . The high levels of expression of AChE in the striatum ( Zhou et al . , 2001 ) , and the extremely fast kinetics of this enzyme ( Quinn , 1987 ) , imply rapid clearance and sensitivity to dynamic fluctuations in CIN firing rate . However , current techniques for measuring acetylcholine concentration have insufficient temporal resolution to detect dips during pauses ( Mattinson et al . , 2011 ) . The observed effects of a pause in CIN firing on SPNs , are thus the first direct evidence that pauses in CIN firing are meaningful signals to the postsynaptic SPNs . Analysis of the firing of SPNs in the period immediately after the pause in CIN firing did not show an additional inhibition in the present study . Thus , the rebound firing of CINs that occurred in this period appeared to cause no additional inhibition relative to the firing rate reached during the exposure to light . The recovery to baseline levels after the 1 s pause , however , was relatively slow compared to the recovery after the 0 . 5 s pause . After the 1 s pause there was no significant increase in firing until 700 ms after the end of the light exposure , consistent with a second inhibitory process . During longer pauses ( 1 s ) , acetylcholine levels would be expected to have decreased below baseline levels , leading to the inhibition reported in the present study . Inhibitory effects of the rebound burst might be expected to add to the pause-related inhibition . However , these might not be apparent in the measurements due to a floor effect . That is , the firing rate of the SPNs is already low at the end of the longer pause , and further inhibition at this point would not produce significant effects . An implication of our findings is therefore that the effects of a burst following a pause may depend on the duration of the pause . During natural firing patterns , CIN excitation preceding and following a pause can be expected to result in a complex interplay of the temporal dynamics of acetylcholine level changes with the dynamics of muscarinic and nicotinic receptors in SPNs , presynaptic glutamatergic inputs , or GABAergic interneurons . In vivo , there is often an excitation before the pause ( Matsumoto et al . , 2001 ) . A burst preceding a pause was not studied in the current experiments , which started with the optogenetically-induced pause . Previously , Witten et al . ( 2010 ) used optogenetically induced excitation of CINs and found that a 10 Hz burst caused inhibition of SPNs . In that case the first burst in the burst-pause sequence would be expected to initiate a complex sequence of events . The burst would have excitatory effects on neuropeptide Y–expressing neurogliaform neurons mediated by nicotinic receptors , causing fast GABA inhibition ( English et al . , 2011; Faust et al . , 2015; Faust et al . , 2016 ) and excitatory presynaptic effects leading to release of GABA from dopamine terminals on a slower timescale ( Nelson et al . , 2014 ) . During the initial burst , muscarinic receptors would also be activated causing presynaptic inhibition of cortical input by M2 ( Ding et al . , 2010 ) and increased excitability by M1 activation ( Lv et al . , 2017 ) Activation of nicotinic receptors on dopamine terminals would also cause dopamine release ( Threlfell et al . , 2012 ) . Using a standard protocol to assess the short term synaptic changes typical of corticostriatal synapses , we observed that during a pause such protocol produced an increase in the first EPSP of the train , while the total facilitation measured as the ratio between the last to first pulse was decreased , thus defining a temporal window in which corticostriatal integration was attenuated . We did not investigate the mechanism underlying this effect . It is possibly related to muscarinic receptors that are expressed at pre-and postsynaptic sites on corticostriatal synapses on SPNs . A brief pause of tonic activation of presynaptic M2/M4 receptors controlling glutamate release ( Pakhotin and Bracci , 2007; Pancani et al . , 2014 ) might lead to the increased amplitude of the first EPSP . Another possibility is a postsynaptic effect of reduced M1 receptor activation ( Ding et al . , 2010; Shen et al . , 2007 ) . Although other mechanisms can be at play , the phenomenon observed suggests that the pause might activate a filter , enhancing the signal-to-noise ratio , effectively reducing the ability of spatial summed excitatory inputs to reach action potential threshold and activate the postsynaptic SPN . However , further experiments would be required to determine the physiological basis for the pause-related perturbation in synaptic integration . Our recordings from acute slices show that the pause induced an increase in spike threshold and a decrease in excitability of SPNs . Both effects are occluded by pirenzepine , an inhibitor of M1 and M4 receptors . Thus , reduced muscarinic receptor activation due to decreased acetylcholine concentration during a pause in CIN firing may be one mechanism for the expression of this inhibition on SPNs activity . Acetylcholine activation of the M1 receptor excites SPNs by reducing membrane K+ conductances ( Hsu et al . , 1996 ) , modulating a transient K+ conductance ( Akins et al . , 1990 ) , and increasing NMDA receptor mediated currents and persistent Na+ currents ( Calabresi et al . , 1998 , 2000 ) . Conversely , a decrease in M1 receptor activation as occurs during a pause , would be expected to increase the membrane K+ conductances , hyperpolarizing the membrane potential , and decrease NMDA currents , causing shorter duration of the UP state ( Plotkin et al . , 2011 ) . Hence , the present findings are compatible with known actions of muscarinic M1 receptors on SPNs . The inhibitory effect of the pause suggests a novel mechanism for the modulatory function of CINs in striatal function . The activity of CINs in awake animals is related to behavioral contexts such as reward probability ( Morris et al . , 2004; Shimo and Hikosaka , 2001 ) , stimulus location ( Ravel et al . , 2006 ) or behavioral context and current state ( Lee et al . , 2006; Stalnaker et al . , 2016 ) . Ensembles of SPNs are recruited by cortical inputs causing transitions to firing activity . A pause in CIN firing would inhibit the SPNs within its axonal arbor , and favour a transition to a more hyperpolarized membrane potential , transiently decreasing the neuronal output of the ensemble of SPNs . In the context of inhibitory interactions among SPNs ( Tunstall et al . , 2002 ) , this might favour cortical information transfer in a different ensemble of SPNs in the domain of tonically firing CINs . Acetylcholine released by a burst preceding a pause would have presynaptic effects on dopamine terminals ( Cragg , 2006 ) and experiments using optogenetic stimulation show that additional CIN spikes cause phasic release of dopamine by activating presynaptic nicotinic receptors on dopamine terminals ( Threlfell et al . , 2012 ) . These nicotinic receptors rapidly desensitize in increased acetylcholine and may be expected to regain their sensitivity to acetylcholine after a pause , facilitating further dopamine release . Recovery from desensitization during the pause depends on the duration of the exposure to acetylcholine ( Reitstetter et al . , 1999 ) . Partial recovery begins on a subsecond timescale ( Yu et al . , 2009 ) consistent with the duration of a pause . During the pause , acetylcholine clearance occurs allowing recovery of nicotinic receptors to continue . At the same time , the stimulation is removed from muscarinic receptors leading to the inhibition that we describe . The net effect will depend on the acetylcholine levels reached during the burst , the rate of clearance , and the duration of the subsequent pause . The sequence would enable dopamine-dependent plasticity of the corticostriatal synapses active at the time of the pause . Such a mechanism may be important for understanding set shifting in behavioral flexibility , and loss of flexibility in neuropsychiatric diseases such as Parkinson’s disease ( Ztaou et al . , 2016 ) , dystonia ( Pisani et al . , 2007 ) and schizophrenia ( Lieberman et al . , 2008 ) .
Male heterozygous B6;129S6-Chattm2 ( cre ) Lowl/J mice ( ChAT-cre; The Jackson Laboratory ) were used in the present study . All experimental procedures were performed in accordance with and approved by the Okinawa Institute of Science and Technology Animal Care and Use Committee . Adeno-associated virus ( AAV ) encoding the neuronal silencer NpHR ( AAV5-DIO-NpHR-eYFP ) was obtained from the University of Pennsylvania ( Gene Therapy Program , Pennsylvania , USA ) or University of North Carolina ( UNC vector core , USA ) . At postnatal day 14 ( P14 ) we made a 300 nl unilateral stereotaxic guided injection in ChAT-cre mice . The injection volume and flow rate were controlled with an injection pump ( World Precision Instruments , Sarasota , USA ) . Viral injections were targeted to the dorsal striatum ( AP +0 . 7 , ML ±1 . 5 relative to bregma , DV −1 . 7 relative to dura ) . Mice were anesthetized with vaporized isoflurane ( Muromachi , Japan ) and returned to the home cage with the doe after the end of the surgery . Post-surgery , mice were given a single dose of Carprofen ( Rimadyl , 0 . 5 mg/kg IP ) and Bupurenorphin ( Lepetane , 0 . 05 mg/kg IP ) for pain alleviation . Injected ChAT-cre mice aged 7–10 weeks old were anesthetized with isoflurane and decapitated . The brain was quickly removed and rested for 30 s in ice-cold oxygenated NMDG cutting solution containing ( in mM ) : 92 NMDG , 2 . 5 KCl , 1 . 25 NaH2PO4 , 30 NaHCO3 , 20 HEPES , 25 glucose , two thiourea , 5 Na-ascorbate , 3 Na-pyruvate , 0 . 5 CaCl2· , 10 MgCl2 , and titrated to 7 . 2–7 . 4 pH with HCl . Slices ( 300 μm thick ) containing the striatum were cut on a vibratome ( VT1200S , Leica , Germany ) on an oblique plane , at 45° between coronal and horizontal cuts to preserve corticostriatal connections . Slices were mounted on a porous membrane and incubated for 30 min at 34° C in oxygenated ACSF containing the following ( in mM ) : 126 . 0 NaCl , 2 . 5 KCl , 2 . 0 CaCl2 , 2 . 0 MgCl2 , 18 . 0 NaHCO3 , 1 . 25 NaHPO4 , 10 . 0 glucose , then allowed to recover for at least 1 hr at room temperature before recording . Whole cell recordings were obtained from SPNs identified by size in the dorsal striatum in an area where fluorescent CINs expressing NpHR-eYPF could be visualized ( OlympusBX51WI , DAGE-MTI IR-1000 , and Olympus DP72 ) . Pipets ( 4–6 MΩ ) were pulled from P-97 ( Sutter Instruments , Novato , CA ) and filled with an intracellular solution containing the following ( in mM ) : 119 K-MeSO4 , 12 KCl , 1 MgCl2 , 0 . 1 CaCl2 , 10 HEPES , 1 EGTA , 0 . 4 Na-GTP , 2 Mg-ATP , ( 280–300 mOsm , pH 7 . 3 adjusted with KOH ) . Recordings were performed in a chamber perfused with ACSF at a rate of 3 ml/min and maintained at 32°C . Acquisition was done using Clampex 10 . 4 , MultiClamp 700B amplifier and Digidata 1440A ( Molecular Devices , San Jose , CA ) . For optogenetic manipulations , CINs expressing NpHR-eYFP were activated with short pulses of green light generated by a 530 nm LED ( CAIRN , UK ) under the control of the digital output of the amplifier . Current steps of 500 ms at increments of 20 pA were used to assess rhoebase current as well as the relationship between AP number and increasing current injection . Pirenzepine was bath applied for at least 5 min before recording the membrane responses to current step injections . Mice between 5 and 8 weeks old were anesthetized with a ketamine/xylazine mix ( 100 and 10 mg/kg i . p . , respectively , supplemented as needed injection during the course of the experiment ) and fixed on a Kopf stereotaxic apparatus . The animals were kept warm ( ~37°C ) for the whole duration of the surgery via a heating pad connected to a DC temperature controller provided with a feedback system ( FHC Inc . , Bowdoin , ME ) . An eye lubricant was applied to prevent corneal drying during the surgery . A stereomicroscope ( LEICA M651 ) was used to locate the atlas coordinates , using bregma as a reference ( Paxinos and Franklin , 2001 ) . Two craniotomies were drilled between 0 and 1 . 5 mm anterior from bregma , to allow the insertion of the optic fiber and recording electrode , and between 1 . 8 and 2 . 2 mm for insertion of a bipolar stimulating electrode for experiments of corticostriatal plasticity . Electrophysiological recordings were obtained from dorsal striatum between 1 , 9 and 2 . 5 mm from dura using a Multiclamp 700B amplifier connected to a Digidata 1440A . Data were acquired with pClamp 10 ( Molecular Devices ) , digitized at 20 kHz , filtered at 10 kHz , and analyzed offline with Clampfit 10 . 4 ( Molecular Devices ) or MATLAB as indicated . Activation of NpHR was achieved with a 100 mW , 593 nm laser ( Shanghai Dream Lasers Technology Co . , Ltd . , China ) via an optic fiber ( 200 μm , 0 . 22 NA ) inserted into the dorsal striatum ( +1 mm AP , 0 mm ML , −2 . 1 mm DV from dura ) with a 26° angle . Light pulses were externally triggered using pClamp ( Molecular Devices ) . The intensity of light stimulation was between 4–10 mW , measured at the tip of the fiber , outside of the brain . To measure the intensities of light we used a an optical power meter ( model 1936 C , Newport , Irvine , CA , USA ) equipped with a photodiode sensor ( model 918D-UV-OD3 , Newport , Irivine , CA , USA ) . For the activation of NpHR we used a square pulse with the following durations: 0 . 5 s , 1 s , 5 s . Whole-cell patch-clamp recordings were achieved using a standard blind-patch approach as previously described ( Margrie et al . , 2002 ) . Electrodes used for whole-cell recordings had a resistance between 4 . 5 and 6 MΩ and were filled with a solution containing ( in mM ) 135 K-gluconate , 4 KCl , 10 HEPES , 10 sodium phosphocreatine , 4 Mg-ATP and 0 . 3 Na-GTP ( 290–300 mOsm , pH 7 . 2–7 . 3 adjusted with KOH ) . The intracellular solution was supplemented with 0 . 2% ( wt/vol ) biocytin for post hoc cellular identification and morphological reconstruction . Only SPNs that showed a typical bimodal distribution of the membrane potential and initial membrane potentials values ≤ −60 mV in vivo were included in the study . At the end of each recording , injection of biocytin was achieved by applying positive current pulses ( 0 . 5–1 nA , 500 ms , for 5–10 min ) through the bridge circuitry of the amplifier . Juxtacellular recordings of striatal neurons were obtained using glass electrodes ( 5–7 MΩ ) filled with Ringer’s solution containing ( in mM ) : 135 NaCl , 5 . 4 KCl , 5 HEPES , 1 . 8 CaCl2 , 1 MgCl2 , or intracellular solution . Juxtacellular units were high pass filtered at 300 Hz . Light durations of 0 . 5 and 1 sec s , were delivered at 50 s interval and repeated 10 times . Spontaneous firing properties of striatal neurons in juxtacellular configuration were estimated during a 3–5 min period before any stimulation ( Figure 1—figure supplement 2 ) . Optically tagged single unit showing patterned pause-burst responses were classified as CINs . Cluster separation between CINs and SPNs was done using classically defined electrophysiological characteristics ( Isomura et al . , 2013; Sharott et al . , 2012 ) : spike waveform ( half-width ) , firing frequency and regularity ( see Figure 1—figure supplement 2 ) . Firing frequency , interspike intervals and half-width were calculated using a detection threshold function in Clampfit . The units obtained in awake head fixed configuration showed similar clustering ( Figure 1—figure supplement 2C and D ) and were pooled together . For plasticity experiments , a bipolar electrode was used to stimulate cortical inputs from the motor cortex area which send dense projections to the dorsal striatum ( Wall et al . , 2013 ) . EPSPs were electrically evoked with trains of 10 stimuli delivered at 25 Hz ( repeated every 30 s ) . To test the effect of a pause each train started 500 ms after the onset of the light pulse ( total duration: 1 s ) . The EPSP amplitude was quantified by measuring the peak of the synaptic response obtained from the average trace of all the sweeps recorded . Values were normalized for the first EPSP amplitude . UP and DOWN states were algorithmically detected using the strategy proposed for characterizing membrane potential fluctuations in electrophysiological data ( Reig and Silberberg , 2014; Seamari et al . , 2007 ) . UP states that were at least 0 . 2 s in duration were considered . Data were further analyzed with MATLAB 2014 . Spontaneously occurring UP and DOWN states were analyzed before , during and after a pause in CINs firing , by sampling a window of 5 s in each condition . The duration of the light was 5 s , delivered at 30 s interval and repeated at least five times . UP and DOWN states occurring during a pause were compared to those occurring in the preceding 5 s and 10 s after the end of the light pulse , when CINs tonic firing recovered to baseline rate , based on observations from juxtacellular recordings . The average of all UP states occurring before ( n = 334 ) , during ( n = 319 ) and after a pause ( n = 325 ) revealed a slower slope of the UP states during light . To quantify the slope of the transitions from DOWN to UP state , UP states were center aligned and scaled to have equal baseline values . The slopes of the transition from DOWN to UP state was calculated from the best-fit line of a linear regression on the membrane potential values in the region between 10% to 90% of the peak of the UP state . The average peak potential was calculated by taking the middle 50% of the membrane potential values for each single UP state event . Head-fixed awake in vivo electrophysiology experiments required a second surgery for implanting the head-bar after at least 6 weeks from virus injection . Mice were anaesthetized with gaseous isofluorane ( Muromachi , Japan ) and placed in a stereotaxic frame . The skull was cleaned to locate and mark bregma with an oil based marker and a head-bar ( Phenosys , Germany ) was glued to the skull with Super-bond ( Sun Medical , Japan ) . Post-surgery , mice were given a single dose of Carprofen ( Rimadyl , 0 . 5 mg/kg IP ) and Buprenorphin ( Lepetane , 0 . 05 mg/kg IP ) for pain alleviation . 3 days after implantation , mice were handled for at least three sessions as they habituated to head-fixation and jet-ball ( PhenoSys , Germany ) . Mice were able to freely moving over the spherical ball and they all exhibited grooming , resting , or running behaviors during each session . After habituation , on the day of electrophysiology recording , a craniotomy was performed under ketamine/xylazine mix . Mice recovered from the anaesthesia before initiation of Juxtacellular recordings and optical stimulation of CINs expressing NpHR-eYFP were obtained using same procedure as reported for anesthetized experiments . To recover the morphology of neurons filled with biocytin during in vivo whole-cell patch clamp recordings , brains were removed after completion of electrophysiology experiments and fixed for at least one week at 4°C in a solution containing 4% paraformaldehyde ( PFA ) in 1X phosphate buffer saline ( PBS ) , pH 7 . 0 . Coronal slices ( 60 μm thick ) were prepared with a vibratome ( VT1000S , Leica ) , washed with 0 . 3% Triton-X-100 in 1X PBS and incubated overnight at 4°C in Alexa Fluor 594 streptavidin conjugate ( diluted 1:1000; Molecular Probes , Japan , S11227 ) in the same solution . To analyse the extent of the viral spread in cholinergic interneurons , slices were incubated overnight with the primary antibody goat anti-Chat ( diluted 1:1000; Millipore , AB144P ) . Slices were then washed in 1X PBS and incubated overnight at 4°C in 1:1000 Alexa Fluor 594 ( diluted 1:1000; Invitrogen , Japan , A11058 ) in the same solution . Immunofluorescence data of mounted coverslips were acquired with a confocal laser scanning microscope ( Zeiss LSM 780 ) using Argon 488 nm and DPSS 561 nm lasers and 10x/0 . 45 and 20x/0 . 8 Plan-Apochromat lens controlled by Zen 2012 software . Quantification of colocalization of ChAT and NpHR-eYFP was performed using four alternate sections per mouse ( N = 4 ) , in a range between +1 . 2 to+0 . 8 mm ( AP ) from Bregma , corresponding to the coordinates used for the placement of the fiber optic . Given the lack of a clear boundary between dorsal and ventral striatum we objectively defined the dorsal region of the striatum by drawing a horizontal line at ~50% of the area of the striatum . Within this region we identified and counted a total of 903 neurons: ChAT+ , NpHR-eYFP-=387; ChAT- , NpHR-eYFP+=5; and ChAT+ , NpHR-eYFP+=511 . The following drugs were used: pirenzepine dihydrochloride ( 1071 , Tocris ) . Aliquots of stock solutions were prepared and frozen at −20°C . All other chemicals were from either Sigma-Aldrich , Wako , or Nacalai Tesque , Japan . Statistical analysis was performed with GraphPad Prism seven software . Values are given as mean ± SEM of n experiments . D’Agostino and Pearson normality test , Shapiro-Wilk normality test and KS normality test were used to determine if the values in our data came from a Gaussian distribution . All data sets passed at least two out of the three normality test , and thus we used parametric tests . Repeated measure two-way ANOVA for analysis of light-drug interaction in acute slices experiments were matched by sub-column and spread across a row with Sidak’s multiple comparisons test . Repeated measure with one-way ANOVA with Tukey’s multiple comparison test were used to compare before , during and after pause in Figure 2D , Figure 2E , Figure 2H as well as Figure 4B . For Figure 4E , we also used a repeated measure with one-way ANOVA followed by Tukey’s multiple comparison test to compare rheobase changes between treatments ( control , light , pirenzepine , pirenzepine with light ) . We used an Ordinary One-way ANOVA for Figure 2F . Student’s paired t-test were used for all other comparisons between two groups . No statistical tests were run to predetermine the sample size , and blinding and randomization were not performed . Probability threshold for statistical significance in these analyses was p<0 . 05 . To analyse the effects of light exposure on the firing rate of SPNs we used a minimal criterion of two S . D . difference from the mean . To take in account the low firing rate and high coefficient of variation in the spike timings , which presents a challenge for analyses to test for statistical significance , we used a standard method of analysis based on the cumulative sum of residuals ( CUSUM ) , to identify the point at which deviations from baseline became significant ( Ellaway , 1978; Page , 1957 ) . To calculate the CUSUM , spikes were binned into a PSTH . The mean firing rate ( µ ) within a reference window of n1 bins immediately before the onset of the light , and the S . D . of the firing rate ( σ ) were calculated . The CUSUM was calculated by subtracting µ from each bin . The cumulated successive differences were calculated using the formula for the rth CUSUM ( Davey et al . , 1986; Ellaway , 1978 ) of:Sr=∑i=1r ( xi−μ ) A change in firing rate was considered significant if the observed deviation of the CUSUM from the mean had a probability predicted from a Poisson distribution of p<0 . 01 ( for group data ) or p<0 . 05 ( for single neurons ) . The critical values for the ith bin were calculated using the derivation of Imamura and Onoda ( 1983 ) :Pr{Si≥2 . 33 σi−n1}=0 . 01 , orPrSi≥1 . 65 σi-n1=0 . 05 The CUSUM was calculated using a reference window of 1 s immediately prior to the light on period . To analyse the recovery after the light was turned off , in the group exposed to a 1 s illumination , the CUSUM was calculated using a reference window of 0 . 5 s immediately prior to the light off . Bins with frequencies more than 2 s . D . s below the mean firing rate of the baseline were considered significantly different from baseline if they occurred after the CUSUM crossed the significance level . | Nerve cells or neurons communicate with one another using electrical impulses and chemical messengers called neurotransmitters . Additional molecules known as neuromodulators regulate the communication process . In contrast to neurotransmitters , neuromodulators do not send messages directly from one neuron to the next . Instead they change the way that neurons respond to neurotransmitters . For example , the neuromodulator acetylcholine is most abundant in a region called the striatum . Located deep within the brain , the striatum contributes to learning and memory , motivation , and movement . Studies in rodents show that neurons within the striatum called cholinergic interneurons are almost continuously active . Each time these cells fire , they release acetylcholine . But whenever an animal experiences something unusual or important , the interneurons temporarily stop firing . Zucca et al . wanted to know whether these pauses in firing also act as a signal within the striatum . To find out , Zucca et al . inserted a light-sensitive ion channel into cholinergic interneurons in the mouse striatum . Activating the ion channels with a laser beam stopped the interneurons from firing . Zucca et al . showed that these pauses in firing reduced the activity of another group of neurons , the spiny projection neurons . These are the major output neurons of the striatum . They send messages from the striatum to other parts of the brain . The results thus suggest that cholinergic interneurons signal notable events by temporarily blocking output from the striatum . Understanding how cholinergic interneurons work will help reveal how the striatum drives behavior . It may also lead to treatments for diseases caused by cholinergic system dysfunction . Many patients with Parkinson’s disease or schizophrenia take medicines to block the effects of acetylcholine . Understanding how acetylcholine affects the striatum may help clarify how these treatments work . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Pauses in cholinergic interneuron firing exert an inhibitory control on striatal output in vivo |
Temperature-sensitive transient receptor potential ( TRP ) ion channels are members of the large tetrameric cation channels superfamily but are considered to be uniquely sensitive to heat , which has been presumed to be due to the existence of an unidentified temperature-sensing domain . Here we report that the homologous voltage-gated potassium ( Kv ) channels also exhibit high temperature sensitivity comparable to that of TRPV1 , which is detectable under specific conditions when the voltage sensor is functionally decoupled from the activation gate through either intrinsic mechanisms or mutations . Interestingly , mutations could tune Shaker channel to be either heat-activated or heat-deactivated . Therefore , high temperature sensitivity is intrinsic to both TRP and Kv channels . Our findings suggest important physiological roles of heat-induced variation in Kv channel activities . Mechanistically our findings indicate that temperature-sensing TRP channels may not contain a specialized heat-sensor domain; instead , non-obligatory allosteric gating permits the intrinsic heat sensitivity to drive channel activation , allowing temperature-sensitive TRP channels to function as polymodal nociceptors .
Sensitive detection and discrimination of temperature cues are fundamental to the survival and prosperity of humans and animals . While Hodgkin and Huxley showed more than 60 years ago that temperature could profoundly influence membrane excitability ( Hodgkin and Huxley , 1952 ) , it is in general not well understood how heat affects membrane excitability and how temperature-dependent changes in neuronal activity contribute to physiology . Neuronal action potentials are generated and modulated by a precise combination of ionic currents produced by the activity of ion channels . Activation of ion channels in turn is the result of complex conformational rearrangements in channel protein controlled by specific physical or chemical stimuli ( Hille , 2001 ) . Heat contributes to activation energy for channel conformational changes , through which it regulates channel activity and the shape of action potential ( Rodriguez et al . , 1998; Liang et al . , 2009 ) . For most ion channels , thermal energy affects the rate of conformational transitions by 2-to-5 folds per 10°C ( Pusch et al . , 1997; Rodriguez et al . , 1998 ) . Within the physiological temperature range , thermal energy alone was found to be generally insufficient to initiate channel activation . Among the exceptions are a group of transient receptor potential ( TRP ) channels including TRPV1-4 , TRPM8 , TRPM3 , TRPM4 , TRPM5 , TRPC5 , and TRPA1 . These channels are potently activated by heat in the absence of another stimulus and hence serve as key cellular temperature sensors ( Clapham , 2003; Zheng , 2013 ) . How these TRP channels respond to heat with exquisite sensitivity , however , remains mysterious . To understand the molecular mechanism underlying high temperature sensitivity of the TRP channels , we conducted a comparative investigation of the voltage-gated potassium ( Kv ) channels . As members of the tetrameric cation channel superfamily , TRP channels and Kv channels are structurally similar . They all have six transmembrane segments , with the S1 to S4 segments forming an isolated peripheral domain surrounding the ion-permeating pore composed of S5 , S6 , and the loop between them . Comparison of the crystal structures of Kv channels ( Long et al . , 2005 , 2007 ) and the cryo-EM structures of TRPV1 ( Cao et al . , 2013; Liao et al . , 2013 ) further shows that detailed structural features of these two types of channels are also very similar . However , the two types of channels are functionally distinct . Kv channels are activated by membrane depolarization with a steep voltage dependence ( Sigworth , 1994 ) , while TRP channels are only weakly activated at highly depolarized voltages ( Zheng , 2013 ) . More importantly , unlike the temperature-sensitive TRP channels , Kv channels cannot be directly activated by changes in temperature , and the activity of Kv channels are generally considered to be not very heat-sensitive ( Rodriguez et al . , 1998; Cui et al . , 2012 ) . The differences lead to the widely accepted assumption that through evolution some TRP channels have acquired specific protein structures that serve as a ‘heat sensor’ . In the present study we tested this hypothesis by examining the temperature response of Kv channels .
TRPV1 is an archetypical temperature-sensitive TRP channel ( Caterina et al . , 1997 ) . Heat strongly activates the channel , which could be observed at a broad voltage range using a ramp protocol ( Figure 1A , upper panel ) . A common way to characterize temperature sensitivity is the Q10 value ( defined as the folds increase in current amplitude upon a 10°C increase in temperature ) . For TRPV1 , the Q10 value is above 20 over a more than 200 mV voltage range ( Figure 1A , lower panel ) ( Benham et al . , 2003 ) , reflecting outstanding sensitivity of channel activation to heat . Having high temperature sensitivity at a wide voltage range is crucial for the channel's physiological role as a temperature sensor—it allows TRPV1-expressing sensory neurons to detect heat no matter the neurons are in the resting state or excited state . 10 . 7554/eLife . 03255 . 003Figure 1 . Heat sensitivity of TRPV1 and Shaker channel exhibits distinct voltage dependence . ( A ) TRPV1 current was elicited by a voltage ramp at varying temperatures ( upper panel ) . Q10 , quantified between 36 . 6°C and 46 . 0°C , remains high across the entire voltage range from −100 mV to +150 mV ( lower panel ) . ( B ) Shaker channels exhibit a low heat sensitivity at most voltages as quantified by Q10 between 22 . 4°C and 27 . 7°C . However , Q10 transiently peaks ( red arrow ) around the voltage ( −70 mV ) where the channel just starts to open . Simplified gating schemes involving two transitions are shown on the bottom . The C→C′ transition is weakly voltage-dependent for TRPV1 but highly voltage-dependent for Shaker . ( C ) TRPV1 and Shaker channels show distinct voltage-dependent activation behaviors . TRPV1 conductance ( open circle ) has a shallow voltage dependence ( with an apparent gating charge of 0 . 76 ± 0 . 06 e0 , n = 4 ) that occurs in a highly depolarized range ( V1/2 = 114 . 9 ± 10 . 9 mV , n = 4 ) . Shaker activation has a steep voltage dependence ( with an apparent gating charge 5 . 1 ± 1 . 3 e0 , n = 5 ) that occurs at hyperpolarized voltages ( V1/2 = −54 . 2 ± 4 . 7 mV , n = 5 ) . The Q-V curve for Shaker ( dotted curve ) is reproduced from a published study ( Schoppa and Sigworth , 1998 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03255 . 003 Activity of Shaker potassium channel , in contrast , exhibited much lower temperature sensitivity ( Figure 1B , upper panel ) . As a voltage-gated channel , Shaker activates upon depolarization to about −60 mV ( Figure 1C ) . The threshold voltage for activation was only slightly shifted by raising temperature . The average Q10 value remained low , at below 4 ( Figure 1B , lower panel ) . Low temperature sensitivity is anticipated for Shaker and many other Kv channels , because opening and closing of the ion permeation pore in these channels is obligatorily coupled to movement of the voltage-sensor controlled by the membrane potential ( Sigworth , 1994 ) . At hyperpolarized voltages , the channel is locked in the initial closed state ( C , Figure 1B ) , in which the voltage-sensor is kept in the down conformation . A strongly voltage-dependent transition , involving the movement of ∼13 e0 gating charges across the transmembrane electric field ( Schoppa et al . , 1992; Zagotta et al . , 1994; Aggarwal and MacKinnon , 1996 ) , moves the channel to another closed state , C′ , from which it can transition to the open state , O , with little voltage dependence . Since thermal energy is insufficient to supply the activation energy for voltage-sensor to overcome transmembrane voltage , opening of the channel is dictated by the membrane potential . The high fidelity of Shaker and other voltage-gated channels in reporting changes in membrane potential at variable environmental conditions is the basis for reliable electrical signaling of the nervous system . A closer inspection of the Shaker channel Q10 measurement , however , revealed that it did increase modestly around −80 to −60 mV ( arrow in Figure 1B , lower panel ) , approaching 10 at its peak . It is intriguing that this is the voltage range at which the voltage-sensor starts to move , permitting the C→C′ transition . This can be seen in the voltage dependence of gating charge movement ( Q-V curve , Figure 1C ) . Since it has been previously suggested that the voltage-dependent transition in TRPV1 is highly temperature-sensitive ( Voets et al . , 2004 ) , we wondered whether the transient Q10 increase in Shaker might reflect temperature sensitivity of the voltage-sensor movement . To test this possibility , we conducted similar measurements with the voltage-gated Ca2+-modulated BK potassium channel , because for this channel the separation between the G-V curve and the Q-V curve can be conveniently controlled by intracellular Ca2+ . We observed that , like Shaker , BK in the presence of intracellular Ca2+ also exhibited a transient Q10 increase at the voltage range where voltage-sensor started to be activated by depolarization , around −80 mV ( Figure 2A ) . Removing Ca2+ shifts the voltage range for channel activation , substantially increasing the separation between G-V and Q-V curves ( Horrigan and Aldrich , 1999 ) . The change is achieved mainly through a dual-allosteric coupling between the C′→O transition and the Ca2+ and voltage induced transitions ( Horrigan and Aldrich , 1999 ) . We observed two interesting effects of Ca2+ removal on the transient Q10 increase . First , it substantially shifted the voltage range at which the transient Q10 increase occurred , to around +100 mV ( Figure 2B ) . This large shift mirrored the substantial shift of the G-V curve . More importantly , in the absence of Ca2+ the extent of the Q10 increase was enhanced dramatically even though the total gating charge movement remained unchanged ( Horrigan and Aldrich , 1999; Figure 2B ) . These observations suggest that the Q10 increase may not be associated with the voltage-sensor movement ( as reflected by the Q-V curve ) but rather associated with the activation gate opening ( as reflected by the G-V curve ) , that is , instead of the C→C′ transition , it seemed to be the C′→O transition that yielded the Q10 increase . 10 . 7554/eLife . 03255 . 004Figure 2 . Voltage dependence of BK channel temperature sensitivity shifts with intracellular Ca2+ . Q10 ( blue trace , right axis ) calculated from temperature-dependent changes in the G-V curves at labeled temperatures ( left axis ) measured in the absence of Ca2+ ( lower panel ) are larger in value and shifted more to the right in voltage dependence compared to those measured in presence of saturating Ca2+ ( 200 μM , upper panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03255 . 004 The possibility of a voltage-sensor associated Q10 increase was further ruled out when we tested additional Kv channels . We found that both Kv2 . 1 and Kv4 . 3 exhibited a surprisingly prominent Q10 transient increase that peaked in the 20-to-30 range ( Figure 3 ) , a level comparable to that of TRPV1 . Highly temperature-sensitive activation could be observed when either a voltage ramp or a voltage step was applied . As these channels have similar voltage dependence to Shaker and BK channels ( Islas and Sigworth , 1999; Dougherty et al . , 2008 ) , the substantially higher Q10 values cannot come from the voltage-sensor movement . This conclusion is reinforced by the fact that the activity of TRPV1 ( and other temperature-sensitive TRP channels ) is minimally voltage-sensitive ( with less than 1 e0 gating charge; Figure 1C ) ( Voets et al . , 2004 ) . Indeed , we observed that the energetic effects of heat on the voltage-dependent activation per se are similar for TRPV1 and most Kv channels ( Figure 4A , B ) , and the peak Q10 values exhibited no correlation with the temperature-dependent shifts in the half-activation voltage ( Figure 4C ) . 10 . 7554/eLife . 03255 . 005Figure 3 . Both Kv2 . 1 and Kv4 . 3 channels are highly heat-sensitive within a narrow voltage range . ( A ) and ( E ) Large increases in current were observed from both Kv2 . 1 and Kv4 . 3 when temperature was raised . ( B ) Kv2 . 1 channels were substantially activated by heat at −50 mV ( filled circles ) but not at −80 mV ( open circles ) . ( C ) Representative current traces at time points shown in ( B ) . ( F ) and ( G ) Similar behaviors as Kv2 . 1 were observed with Kv4 . 3 . ( D ) and ( H ) Voltage dependence of G-V curves ( left axis ) and Q10 ( right axis ) for Kv2 . 1 ( D ) and Kv4 . 3 ( H ) . G-V curves are color-coded as in ( A ) and ( E ) , respectively . Q10 peaks in the range where the channel just starts to be voltage-activated . DOI: http://dx . doi . org/10 . 7554/eLife . 03255 . 00510 . 7554/eLife . 03255 . 006Figure 3—figure supplement 1 . High heat sensitivity of Kv2 . 1 channels was robustly observed when the length of voltage ramp and extracellular K+ concentration were changed . ( A ) The length of voltage ramp from −100 mV to +50 mV was increased from 200 ms to 1 s . With this slowed voltage ramp large increases in current were still observed from Kv2 . 1 when temperature was raised . ( B ) Voltage dependence of G-V curves ( left axis ) and Q10 ( right axis ) for Kv2 . 1 with the slow voltage ramp . ( C ) 130 mM extracellular K+ was replaced with 130 mM Na+ . Large increases in current were still observed from Kv2 . 1 when temperature was raised . ( D ) Voltage dependence of G-V curves ( left axis ) and Q10 ( right axis ) for Kv2 . 1 with 130 mM extracellular Na+ . DOI: http://dx . doi . org/10 . 7554/eLife . 03255 . 00610 . 7554/eLife . 03255 . 007Figure 4 . Heat-induced shift in V1/2 is not responsible for high temperature sensitivity . ( A ) Changes in V1/2 were induced by raising temperature from the room temperature to a level that maximally activated the channels ( n = 3-to-4 ) . ( B ) Corresponding changes in free energy were calculated as z⋅F⋅ΔV1/2 , where z is the total gating charge , F is Faraday's constant . n = 3-to-4 . Values of z for Kv channels are based on published work ( Schoppa and Sigworth , 1998; Islas and Sigworth , 1999; Dougherty and Covarrubias , 2006 ) . **p<0 . 01; ***p<0 . 001; n . s . , no significance . ( C ) Q10 value of each channel type was plotted against the temperature-induced shift in V1/2 . At similar levels of shift in V1/2 , both large and small Q10 values were observed , suggesting that shift in V1/2 is not responsible for high temperature sensitivity . n = 3-to-4 . DOI: http://dx . doi . org/10 . 7554/eLife . 03255 . 007 What makes Kv2 . 1 and Kv4 . 3 highly temperature-sensitive at a narrow hyperpolarized voltage range ? One noticeable common property of these two channels is a prominent closed state inactivation ( CSI ) process ( Klemic et al . , 1998; Amadi et al . , 2007; Figure 5 ) , during which the voltage-sensor is decoupled from the activation gate ( Klemic et al . , 1998; Barghaan and Bahring , 2009 ) . This ‘slippage’ in coupling relaxes the obligatory gating by voltage , increasing the probability of transitioning from C into C′ at mild depolarization ( Shin et al . , 2004 ) . In other words , CSI simultaneously makes Kv2 . 1 and Kv4 . 3 more TRPV1-like in both voltage- and heat-dependent activation . The Q10 increase peaked at roughly −70 to −50 mV , where the probability of the channels being not open but in the CSI state was high ( Figure 5B , C ) . Under such condition , both channels could be strongly heat-activated ( Figure 3 ) . In contrast to Kv2 . 1 and Kv4 . 3 , the less temperature-sensitive Shaker channel does not undergo CSI ( Aldrich , 2001; Zhou et al . , 2001; Fineberg et al . , 2012 ) , so that its voltage-sensor is always obligatorily coupled to the activation gate , preventing the gate being activated by heat . To rule out the possibility that transient Q10 increase was produced by artificial non-equilibrium gating generated by the voltage ramp protocol , we repeated the experiment with Kv2 . 1 using a much slower ramp protocol ( from 200 ms to 1 s duration ) ( Figure 3—figure supplement 1A , B ) . We also replaced extracellular potassium with sodium in order to observe channel activation as an outward current instead of an inward current ( Figure 3—figure supplement 1C , D ) . Neither operation eliminated the large transient Q10 increase . 10 . 7554/eLife . 03255 . 008Figure 5 . Closed state inactivation ( CSI ) and channel activation overlap within specific voltage ranges . ( A ) Representative current recordings of Kv2 . 1 CSI . Length of the P2 segment was 10 s . Current measured at P3 was normalized to that measured at P1 . The normalized current was then plotted against P2 voltage for voltage dependence of CSI . ( B ) For Kv2 . 1 , the voltage dependence of both CSI ( open squares ) and activation ( filled circles ) follows a single-Boltzmann function ( left axis ) . After normalization the two Boltzmann functions were multiplied to yield the red curve ( right axis ) . This combined probability curve surges in the voltage range where the channels are only slightly activated while significant CSI occurs , which overlaps with the voltage range where high heat sensitivity is observed . ( C ) Similar voltage dependences of CSI and activation were observed for Kv4 . 3 . n = 3-to-5 . DOI: http://dx . doi . org/10 . 7554/eLife . 03255 . 008 If the Q10 increase in Kv2 . 1 and Kv4 . 3 is due to relaxation of the obligatory coupling between the voltage-sensor and the activation gate through CSI , one would expect to be able to produce a larger Q10 value in Shaker by decoupling its voltage-sensor from the activation gate . Two well-characterized Shaker mutants are known to exhibit this property , ILT ( Smith-Maxwell et al . , 1998 ) and V2 ( Schoppa and Sigworth , 1998 ) . The mutations in ILT generate a prominent separation between the Q-V curve and the G-V curve ( Figure 6A ) . When we examined the ILT channel , we indeed observed a large Q10 transient that peaked above 20 , making activation of this channel much more temperature-sensitive than the wild-type channel ( Figure 6B–E ) . Importantly , the Q10 transient started near −20 mV where the voltage-sensor has been mostly elevated by depolarization . Thus , the Q10 transient seen in ILT and other Kv channels is not originated from the voltage-sensor movement . On the contrary , our observations argue strongly that it is the release of voltage-sensor control—through depolarization , CSI , and mutations—that revealed an intrinsically highly temperature-sensitive process in Kv channels that occurs during the C′→O transition . 10 . 7554/eLife . 03255 . 009Figure 6 . Shaker ILT and V2 mutants exhibit large but opposite heat responses . ( A ) and ( F ) Conductance-voltage ( G-V; n = 3-to-6 ) and gating charge-voltage ( Q-V ) curves for ILT ( A ) and V2 ( F ) . Q-V curves for ILT and V2 are reproduced from published studies ( Schoppa and Sigworth , 1998; Ledwell and Aldrich , 1999 ) . ( B ) Large increases in current are observed from ILT when temperature was raised . To increase current amplitude when driving voltage was small , extracellular solution contained 130 mM KCl , while intracellular solution contained 130 mM NaCl . ( C ) Voltage dependence of Q10 for ILT based on G-V curves derived from ( B ) . ( D ) and ( E ) ILT channels were substantially activated by heat at +40 mV . ( G ) In contrast to the ILT current , V2 current decreases upon temperature rise . ( H ) Voltage dependence of Q10 for V2 based on G-V curves derived from ( G ) . Q10 drops below 1 when voltage reaches a level that the channel starts to open , as increasing temperature deactivates the channel . ( I ) and ( J ) V2 channels were substantially deactivated by heat at +20 mV . DOI: http://dx . doi . org/10 . 7554/eLife . 03255 . 009 The Shaker V2 mutant also has a large separation between Q-V and G-V curves ( Figure 6F ) . Consistent with the conclusions we postulate above , V2 also exhibited a noticeable Q10 change at the voltage range for channel opening . Interestingly , we observed that instead of heat activation , the channel exhibited a heat-induced deactivation behavior ( Figure 6G–J ) . At voltages where the voltage-sensor has been fully elevated , V2 current decreased substantially upon heating . This yielded Q10 values less than 1 , as seen in the cold-activated TRPM8 channel ( McKemy et al . , 2002 ) . It remains to be determined whether the heat-induced deactivation process in V2 reflects a reversal of the heat-induced activation process in wild-type Kv channels and ILT . It is nonetheless important to note that V2 and ILT differ by only a few amino acids ( V2 contains L382V in the S4–S5 linker; ILT contains V369I , I372L , and S376T in the S4 segment ) . Given that these mutations appear to alter gating transition rates and equilibriums without changing the fundamental voltage-dependent activation mechanism ( Schoppa and Sigworth , 1998; Smith-Maxwell et al . , 1998 ) , we speculate that the same gating process may yield the opposite Q10 changes . While future investigation is needed to examine this particular speculation and identify the gating transition involved , it is important to note that V2 is also intrinsically heat-sensitive , which is revealed by decoupling the voltage-sensor from the activation gate . In summary , we found that gating of Kv channels can be highly temperature-sensitive ( Figure 7 ) . Potentially this transient increase in temperature sensitivity may have important functional consequences , especially since a change in the Kv channel activity at slightly depolarized membrane potentials would substantially alter action potential frequency and neuronal coding . Similarly , any temperature-dependent change in activity of the voltage-gated sodium and calcium channels may also have important consequence in membrane excitability . However , while it has been demonstrated that temperature-dependent changes in Kv channel activity profoundly affect neuronal excitability and coding ( Vandenberg et al . , 2006; Graham et al . , 2008 ) as well as secretion ( MacDonald et al . , 2003 ) , how dynamic changes in temperature sensitivity of ion channels affect physiology remains poorly understood , partially because most studies of ion channel behaviors have been carried out at the room temperature instead of around the body temperature . 10 . 7554/eLife . 03255 . 010Figure 7 . Comparison of Q10 ( A ) and thermodynamic measurements ( B ) between TRPV1 , Kv and CLC-0 channels . Data for CLC-0 are reproduced from published results ( Yang et al . , 2010 ) . n = 3-to-6; *p<0 . 05; **p<0 . 01; ***p<0 . 001; n . s . , no significance . DOI: http://dx . doi . org/10 . 7554/eLife . 03255 . 010
A central observation of the present study is that there is a transient increase in temperature sensitivity of Kv channels , which occurs when the voltage-sensor is partially decoupled from the activation gate . Our study revealed an intrinsically heat-sensitive gating process in these voltage-gated channels . Interestingly , a previous study of Shaker channel activation indeed identified a late transition preceding channel opening that exhibits higher temperature sensitivity than all the other transitions ( including the voltage-sensor movements ) ( Rodriguez et al . , 1998 ) . In agreement of Clapham and Miller ( 2011 ) , these observations indicate that there may not be a specialized ‘heat sensor’ to support the heat activation process . Generally speaking , an equilibrium process A←→B would be highly temperature-sensitive if the transition is associated with a large enthalpic change . This can be seen from the temperature dependence of the probability functionPB=11+exp ( ΔHRT−ΔSR ) , in which PB is the probability of the system to reside in the B state , ΔH and ΔS are enthalpic and entropic changes associated with the transition , respectively , T is temperature , and R is the gas constant . For temperature-sensitive TRP channels , a large enthalpic change during channel activation was indeed observed ( Brauchi et al . , 2004; Voets et al . , 2004; Yang et al . , 2010 ) . The observation of an associated large entropic change comes from the requirement for the heat-induced transition to occur at physiological temperatures ( Yang et al . , 2010 ) . A simple interpretation of the large enthalpic and entropic changes is that they represent the occurrence of a substantial conformational rearrangement in the channel protein . We showed in the present study that in Kv channels this conformational rearrangement is not the voltage-sensor movement but rather the process following voltage-sensor movement . Kv channels and TRP channels are known to function as an allosteric protein . Observations from the present study can be understood using the following simple allosteric gating scheme:in which the vertical transition represents voltage-sensor movement driven by membrane depolarization , and the horizontal transition represents pore opening . Because for Shaker and many other Kv channels voltage-sensor movement is very strongly coupled to pore opening , activation gating occurs in the sequence C←→C′←→O′ . At hyperpolarized voltages , the strong voltage dependence of the C←→C′ equilibrium ensures that the channel resides in the initial C state , so that the temperature-sensitive transition C′→O′ cannot occur . ( According to allosteric principles , the C′→O′ transition is strongly favored over the C→O transition because of the large activation energy associated with the vertical transition . ) Only when the channel moves from C to C′ upon mild depolarization can the C′→O′ transition be observed upon temperature changes . CSI in Kv2 . 1 and Kv4 . 3 as well as mutations in Shaker ILT and V2 weaken the allosteric coupling between the voltage-sensor and the pore , allowing the temperature-sensitive horizontal transition to occur . Realizing that Kv channels are intrinsically highly temperature-sensitive has important mechanistic implications for how some TRP channels serve as temperature sensors and nociceptors . The high temperature sensitivity of TRP channels originates from exceptionally large enthalpic and entropic changes associated with the activation process ( Brauchi et al . , 2004; Voets et al . , 2004; Yang et al . , 2010 ) . From the Gibbs equation , ΔG = ΔH − TΔS , it is clear that large ΔH and ΔS values make free energy change associated with the activation process , ΔG , to be highly temperature-dependent . It is generally assumed that the large enthalpic and entropic changes come from a substantial conformational rearrangement in the TRP channels . The existence of large conformational rearrangements has been directly detected using fluorescent tags in TRPV1 as well as the CLC chloride channels ( whose common gating process is also highly temperature-sensitive ) ( Pusch et al . , 1997; Bykova et al . , 2006; Yang et al . , 2010; Ma et al . , 2011 ) . Nonetheless , the structural and mechanistic basis for large conformational rearrangements in temperature-sensitive TRP channels remains unknown . Consequently , it has been highly controversial how these TRP channels sense heat . Here we show that in Kv channels the voltage-dependent C→C′ transition , representing movements of the voltage-sensor , is not specially temperature sensitive , in agreement with a previous study of the Shaker channel ( Rodriguez et al . , 1998 ) . Instead , the C′→O′ transition , representing later activation transitions of the channel protein , yields the transient increase in Q10 when the obligatory voltage-sensor control is alleviated . We and others showed that temperature-sensitive TRP channels also behave like an allosteric protein ( Latorre et al . , 2007; Matta and Ahern , 2007; Jara-Oseguera and Islas , 2013; Yang et al . , 2014 ) . High temperature sensitivity of TRP channels may also originate from the C′→O′ transition . A previous study demonstrated , based on a simple two-state C←→O gating scheme , that heat strongly shifts the gating equilibrium by deferentially affecting the forward and backward transitions which are also weakly voltage-sensitive ( Voets et al . , 2004 ) . The origin of the voltage-sensitive process in TRPV1 and other TRP channels remains unclear . It is nonetheless becoming clear that an S4-type voltage-sensor seen in Kv channels does not exist in TRP channels . Since voltage is a poor activator for TRPV1 but heat is a strong activator , it is unlikely heat solely works through the voltage-dependent process to open TRPV1 ( Matta and Ahern , 2007; Yang et al . , 2010 ) . In the context of Scheme I , it appears that the observation by Voets et al . ( 2004 ) is applicable to the horizontal transition instead of the vertical transition . Our study indicates that TRPV1 is heat-sensitive at a broad voltage range because the C→C′ transition is weakly voltage-dependent and not obligatory . A recent cryo-EM structural study of TRPV1 provided direct evidence that the peripheral S1–S4 domains remain stationary while the pore domains undergo substantial structural rearrangements when the channel is activated by agonists ( Cao et al . , 2013; Liao et al . , 2013 ) . It is thus possible that the highly temperature-sensitive process of TRPV1 may come from the pore opening process . Indeed , nearly all the known TRPV1 agonists interact directly or indirectly with the pore domain: H+ ( Jordt et al . , 2000 ) , spider toxin ( Bohlen et al . , 2010 ) , and divalent cations ( Ahern et al . , 2005; Luebbert et al . , 2010; Cao et al . , 2014; Yang et al . , 2014 ) from the extracellular side , and capsaicin and resiniferatoxin ( Szallasi et al . , 1999; Jordt and Julius , 2002 ) from the intracellular side . While capsaicin and resiniferatoxin bind to sites adjacent to the pore-forming domains , the cryo-EM data indicate that they induce channel activation by altering the pore conformation while leaving the S1–S4 peripheral domains mostly unchanged ( Cao et al . , 2013; Liao et al . , 2013 ) . Results from previous studies on mutant channels ( Myers et al . , 2008; Grandl et al . , 2010; Cui et al . , 2012 ) and site-directed fluorescence recordings ( Yang et al . , 2010 , 2014 ) further support the involvement of the outer pore in heat activation . Can conformational rearrangement of the channel pore produce the large ΔH and ΔS required for high temperature sensitivity ? Results from the cryo-EM study ( Cao et al . , 2013 ) and fluorescence study ( Yang et al . , 2010 , 2014 ) of TRPV1 clearly support this possibility . To serve as biological heat sensors , temperature-sensitive TRP channels need to fulfill two fundamental requirements . First , its activity must be highly sensitive to changes in temperature . Second , heat-induced activity cannot be dependent on the presence of another stimulus; specifically , heat activation must be able to occur when the sensory neuron is in the resting state ( at hyperpolarized membrane potentials ) . As discussed above , the first requirement is fulfilled because channel activation is associated with large ΔH and ΔS originated from a substantial conformational rearrangement . Results from the present study argue that the second requirement is fulfilled because the occurrence of heat-induced conformational rearrangements is not controlled by the voltage-dependent transition ( or other stimulus-induced transitions ) . Allosteric but non-obligatory coupling of multiple stimuli to the opening of channel pore allows TRP channels to serve as polymodal cellular sensors .
The following cDNA constructs were used in this study: murine TRPV1 ( a gift from Dr Michael X Zhu , University of Texas Health Science Center at Houston ) , Kv2 . 1 , Shaker IR channel ( with N terminal inactivation ball removed , used as WT ) and Shaker IR L382V ( V2 mutant ) ( all gifts from Dr Jon Sack , University of California at Davis ) , Shaker IR-based V369I , I372L , and S376T triple mutant ( ILT mutant ) ( a gift from Dr Kenton Swartz , National Institutes of Health at Bethesda ) , Kv4 . 3 ( a gift from Dr Kewei Wang , Peking University in China ) , and murine Slo1 construct for BK channel pore-forming subunit ( a gift from Dr Jim Trimmer , University of California at Davis ) . To facilitate identification of channel-expressing cells , fluorescence protein eYFP was fused in frame to the C-terminus of TRPV1 , while GFP was attached to the C-terminus of Kv4 . 3 and mSlo1 and N-terminus of Kv2 . 1 . Fusion of these fluorescence proteins did not change channel function , as previously described ( Antonucci et al . , 2001; Giraldez et al . , 2005; Cheng et al . , 2007; Cui et al . , 2008 ) . HEK293 cells were cultured at 37°C with 5% CO2 in a DMEM medium with 10% FBS , 100 U/ml penicillin and 100 mg/ml streptomycin . Cells were transiently transfected with ∼0 . 8 μg cDNA using Lipofectamine 2000 according to the manufacturer's instruction ( Invitrogen , Grand Island , NY ) . Electrophysiogical experiments were performed 24 to 48 hr after transfection . Macroscopic currents were recorded in the inside-out configuration using a HEKA EPC10 amplifier with PatchMaster software ( HEKA , Germany ) . Patch pipettes were pulled from borosilicate glass and fire-polished to a resistance of ∼2 МΩ . The conductance-voltage ( G-V ) curve was determined from currents in response to a series of voltage steps starting from a deep hyperpolarized voltage ( in most cases −100 mV ) . To record heat responses , both voltage-step and voltage-ramp protocols were applied . For voltage-step protocols , membrane potential was first clamped at a deep hyperpolarized voltage to fully close all channels . For Kv2 . 1 , Kv4 . 3 and Shaker ILT channels , it was then stepped to a depolarized voltage that activated the channels to 1-to-3% of their maximum voltage activation according to the G-V curves determined previously . For Shaker V2 , transmembrane voltage was stepped to a level that generated ∼60% maximum response , as heating would deactivate the channel . The transmembrane voltage was then stepped back to the original hyperpolarized level to close the channels . For voltage-ramp protocols , membrane potential was again first clamped at a hyperpolarized voltage , from which the voltage was linearly ramped up to a depolarized voltage that maximally activated the channels according to the G-V curves . Data were filtered at 2 . 8 kHz and sampled at 10 . 0 kHz . For TRPV1 channel , both bath solution and pipette solution contained 130 mM NaCl , 0 . 2 mM EDTA and 3 mM Hepes ( pH 7 . 2 ) . For potassium channels and their mutants , both bath solution and pipette solution contained 130 mM KCl , 0 . 2 mM EDTA and 3 mM Hepes ( pH 7 . 2 ) unless otherwise stated . To heat the membrane patch containing ion channels , the bath solution heated by an SHM-828 eight-line heater driven by a CL-100 temperature controller ( Harvard Apparatus , Holliston , MA ) was perfused to the pipette tip . A custom-made manifold was attached to the output ports of the heater to increase flow volume and to provide heat insulation . To accurately monitor local temperature at the pipette tip , we placed the IT-24P ultrafine thermocouple bead of a BAT-12 microprobe thermometer ( Physitemp , Clifton , NJ ) less than 1 mm from the pipette tip . HEKA patch-clamp amplifier registered temperature readings from the thermometer simultaneously with current recording . The speed of temperature change was set at a moderate rate of about 0 . 3°C/s . This rate ensured that heat-driven gating transitions of the channels reached equilibrium during the course of temperature change , so that the channels were recorded at the equilibrium state . When the experimental temperature was not controlled , recordings were conducted at room temperature at 24°C . Leak current was subtracted from raw current recordings before any data analysis was performed . Leak conductance was determined as the current measured at −100 mV divided by the driving voltage . The level of leak current at each testing voltage was calculated as the product of this leak conductance and the driving voltage . G-V curves were fitted to a single-Boltzmann function: ( 1 ) GGmax=11+e−qFRT ( V−V1/2 ) , where G/Gmax is the normalized conductance , V1/2 is the half-activation voltage , q is the apparent gating charge and F is Faraday's constant . Heat sensitivity of each channel type was quantified by the enthalpic change , ΔH , and the entropic change , ΔS , associated with the heat-induced activation process . To measure these thermodynamic parameters , we constructed a Van't Hoff plot from highly temperature-sensitive current changes and fitted it to the following equation: ( 2 ) lnKeq=−ΔHRT+ΔSR , where Keq is the equilibrium constant for heat-driven activation calculated from the channel open probability , R is the gas constant , and T is the temperature in Kelvin . Current was measured at a voltage that induced 1-to-3% channel open probability according to its G-V curve . The Q10 value of temperature-dependent single-channel conductance increase was set at 1 . 5 , with which the macroscopic current amplitude at different temperatures was corrected as if the temperature were 22°C ( Yang et al . , 2010 ) . For Kv channels , the conductance-corrected G-V curves were normalized to published maximum open probability . For TRPV1 , corrected-conductance was normalized to the conductance elicited by saturating concentration ( 3 μM ) of capsaicin . Open probability determined this way was further used to calculate Keq . Another way to quantify heat response was Q10 measurement . We first measured current amplitude ( I1 ) at the threshold temperature ( T1 ) where heat activation started . The current amplitude ( I2 ) at a temperature approximately 10°C higher ( T2 ) was measured for TRPV1 , Kv2 . 1 , Kv4 . 3 , Shaker ILT and Shaker V2 channels . For Shaker WT channel I2 was measured at a temperature 5°C higher than T1 because further elevation of the experimental temperature led to channel inactivation . Q10 was calculated as: ( 3 ) Q10= ( I2I1 ) 10T2−T1 . All statistical data are given as mean ± SEM . Student's t test was used to examine the significance of statistical differences: *p<0 . 05; **p<0 . 01; ***p<0 . 001; n . s . , no significance . | If you touch something too hot , it can cause you pain and damage your skin . Sensing the heat given off by an object or the temperature of the environment is possible , at least in part , because of proteins called temperature-sensitive TRP ion channels . These proteins are found in the cell membranes of nerve endings that are underneath the skin; and they open in response to heat , allowing ions to flow into the nerve cell . This in turn triggers a nerve impulse that is sent to our central nervous system and is perceived as heat and/or pain . The ability to sense heat was thought to be unique to these TRP ion channels , and it was believed that these ion channels contained an as-yet unidentified temperature-sensing domain . However , Yang and Zheng now report that similar ion channels , which open in response to changes in the voltage that exists across a cell's membrane , are also sensitive to changes in temperature . The temperature response of these ‘voltage-gated channels’ had largely eluded the attention of researchers in the past . This is because parts of the ion channel—which act like a ‘voltage sensor’ and only shift when the membrane voltage changes—normally keep the channel closed and directly open the channel when they move . Like all other proteins , ion channels are made from smaller building blocks called amino acids; and by changing some of the amino acids in the voltage-gated channel Yang and Zheng could decouple these normally linked actions . The changes to the channel meant that it did not immediately open when the voltage sensor moved; and decreasing the concentration of calcium ions inside the cell had the same effect as changing these amino acids . Both approaches revealed that , after a change in membrane voltage caused the voltage sensor to move , the ion channel remained closed until a high temperature caused it to open . Yang and Zheng revealed that the response of the modified voltage-gated channel to temperature was comparable to that of a typical heat-sensitive TRP ion channel . Further experiments showed that replacing some of the amino acids in the voltage-gated potassium ion channel with different amino acids could cause the channel to be either opened or closed by heat . The findings of Yang and Zheng indicate that temperature-sensing TRP channels may not contain a specialized heat-sensor domain . Instead , as these TRP ion channels do not require other parts of the protein to move in order to open the channel , they can be activated by their own inherent sensitivity to heat . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | High temperature sensitivity is intrinsic to voltage-gated potassium channels |
Uropathogenic E . coli ( UPEC ) , which cause urinary tract infections ( UTI ) , utilize type 1 pili , a chaperone usher pathway ( CUP ) pilus , to cause UTI and colonize the gut . The pilus rod , comprised of repeating FimA subunits , provides a structural scaffold for displaying the tip adhesin , FimH . We solved the 4 . 2 Å resolution structure of the type 1 pilus rod using cryo-electron microscopy . Residues forming the interactive surfaces that determine the mechanical properties of the rod were maintained by selection based on a global alignment of fimA sequences . We identified mutations that did not alter pilus production in vitro but reduced the force required to unwind the rod . UPEC expressing these mutant pili were significantly attenuated in bladder infection and intestinal colonization in mice . This study elucidates an unappreciated functional role for the molecular spring-like property of type 1 pilus rods in host-pathogen interactions and carries important implications for other pilus-mediated diseases .
To mediate colonization of host and/or environmental habitats , Gram-negative bacteria encode a highly conserved family of adhesive pili called chaperone-usher pathway ( CUP ) pili . Notably , CUP pili are critical virulence factors in a wide range of pathogenic bacteria , including Escherichia , Klebsiella , Pseudomonas , Haemophilus , Salmonella and Yersiniae genera ( Nuccio and Bäumler , 2007 ) . To date , 38 distinct CUP pilus types have been identified in Escherichia and Shigella genomes and plasmids , each of which is hypothesized to promote bacterial colonization of a distinct habitat ( Nuccio and Bäumler , 2007; Wurpel et al . , 2013 ) . Interestingly , single Escherichia coli genomes carry up to 16 distinct CUP operons , suggesting that the retention of an assortment of CUP pilus types by a single strain may be necessary to accommodate the complex lifecycle of E . coli ( Wurpel et al . , 2013 ) . Arguably , the best-studied CUP pili are those encoded by uropathogenic E . coli ( UPEC ) , which is the causative agent of the majority of urinary tract infections ( UTIs ) . UTIs affect 150 million people annually worldwide and are associated with significant morbidity and economic impact ( Foxman , 2014 ) . UTI treatment failure is common , with ~25% of woman suffering from recurrent UTI ( rUTI ) , and is due , in part , to the increasing prevalence of drug-resistant UPEC strains ( Scholes et al . , 2000; Zowawi et al . , 2015 ) . UPEC that infect the urinary tract often originate from the host gastrointestinal tract . After being shed from the gut in the feces , UPEC can colonize peri-urethral or vaginal areas and subsequently ascend through the urethra to the bladder and/or kidneys , instigating UTI . In mice , type 1 pili , which promote binding to mannosylated proteins , play critical roles in both the gut and urinary tract . Recent work has revealed that type 1 pili help mediate UPEC intestinal colonization , thus promoting the establishment and/or maintenance of the UPEC reservoir in the gut that can eventually seed UTI ( Spaulding et al . , 2017 ) . Upon entering the bladder , type 1 pili facilitate bacterial colonization and subsequent invasion into epithelial cells lining the bladder lumen ( Mulvey et al . , 1998; Martinez et al . , 2000 ) . Bladder invasion is a critical step in UPEC pathogenesis , allowing the bacteria to replicate in a niche protected from innate immune defense mechanisms , antibiotics , and expulsion during urination . UPEC that cannot invade the urothelium , like those lacking type 1 pili or its associated adhesin , FimH , are quickly cleared from the bladder , emphasizing the importance of type 1 pili mediated host-pathogen interactions on the fitness of UPEC during cystitis ( Wright et al . , 2005 ) . After invading into a bladder cell , UPEC escape into the host cell cytoplasm and replicate to form biofilm-like intracellular bacterial communities ( IBCs ) ( Anderson et al . , 2003; Justice et al . , 2004 ) comprised of ~104 cells ( Wright et al . , 2007 ) . Mouse models of UTI have revealed that while some mice are capable of self-resolving acute UPEC infection , others progress to chronic cystitis , which is characterized by persistent high titer bacteriuria ( >104 CFU/ml ) and high bacterial bladder burdens ( >104 CFU ) two or more weeks after inoculation ( Hannan et al . , 2010 ) . In the absence of antibiotic treatment , chronic cystitis can also be observed in women ( Mabeck , 1972; Ferry et al . , 2004 ) . E . coli strains , including UPEC , are grouped into distinct clades ( e . g . , clades A , B1 , B2 , D , and E ) based on their genetic relatedness ( Tenaillon et al . , 2010 ) . While UPEC strains tend to be genetically heterogeneous the majority of UPEC strains isolated from women with UTI in the USA reside in the B2 clade ( Schreiber et al . , 2017 ) . While the types of CUP pilus operons encoded in E . coli genomes varies between different clades and individual strains , the vast majority of sequenced E . coli strains , including nearly all sequenced UPEC clinical isolates , carry an intact copy of the type 1 pilus ( Wurpel et al . , 2013; Schreiber et al . , 2017 ) . Type 1 pili are encoded by the fim operon which , like the gene clusters encoding other CUP pili , encodes all the dedicated proteins necessary to assemble a mature pilus onto the bacterial surface , including: an outer-membrane pore-forming usher protein , a periplasmic chaperone protein , pilus subunits , and the tip adhesin protein . Most pilus tip adhesins , including the FimH adhesin , are made up of two domains , an N-terminal lectin domain , which is responsible for recognition and attachment to a ligand ( s ) , and a C-terminal pilin domain , which connects the adhesin to the bulk of the pilus ( Jones et al . , 1992 ) . Pilus subunits , including the pilin domain of the adhesin , are comprised of an incomplete immunoglobulin ( Ig ) -like fold , which lacks the C-terminal β-strand and require the action of the dedicated periplasmic chaperone for proper folding ( Sauer et al . , 2002 ) . In a process known as donor strand complementation ( DSC ) , a periplasmic chaperone templates subunit folding by transiently providing one of its β-strands ( Sauer et al . , 1999 ) . Chaperone-subunit complexes are then delivered to the outer membrane usher , which catalyzes pilus assembly via a reaction known as donor strand exchange ( DSE ) . In DSE , the strand donated to a nascent subunit by the chaperone is replaced by the N-terminal extension ( Nte ) of an incoming subunit ( Sauer et al . , 2002 ) . Following this pattern , the type 1 pilus usher ( FimD ) and chaperone ( FimC ) help type 1 pili assemble into a composite pilus structure consisting of a short tip fibrillum made up of the adhesin protein ( FimH ) and two minor subunits ( FimG and FimF ) that is joined to the pilus rod , a homopolymer of ~1000 FimA subunits . Once extruded to the extracellular surface , the type 1 pilus coils into a rigid right-handed helical structure that is capable of unwinding into a flexible linear fiber ( Abraham et al . , 1992; Jones et al . , 1995; Saulino et al . , 2000; Aprikian et al . , 2011 ) . This ability to transition between a coiled , helical rod and a linear fiber has been proposed to allow the type 1 pilus to act as a ‘molecular spring’ to maintain adherence in the face of fluid shear forces ( Zakrisson et al . , 2012 ) . Specifically , we hypothesize that in the absence of urine voiding , the type 1 pilus rod is maintained in the coiled helical state that permits subsequent contact and invasion into bound epithelial cells . However , upon encountering the shear forces associated with micturition , the pilus extends to the linear form , absorbing the shear force and thus preventing the expulsion of the bacteria from the bladder . Here , using high-resolution cryo-electron microscopy ( cryo-EM ) , we solved the structure of the type 1 pilus rod . Residues involved in critical FimA-FimA interactions were identified that when mutated reduced the force required to unwind the helix , despite not altering the ability of the FimA protein to be incorporated into the pilus rod . When introduced into the chromosome of UTI89 , a human UPEC clinical cystitis isolate , these point mutations dramatically reduced the ability of UTI89 to establish an intestinal reservoir and cause acute and chronic cystitis . In contrast , these point mutant strains did not result in large scale differences in levels of piliation or ability to agglutinate mannose-expressing guinea pig red blood cells in vitro . Taken together our results show that the identity of residues within the FimA rod , are critical for the type 1 pili mediated virulence of UPEC in the urinary tract and suggest that the helical pilus rod has an important functional role , beyond serving as a platform to present FimH , in promoting colonization in the gut and infection of the bladder .
To characterize FimA polymers that form the type 1 pilus rod , we solved the cryo-EM structure of native type 1 pili appearing in a preparation of recombinantly expressed Type IV pili ( T4P ) from the E . coli K12 strain BW25113 . As shown in Figure 1A , three types of filaments could be separated by eye: T4P , flagellar filaments , and a third class that were thicker and more rigid than T4P but thinner than the flagellar filaments . Sequencing the fimA PCR product showed that the encoded amino acid sequence was identical to the FimA protein from BW25113 ( GenBank AIN34588 . 1 ) and MG1655 ( GenBank NP_418734 . 1 ) . There was no possibility of cross-contamination of the FimA and T4P filament images , as each has a very different helical symmetry . Averaged power spectra from FimA filaments ( Figure 1—figure supplement 1 ) showed no trace of the Type IV pilus spectrum ( Figure 1—figure supplement 1 ) . The FimA reconstruction had an overall resolution of 4 . 2 Å ( Figure 1—figure supplement 1 ) , which is sufficient to build an atomic model of the structure ( DiMaio et al . , 2015 ) . There were no ambiguities in threading the known FimA sequence through the density map ( Supplementary file 1 ) , even though the electron density observed on the outside of the rod is less well defined ( Figure 1B ) . As a result , the first two residues and the last residue of FimA , which are located close to each other on the outside of the rod , are not resolved in the density map . The 4 . 2 Å estimate of the resolution is consistent with the well-separated β-strands and the density present for certain bulky side chains in the central lumen ( Figure 1C–E ) . Further , in the structure we observe the N-terminal donor strand of one subunit completing the β-sheet of the adjacent subunit . Overall , the type 1 pilus is a 70 Å diameter rod where each adjacent subunit rotates around the helical axis by 115° and translates along the axis by 7 . 7 Å ( Figure 2A ) . If we label each subunit along the 1-start helix by N , a subunit N0 interacts with six adjacent subunits ( N-1 , N-2 , N-3 , N+1 , N+2 , N+3 ) ( Figure 2A , B ) . The donor strand complementation involves mostly hydrophobic interactions . Most prominently , the hydrophobic amino acids Val5 , Val10 and Phe12 in the N-terminal β-strand ( Nte ) of one subunit are inserted into the next subunit’s hydrophobic groove created by a missing β-strand ( Figure 2—figure supplement 1 ) . Further , residues corresponding to FimABW2511322–33 , which are part of the interior surface of the rod helix , show structural differences between our cryo-EM rod structure and the previously solved crystal structures of FimA ( Crespo et al . , 2012 ) ( Figure 2—figure supplement 1 ) . In the DSC interaction of the FimA-FimC complex FimA residues 22–25 form a β-strand interaction with residues 59–56; and residues 31–33 form β-strand interactions with the FimC chaperone residues 102–104 ( Crespo et al . , 2012 ) . Further , within a self-complemented DSE structure FimA residues 29–33 form a β-strand interaction with the appended Nte residues ( 9-13 ) ( Walczak et al . , 2014 ) . Within the cryo-EM rod FimA structure , residues 25–29 adopt a helical conformation and residues 30–33 form a β-strand interaction with residues 10–13 from the Nte of the next FimA subunit . These residues form the center strand within the hollow helical core , and thus the ability of this region to adopt multiple conformations may be necessary in order for FimA to function as the major rod subunit . The interface between subunits N0 and N+3 is extremely important in maintaining the FimA helical rod . Previously , two atomic models ( PDB ID: 2N2H and 2MX3 ) for the type 1 pilus have been generated using the same solid-state NMR data ( Habenstein et al . , 2015 ) , with the helical symmetries for these models shown in Supplementary file 2 . The results were described as being FimA from uropathogenic E . coli ( UPEC ) ( Habenstein et al . , 2015 ) , however , the sequence corresponds to that of E . coli K12 strains , as does the recent FimA cryo-EM structure ( Hospenthal et al . , 2017 ) that is also described as being from UPEC . When subunit N0 is superimposed between our cryo-EM model with a subunit in 2N2H and 2MX3 , significant differences can be seen in the interface with subunit N+3 in the different models ( Figure 2—figure supplement 2 ) . The overall RMSD between Cα atoms of subunit N+3 in our cryo-EM model and 2N7H is 5 . 7 Å ( Figure 2—figure supplement 2 ) , while it is 11 . 0 Å with 2MX3 ( Figure 2—figure supplement 2 ) . Between 2MX3 and 2N2H , the overall RMSD is 5 . 6 Å for subunit N+3 when subunits N0 are aligned ( Figure 2—figure supplement 2 ) . Comparison of our type 1 pilus model with the P pilus structure ( Hospenthal et al . , 2016 ) shows an interesting difference: in the P pilus , residues 1–5 are fully ordered and make a 90° bend to form a ‘staple’ which involves contacts with two other subunits ( Figure 2—figure supplement 3 ) . In contrast , the first two residues are disordered in our type 1 pilus rod , and the remaining three residues in this region project straight out of the structure and make no contacts with other subunits ( Figure 2—figure supplement 3 ) . To explore the variation and evolution of FimA , we examined a set of 1 , 872 FimA protein sequences in the Ensembl Bacteria database ( Kersey et al . , 2016 ) ( Supplementary file 3 ) . After filtering to remove protein sequences that are expected to be non-functional and trimming of the signal sequence , the remaining 1 , 828 protein sequences were aligned . Here , we found that many of the FimA sequences were comprised of 159 residues , including BW25113 , while other FimA sequences , such as the one from UTI89 , a prototypical UPEC isolate , contained an additional two amino acid residues at the N-terminus of the mature protein ( Figure 3A ) . Global alignment revealed a consensus sequence of 161 residues where half of the FimA residues in the mature protein are invariant ( 79/161 residues ) and that 83 . 9% are highly conserved ( 139/161 residues with >95% sequence identity ) ( Figure 3A ) . We found the FimA sequences to be significantly more variable than other parts of the type 1 pilus machinery . For example , the mature forms of the FimC chaperone and FimH adhesin ( n = 1 , 760 and n = 1 , 943 , respectively; Supplementary file 4 ) obtained from the same database contain 97 . 6% and 97 . 5% highly conserved residues , respectively , as compared to 83 . 9% in FimA . In general , most E . coli genes evolve slowly ( Lee et al . , 2012 ) , so the degree of divergence between alleles of FimA is abnormally high . This suggests that the selection forces acting on the FimA protein are different from the forces acting on the rest of the fim operon . Interestingly , the non-conserved residues within the FimA sequences are all located on the exterior surface of the solved cryo-EM model ( Figure 3B ) , which is likely indicative of immune pressures selecting for antigenic diversification ( Wildschutte et al . , 2004 ) . In particular , the first few amino acids of the N-terminus of the mature protein , disordered in our reconstruction , have some of the highest variability throughout the entire protein , suggesting that these amino acids do not play a functional role in the dynamics of unwinding in the type one pilus rod . In contrast , almost every residue involved in subunit-subunit interactions in the rod is either highly conserved or invariant ( Figure 3B , C ) . This includes every residue lining the lumen ( Figure 3C ) . The conservation of residues lining the lumen might be due to the transport of some product through this opening that is ~15 Å across ( the height of the triangular lumen that rotates as one travels along the pilus ) . More likely , however , every residue in the lumen is either involved in making a subunit-subunit contact ( including the DSE contacts with the Nte from an adjacent subunit ) or packed tightly between residues that are making such contacts . To examine the importance of FimA-FimA interactions on bacterial pathogenesis , we chose to investigate how mutations in the fimA sequence of BW25113 altered functional and mechanical characteristics of type one pili . In addition , to assess whether FimA mutants altered the type-1 mediated ability of UPEC to infect the bladder and colonize the gut , we made mutations in the chromosomal fimA gene of UTI89 . Thus , we do note that FimAUTI89 differs from our cryo-EM rod FimABW25113 sequence in 16 residues , with all of the differences lying within highly variable residues located on the exterior of the pilus rod , including two additional amino acids at the N-terminus . These two additional N-terminal residues in UTI89 generate a shift in the numbering of UTI89 residues compared to FimABW25113 ( e . g . , A22R in BW25113 corresponds to A24R in UTI89 ) . To avoid confusion herein all residues are discussed using the BW25113 numbering system . To evaluate the evolutionary pressures shaping FimA and thus the type one pilus rod , we obtained gene sequences encoding the 1 , 828 FimA proteins from the Ensemble database to measure the selection pressures acting on the gene . After removing incomplete and duplicate sequences from the analysis , we were able to compare a total of 191 unique fimA sequences ( Supplementary file 3 ) , which were then trimmed to remove the signal sequence . These sequences encoding the mature protein were examined for evidence of recombination , which can result from horizontal transfer of genetic sequences between distinct strains and confound evolutionary analyses that assume all sequences were vertically inherited ( Anisimova et al . , 2003 ) . Phylogenetic trees were thus corrected for use in subsequent analyses ( Kosakovsky Pond et al . , 2006 ) . We then assessed the ratio of the rates of nonsynonymous ( dN ) to synonymous nucleotide mutations ( dS ) in each codon in the alignment ( dN/dS ) to estimate the selection pressures acting on each residue in the FimA protein using a fixed-effects likelihood measurement algorithm ( Kosakovsky Pond and Frost , 2005 ) . In general , a dN/dS ratio >1 is indicative of selection pressure favoring change in amino acid identity at the position ( i . e . , adaptive or positive selection ) while a dN/dS ratio <1 indicates conservation at the codon ( i . e . , purifying selection or negative selection ) . Here , we found that 66 codons were under purifying selection , five codons were under adaptive selection , and an additional 27 codons were too conserved to be included in analysis ( Supplementary file 5 ) . The remaining 63 codons showed no evidence of statistically significant selection in either direction . All of the positively selected sites , N64 , A109 , T117 , S120 and F138 , encode residues located on the exterior surface of the FimA rod and away from any subunit-subunit interface ( Figure 3—figure supplement 1 ) . The 66 codons under purifying selection combined with the 27 codons too conserved for analysis encode many of the highly conserved residues found throughout subunit-subunit interfaces with the subunits above , below , and alongside ( Figure 3B , C ) . This strongly suggests that evolutionary pressures are working to conserve the identity of residues that interact within the helical rod , while diversifying amino acid residues that are exposed to the host and susceptible to immune recognition . To more specifically determine if variation in FimA correlated with E . coli clades or pathogenic lifestyles , such as uropathogenicity , we examined the diversity and distribution of the fimA gene in a curated set of 67 E . coli genomes . This dataset included 21 distinct UPEC strains isolated from a cohort of women with frequent , recurrent UTI and 46 reference E . coli strains that included lab and commensal strains , as well as a variety of intestinal and extra-intestinal pathogens ( Schreiber et al . , 2017 ) ( Supplementary file 6 ) . The fimA gene was carried by the majority of E . coli strains analyzed ( 57/67 ) ; including nearly every UPEC strain ( 96 . 3% or 26/27 strains ) . Importantly , the fimA sequences from UPEC strains were spread through the phylogenetic tree , indicating that there was not a single variant of FimA found in all UPEC strains ( Figure 3—figure supplement 1 ) . To determine if these different variants were under different selection pressure , we measured the dN/dS ratio of each branch in this fimA phylogeny to see if there were branches that were under different selection pressures than the rest of the branches in the phylogeny ( i . e . , the ‘background’ rate of selection ) ( Kosakovsky Pond et al . , 2011 ) . Here , we identified three branches with statistically significant evidence of episodic diversifying selection , including two branches carrying most of the clinical UPEC strains ( branches labeled A and B ) as well as enterohemorrhagic and enterotoxigenic E . coli strains in the branch labeled C ( Figure 3—figure supplement 1 ) . This pattern of evolution is indicative of strong adaptive selection acting on some , but not all , branches in a phylogeny . Taken together , we find that fimA is much more diverse than other parts of the type 1 machinery , owing to high rates of nucleotide polymorphisms and genetic recombination , and that fimA has undergone repeated rounds of strong selective pressures that have conserved residues responsible for subunit-subunit interfaces while varying the external surface of the protein that is exposed to the host milieu . To determine which FimA residues are required to form the helical rod , we constructed single amino acid codon mutations in the BW25113 fimA gene , which were subsequently cloned into the expression vector , pTRC99a ( Amann et al . , 1988 ) . Residues were changed to either Arginine or Glutamine to insert large charged residues that would promote disruption of FimA interactions without making the surface more hydrophobic . We expressed these variant fimA genes in trans in UTI89-LON∆fimA , a strain with a chromosomal deletion of the fimA gene in a UTI89 strain where the phase-variable fimS promoter is locked in the ON orientation ( LON ) by altering the left inverted repeat necessary for promoter inversion ( Kostakioti et al . , 2012 ) . This strain transcribes the fim operon constitutively , which removes the possibility that differences in type 1 pili assembly/function are due to effects on phase variation by the mutants . We assessed the function of the type 1 pilus in each fimA mutant by measuring the agglutination of guinea pig red blood cells ( GP-RBCs ) . Most of our fimA mutants showed hemagglutination ( HA ) titers similar to the wild-type ( WT ) fimA ( Supplementary file 7 ) . However , several mutations ( A25R , V32R , V65R , V85R , and P145R ) abolished the production of adhesive pili ( Supplementary file 7 ) . This suggests that these mutations , which are located in areas of high conservation throughout the FimA monomer , may disrupt critical interactions with FimC or FimD or may not be able to correctly fold , thus preventing the assembly of the pilus . Notably , each of these residues was either under strong purifying selection pressure ( A25 , V85 and P145 ) or was too conserved for evolutionary analysis ( codons for V32 and V65 displayed no synonymous mutations and extremely limited non-synonymous mutation rates ) , which emphasizes their importance in assembly of the type 1 pilus rod ( Figure 3A ) . Expression of some mutant fimA genes , particularly V5R , E45R , E121R and A142R , resulted in a bacterial clumping phenotype when grown in static culture . D62R and D114R also instigated bacterial clumping , but to a lesser extent . Three of these substitutions , E45R , E121R and A142R , are found in residues with relatively high variability in our analysis of FimA sequences , with just 33 . 9% , 62 . 3% , and 52 . 1% sequence identity , respectively . Further , we found that two of these residues ( V5 and D114 ) showed statistically significant evidence of strong purifying selection ( Figure 3A , Supplementary file 6 ) . To measure the impact of FimABW25113 variants produced by the UTI89-LON∆fimA strain on pili force-extension responses , we chose six FimA mutant strains that did not disrupt pilus formation or produce severe bacterial clumping in vitro ( A22R , A92R , D114R , D62R , K155R and P132R ) and applied optical tweezers ( Figure 4—figure supplement 1 ) . Since type 1 pili are assembled from FimA subunits into a helical coil , they normally extend in three force phases: linearly increasing , constant force , linearly increasing – where the characteristic constant force originates from sequential unwinding of subunits ( region between the dashed lines in Figure 4A ) . The characteristic constant force extension enables the clear identification of instances where multiple pili are attached to a single microsphere , thus enabling us to determine the unwinding force of a single pilus ( Figure 4—figure supplement 2 ) . The unwinding force directly relates to the strength of layer-to-layer interactions ( e . g . , between subunits N0 and N+3 ) , thus any changes in these interactions caused by mutations would shift the magnitude of the force plateau . To investigate this , we first measured unwinding force of single pili in the strain expressing WT FimA and subsequently examined the force response for each fimA mutant . The unwinding force of WT pili was 30 . 3 ± 0 . 2 pN ( mean ±standard error ( SE ) ) whereas all mutants showed a reduction in the force required for pilus unwinding ( Figure 4B ) . The unwinding force measured for WT type 1 pili in this study is similar to what has been reported in the literature ( Andersson et al . , 2007 ) . In particular , the A22R variant showed the largest reduction in the unwinding force , 11 . 2 ± 0 . 2 pN . We can explain this by the fact that Ala22 in our rod model is tightly packed against Ala93 from another subunit and Val37 from a third subunit ( Figure 4C ) . Thus , replacing a small alanine with a long arginine side chain is expected to disrupt this interface . In the bladder , type 1 pili are required for binding and invasion of UPEC into superficial facet cells that line the bladder lumen and for the formation of intracellular bacterial communities ( IBCs ) during the first 6–18 hr of acute infection . This has been shown to be critical for ongoing infection in both humans and mouse models of cystitis ( Spaulding and Hultgren , 2016 ) . To determine if mutations in the FimA sequence altered UPEC pathogenesis in relevant mouse models , we constructed UTI89 strains with clean single codon mutations in the chromosomal fimA gene ( with the phase-variable fimS promoter intact ) . For these studies , we chose to investigate the effects of four FimA mutants that were examined via optical tweezers . Each of these mutations were made in the codon of a highly conserved amino acid positions of FimAUTI89 ( A22 , D62 , D114 , and P132 ) . When compared to a strain with the reintegrated WT UTI89 fimA sequence , we found that all reintegrated FimA mutant strains demonstrated similar levels of GP-RBC agglutination , in vitro ( Figure 5—figure supplement 1 ) , despite some variability in the level of piliation between FimA mutant strain ( Figure 5—figure supplement 1 ) . Interestingly , those mutants that show some increased hemagglutination in the presence of exogenous mannose compared to the WT strain correlated with those that showed some clumping in culture . Overall , these findings indicate that the FimA mutations , which reduced the unwinding force of the rod , do not prevent the expression or function of type 1 pili in vitro . We next investigated how each mutant altered the kinetics of bladder infection during competitive infections with WT UTI89 over 28 days . In mice that developed chronic cystitis ( defined as the development of persistent high titer ( >104 cfu/ml ) bacteriuria and high titer ( >104 cfu/ml ) bladder bacterial burdens at sacrifice >4 weeks post-infection ) , strains producing FimAUTI89 variants with the D62R , D114R , and A22R substitutions were outcompeted by up to six logs by the reintegrated WT strain ( Figure 5A–E ) . The FimA P132R variant had no effect on the ability of the strain to compete with the WT strain in chronically infected mice ( Figure 5B ) . Mice that resolved their infections ( defined as any animal whose urine or bladder titers dropped below 104 cfu/ml at least once during the 4 week infection ) in this experiment are shown in Figure 5—figure supplement 2 . In mice infected with a single UTI89 strain , the WT and P132R variant caused chronic cystitis at similar rates ( 45% and 30% , respectively ) ( Figure 5F , Figure 5—figure supplement 3 ) . However , the D114R and D62R mutant strains had reduced rates of chronic cystitis of 20% and 10% , respectively . Interestingly , 100% of mice infected with the A22R variant resolved their infections over the 4 week experiment , with half of the mice ( 10/20 ) exhibiting sterile urines by 10 dpi compared to just 15% ( 3/20 ) of mice infected with the WT strain at the same time point ( Figure 5—figure supplement 3 ) . The defect in chronic infection caused by fimA variants is likely due to pathogenic deficiencies during acute UTI . Accordingly , we found that two fimA variants ( A22R and D114R ) significantly altered the ability of UTI89 to form IBCs at 6 hr post infection ( hpi ) ( Figure 5G ) , with 8/10 mice infected with the D114R variant forming <15 IBCs . Even more strikingly , 70% of mice infected with the A22R variant formed no IBCs at 6 hpi and the other 30% formed three or less . This is in stark contrast to WT and P132R strains , which formed between 50–100 IBCs on average by 6hpi . Interestingly , the A22R mutant was severely attenuated in its ability to invade into bladder cells starting as early as 1 hr post infection ( Figure 5H ) . The D62R variant was defective in chronic cystitis in both the competitive and single bladder infections , but no acute fitness defects were observed . Further , mice infected with D62R formed a similar number of IBCs as the WT strain at 6 hpi , suggesting that the defect occurs at a later time-point . Accordingly , in mice singly infected with fimA D62R bacterial clearance is delayed , starting between 10–14 days ( Figure 5—figure supplement 3 ) . In addition to playing a pivotal role in the urinary tract , a recent study found that type 1 pili also promote the establishment and/or maintenance of the UPEC intestinal reservoir ( Spaulding et al . , 2017 ) . Deletion of the operon encoding the type 1 pilus or the FimH adhesin impedes intestinal colonization by UTI89 ( Spaulding et al . , 2017 ) . Therefore , we tested the impact of each of our four chromosomal FimA genetic variants on UPEC intestinal colonization levels . We found that strains producing FimAUTI89 variants D62R , D114R , and A22R showed lower levels of intestinal colonization ( by up to 2 logs ) in the feces , cecum , and colon of mice compared to those colonized by WT fimA ( Figure 6 ) . This 2-log decrease mirrors the defect observed in UTI89Δfim or UTI89ΔfimH strains , suggesting that these mutations prevent type 1 pilus-dependent gut colonization in UTI89 ( Spaulding et al . , 2017 ) . While not statistically significant , fimAUTI89 P132R variant showed lower colonization ( by up to one log ) in the feces and cecum than the WT strain ( Figure 6 ) . Together , our data indicates that the sequence of the FimA major subunit is critical for pilus function in the bladder and gut and thus has a major impact on the outcome of UTI .
Tremendous advances in cryo-EM within the past four years ( Egelman , 2016; Subramaniam et al . , 2016 ) largely driven by the introduction of direct electron detectors ( Li et al . , 2013 ) has meant that many complexes that were recalcitrant to crystallization can now be readily solved at near-atomic resolution by cryo-EM . In particular , it is exceedingly difficult to crystallize most helical polymers , as unless such a polymer has exactly two , three , four or six subunits per turn , it cannot be packed in any crystal space group so that all subunits are in equivalent environments . The type 1 pilin , FimA , has been extensively studied by x-ray crystallography and solution NMR , while the type 1 pilus has only been studied at high resolution by solid-state NMR ( Habenstein et al . , 2015 ) . We show here that type 1 filaments , present as a background in a preparation of T4P , allow us to reach a near-atomic level of resolution and build an atomic model for the FimA rod . The ability of cryo-EM to separate out multiple conformations among particles ( Gui et al . , 2017; Vonck and Mills , 2017 ) or even of subunits within the same particle ( Roh et al . , 2017 ) has been one of the greatest strengths of cryo-EM , allowing for multiple states to be solved from the same micrographs . Biochemically heterogeneous preparations , such as when virions are a mixture of empty particles and those containing DNA ( Dong et al . , 2017 ) are now routinely sorted out into homogeneous structural classes . We have taken advantage of that strength here to show that filament images which might otherwise have been discarded as a background contaminant can actually be used to build an atomic model . Our type 1 pilus rod model provides new insights into structure-function relationships in type 1 pili . FimA is critical for proper assembly of the type 1 pilus and , as the major subunit that makes up the pilus rod , is also critical for the proper display of the FimH adhesin . However , here we have uncovered a more complex and previously unappreciated role for FimA and the type 1 pilus rod in host-pathogen interactions . We identified three mutations in FimAUTI89 ( A22R , D114R , and D62R ) that expressed adhesive pili in vitro but reduced the ability of UTI89 to colonize the bladder , acutely and chronically . One of these mutants , FimAUTI89 A22R , was severely attenuated during acute infection , forming almost no IBCs and thus was unable to chronically infect the mouse bladder . We hypothesize that the identified FimA point mutants handicap UPEC pathogenicity in vivo by altering the properties of the pilus rod . The pilus rod is hypothesized to act as a ‘molecular spring’ transitioning between a flexible , linear fiber and a coiled helix . This spring-like property is thought to prevent the pilus from breaking or detaching from the host surface by temporarily expanding to a linear form after encountering shear forces , which can occur in the bladder during urine voiding or in the gut during fecal or mucus shedding . Such a transition between the helical and unwound form of the FimA homopolymer has been predicted to significantly dampen the force experienced by the adhesion-receptor complex at the tip ( Zakrisson et al . , 2012 ) . Since FimA-FimA interactions create the bulk of the pilus rod , mutations that reduce the stability of these protein-protein interactions can alter the ability of the pilus rod to withstand shear forces , which we discovered has detrimental effects on pilus function and pathogenesis . This is consistent with ~50% of the residues in the mature FimA protein ( 79/161 ) being invariant ( Figure 3 ) , which includes almost every residue involved in a subunit-subunit interface in the rod . Pili formed with the WT FimABW25113 protein require more than twice the force to unwind than the pili formed by the FimA BW25113 variant with an A22R mutation ( ~30 pN vs . 11pN ) , likely due to steric clashes caused by this substitution affecting the tightly packed interface between subunits . Thus , we hypothesize that the pilus formed by the FimA A22R mutant variant may not have the strength to absorb and withstand the shear forces experienced during urination and thus allow the bacteria to be swept out of the bladder , preventing infection . In the gut , the pilus rod may play a similar role during mucus shedding . A recent study suggests that type 1 pili may promote UPEC colonization of the upper crypts ( Spaulding et al . , 2017 ) . These bacteria would likely experience constant , low levels of shear force during mucus turnover and thus need the pilus to withstand some force to enable the bacteria to maintain their intestinal niche . Accordingly , three of the fimAUTI89 mutants ( A22R , D114R , and D62R ) that were attenuated in bladder colonization also displayed significantly reduced intestinal colonization . However , in general , the phenotypes of the fimAUTI89 mutants in gut colonization were not as severe as in the bladder . Two mutations in FimAUTI89 ( D62R and D114R ) resulted in a mild clumping phenotype when bacteria were grown in vitro . This clumping phenotype was also seen in the equivalent mutants in FimABW25113 ( D62R , D114R ) . These residues are located on the exterior surface of the rod structure and do not have as strong an effect on the force needed for unwinding as FimABW2511 A22R . However , the clumping phenotype suggests that these mutations may alter side-to-side interactions between different pilus rods promoting pilus-pilus interactions within and between bacteria , thus inhibiting the pilus mediated interactions needed in vivo for attachment , invasion and/or IBC formation . Interestingly , we do not observe a defect in bladder colonization in mice infected with FimAUTI89 D62R until 10dpi , suggesting that normal rod function is needed during both chronic and acute infection . Together , our data suggest that the type 1 pilus rod may mediate colonization phenotypes through damping of shear forces or through other mechanism ( s ) outside of scaffold support for FimH-binding . A possibility that we cannot exclude at the moment is that the lower uncoiling force of the A22R mutant exposes parts of the rod that would be largely buried in the wt filament to host proteases . This would still suggest that the mechanical properties of these rods have been ‘tuned’ to optimize for particular environments , as a wt rod that could never uncoil would be susceptible to breakage by the shear forces , while a rod that uncoils too easily would be unnecessarily exposed to digestion by proteases . Our analyses indicate that FimA is undergoing selection pressures due to as-yet undefined host-pathogen interactions , which may explain some colonization defects seen here . FimA displays patterns of episodic , divergent selection on surface exposed amino acids in a pattern that is similar , though much less robust , to what is seen in the flagellin protein in Salmonella ( Li et al . , 1994 ) . In Salmonella , the flagellin subunits are split into several domains where the domains responsible for subunit-subunit assembly are highly conserved while the more external domains of the protein show high rates of variability ( Andersen-Nissen et al . , 2005; Galkin et al . , 2008 ) . The Salmonella flagellin protein is immunogenic , but the regions that induce inflammatory response are found in the conserved domains of the protein ( Wildschutte et al . , 2004 ) . While it is known that flagellin are bound by Toll-like receptor 5 ( TLR5 ) resulting in the induction of the innate immune response ( Smith and Ozinsky , 2002 ) and that this recognition is targeted towards conserved features of flagellin ( Andersen-Nissen et al . , 2005 ) , it is currently unknown which , if any , host immune receptor are capable of recognizing the FimA rod or subunits or which structural features are targeted by the host immune system . Further research is needed to fully elucidate the host pressures and responses that are influencing the evolution of the FimA rod structure . Evolutionary and structural analysis of FimA , in combination with our in vitro and in vivo phenotyping , yielded several important insights into the selection pressures faced by UPEC as well as the evolutionary trajectories that pathogens follow to enhance their colonization of different host niches . Notably , we saw that different phenotypes caused by mutations in the FimA protein were associated with different classes of selection pressure . For example , the bacterial clumping phenotype is associated with mutation of three codons under purifying selection , but an equal number of the mutations that resulted in clumping are in codons with little to no evidence of selective pressure . In contrast , all mutations that failed to complement a fimA gene deletion were made in codons that are under very strong purifying selection or are too conserved for analysis . Together , the difference in selection pressure suggests that the mis-assembly of the pilus is much more harmful and/or toxic to E . coli than bacterial clumping . Further , the integration of evolutionary analysis with in vivo and in vitro functional analysis allowed us to decouple the selection pressures acting to preserve amino acid sites necessary for pilus assembly ( such as P145in FimAUTI89 ) from the selection pressure maintaining the codons that were necessary for pilus function ( such as A22 in FimAUTI89 ) . Given the intricacy of pilus assembly , one could expect that most of the codons under purifying selection would be related to pilus construction . Instead , we found that many codons under purifying selection are involved in keeping the force needed to unwind the type 1 rod within a narrow range . This leads to reasoning that a ‘weak’ or ‘loose’ FimA rod may be just as detrimental for E . coli as having no rod at all , at least in the eyes of evolution . In summary , by combining structural studies , force spectroscopy , genetic analysis , and relevant mouse models of UTI and gut colonization , we conclude that the mechanical properties of the type 1 pilus rod are essential for its functional role in mediating E . coli pathogenesis and persistence and appear to have been carefully ‘tuned’ by evolution . Further studies of the hundreds of CUP pili encoded in Gram-negative bacteria are needed to further understand the unique and general aspects of the evolution of CUP pilus fibers . In addition , other bacterial pili , such as T4P , which have arisen independently of CUP pili but can play similar roles in pathogenesis , can also elongate under force ( Biais et al . , 2010 ) and thus it remains an interesting question as to how the physical properties of other pili have been selected for particular environments and how these properties impact bacterial pathogenesis .
The Washington University Animal Studies Committee approved all procedures used for the mouse experiments described in the present study . Overall care of the animals was consistent with The Guide for the Care and Use of Laboratory Animals from the National Research Council and the USDA Animal Care Resource Guide . The BW25113 fimA gene sequence was cloned between the EcoRI and BamHI restriction sites in pTRC99A using standard PCR cloning techniques to create plasmid pTRC-fimA . Mutations were made within this plasmid using appropriate complementary primers to engineer codon changes in the template , pTRC-fimA , using Pfx polymerase and manufacturers instructions for PCR , followed by DpnI treatment of the resulting products to remove the methylated template before transformation into C600 . Mutations were verified by sequencing . Mutant plasmids were transformed into UTI89-LON , ∆fimA for expression and functional studies as indicated . In order to construct point mutations in the fimA allele in the UTI89 chromosome , the UTI89 fimA gene was deleted using a previously published technique that allows for flawless integration ( Khetrapal et al . , 2016 ) . Briefly , fimA was deleted by homologous recombination using pSLC- 217 as a template and primers containing 50 bp of homology to flanking regions of fimA . A deletion was then constructed using the previously described Red Recombinase method that would allow for reinsertion of constructs into the fimA site . Concurrently , a copy of UTI89 fimA was cloned into pTRC99a . Point mutations were then introduced into this construct using site directed mutagenesis . PCR fragments from confirmed mutants , and the WT allele , were reintegrated into the UTI89-LON , ∆fimA mutants constructed above at the original deletion site . Successful reintegration events were sequenced to confirm flawless integration and mutation presence . Animals were maintained in a single room in our vivarium . Prior to and after infection all animals received PicoLab Rodent Diet 20 ( Purina ) ad libitum . All animals were maintained under a strict light cycle ( lights on at 0600 hr , off at 1800 hr ) . Mice were acquired from indicated vendors and randomly placed into cages ( n = 5 mice/cage ) by employees of Washington University’s Division of Comparative Medicine ( DCM ) ; no additional methods for randomization were used to determine how animals were allocated to experimental groups . Investigators were not blinded to group allocation during experiments . For bladder infections , 6 week old female C3H/HeN mice were obtained from Envigo and were maintained in our vivarium for one week prior to infection . Bladder infections were performed via transurethral inoculation ( Hung et al . , 2009 ) . UPEC strains were prepared for inoculation as described previously ( Hung et al . , 2009 ) . Briefly , a single UTI89 colony was inoculated in 20 mL of Luria Broth ( LB ) and incubated at 37°C under static conditions for 24 hr . Bacteria were then diluted ( 1:1000 ) into fresh LB and incubated at 37°C under static conditions for 18–24 hr . Bacteria were subsequently washed three times with PBS and then concentrated to ~1×108 CFU per 100 μL for intestinal infections and ~1×108 CFU per 50 μL for bladder infections . Bacteria were subsequently washed three times with PBS and then concentrated to ~1×108 CFU per 50 μL for bladder infections . For intestinal colonization experiments , 6 week old female C3H/HeN mice were obtained from Envigo and were maintained in our vivarium for no more than 2 days prior to intestinal colonization . Mice received a single dose of streptomycin ( 1000 mg/kg in 100 μL water by oral gavage ( PO ) ) followed 24 hr later by an oral gavage of ∼108 CFU UPEC in 100 μL phosphate-buffered saline ( PBS ) ( Spaulding et al . , 2017 ) . Bacteria were subsequently washed three times with PBS and then concentrated to ~1×108 CFU per 100 μL for intestinal infections . In all cases , fecal and urine samples were collected directly from each animal at the indicated time points . Fecal samples were immediately weighed and homogenized in 1 mL PBS . Urine samples were immediately diluted 1:10 prior to plating . Mice were sacrificed via cervical dislocation under isofluorane anesthesia and their organs were removed and processed under aseptic conditions . Intestinal segments ( cecum and colon ) were weighed prior to homogenization and plating on LB supplemented with the appropriate antibiotic . Exclusion criteria for mice were pre-established; ( i ) both introduced strains in competitive infections became undetectable during the course of a 14 day experiment , and ( ii ) mice died or lost >20% of their body weight . No mice in this study met these criteria . Each experiment was conducted with both technical ( i . e . , a single inoculum of bacteria ) and biological ( i . e . , separate bacterial cultures of the same strain ) replicates . 6 week old female C3H/HeN mice were given a transurethral inoculation with WT UTI89 or a fimA mutant strain . To accurately count the number of IBCs , mice were sacrificed 6 or 12 hr after infection . Bladders were removed aseptically , bi-sected , splayed on silicone plates and fixed in 4% ( v/v ) paraformaldehyde . IBCs , readily discernable as punctate violet spots , were quantified by LacZ staining of bladder wholemounts ( Justice et al . , 2006; Cusumano et al . , 2011 ) . Bacterial invasion assays were performed at 1 , 3 , and 6 hr post infection as previously described ( Mulvey et al . , 1998 ) . Bacteria were grown under type 1 pilus-inducing conditions ( Greene et al . , 2015 ) , with appropriate antibiotics and . 01- . 02mM IPTG induction , if indicated . Pilus expression was assessed by hemagglutination assays ( HA ) as previously described ( Greene et al . , 2015 ) using bacterial cultures normalized to an optical density at 600 nm ( OD600 ) of 1 and guinea pig erythrocytes normalized to an OD640 of 2 . The experiment was conducted in parallel in PBS with 2% w/v methyl-α-D-mannopyranoside . Electron micrographs ( EM ) were taken of UTI89 or UTI89 isogenic mutants after growth under type 1 pilus-inducing conditions . A total of 300 bacterial cells were counted for each condition , and piliation on those cells was classified as bald ( no pili ) , low ( 1 to 20 pili/cell ) , moderate ( 20 to 200 pili/cell ) , or abundant ( >200 pili/cell ) . For expression of type 1 pili the strains were grown in Luria Broth ( LB ) supplemented with carbenicillin ( 100 µg/mL ) and IPTG ( 50 µM ) , at 37°C overnight . The optical tweezers ( OT ) setup is built around an inverted microscope ( Olympus IX71 , Olympus , Japan ) equipped with a high numerical aperture oil immersion objective ( model: UplanFl 100X N . A . = 1 . 35; Olympus , Japan ) and a 1292 × 964 pixel camera with a cell size of 3 . 75 × 3 . 75 μm ( model: StingRay F-125 , Allied Vision ) ( Mortezaei et al . , 2013 ) ( Figure 4—figure supplement 1 ) . The OT stands in a temperature controlled room with computers and controllers isolated from the room to reduce noise and vibrations . We use a continuous wave Nd:YVO4 laser ( Millennia IR , Spectra Physics , Santa Clara , CA ) operating at 1064 nm for trapping a single bacterium or microspheres . A probe laser ( low power HeNe-laser operating at 632 . 8 nm ) is merged with the trapping laser using a polarizing beam splitter cube ( PBSC ) . The light from the probe laser is refracted by the trapped object and collected by the condenser and thereafter imaged onto a 2D position sensitive detector ( PSD , L20 SU9 , Sitek Electro Optics , Sweden ) . The PSD convert the incoming light to a photocurrent and thereafter to a voltage that is sent to a programmable low pass filter ( SR640 , Stanford research systems ) , later collected by a computer and processed with an in-house LabVIEW program . We minimized the amount of noise in the setup and optimized the measured time series using the Allan variance method described in ( Andersson et al . , 2011 ) . To prepare a sample we suspended bacteria in 1xPBS to a concentration ( 1:1000 of OD600 = 1 ) suitable for single cell analysis using optical tweezers ( OT ) . Surfactant-free 2 . 5 µm amidine polystyrene microspheres ( product no . 3–2600 , Invitrogen , Carlsbad , CA ) were similarly suspended in Milli-Q water , these microspheres were trapped and used as force probes . To mount bacteria and reduce the influence of surface interactions we prepared a 1:500 suspension of 9 . 5 mm carboxylate-modified latex microspheres ( product no . 2–10000 , Interfacial Dynamics , Portland , OR ) in Milli-Q . We dropped ten microliters of the microsphere-water suspension onto 24 × 60 mm coverslips ( no . 1 , Knittel Glass , Braunschweig , Germany ) and placed these in an oven for 60 min at 60°C to immobilize the microspheres to the surface . To firmly adhere bacteria to the microspheres , we added a solution of 20 mL of 0 . 01% poly-L-lysine ( catalog no . P4832 , Sigma-Aldrich , St . Louis , MO ) to the coverslips , which , after 45 min incubation at 60°C , were stored until use . A free-floating bacterium was trapped by the optical tweezers run at low power to avoid cell damage . The bacterium was thereafter mounted on a large 9 . 5 µm microsphere coated with poly-L-lysine . We trapped a small free-floating 2 . 5 µm microsphere by the optical tweezers with normal power ( a few hundreds of mW ) and brought it close to ( within tens of µm ) but not in direct contact with , the bacterium . To calibrate the trap stiffness we used the Power spectrum method by sampling the microspheres position at 131 , 072 Hz and average 32 consecutive data sets acquired for 0 . 25 s each ( Tolić-Nørrelykke et al . , 2006 ) . Typically , the trap constant was found to be ~140 pN/µm for an output laser power of 800 mW . After calibration , the small microsphere was gently brought close to the bacterium in order to attach a pilus with the microsphere ( Figure 4—figure supplement 1 ) . To extend a single pilus ( Figure 4 ) the piezo stage was moved at a constant speed of 10 nm/s and the sampling frequency was set to 10 , 000 Hz that was downsampled by 800 . Occasionally , we measured the responses of multiple pili attached to the bead resulting in a force-extension response as the sum of all attached pili . This was , however , not a problem in general since the shorter pili detached from the bead in a sequential order , leaving only the single , longest pili attached to the bead for measurement ( Figure 4—figure supplement 2 ) . Finally , we controlled the piezo-stage and sampled the data using an in-house LabView program ( Andersson , 2018; copy archived at https://github . com/elifesciences-publications/ot-control ) . To make a flow chamber , we added a ring of vacuum grease ( Dow Corning , Midland , MI ) around the area containing the poly-L-lysine-coated microspheres on one of the coverslips . Carefully , we dropped a 3 mL suspension of bacteria and a 3 mL suspension of probe microspheres ( surfactant-free 2 . 5 mm white amidine polystyrene latex microsphere , product no . 3–2600 , Invitrogen , Carlsbad , CA ) onto the area and sealed the flow chamber by placing a 20 × 20 mm coverslip ( no . 1 , Knittel Glass ) on top . Thereafter , we mounted the sample in a sample holder that is fixed to a piezo-stage ( Physik Instrument , P-561 . 3CD stage ) in the OT instrumentation . To get a reliable OT calibration parameter values we measured the temperature using a thermocouple in the sample chamber , 23 . 0 ± 0 . 1°C and the suspension viscosity was assumed to only vary with temperature , thus , the viscosity was set to 0 . 932 mPas ± 0 . 002 mPas . To prepare the pilus extracts , bacteria of the E . coli strain BW25113 ( Datsenko and Wanner , 2000 ) were inoculated by dense streaking on eight M9 minimal agar plates containing 0 . 5% glycerol ( vol/vol ) . After a 72 hr incubation at 30°C , bacteria were harvested in 30 mL of LB medium , vortexed vigorously for 5 min and passed eight times through a 26-Gauge needle , to detach pili from the cells . Bacteria were removed at 4°C by three successive 10 min centrifugation steps at 16 , 000 x g . To collect the pili , cleared supernatants were centrifuged for 1 hr at 100 , 000 g in a cold Beckman Ti60 ultracentrifuge rotor . Pellet containing the crude pilus fraction was taken up in 200 μL of 50 mM HEPES , 50 mM NaCl pH 7 . 4 , and maintained at 4°C for further analysis . 3 μL of sample was applied to glow discharged lacey carbon grids ( TED PELLA , Inc . , 300 mesh ) . Then the grids were plunge-frozen using a Vitrobot Mark IV ( FEI , Inc . ) , and subsequently imaged in a Titan Krios at 300keV with a Falcon II direct electron detector ( pixel size 1 . 05 Å/pixel ) . A total of 6803 images , each of which was from a total exposure of 2 s dose-fractionated into seven chunks , were collected at a range of underfocus between 0 . 5 ~ 3 μm . Images were motion corrected using MotionCorr ( Li et al . , 2013 ) , and the program CTFFIND3 ( Mindell and Grigorieff , 2003 ) was used for determining the defocus and astigmatism . Images with poor CTF estimation as well as defocus >3 μm were discarded . The SPIDER software package ( Frank et al . , 1996 ) was used for most other operations with the first two-chunk sums ( containing a dose of ~20 electrons/ Å2 ) of the motion-corrected image stacks . The CTF was corrected by multiplying the images from the first two-chunk sums with the theoretical CTF , which is a Wiener filter in the limit of a very poor signal-to-noise ratio ( SNR ) . This both corrects the phases which need to be flipped and improves the SNR . The e2helixboxer routine within EMAN2 ( Tang et al . , 2007 ) was used for boxing the filaments from the images . A total of 72 , 627 overlapping segments ( 384 px long ) , with a shift of 11 px between adjacent segments ( ~97% overlap ) , were used for the IHRSR ( Egelman , 2000 ) reconstruction . With a featureless cylinder as a starting reference , 72 , 627 segments were used in IHRSR cycles until the helical parameters ( axial rise and rotation per subunit ) converged . Analysis of the population suggested that the axial rise was fairly fixed , but that the twist was variable . Using a reference-based sorting with models having a fixed rise but a variable twist , approximately 55% of the segments were excluded , having a twist outside of the range 114 . 4° to 115 . 6° . A sub-set of 32 , 726 segments were used for a few more cycles of IHRSR . The resolution of the final reconstruction was determined by the FSC between two independent half maps , generated from two non-overlapping data sets , which was 4 . 2 Å at FSC = 0 . 143 . We used a previous FimA NMR model ( PDB id: 2JTY , a single chain ) as an initial template to dock into the cryo-EM map by rigid body fitting , and then manually edited the model in Chimera ( Pettersen et al . , 2004 ) and Coot ( Emsley et al . , 2010 ) . We then used the combined model ( 1–19 and 21–159 ) as the starting template to re-build a single chain of the FimA protein using the RosettaCM protocol ( Wang et al . , 2015 ) . Next , the full length model of FimA missing the first two residues and the last residue was iteratively refined by Phenix real-space refine ( Adams et al . , 2010 ) and manually adjusted in Coot . The refined single chain of the FimA model was then re-built by RosettaCM ( Wang et al . , 2015 ) with helical symmetry and refined by Phenix to improve the stereochemistry as well as the model map coefficient correlation . The FimA model was validated with MolProbity ( Chen et al . , 2010 ) and the coordinates deposited to the Protein Data Bank with the accession code 6C53 ( atomic structure ) . The corresponding cryo-EM map was deposited in the EMDB with accession code EMD-7342 . The refinement statistics are given in Supplementary file 1 . The protein sequences of closely-related homologs of E . coli FimA , FimC , and FimH were obtained by individual searches the Ensembl Bacteria Genome database ( Kersey et al . , 2016 ) using the phmmer web-server ( Finn et al . , 2015 ) with a BLOSUM62 ( FimA ) or a BLOSUM90 ( FimC and FimH ) scoring matrix using the full-length E . coli UTI89 protein sequences as queries . The sequence matches were then filtered to remove low scoring hits and non-functional sequences ( i . e . , those predicted to lack critical sequence features such as complete signal sequences and/or C-terminal tyrosine residues in FimA and FimH ) . The signal sequences were trimmed from the remaining homologs using Geneious v 6 . 1 . 7 ( Kearse et al . , 2012 ) and the protein sequences were aligned with the MAFFT program using two iterations of the FFT-NS-i algorithm based on the PAM200 scoring matrix ( Katoh and Standley , 2013 ) . Conservation at each amino acid position was calculated using custom Python scripts and a sequence logo was created using the ggseqlogo package in R ( R Core Team , 2017 ) using RStudio ( RStudio Team , 2015 ) . To estimate selection pressures on each codon in fimA , we obtained all available gene sequences encoding the protein sequences described above from the Ensembl Bacteria Genomes database ( n = 1 , 825 , three were removed by submitter’s request ) using custom bash scripts ( Supplementary file 3 ) . A total of 191 unique sequences were identified using Geneious v 6 . 1 . 7 and trimmed to remove the signal sequence . Evolutionary model selection was performed using maximum likelihood ratio testing on the Datamonkey webserver ( Pond and Frost , 2005; Delport et al . , 2010 ) , which identified the TIM2 model ( model 010232 ) as the most likely model of nucleotide substitution for the fimA homologs . These sequences were then were scanned for evidence of recombination using single breakpoint analysis ( Kosakovsky Pond et al . , 2006 ) and phylogenetic trees with a correction for a breakpoint found in codon 107 ( position 320 ) were generated . These phylogenetic trees and evolutionary model were then used to measure the ratio of the rates non-synonymous ( dN ) to synonymous ( dS ) mutation in each codon ( i . e . , a dN/dS ratio ) using a fixed-effects likelihood test to identify statistical significance ( Kosakovsky Pond and Frost , 2005 ) . Using a collection of 67 curated , reference genomes from a previous study ( Schreiber et al . , 2017 ) , we examined the carriage and phylogenetic context of fimA carriage using a BLAST-based search ( Camacho et al . , 2009 ) with the UTI89 fimA gene as a query . Full-length sequences were extracted from the genomes using Geneious v 6 . 1 . 7 , trimmed to remove signal sequences , and aligned using the MUSCLE program ( Edgar , 2004 ) . A phylogenetic tree was estimated using the RAxML program ( Stamatakis , 2006 ) with the GTRCAT model and supported with 1000 bootstraps ( Stamatakis et al . , 2008 ) . Evidence for episodic , diversifying selection was then identified using a random effects likelihood ratio test for each branch of the fimA phylogenetic tree ( Kosakovsky Pond et al . , 2011 ) using unique sequences from the genomes ( 32 duplicates removed , n = 25 ) . Branches showing statistically significant evidence for episodic , diversifying selection , as measured by a chi-squared test were then indicated on the corresponding branches of the phylogenetic tree constructed using RAxML . | Escherichia coli , or E . coli for short , is a type of bacteria commonly found in the guts of people and animals . Certain types of E . coli can cause urinary tract infections ( UTIs ) : they travel from the digestive tract up to the bladder ( and sometimes to the kidneys ) where they provoke painful symptoms . To cause the infection , the bacteria must become solidly attached to the lining of the bladder; otherwise they will get flushed out whenever urine is expelled . Pili are hair-like structures that cover a bacterium and allow it to attach to surfaces . E . coli has many different types of pili , but one seems particularly important in UTIs: type 1 pili . These pili are formed of subunits that assemble into a long coil-shaped rod , which is tipped by adhesive molecules that can stick to body surfaces . The current hypothesis is that the pili act as shock absorbers: when the bladder empties , the pili’s coil-like structure can unwind into a flexible straight fiber . This would take some of the forces off the adhesive molecules that are attached to the bladder , and help the bacteria to remain in place when urine flows out . However , the exact structure of type 1 pili is still unclear , and the essential role of their coil-like shape unconfirmed . Here , Spaulding , Schreiber , Zheng et al . use a microscopy method called cryo-EM to reveal the structure of the type 1 pili at near atomic-level , and identify the key units necessary for their coiling properties . The experiments show that pili with certain mutations in these units unwind much more easily when the bacteria carrying them are ‘tugged on’ with molecular tweezers . The bacteria with mutant pili are also less able to cause UTIs in mice . The coiling ability of the type 1 pili is therefore essential for E . coli to invade and colonize the bladder . Every year , over 150 million people worldwide experience a UTI; for 25% of women , the infection regularly returns . Antibiotics usually treat the problem but bacteria are becoming resistant to these drugs . New treatments could be designed if scientists understand what roles pili play in the infection mechanisms . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"microbiology",
"and",
"infectious",
"disease"
] | 2018 | Functional role of the type 1 pilus rod structure in mediating host-pathogen interactions |
FK506 ( Tacrolimus ) is a potent inhibitor of calcineurin that blocks IL2 production and is widely used to prevent transplant rejection and treat autoimmunity . FK506 treatment of dendritic cells ( FKDC ) limits their capacity to stimulate T cell responses . FK506 does not prevent DC survival , maturation , or costimulatory molecule expression , suggesting that the limited capacity of FKDC to stimulate T cells may be due to inhibition of calcineurin signaling in the DC . Instead , we demonstrate that DC inhibit T cells by sequestering FK506 and continuously releasing the drug over several days . T cells encountering FKDC proliferate but fail to upregulate the survival factor bcl-xl and die , and IL2 restores both bcl-xl and survival . In mice , FKDC act in an antigen-specific manner to inhibit T-cell mediated autoimmune arthritis . This establishes that DCs can act as a cellular drug delivery system to target antigen specific T cells .
FK506 ( Tacrolimus ) has a long record of clinical success in preventing transplant rejection and treating autoimmunity , but its use is limited by side effects including diabetes , hypertension , nephrotoxicity and neurotoxicity . FK506 is a potent inhibitor of calcineurin , a phosphatase stimulated by T cell receptor signaling to regulate the transcription factor nuclear factor of activated T cells ( NFAT ) . Through this mechanism , FK506 inhibits IL2 production ( Bierer et al . , 1990 ) and triggers activated T cell death ( Horigome et al . , 1998; Migita et al . , 1999 ) . FK506 also limits the capacity of dendritic cells ( DC ) to stimulate allogeneic T cell responses in vitro ( Woltman et al . , 2000; Szabo et al . , 2001; Duperrier et al . , 2005 ) . While other immunosuppressants , including rapamycin and mycophenolate , are known to inhibit DC maturation ( Hackstein and Thomson , 2004; Popov et al . , 2006 ) and thereby ameliorate autoimmunity ( Popov et al . , 2006 ) or promote long term transplant tolerance when transferred in vivo ( Turnquist et al . , 2007 ) , FK506 does not inhibit DC maturation ( Woltman et al . , 2000; Szabo et al . , 2001; Duperrier et al . , 2005 ) . These observations have suggested that FK506-mediated blockade of calcineurin signaling in DCs impacts T cell stimulation ( Hackstein and Thomson , 2004 ) . Instead , we find that FK506 treated DC ( FKDC ) sequester and release the drug itself , at doses sufficient to block T cell activation , to target antigen specific immune responses and to prevent collagen induced arthritis in mice . This demonstrates that DC can act as a biologic agent for drug delivery , with the potential to reduce drug dose and increase specificity in the treatment of autoimmune disease .
To examine the mechanism by which FK506 acts upon DCs , we treated human DC with FK506 ( FKDC ) for 40 hr during maturation . Despite washing to remove residual drug from the supernatants , FKDC were less effective stimulators of allogeneic T cell proliferation than DC ( Figure 1A ) . Similarly , when memory T cell responses to influenza were measured by IFNγ Elispot assay ( Albert et al . , 1998; Orange et al . , 2004 ) , FKDC were found to be markedly less stimulatory than untreated DCs ( Figure 1B ) . Moreover , in agreement with previous work ( Woltman et al . , 2000; Duperrier et al . , 2005 ) , DC survival , maturation , expression of MHC and costimulatory molecules were unaffected by FK506 treatment ( Figure 1C ) and therefore could not account for their reduced stimulatory capacity . We next exposed DC to FK506 for various times . Unexpectedly , FK506 pretreatment of mature DC for as little as 20 min led to the same level of T cell inhibition in the IFNγ Elispot assay as pretreatment for 40 hr ( Figure 1D ) , reminiscent of previous work demonstrating maximal 3H-FK506 uptake by Jurkat T cells in only 20 min of FK506 treatment ( Siekierka et al . , 1989b ) . The allogeneic mixed lymphocyte reaction ( allo-MLR ) and Elispot assay require prolonged co-culture ( 5 days and 40 hr , respectively ) . To better define the molecular nature of the FKDC effect , T cells and DC were co-cultured for only 6 hr and IFNγ mRNA induction was measured by qRT-PCR . Consistent with the results from Elispot assays , FKDC were poor inducers of T cell IFNγ mRNA responses to influenza ( Figure 1E ) . To determine if FKDC lack an activating signal or produce an inhibitory signal , untreated DC were mixed with FKDC . Combining FKDC with untreated DC inhibited the T cell stimulatory capacity of the untreated DC in a dose dependent manner ( Figure 1E ) . To test whether inhibition was mediated by a secreted factor , T cells were resuspended in DC or FKDC supernatant and stimulated with anti-CD3/28 . FKDC supernatants , but not those of untreated DC , were potent inhibitors of IFNγ , IL17A , IL4 and IL2 induction ( Figure 1F ) . Taken together , these data provide strong evidence that FKDC produce a T cell inhibitor . 10 . 7554/eLife . 00105 . 003Figure 1 . FK506 treated DC secrete a direct T cell inhibitor . ( A ) T cell proliferative responses to allogeneic DC or FKDC after 5-day coculture . DC wash control indicates untreated DCs suspended in supernatant of the last wash of FKDC . Data are mean counts per minute ± standard deviation ( SD ) of triplicate wells . ( B ) T cell IFNγ Elispot response to DC or FKDC presenting influenza infected ( FLU+ ) or control ( FLU− ) apoptotic 3T3 cells . Data are mean spot forming cells ( SFC ) per million cells + SD of triplicate wells . ( C ) Phenotype of FK506 treated or untreated DC . ( D ) CD8 T cell IFNγ Elispot response to DC treated with 40 hr , 30 hr , 3 hr or 20 min of FK506 . Data are mean SFC per million cells + SD of triplicate wells . ( E ) CD8 T cell induction of IFNγ mRNA in response to 6-hr stimulation with DC or FKDC . 1 × 106 T cells cultured with various ratios of syngeneic DC and FKDC ( x-axis ) . Data are mean IFNγ mRNA induction of T cells cultured with DC presenting influenza vs DC presenting control cell + SD of triplicate wells . ( F ) Anti-CD3/28 stimulated T cells cultured in untreated DC or FKDC supernatants . Data are mean mRNA induction of stimulated vs unstimulated groups + SD of triplicate wells . p values were obtained using two-tailed unpaired t-test . Data are representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 00105 . 003 To test if the inhibitory action of FKDC requires new gene transcription , translation or translocation through the secretory pathway , FKDC were pretreated with actinomycin , emetine , or brefeldin A prior to FK506 treatment . After washing and co-culture with T cells stimulated with anti-CD3/28 , none of these treatments affected FKDC secretion of the T cell inhibitor ( Figure 2A ) . We hypothesized that FK506 diffuses passively from FKDC , despite adequate washout from initial cultures . Two approaches were used to test this hypothesis: a direct measurement of FK506 from FKDC culture supernatants and a functional assay to determine if the inhibitory effect of FKDC supernatants could be abolished by removal of FK506 from the media . First , DC were treated with FK506 , washed extensively , and cultured for 1–4 days , during which time they were washed daily and supernatants collected 6 hr later . FK506 was maintained in the supernatants at levels sufficient to block T cell activation for 72 hr after initial washout ( Figure 2B ) . Next , to assess whether FK506 is the only inhibitor , supernatants were treated with magnetic beads coated with FK-1 antibody that binds with high affinity to FK506 and then immunoadsorbed on a magnetic column . T cells were resuspended in FK-1 depleted or control supernatants and stimulated with anti-CD3/28 . Treatment of FKDC supernatants with FK-1 depleted supernatant specifically abolished the inhibitory factor completely ( Figure 2C ) . To determine how FKDC might function as a repository for FK506 , we evaluated DC gene expression profiles for expression of FK binding proteins . DCs express several FKBP , with FKBP12 ( also known as FKBP1A ) being the most robustly expressed ( Table 1 ) , suggesting that these proteins act in concert as an FK506 drug sink . Rapamycin is another immunomodulatory drug , which binds FKBP12 , and it binds FKBP12 with higher affinity than FK506 ( FK506 Kd = 0 . 4 nM , Rapamycin Kd = 0 . 2 nM ) ( Bierer et al . , 1990 ) . Unlike FK506 , which blocks IL2 cytokine production , Rapamycin inhibits the downstream effects of IL2R signaling . The binding of FK506 and Rapamycin to a common intracellular protein was originally demonstrated in Jurkat T cells , which were rendered resistant to FK506 mediated blockade of IL2 induction when pretreated with Rapamycin ( Bierer et al . , 1990 ) . We used this strategy to test if FKBP12 is specifically important for the accumulation of FK506 in DCs . Similar to the reported results in Jurkat cells , we found that pretreatment with Rapamycin renders FKDC unable to block IL2 induction in T cells ( Figure 2D ) . This data indicates FKBP12 plays a physiologically relevant role in mediating the action of FKDC on T cell inhibition . 10 . 7554/eLife . 00105 . 004Figure 2 . FK506 derived from FK506 treated DC blocks T cell activation . ( A ) Transcription , translation and translocation through the secretory pathway are not required for FKDC to produce a T cell inhibitor . DC were pretreated with actinomycin , emetine , brefeldin A or media prior to treatment with FK506 or media control and washed extensively . T cells were stimulated for 4 hr with anti-CD3/28 or media and mixed with treated DC . Data are mean IFNγ mRNA induction of stimulated vs unstimulated control T cell groups + standard deviation ( SD ) of triplicate wells . ( B ) FKDC were washed and supernatants measured daily via ELISA . Levels after last wash were undetectable . Dashed line indicates FK506 concentration that inhibits 50% induction of IL2 . Data are nM ± SD of triplicate wells . ( C ) FKDC or untreated DC supernatants treated with FK-1 ( anti-FK506 ) or anti-KLH ( isotype control ) antibody depletion . Depleted supernatants were added to anti-CD3/28 stimulated T cells . Data are mean IFNγ mRNA induction + SD of triplicate wells . ( D ) DCs treated with 0 . 5 μM rapamycin or media control for 18 hr prior to treatment with 0 . 5 μM FK506 or media control for 1 hr and washed extensively . Syngeneic CD4+ T cells were cultured with various treated DCs and CD3/28 beads for 4 hr . Data are mean fold induction of IL2 mRNA of stimulated vs unstimulated cells ± SD . p values were obtained using two-tailed unpaired t-test . Data are representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 00105 . 00410 . 7554/eLife . 00105 . 005Table 1 . Mature , monocyte derived DC express mRNA of several FKBPDOI: http://dx . doi . org/10 . 7554/eLife . 00105 . 005Accession numberGene nameMean affymetrix expression levelStandard deviationNM_000801 . 1FK506-binding protein 1A ( 12kD ) ( FKBP1A ) 1210 . 0178 . 4NM_004470 . 1FK506-binding protein 2 ( 13kD ) ( FKBP2 ) 370 . 823 . 7NM_003602 . 1FK506-binding protein 6 ( 36kD ) ( FKBP6 ) 289 . 051 . 1NM_012181 . 1FK506-binding protein 8 ( 38kD ) ( FKBP8 ) 211 . 882 . 9NM_004117 . 1FK506-binding protein 5 ( FKBP5 ) 203 . 640 . 8NM_002014 . 1FK506-binding protein 4 ( 59kD ) ( FKBP4 ) 121 . 847 . 1NM_002013 . 1FK506-binding protein 3 ( 25kD ) ( FKBP3 ) 116 . 521 . 2AF322070 . 1FK506-binding protein FKBP9115 . 611 . 6NM_004116 . 1FK506-binding protein 1B ( 12 . 6 kD ) ( FKBP1B ) 90 . 725 . 3FK506 binding proteins expressed in mature human DCs are ranked by mean Affymetrix expression level; values over 200 are considered to be robustly present . Mean expression and standard deviation are derived from four biologic replicates . Taken together , these results demonstrate that DCs can absorb significant amounts of FK506 , and then release drug into the supernatant in concentrations sufficient to inhibit T cell activation in a sustained manner , for at least 72 hr after a single brief treatment of DCs . We next evaluated the mechanism by which FKDC inhibit T cells . Previous work suggests that T cell activation when calcineurin is inhibited leads to antigen specific T cell death ( Vanier and Prud'homme , 1992; Migita et al . , 1995 , 1997; Horigome et al . , 1998 ) . To examine the effect of FKDC on T cell survival , CFSE labeled T cells were cultured with allogeneic FKDC or untreated DC and apoptosis was measured by annexin V staining of undivided and postmitotic cells . There was no difference in the amount of cell death in undivided T cells ( Figure 3A ) . As expected , proliferating T cells stimulated with untreated DC divided , and adding IL2 increased cell death consistent with overstimulation of the IL2 pathway and induction of activation induced cell death ( AICD ) ( Lenardo , 1991 ) . In contrast , T cells proliferating in response to FKDC were twice as likely to die and IL2 increased survival ( Figure 3B ) , suggesting FK506 blocked normal T cell signaling , including IL2 production , such that cell survival was rescued by additional IL2 . Since activated T cell death can be AICD-independent ( Hildeman et al . , 2002 ) and regulated by bcl-2 family members , we examined this pathway . Fewer proliferating T cells upregulated the survival factor bcl-xl in response to FKDC than untreated DC ( Figure 3C ) , suggesting cell autonomous ( mitochondrial mediated ) cell death ( Akbar et al . , 1996; Hildeman et al . , 2002 ) , consistent with reports that FK506 sensitizes activated T cells to apoptosis by blocking bcl-xl induction ( Migita et al . , 1995 , 1997 ) . Moreover , we found IL2 augments bcl-xl induction and rescues T cells from death ( Figure 3C ) . Furthermore , IL2 rescues IFNγ production in FKDC activated memory T cells ( Figure 3D ) . In conclusion , co-culture of T cells with FKDC initiates a program of activation that leads to increased apoptosis . 10 . 7554/eLife . 00105 . 006Figure 3 . FK506 derived from FK506 treated DC causes activated T cell death in vitro . ( A ) Annexin V stained CFSE labeled allogeneic CD4+ T cells cultured with DC or FKDC ± IL2 for 4 days . Boxes indicate gating strategy for selecting undivided or divided T cells in Figure 3B and C . ( B ) Histogram of annexin V staining of undivided or divided T cells cultured with allogeneic DC or FKDC ± IL2 . Percent annexin V positive divided cells . ( C ) Histogram of bcl-xl staining of undivided or divided T cells cultured with allogeneic DC or FKDC ± IL2 . Percent bcl-xl positive divided cells . ( D ) CD8 T cell IFNγ Elispot response to DC or FKDC presenting FLU infected apoptotic cells supplemented with IL2 . Data are mean SFC + standard deviation of triplicate wells . p values were obtained using two-tailed unpaired t-test . Data are representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 00105 . 006 To test the functional significance and specificity of FKDC-mediated T cell inhibition in vivo , we tested the effect of FKDC in the collagen induced arthritis animal model ( CIA ) . DBA1/J mice were primed with intradermal bovine type II collagen ( CII ) emulsified in complete Freund's adjuvant ( CFA ) , which induces a reproducible and severe arthritis . 7 and 14 days later , mice were intravenously infused with either untreated DC or FKDC loaded with either CII , or an irrelevant antigen , type I collagen ( CI ) . Since some reports suggest FK506 may impair DC antigen presentation ( Lee et al . , 2005 ) , DC were pulsed prior to FK506 treatment . Mice infused with FKDC pulsed with CII , but not with CI or untreated DC pulsed with CII , were significantly protected as measured by mean arthritis severity score ( Figure 4A ) and by incidence of severe arthritis ( Figure 4B ) . Similar trends were observed in two other independent CIA experiments , though no formal statistical tests were performed due to small sample size ( n = 3–5 per group ) . At day 87 after disease induction , all paws of control mice and mice treated with FKDC-CII or DC-CII were assigned histopathology ( Figure 4C ) and radiographic ( Figure 4D ) severity scores , revealing that mean arthritis severity scores were highly correlated with radiographic scores and histopathology severity scores ( r = 0 . 97 and 0 . 96 , respectively ) . Mice treated with FKDC-CII had significantly reduced radiographic ( Wilcoxon rank test p<0 . 05 ) and histopathology severity scores ( Wilcoxon rank test p<0 . 05 ) compared to mice treated with DC-CII . 10 . 7554/eLife . 00105 . 007Figure 4 . FK506 treated DC modulate antigen specific immune responses in vivo . ( A ) DBA1/J bone marrow derived DC pulsed with either type I or II collagen and treated with FK506 or media . DC were transferred intravenously 7 and 14 days after type II collagen/CFA immunization . Data are mean arthritis severity score of each group on each day . N = 8–12 per group . ( B ) Incidence of severe arthritis . ( C ) Histology of normal paw and paw of CIA mouse treated with FKDC-CII or DC-CII ( mouse with median histologic score per group is presented ) . C: Cartilage damage; R: Bone resorption; I: inflammation; P: pannus; O: Osteophyte . ( D ) Radiographs from paws presented in ( C ) . ( E ) CD4+ splenocytes from an untreated mouse were cultured with serum drawn from PBS , FK506 or FKDC treated mice after the last dose of treatment ( day 14 ) , and the CD4+ T cells then stimulated for 4 hr with CD3/28 beads . Data are mean fold induction of IL2 mRNA of stimulated vs unstimulated groups + standard deviation . N = 6–10 mice per group . p value was calculated using a two tailed t-test to compare FK506 and FKDC groups . ( F ) Serum chemistries of mice treated with 14 days , as indicated , with PBS , FK506 or FKDC , as described in ( E ) . N = 6–10 mice per group . Gray circles are serum amylase values for individual mice . Black bars represent median values per group . p value was calculated using a two-tailed Mann–Whitney test comparing FK506 and FKDC treated groups . Data from ( E ) and ( F ) are representative of two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 00105 . 007 Systemic treatment with FK506 is also effective for preventing CIA ( Takagishi et al . , 1989 ) . In work by Takagishi et al , groups of 10–16 DBA/1 mice were immunized with intradermal bovine CII emulsified in CFA ( similar to the method of arthritis induction in this manuscript ) . They compared 1 , 2 , 3 or 4 mg/kg FK506 or saline injected subcutaneously each day between day 0 and day 13 . Mice treated with 4 mg/kg FK506 do not develop any arthritis; however mice treated with 1 mg/kg FK506 receive no protection from arthritis . Mice treated with 2 mg/kg on day 0–13 led to a 50% reduction in arthritis severity compared to placebo . We compared the toxicity of our FKDC therapeutic regimen with toxicity seen using the lowest therapeutically efficacious dose ( 2 mg/kg ) of systemic FK506 . Six to 10 mice were allocated to one of three treatment groups: subcutaneous 2 mg/kg FK506 or PBS control daily for 14 days , or 0 . 5 million FKDC-CII intravenously on day 7 and 14 . Maximal effects of either treatment regimen would be expected shortly after the last dose . Serum collected on day 14 , after the last dose of either treatment regimen , was assayed for its ability to inhibit T cell activation , a standard assay for FK506 toxicity . We found that serum from mice treated with 2 mg/kg FK506 , but not mice treated with FKDC , potently inhibits T cell activation ( Figure 4E ) . This is consistent with higher serum levels of FK506 and systemic immunosuppression in mice treated with conventional FK506 treatment compared to DC based FK506 delivery . We also screened mice in the above treatment groups on day 14 for serum fasting glucose , amylase , lipase , BUN , AST and ALT . Amylase was increased in the systemic FK506 treatment group but not in the FKDC treatment group ( Figure 4F ) . This result is consistent with prior studies revealing toxicity to the exocrine pancreas in rats ( Doi et al . , 1992; Ito et al . , 1994 ) . There were no significant differences in the glucose , BUN , lipase , AST and ALT between any treatment groups ( data not shown ) . We conclude that in vivo transfer of antigen pulsed FKDC modulate immune outcomes in an antigen specific manner and protect from systemic immunosuppression and off target organ toxicity .
We demonstrate that DCs treated with a brief pulse of FK506 sequester and then release sufficient drug to inhibit T cells for an extended period of days . The capacity of DCs to accumulate FK506 likely relates to their high expression of binding proteins such as FKBP12 , which have high affinity for FK506 ( Siekierka et al . , 1989a ) . This suggestion is supported by the observation that T cells , which also express high levels of FKBP12 , have the ability to concentrate radiolabeled FK506 500–1000 fold ( Dumont et al . , 1994 ) . These observations also suggest the possibility that the use of such cellular drug sinks could be extended to other cell types with abundant FKBP expression , or even extrapolated to other agents with analogous intracellular binding proteins . The actions of FKDC are potentially clinically relevant , as they prevent autoimmune arthritis in mice . The lack of efficacy of FKDC loaded with irrelevant antigen indicates that the amount of drug delivered by FKDC is not by itself sufficient to elicit a systemic effect . Upon encounter with antigen-specific FKDC , T cells initiate a program of activation but then undergo apoptosis , explaining the antigen specific effect of FKDC transferred in vivo . Systemic treatment of CIA with FK506 requires a minimum of 2 mg/kg or approximately 40 μg subcutaneously to elicit any therapeutic effect ( Takagishi et al . , 1989 ) . In contrast , the amount of FK506 measured in culture supernatants ( Figure 2B ) predicts the dose delivered by 0 . 5 × 106 FKDC is a maximum of 5 ng; this means 8000-fold less drug is required to elicit a therapeutic effect when delivered by a DC . Since DC can be pulsed with antigen of interest prior to drug loading , DC based drug delivery has the potential to limit adverse events by substantially limiting required doses by precisely targeting antigen specific T cells . Antigen-loaded DC have been widely used with excellent safety profiles ( Sabado and Bhardwaj , 2010 ) , and this platform may be of interest to explore in clinical studies .
Human CD14+ cells cultured with GM-CSF and IL4 for 6 days were matured with TNF-α and PGE-2 . DC were cultured with 0 . 5 μM FK506 ( Astellas Pharmaceuticals , Killorglin , Ireland ) or media control at 37°C for various intervals and washed . DC or FKDC were cultured in triplicate wells with 2 × 105 allogeneic CD4 T cells for 5 days and loaded with 4 Ci/ml H-thymidine ( Migita et al . , 1999 ) for 18 hr . T cells were assayed for beta emission and c . p . m . were measured . Uninfected or influenza infected 3T3 cells were exposed to ultraviolet irradiation to induce apoptosis and subsequently cultured with immature DC and maturation stimulus as above . DC were treated with FK506 or media for various durations . DC were washed extensively and co-cultured with syngeneic CD4 or CD8 T cells for 40 hr ( 1 DC: 30 T cells ) in a IFNγ antibody coated 96-well microtiter plate . mRNA was isolated from T cells cultured with DC for 6 hr with an RNeasy kit ( Qiagen , Valencia , CA ) , cDNA was synthesized with Quantitect reverse transcription kit ( Qiagen ) , including genomic DNA removal . Primer pair sequences: Human IFNγ Forward: TCAGCTCTGCATCGTTTTGGGTTC , Reverse: TCCGCTACATCTGAATGACCTGCAT; Human IL17A Forward: CGGACTGTGATGGTCAACCTGA , Reverse: GCACTTTGCCTCCCAGATCACA; IL4 Forward: CCGTAACAGACATCTTTGCTGCC , Reverse: GAGTGTCCTTCTCATGGTGGCT; IL2 Forward: AGAACTCAAACCTCTGGAGGAAG , Reverse: GCTGTCTCATCAGCATATTCACAC; HRP14 Forward: CGGAGCTGACCAGACTTTTC , Reverse: GGTTCGACCGTCATACTTCTTC . Specificity ( melting-curve analysis ) and priming efficiency was confirmed . Stratagene Mx3000P system and PerfeCTa SYBR Green SuperMix ( Quanta Biosciences , Gathersburg , MA ) were used for real-time PCR . Magnetic dynabeads ( Dynal , Life Technologies , Carlsbad , CA ) , coated with 100 nM FK-1 ( Abcam , Cambridge , MA ) or anti-KLH antibody , were incubated with supernatants for 2 hr and depleted on magnetic columns . DC were pretreated with actinomycin ( 5 μM; Sigma , Milwaukee , WI ) , emetine ( 1 μM; Sigma ) , or brefeldin A ( 10 μg/ml; Sigma ) for 1 hr prior to FK506 treatment . After 2 hr of treatment with FK506 or media control , FKDC or DC were washed and added to T cells stimulated with anti-CD3/CD28 stimulator beads ( Invitrogen ) ( 25 μl per 1 × 106 cells ) at 37°C for 4 hr . mRNA was isolated and qRT-PCR performed . Fold induction of cytokine mRNA was calculated as 2^- ( delta CT of T cells stimulated with anti-CD3/28 and mixed with DC − delta CT of unstimulated T cells mixed with DC ) . FKDC were washed extensively . An aliquot of the last wash was harvested and cells were cultured at 1 × 106 live cells/ml to generate supernatants 6 hr later . Cells were re-washed daily and 1 × 106 live cells/ml were replated in culture to generate supernatants 6 hr later . FK506 was measured using ELISA ( USCNlife , Houston , TX ) . Absorbance was measured in a microtiter plate reader ( Bio-Rad , Hercules , CA ) ( 450 nm ) and converted into units ( ng/ml ) by plotting against autoantibody titer of calibrators/standards ( detection range 0 . 156–100 ng/ml ) . Data were obtained from a previously reported ( Longman et al . , 2007 ) Affymetrix array of monocyte-derived mature DCs , and analyzed using Microarray Suite 5 . 0 ( Affymetrix , Santa Clara , CA ) . Monocyte derived DC were treated with 0 . 5 µM Rapamycin overnight followed by 0 . 5 µM FK506 or media control for 2 hr and washed extensively . Treated or untreated control DC were cocultured with CD4+ T cells and CD3/28 beads for 4 hr . PCR was performed as above . CD4+ T cells were stained with 2 . 5 μM CFSE and cultured with allogeneic DC or FKDC . On the fourth day , cells were harvested , washed , and stained with Annexin V or anti-bcl-xl antibody ( Southern Biotech , Birmingham , AL ) . 8- to 10-week-old , male DBA1/J mice ( The Jackson Laboratory , Bar Harbor , ME ) were immunized with 100 μg bovine Type II Collagen ( Chondrex , Redmond , WA ) emulsified in CFA . Bone marrow derived DC ( Inaba et al . , 2001 ) pulsed with 50 μg/ml bovine type II collagen or bovine type I collagen ( Chondrex ) were subsequently treated with media or 0 . 5 μM FK506 for 2 hr and washed . 0 . 5 × 106 cells were infused intravenously 7 and 14 days post immunization . Arthritis severity score ( Brand et al . , 2007 ) ( four paws/mouse ) was rated by a blinded assessor three times weekly ( n = 8–12 mice/group ) . On day 87 of one of three experiments , limbs were fixed in 10% buffered formalin , paraffin embedded . Radiographs ( Faxitron MX-20 cabinet x-ray system ) and hematoxylin and eosin stained sections were scored by a board-certified blinded veterinary pathologist as previously described ( Clark et al . , 1979; Bendele et al . , 1999 ) . DBA/1J mice were treated with either 500 µL PBS control or 2 mg/kg FK506 ( Astellas ) subcutaneously for 14 days , or 0 . 5 × 106 FKDC-CII ( prepared as above ) intravenously on days 7 and 14 . 4 hr after the last treatment , serum from individual mice was added to CD4+ splenocytes stimulated with CD3/28 dynabeads or media control for 4 hr at 37°C . mRNA was isolated and qRT-PCR performed . Fold induction of IL2 mRNA was calculated as 2^- ( delta CT of T cells stimulated with anti-CD3/28 − delta CT of unstimulated T cells ) . Serum was also sent to the Memorial Sloan-Kettering Cancer Center of Comparative Medicine and Pathology for serum chemistries including BUN , amylase , lipase , glucose , AST and ALT . Two-tailed , unpaired T tests were used to determine the level of significance of differences between T cells stimulated by DC or FKDC in the allo-MLR , Elispot and PCR assays . Each CIA group contained 8–12 mice . Results from pilot experiments indicated that seven mice per group were required to obtain more than 80% power to detect a 50% reduction in arthritis severity score at 0 . 05 level . A linear mixed model was used to model arthritis severity score of mice , then a likelihood ratio test was conducted to compare if arthritis severity score of mice vaccinated with various DC groups differed . The generalized estimating equation modeled incidence of severe arthritis in mice vaccinated with the various DC groups . Spearman's correlation coefficient was used to determine the correlation between clinical mean arthritis severity scores and radiographic scores and histopathology severity scores . The Wilcoxon rank test was used to compare radiographic and histopathology scores of the mice treated with FKDC/CII or DC/CII . | Although our health depends on our immune system's ability to recognize and attack foreign material , this same response can cause the body to reject an organ transplant or even to spontaneously attack itself ( this is called autoimmune disease ) . To help prevent rejection , patients who receive donated organs are given immunosuppressant drugs , with a compound called FK506 , or Tacrolimus , the most commonly used . However , FK506 can have a number of serious side effects , including high blood pressure , kidney damage and diabetes . The job of starting an immune response falls in large part to a type of white blood cell called the dendritic cell , which patrols the body in search of cells in trouble—such as those infected with viruses . Dendritic cells are efficient at engulfing dying cells , which they break down and display fragments of on their cell surface . These fragments—which are known as antigens—are presented directly to T cells , which trigger a cascade of additional immune responses leading ultimately to the destruction of infected cells . In some cases of autoimmune disease , however , T cells begin to mistake the body's own cells for infected cells and to launch attacks against healthy tissue . Evidence suggests that immunosuppressive drugs such as FK506 can help to tone down these inappropriate immune responses . However , the use of FK506 to treat autoimmune disease has been limited by its side effects . Now , Orange et al . have shown that dendritic cells can be exploited to deliver drugs such as FK506 in a targeted and controlled manner . When the researchers loaded dendritic cells with FK506 , they found that the cells sequestered the drug and then released it slowly in quantities that were sufficient to inhibit T-cell responses for at least 72 hr . Using a mouse model of rheumatoid arthritis—an autoimmune disease characterized by inflammation and destruction of joint tissue—Orange and co-workers demonstrated that their novel drug delivery system could be therapeutically useful . They loaded dendritic cells displaying the antigen that triggers the mouse immune system to attack joint tissue , with FK506 , and used the resulting cells to treat arthritic mice . Mice that received these cells showed less severe arthritis than control animals treated with dendritic cells that had not been loaded with FK506 . Moreover , the total dose of FK506 that the mice were exposed to was very low , with the result that they showed no evidence of the side effects typically seen with this drug . This proof-of-concept study suggests that dendritic cells could be used for the gradual and controlled delivery of drugs to specific target cells within the immune system . By precisely targeting relevant immune cells , it should be possible to use much lower drug doses , and thereby reduce side effects . Follow-up studies are now required to determine whether dendritic cells can be used as vehicles for the delivery of other drugs to treat a range of diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"immunology",
"and",
"inflammation"
] | 2013 | Dendritic cells loaded with FK506 kill T cells in an antigen-specific manner and prevent autoimmunity in vivo |
Here , using mouse squamous cell carcinoma cells , we report a completely new function for the autophagy protein Ambra1 as the first described ‘spatial rheostat’ controlling the Src/FAK pathway . Ambra1 regulates the targeting of active phospho-Src away from focal adhesions into autophagic structures that cancer cells use to survive adhesion stress . Ambra1 binds to both FAK and Src in cancer cells . When FAK is present , Ambra1 is recruited to focal adhesions , promoting FAK-regulated cancer cell direction-sensing and invasion . However , when Ambra1 cannot bind to FAK , abnormally high levels of phospho-Src and phospho-FAK accumulate at focal adhesions , positively regulating adhesion and invasive migration . Spatial control of active Src requires the trafficking proteins Dynactin one and IFITM3 , which we identified as Ambra1 binding partners by interaction proteomics . We conclude that Ambra1 is a core component of an intracellular trafficking network linked to tight spatial control of active Src and FAK levels , and so crucially regulates their cancer-associated biological outputs .
Ambra1 ( activating molecule in Beclin1-regulated autophagy ) is a WD40 domain-containing protein that is involved in the development of the central nervous system , adult neurogenesis and vertebrate embryogenesis ( Benato et al . , 2013; Fimia et al . , 2007; Yazdankhah et al . , 2014 ) . It is an autophagy regulator , binding to Beclin1 and playing a role in the initiation of autophagy , which is required for neurogenesis ( Fimia et al . , 2007 ) . When autophagy is not initiated , the Ambra1-Beclin1-Vps34 complex is bound to the dynein motor complex . Upon induction of autophagy , Ambra1 gets phosphorylated by the kinase ULK1 , resulting in its release from the dynein complex ( Di Bartolomeo et al . , 2010; Strappazzon et al . , 2011 ) . Ambra1 function is negatively regulated by mTOR phosphorylation , which suppresses its binding to the E3-ligase TRAF6 and ULK1 ubiquitylation , thus controlling ULK1 stability and function ( Nazio et al . , 2013 ) . During apoptosis , Ambra1 expression is regulated by caspase-mediated cleavage as well as degradation by calpains ( Pagliarini et al . , 2012 ) , and a recent study reported the ubiquitylation and subsequent degradation of Ambra1 by RNF2 ( Xia et al . , 2014 ) . Further , Ambra1 has been reported to support the binding of c-Myc to the phosphatase PP2A , leading ultimately to c-Myc degradation and reduced cell proliferation and tumourigenesis ( Cianfanelli et al . , 2015 ) , whilst a different study implied that Ambra1 overexpression in cholangiocarcinoma has been correlated with invasion and poor survival ( Nitta et al . , 2014 ) . In the present study , we found that Ambra1 is a focal adhesion kinase ( FAK ) - and Src-binding partner . We therefore investigated the role of Ambra1 in tumour-associated phenotypes regulated by the Src/FAK pathway in squamous cell carcinoma ( SCC ) cells derived from the DMBA/TPA model of carcinogenesis ( driven by mutated oncogenic H-Ras ) ( Quintanilla et al . , 1986 ) . We have previously shown that FAK-dependent phenotypes include cancer cell polarisation and directional migration , depending primarily on FAK’s protein scaffolding activities , including binding to actin regulators like Eps8 , Arp3 and RACK1 ( Schoenherr et al . , 2014; Serrels et al . , 2010 , 2007 ) . In particular , genetic depletion of FAK , or detachment of FAK-expressing cells , cause active Src to be trafficked from focal adhesions at the cell periphery to intracellular puncta containing autophagy proteins . We termed this ‘adhesion-stress-induced autophagy’ , and it is FAK and adhesion dependent rather than nutrient dependent , in contrast to the classical , starvation-induced autophagy . It provides a novel mechanism for coping with high levels of active Src , and other oncogenic kinases like Ret , which are not spatially controlled in the usual way by their binding to FAK ( Sandilands et al . , 2012a , 2012b ) . We found that Ambra1 is critically involved in Src/FAK-dependent cancer cell polarisation and chemotactic invasion . In FAK-depleted SCC cells , Ambra1 controls the targeting of active Src to intracellular autophagic puncta . This is mediated by the novel Ambra1-binding proteins Dynactin 1 ( Dctn1; also known as p150Glued ) and interferon-induced transmembrane protein 3 ( IFITM3 ) , which together with Ambra1 , are critical to trafficking processes that control spatial Src activity and Src-dependent phenotypes . A FAK mutant that is not able to bind Ambra1 promotes increased cell adhesion and invasion by retaining more phospho-FAK ( pFAK ) and phospho-Src ( pSrc ) at focal adhesions – showing that Ambra1 can both positively and negatively regulate the amount of active Src at cellular adhesion sites . Overall , these data imply a novel role for the autophagic protein Ambra1 , and its key interacting partners Dynactin 1 and IFITM3 , at the heart of an intracellular trafficking network that , in turn , acts as a ‘spatial rheostat’ for phospho-Src and Src/FAK-mediated cancer processes .
We discovered Ambra1 in a phage display screen for novel FAK binding partners . As Ambra1 regulates autophagy , and we had already published that active Src was trafficked away from adhesions via autophagosomes in the absence of FAK ( Sandilands et al . , 2012a ) , we set out to test the involvement of Ambra1 in the autophagic trafficking of active Src . First , we confirmed Ambra1 as a novel FAK binding partner in SCC cells by reciprocal co-immunoprecipitations ( co-IPs; Figure 1A , B ) . 10 . 7554/eLife . 23172 . 003Figure 1 . Ambra1 interacts with FAK at focal adhesions . FAK or Ambra1 were immunoprecipitated from FAK-WT and FAK -/- cell lysates using ( A ) anti-FAK 4 . 47 agarose or ( B ) anti-Ambra1 , followed by western blot analysis with anti-FAK and anti-Ambra1 . ( C ) FAK was immunoprecipitated from Ambra1 +/+ and Ambra1 -/- MEF cell lysates using anti-FAK 4 . 47 agarose , followed by western blot analysis with anti-FAK and anti-Ambra1 . Anti-β-actin was used as a loading control . Relative ratios of Ambra1 , FAK/Ambra1 and Ambra1/FAK interactions were calculated by densitometry . ( D ) FAK-WT and FAK -/- cells were seeded onto glass coverslips , fixed and stained using anti-Ambra1 and anti-Paxillin . Scale bars , 20 μm . ( E , F ) Focal adhesions were isolated from FAK-WT and FAK -/- cells using hydrodynamic force . ( E ) Focal adhesions ( solid arrows ) were stained with anti-FAK and anti-Ambra1 ( left panels ) and with anti-Ambra1 and anti-Paxillin ( right panels ) . ( F ) Focal adhesions ( solid arrows ) were stained with anti-Ambra1 and anti-Rack1 in SCC FAK-WT ( left panels ) and SCC FAK -/- cells ( right panels ) . Scale bars , 20 μm . Colocalisation ( Costes r value from five cells ) was analysed using the ImageJ plugin JaCoP ( Bolte and Cordelières , 2006 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 00310 . 7554/eLife . 23172 . 004Figure 1—source data 1 . COSTES r values for immunofluorescence images . COSTES mean and s . d . values for Figure 1D–F are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 00410 . 7554/eLife . 23172 . 005Figure 1—figure supplement 1 . Ambra1 +/+ and -/- mouse embryonic fibroblasts ( MEFs ) . ( A ) Representative images of Ambra1 +/+ and Ambra1 -/- MEFs . ( B ) PCR of Ambra1 +/+ and Ambra1 -/- MEFs . B2M served as a control for equal input . ( C ) SCC FAK-WT and FAK -/- cells were grown on glass coverslips for 24 hr , fixed and stained with anti-Ambra1 , anti-CoxIV and DAPI . ( D , E ) Focal adhesions were isolated from FAK-WT and FAK -/- cells using hydrodynamic force . Focal adhesions ( solid arrows ) were stained with anti-Ambra1 and anti-CoxIV ( D ) or anti-FAK and anti-CoxIV ( E ) in SCC FAK-WT ( left panels ) and SCC FAK -/- cells ( right panels ) . Scale bars , 20 μm . Colocalisation ( Costes r value from five cells ) was analysed using the ImageJ plugin JaCoP . ( F , G ) Total Internal Reflection Fluorescence ( TIRF ) microscopy of SCC FAK-WT and -/- cells stained with anti-Ambra1 and anti-FAK ( F ) or anti-Ambra1 and anti-pSrc Y416 . ( G ) Colocalisation ( COSTES r value of five cells ) was analysed using the ImageJ plugin JaCoP . Scale bars , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 00510 . 7554/eLife . 23172 . 006Figure 1—figure supplement 2 . Knockdown of Ambra1 suppresses FAK phenotypes . ( A ) Polarity assay: FAK-WT and FAK -/- cells were transiently transfected with either a pool or two independent Ambra1 siRNAs . A confluent monolayer of cells plated on fibronectin was wounded using a pipette tip , fixed 1 . 5 hr later and stained with anti-GM130 ( Golgi ) , TRITC-phalloidin and DAPI . The orientation of the Golgi towards to wound edge was used to score polarisation . Scale bars , 20 μm . ( B ) Quantification of the polarity assay in SCC FAK-WT and -/- cells . n = 3 . Error bars , s . d . p<0 . 01 . ( C ) Polarity assay in Ambra1 +/+ and -/- MEFs . n = 3 . Error bars , s . d . p<0 . 01 . ( D , E ) Invasion assay: Cells were seeded on growth factor-reduced Matrigel in serum-free conditions . After 72 hr invasion towards a serum gradient , cells were visualised by staining with calcein . ( D ) Quantification of the invasion assay . n = 5 . Error bars , s . e . m . p<0 . 01 ( * ) and p<0 . 05 ( # ) . ( E ) Representative images of the invasion assay . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 006 The IPs were validated using FAK-deficient ( -/- ) SCC cells and Ambra1 -/- mouse embryonic fibroblasts ( MEFs ) ( Figure 1C; Figure 1—figure supplement 1A , B ) . In order to confirm the co-localisation between FAK and Ambra1 , we performed immunofluorescence ( IF ) . In both SCC FAK-WT and FAK -/- cells , Ambra1 displayed diffuse cytoplasmic localisation with some at focal adhesions ( Figure 1D ) . Therefore , we isolated focal adhesions from FAK-WT and FAK -/- cells by hydrodynamic force and found that Ambra1 co-localised with FAK and Paxillin at isolated focal adhesions ( Figure 1E; Figure 1—figure supplement 1C–E ) . Additionally , Ambra1 localised to hydrodynamic force-isolated nascent focal adhesions , marked by the presence of RACK1 ( de Hoog et al . , 2004; Serrels et al . , 2010 , 2007 ) , in both FAK-WT and FAK -/- cells ( Figure 1F ) . However , the non-autophagy , non-adhesion cellular protein CoxIV was not found at isolated focal adhesions , implying that the IF staining was not of general cellular debris ( Figure 1—figure supplement 1D , E ) . In addition , we found that Ambra1 was present at focal adhesions at the cell-surface interface by Total Internal Reflection Fluorescence ( TIRF ) microscopy; there was partial overlap with pSrc Y416 and FAK , suggesting that , as with other autophagy proteins we have reported previously ( Sandilands et al . , 2012a; Schoenherr et al . , 2014 ) , there are pools of Ambra1 discretely present at sub-locations within focal adhesions ( Figure 1—figure supplement 1F , G; Figure 5—figure supplement 2C , D ) . We previously showed that active Src is localised to intracellular puncta containing autophagy proteins in FAK -/- SCC cells , allowing FAK-deficient cancer cells to cope with high levels of ‘untethered’ active Src ( and other FAK-interacting tyrosine kinases , e . g . Ret ) ( Sandilands et al . , 2012a , 2012b ) . Other FAK- and Src-interacting proteins , like the actin regulator Eps8 , are also involved in this spatial localisation of active Src ( Schoenherr et al . , 2014 ) . Since Ambra1 is a recognised autophagy regulator , we addressed whether it was required for the trafficking of Src to autophagosomes in FAK-deficient cells . Ambra1 and Src , including active phospho-Src ( pSrc; pY416 ) , formed a biochemical complex both in FAK-expressing and -deficient SCC cells as shown by co-immunoprecipitation ( Figure 2A , B ) . Using immunofluorescence , we found that Ambra1 localised to intracellular puncta that contain autophagy-regulating proteins like LC3B ( Figure 2C , D ) . We noted that Ambra1 also appeared to localise to nuclei , and although we confirmed this by probing nuclear proteins isolated after cell fractionation and sucrose gradient purification of nuclei , we do not know its significance ( Figure 2—figure supplement 1A , B ) . In FAK-deficient SCC cells , Ambra1 co-localised with pSrc , indicating that active Src and Ambra1 localised to the same intracellular puncta , described previously to contain autophagy proteins ( Figure 2C ) . The localisation of Ambra1 in these autophagosomes was consistent with co-immunoprecipitation with LC3B , providing evidence that they were present in the same biochemical complex ( Figure 2E ) . We note that there was a reduced steady state level of the lipidated form of LC3B ( LC3B II ) in the SCC FAK -/- cells ( the significance of this is unknown; [Sandilands et al . , 2012a] ) . We are not able to discern which isoform of LC3B is binding to Ambra1 or whether there is any difference between them . 10 . 7554/eLife . 23172 . 007Figure 2 . Ambra1 interacts with Src and mediates trafficking of active Src to autophagosomes . ( A , B ) Src ( A ) or pSrc Y416 ( B ) were immunoprecipitated from FAK-WT and FAK -/- cell lysates using anti-Src agarose or anti-pSrc Y416 antibody , followed by western blot analysis with anti-Ambra1 , anti-pSrc Y416 and anti-Src . Relative ratios of Ambra1/Src and Ambra1/pSrc interactions were calculated by densitometry . ( C ) FAK-WT and FAK -/- cells were seeded onto glass coverslips , fixed and stained using anti-pSrc Y416 , anti-Ambra1 and DAPI . Scale bars , 20 μm . ( D ) SCC FAK-WT and FAK -/- cells were grown on glass coverslips , fixed and stained with anti-Ambra1 , anti-LC3B and DAPI . Scale bars , 20 μm . ( E ) LC3B was immunoprecipitated from SCC FAK-WT and FAK -/- cell lysates using anti-LC3B , followed by western blot analysis with anti-Ambra1 and anti-LC3B . Relative ratios of LC3B II/LC3B I as well as the Ambra1/LC3B and Ambra1/LC3B II interactions were calculated by densitometry . ( F–J ) SCC FAK-WT and FAK -/- cells were transiently transfected with either a pool ( F–H ) or two individual siAmbra1 siRNAs ( F , I , J ) . The cells were grown on glass coverslips , fixed and stained with anti-pSrc Y416 , anti-Paxillin and DAPI . ( G , I ) Representative immunofluorescence images . Scale bars , 20 μm . ( H , J ) Quantification of internalised active Src . n = 3 . Error bars , s . d . p<0 . 001 . Colocalisation ( Costes r value from five cells ) was analysed using the ImageJ plugin JaCoP . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 00710 . 7554/eLife . 23172 . 008Figure 2—source data 1 . COSTES r values for immunofluorescence images and percentage of cells with internalised pSrc . COSTES mean and s . d . values for Figures 2C and D are shown . Mean percentage and s . d . values of cells with internalised pSrc upon transient Ambra1 knockdown by siRNA in SCC FAK-WT and -/- cells are shown ( Figures 2H , J ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 00810 . 7554/eLife . 23172 . 009Figure 2—figure supplement 1 . Ambra1 interacts with Src and mediates pSrc trafficking . ( A ) Nuclei from SCC FAK-WT and FAK -/- cells were isolated using sucrose gradient centrifugation . Lysates were immunoblotted for Ambra1 , GAPDH ( cytosolic marker ) and Lamin A/C ( nuclear marker ) . ( B ) Relative ratio of Ambra1/Lamin A/C was calculated by densitometry . WCL , whole cell lysate . Error bars , s . d . ( C ) Colocalisation ( Costes r value from five cells ) of pSrc Y416/Paxillin upon Ambra1 knockdown by siRNAs was analysed using the ImageJ plugin JaCoP . Error bars , s . d . p<0 . 01 ( * ) and p<0 . 05 ( # ) . ( D ) FAK-WT and FAK -/- cells were transiently transfected with either a pool or two independent Ambra1 siRNAs . Cell lysates were subjected to western blot analysis using anti-Ambra1 , anti-pSrc Y416 and anti-Src . Anti-GAPDH was used as a loading control . ( E ) The relative ratios of pSrc/GAPDH were calculated using densitometry . n = 3 . Error bars , s . d . p<0 . 01 ( * ) and p<0 . 05 ( # ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 009 To investigate whether Ambra1 controlled active Src localisation , we suppressed expression of Ambra1 with both a pool of four and two individual siRNAs ( Figure 2F ) . Knockdown of endogenous Ambra1 either by the pool of four siRNAs ( Figure 2G , H ) or two individual siRNAs ( Figure 2I , J ) resulted in the redistribution of active Src from intracellular puncta back to focal adhesions in FAK -/- cells . However , we were unable to determine whether this relocation would also occur in detached cells , which do not have focal adhesions . Taken together , these data imply that Ambra1 is required for the trafficking of active Src from focal adhesions to intracellular autophagic puncta . We next performed Ambra1 , pSrc Y416 , and FAK ( and IgG control ) IPs using lysates from FAK-WT and FAK -/- cells ( in triplicate ) and determined specifically interacting proteins by quantitative label-free mass spectrometry . Amongst the binding partners of all three ‘baits’ were proteins that are involved in regulating intracellular trafficking , raising the possibility that Ambra1 is part of a Src/FAK ‘trafficking network’ ( Figure 3A; Supplementary file 1–3 ) . 10 . 7554/eLife . 23172 . 010Figure 3 . IFITM3 is in the centre of an Ambra1 , FAK and pSrc trafficking network . ( A ) Network analysis of Ambra1- , FAK- and pSrc Y416-interacting proteins that are involved in trafficking processes . Solid lines indicate protein-protein interactions identified in the mass spectrometry datasets used for the interaction map . Dotted lines indicate Ambra1–FAK/pSrc interactions , which have been previously identified and verified by immunoprecipitation . ( B ) Ambra1 was immunoprecipitated from FAK-WT and FAK -/- cell lysates using anti-Ambra1 antibody , followed by western blot analysis with anti-Dctn1 and anti-Ambra1 . ( C–E ) Ambra1 ( C ) , FAK ( D ) or pSrc Y416 ( E ) were immunoprecipitated from FAK-WT and -/- cell lysates , followed by western blot analysis with anti-IFITM3 , anti-Ambra1 , anti-FAK and anti-pSrc Y416 . Anti-GAPDH served as a loading control . Relative ratios of Dctn1/Ambra1 , IFITM3/Ambra1 , IFITM3 and IFITM3/pSrc interactions were calculated by densitometry . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 01010 . 7554/eLife . 23172 . 011Figure 3—figure supplement 1 . IFITM3 influences the Src-FAK complex . ( A ) Focal adhesions from SCC FAK-WT and FAK -/- cells were crosslinked and isolated for western blot analysis with the indicated antibodies . Paxillin served as a marker for focal adhesions , Lamin A/C was used as a nuclear marker and GAPDH represented a cytosolic marker . WCL , whole cell lysate; FA , focal adhesions . SCC FAK-WT and -/- cells were transiently transfected with siIFITM3 and pSrc Y416 ( B ) and FAK ( D ) were immunoprecipitated from FAK-WT and FAK -/- cell lysates using anti-pSrc Y416 and anti-FAK 4 . 47 agarose , followed by western blot analysis with anti-Ambra1 , anti-FAK , anti-IFITM3 , anti-pSrc Y416 and anti-Src . Anti-GAPDH served as a loading control . ( C , E ) Relative ratios of Ambra1/pSrc ( C ) as well as pSrc/FAK and Src/FAK ( E ) interactions were calculated by densitometry . n = 3 . Error bars , s . d . p<0 . 05 ( # ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 011 To test the hypothesis that Ambra1-dependent targeting of pSrc from focal adhesions to intracellular autophagic puncta involved such trafficking proteins , we selected two novel Ambra1 binding partners for further study . These were Dynactin 1 , whose interaction with Ambra1 appeared to be enriched in FAK-deficient cells from proteomics analyses ( Figure 3 ) and IFITM3 , which was positioned at the centre of the trafficking network as it bound to all three of Ambra1 , pSrc and FAK ( Figure 3 ) . Dctn1 ( Dynactin 1 ) is a component of the dynactin complex and is involved in dynein-mediated transport along microtubules , enhancing motor processivity ( Culver-Hanlon et al . , 2006; King and Schroer , 2000; Waterman-Storer et al . , 1995 ) , and mediating trafficking and maturation of EGFR and lysosomes ( Kedashiro et al . , 2015b; Li et al . , 2014 ) . We confirmed the interaction between Ambra1 and Dynactin 1 by co-IP , thereby validating the interaction identified by mass spectrometry ( Figure 3B ) . IFITM3 , which forms a central node linking the Ambra1 , pSrc and FAK trafficking networks ( Figure 3A ) , is known to block cellular entry of viruses ( Amini-Bavil-Olyaee et al . , 2013; Brass et al . , 2009; Feeley et al . , 2011; Huang et al . , 2011; Lu et al . , 2011; Weidner et al . , 2010 ) , to traffic with Rab7 and LAMP1 , and to interact with v-ATPase on endosomes to enable clathrin localisation ( Feeley et al . , 2011; Wee et al . , 2012 ) . IFITM3 is highly expressed in many cancers ( Andreu et al . , 2006; Hu et al . , 2014; Li et al . , 2011; Yang et al . , 2013 ) and has been variously implicated in tumour cell proliferation , migration and invasion ( Hu et al . , 2014; Li et al . , 2011; Yang et al . , 2013; Zhao et al . , 2013 ) . We confirmed the interaction between Ambra1 and IFITM3 ( Figure 3C ) , FAK and IFITM3 ( Figure 3D ) , and pSrc and IFITM3 ( Figure 3E ) by co-IPs . The interactions between IFITM3 and both Ambra1 and pSrc appeared to be enriched in FAK-expressing cells due to increased IFITM3 expression levels , caused by FAK-dependent IFITM3 transcription ( not shown; ( Figure 3A ) , and we believe that IFITM3 is present at focal adhesions , shown by immunoblotting of isolated focal adhesion preparations ( Figure 3—figure supplement 1A ) . To determine whether IFITM3 was required for complex formation between active Src and Ambra1 in FAK-WT and -/- cells , we knocked down IFITM3 using a pool of siRNAs and analysed the interaction of Ambra1 and pSrc by co-immunoprecipitations ( Figure 3—figure supplement 1B , C ) . While knockdown of IFITM3 did not affect the Ambra1-pSrc interaction ( Figure 3—figure supplement 1B , C ) , IFITM3 depletion reduced the interaction between FAK and active pSrc , indicating that IFITM3 is involved in the optimal association of Src to FAK ( Figure 3—figure supplement 1D , E ) , likely through regulating their precise localisation . Thus , interaction network analysis coupled with co-immunoprecipitations suggest that IFITM3 is a central component of a trafficking network linking Ambra1 , FAK and active Src ( Figure 3A ) . Having verified that both Dynactin 1 and IFITM3 bind to Ambra1 , we next addressed whether they were involved in the Ambra1-dependent intracellular trafficking of active pSrc . We transiently knocked down endogenous Dynactin 1 ( Figure 4A ) or IFITM3 ( Figure 4D ) by siRNA , and analysed the localisation of active pSrc in FAK-deficient SCC cells . In both cases , suppression of protein expression resulted in significant redistribution of active pSrc from intracellular autophagic puncta to focal adhesions ( Figure 4B , C; Figure 4—figure supplement 1A and Figure 4E , F; Figure 4—figure supplement 1B respectively ) , while in FAK-expressing cells , knockdown of Dynactin 1 or IFITM3 had no visible effect on their distribution . These data imply that Ambra1 is part of a functional trafficking network that precisely regulates the spatial distribution of Src activity . Mechanistically , Ambra1 functions via interaction with proteins involved in intracellular trafficking , including Dynactin 1 and IFITM3; the latter lies at the centre of linked Ambra1 , pSrc , and FAK interactomes . 10 . 7554/eLife . 23172 . 012Figure 4 . Knockdown of Dynactin 1 and IFITM3 suppresses trafficking of active Src to autophagic puncta . ( A ) FAK-WT and FAK -/- cells were transiently transfected with a pool of Dynactin 1 ( Dctn1 ) siRNAs and lysed 48 hr post transfection . Dynactin 1 expression was determined by western blotting using anti-Dctn1 . Anti-GAPDH was used as a loading control . ( B ) SCC FAK-WT and FAK -/- cells transiently transfected with Dctn1 siRNA were grown on glass coverslips , fixed and stained with anti-pSrc Y416 , anti-Paxillin and DAPI . Scale bars , 20 μm . ( C ) Quantification of internalised active Src . n = 3 . Error bars , s . d . p<0 . 01 . ( D ) FAK-WT and FAK -/- cells were transiently transfected with a pool of IFITM3 siRNAs and lysed 48 hr post transfection . IFITM3 expression was determined by western blotting using anti-IFITM3 . Anti-GAPDH was used as a loading control . ( E ) SCC FAK-WT and FAK -/- cells transiently transfected with IFITM3 siRNA were grown on glass coverslips , fixed and stained with anti-pSrc Y416 , anti-Paxillin and DAPI . Scale bars , 20 μm . ( F ) Quantification of internalised active Src . n = 3 . Error bars , s . d . p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 01210 . 7554/eLife . 23172 . 013Figure 4—source data 1 . Percentage of cells with internalised pSrc . Mean percentage and s . d . values of cells with internalised pSrc upon transient Dctn1 ( Figure 4C ) or IFITM3 ( Figure 4F ) knockdown by siRNA in SCC FAK-WT and -/- cells are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 01310 . 7554/eLife . 23172 . 014Figure 4—figure supplement 1 . Knockdown of Dynactin 1 and IFITM3 suppress pSrc trafficking to autophagic puncta . SCC FAK-WT and FAK -/- cells transiently transfected with siDctn1 ( A ) or siIFITM3 ( B ) were grown on glass coverslips , fixed and stained with anti-pSrc Y416 , anti-Paxillin and DAPI . Colocalisation ( Costes r value from nine or five cells respectively ) of pSrc/Paxillin was analysed using the ImageJ plugin JaCoP . Error bars , s . d . p<0 . 01 ( * ) and p<0 . 05 ( # ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 014 Having established that Ambra1 and FAK interact with each other and co-localise at focal adhesions ( Figure 1 ) , and that they are co-determinants of the intracellular localisation of Src activity , we next addressed whether Ambra1 was required for Src/FAK-mediated cancer-related processes . We therefore knocked down Ambra1 , and found that there was significant loss of polarisation towards a monolayer wound in FAK-expressing SCC cells , as judged by the direction of the Golgi apparatus ( stained with GM130 ) . This was in contrast to FAK-deficient cells , in which knockdown of Ambra1 had no further effect on the already suppressed polarisation of cells towards a wound ( Figure 1—figure supplement 2A , B ) . The role of Ambra1 in polarisation was confirmed using Ambra1 +/+ and -/- MEFs ( Figure 1—figure supplement 2C ) . In addition , we found that Ambra1 was required for chemotactic invasion into growth factor-reduced Matrigel in FAK-expressing cells ( Figure 1—figure supplement 2D , E ) , whilst there was no effect in FAK-deficient cells , which , as we described previously , do not invade into Matrigel ( Serrels et al . , 2012 ) . These results describe a previously unknown role for Ambra1 in cancer cell polarisation and invasion that is Src/FAK dependent . As Ambra1 binds FAK and is found at isolated focal adhesions , we next addressed whether FAK and Ambra1 interact directly . The Ambra1 binding site in FAK was mapped by peptide array binding analysis as described previously ( Schoenherr et al . , 2014; Serrels et al . , 2007 ) , which resulted in identification of two amino acids in FAK that were required for optimal direct binding of Ambra1 , i . e . amino acids P875 and P881 . These prolines ( P ) were mutated to alanines ( A ) , and the resulting FAK P875A/P881A ( AA ) mutant caused reduced binding of Ambra1 to FAK within cells ( Figure 5A , B ) . The interaction between FAK and p130Cas , which also binds FAK at a similar proline-rich region but not the same combination of amino acids ( P715 , P718 , P878 and P881; ref . [Harte et al . , 1996] ) , was more modestly affected by the P875A/P881A mutation ( quantified in Figure 5C ) . 10 . 7554/eLife . 23172 . 015Figure 5 . Ambra1 binding impaired FAK increases cell adhesion and pSrc at focal adhesions . ( A , B ) FAK ( A ) or Ambra1 ( B ) were immunoprecipitated from FAK-WT , FAK P875A/P881A ( AA ) and FAK -/- cell lysates using anti-FAK 4 . 47 agarose or anti-Ambra1 , followed by western blot analysis with anti-FAK , anti-Ambra1 and anti-p130Cas . Anti-GAPDH was used as a loading control . ( C ) Relative ratios of Ambra1/FAK and p130Cas/FAK interactions were calculated by densitometry . ( D ) Adhesion assay: SCC FAK-WT , FAK P875A/P881A and FAK -/- cells were plated in serum-free conditions on fibronectin-coated plates . Samples were normalised to the 6 hr time point and relative adhesion was calculated by setting the FAK-WT values to 1 . n = 3 . Error bars , s . d . p≤0 . 01 . ( E , F ) SCC FAK-WT , FAK P875A/P881A and FAK -/- cells were grown on glass coverslips for 24 hr , fixed and stained with anti-pSrc Y416 , anti-Paxillin and DAPI . Scale bars , 20 μm . The relative intensity of pSrc staining at focal adhesions from five cells ( at least 10 focal adhesions/cell ) was measured using ImageJ . ( E ) Representative immunofluorescence images are shown . ( F ) Quantification of the relative pSrc intensity at focal adhesions . n = 5 . Error bars , s . e . m . p<0 . 01 ( * ) and p<0 . 05 ( # ) . ( G ) Focal adhesions from SCC FAK-WT , FAK P875A/P881A ( AA ) and FAK -/- cells were crosslinked and isolated for western blot analysis with the indicated antibodies . Paxillin and Talin served as markers for focal adhesions and Lamin A/C was used as a nuclear marker . Hsp90 and GAPDH represented cytosolic markers . WCL , whole cell lysate; FA , focal adhesions . The purity of the isolated focal adhesions was determined by the absence of nuclear proteins like Lamin A/C and cytosolic markers like Hsp90 and GAPDH . There was less active Src at focal adhesions in the FAK-deficient SCC cells due to Src’s internalisation to autophagic structures . Additionally , increased pPaxillin Y118 and Talin levels could be detected in the FAK Ambra1-binding mutant compared to SCC FAK-WT and FAK -/- cells . No changes in Ambra1 levels at focal adhesions could be detected . ( H ) The relative ratios of pFAK/FAK , pSrc/Src and Ambra1 at focal adhesions were calculated using densitometry . n = 3 . Error bars , s . d . p<0 . 01 ( * ) and p<0 . 05 ( # ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 01510 . 7554/eLife . 23172 . 016Figure 5—source data 1 . Relative mean values of adhesion , pSrc intensity at focal adhesions and relative ratios at focal adhesions . Mean and s . d . values of relative adhesion on fibronectin of SCC FAK-WT , P875A/P881A and -/- cells are shown ( Figure 5D ) . The relative mean intensity and s . e . m . of pSrc at focal adhesions are shown ( Figure 5F ) . Relative ratios ( mean and s . d . ) of pFAK/FAK , pSrc/Src and Ambra1 of isolated focal adhesions are shown ( Figure 5H ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 01610 . 7554/eLife . 23172 . 017Figure 5—figure supplement 1 . FAK Ambra1-binding mutant promotes adhesion and invasion . ( A ) Adhesion assay: SCC FAK-WT , FAK P875A/P881A and FAK -/- cells were plated in serum-free conditions on uncoated plates ( plastic ) . Samples were normalised to the 6 hr time point and relative adhesion was calculated by setting the FAK-WT values to 1 . n = 3 . Errors bars , s . d . p≤0 . 01 ( * ) and p<0 . 05 ( # ) . ( B ) Quantification of internalised active Src . n = 3 . Error bars , s . d . p<0 . 001 . ( C ) The average number of focal adhesions per cell was counted . n = 3 . Error bars , s . d . ( D ) The average size of focal adhesions was measured using ImageJ . n = 3 . Error bars , s . d . ( E ) The relative intensity of pFAK Y397 staining at focal adhesions from five cells ( at least 10 focal adhesions/cell ) was measured using ImageJ . n = 3 . Error bars , s . e . m . p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 01710 . 7554/eLife . 23172 . 018Figure 5—figure supplement 2 . FAK Ambra1-binding mutant promotes adhesion and invasion . ( A , B ) SCC FAK-WT , FAK P875A/P881A ( AA ) and FAK -/- cells were grown on glass coverslips , fixed and stained with anti-FAK , anti-Paxillin and DAPI ( A ) and anti-Ambra1 , anti-Paxillin and DAPI ( B ) , respectively . Scale bars , 20 μm . There were no visible differences between total FAK or Ambra1 intensity staining . ( C , D ) Total Internal Reflection Fluorescence ( TIRF ) microscopy of SCC FAK P875A/P881A cells stained with anti-Ambra1 and anti-FAK ( C ) or anti-Ambra1 and anti-pSrc Y416 ( D ) . Colocalisation ( COSTES r value of five cells ) was analysed using the ImageJ plugin JaCoP . Scale bars , 10 μm . ( E ) FAK was immunoprecipitated from FAK-WT , FAK P875A/P881A ( AA ) and FAK -/- cell lysates using anti-FAK 4 . 47 agarose , followed by western blot analysis with anti-FAK and anti-Src . The relative ratio of Src/FAK interaction was calculated by densitometry . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 018 When culturing the cells expressing FAK that was impaired in Ambra1 binding , we noticed that these FAK P875A/P881A cells seemed to display greater adherence than the FAK-WT cells . Therefore , we performed adhesion assays on fibronectin-coated dishes ( Figure 5D ) or on plastic ( Figure 5—figure supplement 1A ) . We found that after 20 and 60 min , FAK P875A/P881A and FAK -/- cells attached to a higher degree than FAK-WT cells . Furthermore , in already adhered FAK P875A/P881A cells ( Figure 5E , middle panels ) , we found that there was more intense pSrc staining at focal adhesions when compared to FAK-WT cells ( Figure 5E , top panels ) or FAK -/- cells , in which pSrc was present at intracellular autophagic puncta ( Figure 5E , lower panels ) . Quantification of the relative intensity of pSrc at focal adhesions ( Figure 5F ) and of pSrc in intracellular puncta ( Figure 5—figure supplement 1B ) is shown . While there was more intense staining of active pSrc at focal adhesions in FAK P857A/P881A cells , there were no significant changes in the number or size of focal adhesions ( Figure 5—figure supplement 1C , D ) . The increased staining of pSrc ( and pFAK Y397; Figure 5—figure supplement 1E ) in FAK P875A/P881A cells was confirmed by immunoblotting isolated focal adhesion preparations ( Figure 5G , right panels , red dots; quantified as phospho/total Src ( and phospho/total FAK ) in Figure 5H ) , demonstrating that both active pSrc and pFAK were elevated relative to total Src and FAK , respectively . This was true also of pPaxillin Y118 ( Figure 5G , right panels ) , reflecting specific retention of phospho- and activated- focal adhesion components at focal adhesions when FAK cannot bind to Ambra1 . This implies that the FAK–Ambra1 complex is crucial to orchestrate the selective removal of active focal adhesion components from focal adhesions or to promote their turnover at these adhesion sites . The elevated levels of active components retained at focal adhesions when the binding of FAK to Ambra1 is impaired may contribute to the enhanced adhesion at early times after plating on extracellular matrix ( Figure 5D ) . To determine whether there were other biological consequences of the aberrant accumulation , or retention , of active pSrc and pFAK at focal adhesions when FAK binding to Ambra1 was impaired , we examined chemotactic invasion into growth factor-reduced Matrigel and proliferation in three-dimensional ( 3D ) culture . We found that elevated active pSrc and pFAK at focal adhesions in FAK P875A/P881A cells was associated with enhanced invasive migration ( Figure 6A , B ) and increased proliferation in 3D compared to FAK-WT cells ( Figure 6C , D ) , without affecting the Src-FAK interaction ( Figure 5—figure supplement 2E ) . As we found the Ambra1 binding partners Dynactin 1 and IFITM3 to regulate the trafficking of active Src in FAK-deficient SCC cells ( Figure 4 ) , we asked whether they also influenced mis-regulation of active pSrc and pFAK levels at focal adhesions when FAK can no longer bind to Ambra1 . We therefore transiently knocked down endogenous Dynactin 1 or IFITM3 expression using siRNA , and we found that the ‘over-invasion’ of FAK P875A/P881A-expressing cells was restored to similar , or lower , invasion levels than FAK-WT-expressing cells ( Figure 6E–G ) . Additionally , elevated active pSrc levels in the FAK P875A/P881A SCC cells were restored to normal ( wildtype ) levels by knocking down Dynactin 1 ( not shown ) . These results demonstrate that Ambra1 binding to FAK is required for the maintenance of normal steady-state levels of active pSrc and pFAK at focal adhesions . When this binding is perturbed so that pSrc and pFAK are elevated , the resultant enhanced invasion requires the Ambra1-interacting trafficking proteins Dynactin 1 and IFITM3 , which are therefore crucial mediators of active Src trafficking . 10 . 7554/eLife . 23172 . 019Figure 6 . Ambra1 binding-impaired mutant FAK increases invasion and 3D proliferation , which is rescued by reducing Dynactin 1 or IFITM3 expression levels . ( A ) Invasion assay: SCC FAK-WT , FAK P875A/P881A ( AA ) and FAK -/- cells were seeded on growth factor-reduced Matrigel in serum-free conditions . After 72 hr , invasion towards a serum gradient was visualised by staining the cells with calcein . n = 8 . Error bars , s . e . m . p≤0 . 01 . ( B ) Representative images of the invasion assay . ( C , D ) 3D proliferation assay: SCC FAK-WT , FAK P875A/P881A and FAK -/- cells were resuspended in methylcellulose solution in growth medium on a layer of agarose . After nine days , images were taken from 6–10 random fields and colonies were counted . ( C ) Quantification of the 3D proliferation assay . n = 3 . Error bars , s . d . p<0 . 01 . ( D ) Representative images of the 3D proliferation assay . ( E–G ) Invasion assay: SCC FAK-WT , FAK P875A/P881A and FAK -/- cells transiently transfected with Dctn1 siRNA were seeded on growth factor-reduced Matrigel in serum-free conditions . After 72 hr , invasion towards a serum gradient was visualised by staining the cells with calcein . ( E ) Representative images of the invasion assay . ( F ) Quantification of the invasion assay . n = 6 . Error bars , s . e . m . p≤0 . 01 . ( G ) Quantification of the invasion assay with cells transiently transfected with IFITM3 siRNA . n = 6 . Error bars , s . e . m . p≤0 . 01 ( * ) and p<0 . 05 ( # ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 01910 . 7554/eLife . 23172 . 020Figure 6—source data 1 . Mean values of invasion and number of colonies . Mean percentage and s . e . m . values of the relative invasion of SCC FAK-WT , P875A/P881A and -/- cells ( Figure 6A ) , as well as upon Dctn1 ( Figure 6F ) and IFITM3 ( Figure 6G ) knockdown by siRNA are shown . The mean number of colonies and s . d . are shown ( Figure 6C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 020
Our results define Ambra1 as a crucial regulator of Src/FAK-mediated cancer phenotypes , acting as a ‘spatial rheostat’ to maintain steady-state levels of active pSrc and pFAK at adhesions ( see model in Figure 7 ) . When FAK is present , Ambra1 forms a complex with FAK and Src and plays a key a role in Src/FAK-regulated cancer cell adhesion , polarisation and invasion ( Figure 7A ) . In FAK-deficient cells , Ambra1 controls the targeting of active Src away from focal adhesions into autophagic structures that the cells use to cope with toxic levels of active Src when it is not ‘tethered’ at adhesion sites by FAK ( Figure 7B; Sandilands et al . , 2012a ) . When Ambra1 is unable to bind to FAK ( as is the case in FAK P875A/P881A-expressing cells ) , this causes over-retention ( lack of removal or turnover ) of active Src ( and FAK ) from focal adhesions , promoting cell adhesion and enhanced invasive migration ( Figure 7C ) . Ambra1 is therefore a novel spatial regulator of the active Src/FAK complex at sites of cell-matrix adhesion , controlling downstream biological effects in cancer cells . It does this by scaffolding trafficking proteins , for example , Dynactin 1 and IFITM3 , which themselves lie at the heart of a network of scaffolding proteins that are needed for optimal trafficking of active Src ( and FAK ) , maintaining appropriate and tolerated levels . 10 . 7554/eLife . 23172 . 021Figure 7 . Model for Ambra1’s role in pSrc trafficking and cancer cell phenotypes . In mouse SCC cells , Ambra1 is localised at autophagosomes and focal adhesions . ( A ) In FAK-WT expressing ( ‘normal’ ) SCC cells , Ambra1 binds to FAK and Src , regulating Src/FAK-mediated cancer processes like adhesion , invasion , polarisation and 3D proliferation . ( B ) In FAK -/- SCC cells , Ambra1 regulates the trafficking of active Src from focal adhesions to autophagosomes . Dynactin 1 ( Dctn1 ) and IFITM3 are involved in this Ambra1-regulated trafficking process . ( C ) In cells expressing the FAK P875A/P881A mutant , Ambra1 binds to a lesser extent to FAK , but still to Src . This impaired FAK–Ambra1 interaction results in increased active Src levels at focal adhesions , resulting in enhanced adhesion , invasion , polarisation and 3D proliferation . Most likely this is due to perturbed trafficking of focal adhesion components , as knockdown of proteins involved in trafficking processes , like Dynactin 1 and IFITM3 , rescues the phenotypes in FAK P875A/P881A expressing SCC cells . DOI: http://dx . doi . org/10 . 7554/eLife . 23172 . 021 Ambra1 is crucial for the development of the central nervous system , neurogenesis and embryogenesis ( Benato et al . , 2013; Fimia et al . , 2007; Yazdankhah et al . , 2014 ) , and is an autophagy regulator , via its interaction with Beclin1 ( Fimia et al . , 2007 ) . The role of Ambra1 in cancer is less clear . One study suggested that Ambra1 may have a tumour suppressor role , as it is involved in the degradation of c-Myc induced by dephosphorylation by PP2A , resulting in reduced proliferation and tumourigenesis ( Cianfanelli et al . , 2015 ) . However , another study implied that Ambra1 may have tumour-promoting functions , since Ambra1 overexpression correlated with invasion and poor survival in cholangiocarcinoma ( Nitta et al . , 2014 ) . The new findings we present here show that Ambra1 contributes to SCC cell behaviour by scaffolding both Src and FAK , and intracellular trafficking proteins , controlling steady-state levels and spatial distribution of the activated forms of Src and FAK at focal adhesions . In this context , therefore , Ambra1 is a scaffold protein that promotes cancer cell phenotypes driven by the Src/FAK pathway , including cancer cell polarisation and chemotactic invasion . This central trafficking scaffold function of Ambra1 , and the discovery that it is a crucial ‘spatial rheostat’ for Src/FAK signalling , adds to the number of important physiological functions for Ambra1 , and defines a new level of regulation of the Src/FAK signalling axis . It is interesting to note that while knockdown of Ambra1 caused reduced cancer cell polarisation towards the denuded area of a wounded monolayer , and suppressed chemotactic invasive migration in FAK-expressing cancer cells , there was no effect in FAK-deficient cells , which display intrinsically low polarisation and invasion capacity . We conclude that Ambra1 is a FAK scaffold required for FAK-dependent cancer cell phenotypes , most likely by controlling the localisation of active FAK , and its upstream regulator Src , at focal adhesions . Ambra1 is reported to be an autophagy regulator ( Fimia et al . , 2007 ) , and we found that it is crucially involved in the redistribution of active Src away from focal adhesions into intracellular autophagic puncta in FAK-deficient cancer cells ( described previously in Sandilands et al . , 2012a ) . Ambra1 is localised to intracellular puncta containing the autophagosomal marker LC3B , and it interacts with LC3B . These data implicate Ambra1 in the selective autophagic trafficking of active Src when cancer cells are under severe ‘adhesion stress’ , such as when FAK is absent . We found that Ambra1-interacting proteins ( identified by mass spectrometry ) included the dynactin complex component Dynactin 1 and IFITM3 , in addition to other proteins considered to be intracellular trafficking proteins . As Ambra1 has been reported to interact with the dynein complex , we thought it likely that Ambra1 may regulate the trafficking of active Src to intracellular autophagic puncta via dynein-dependent processes ( Di Bartolomeo et al . , 2010 ) , and we found that knockdown of endogenous Dynactin 1 by siRNA inhibited the trafficking of active Src from focal adhesions to autophagosomes . Dynactin 1 is thought to be involved in retrograde trafficking and maturation of trafficking vesicles ( Jovasevic et al . , 2015; Kedashiro et al . , 2015a; Li et al . , 2014; Ohbayashi et al . , 2012; Verissimo et al . , 2015 ) , implying that these may be involved in the redistribution of active Src from focal adhesions to autophagosomal puncta in FAK-deficient cells . We noted that Dynactin 1 knockdown resulted in enlarged late endosomes ( not shown ) , while the number of late endosomes was not altered ( not shown ) , implying that impaired trafficking/maturation of vesicles may be responsible for the block to autophagic targeting of active Src upon Ambra1 and Dynactin 1 knockdown . Further , IFITM3 has been reported to localise to late endosomes/lysosomes and overexpression expands Rab7- and LAMP1-positive structures ( Feeley et al . , 2011 ) , and loss of IFITM3 blocks clathrin-mediated phagocytosis by removing clathrin from membranes ( Wee et al . , 2012 ) . Knockdown of IFITM3 also significantly impaired the trafficking of active Src to autophagic puncta . In proteomic network analyses , we found that IFITM3 interacts both with FAK and active Src , as well as Ambra1 , thereby linking the Ambra1- , FAK- and pSrc-interacting trafficking protein networks together . Furthermore , IFITM3 appears to be important for the interaction of active pSrc with FAK as well as to a lesser extent of Ambra1 with FAK ( not shown ) , but not Ambra1 with pSrc . These findings suggest that IFITM3 is involved in regulating FAK’s interaction with Ambra1 and pSrc . Overall , these data place IFITM3 at the centre of a larger trafficking protein network linking Ambra1 , FAK and active Src sub-networks . The full range of the Src/FAK/Ambra1-interacting trafficking proteins from the interactome network ( Figure 3 ) that play a role in pSrc trafficking are unknown at present . Expression of a mutant FAK protein ( in otherwise FAK-deficient cancer cells ) that was hugely impaired in its ability to bind Ambra1 ( FAK P875A/P881A ) , demonstrated the importance of the complex between FAK and Ambra1 in promoting the removal of phosphorylated components from focal adhesions , including active pSrc and pFAK . This Ambra1 binding-impaired mutant FAK protein caused increased steady-state levels of active pSrc and pFAK at focal adhesions , and enhanced adhesion , proliferation in 3D and chemotactic invasive migration , without affecting Src/FAK binding . Enhanced invasion was reversed by knockdown of the Ambra1-binding trafficking proteins Dynactin 1 and IFITM3 , demonstrating that these Ambra1-scaffolded proteins are involved in the dynamic regulation of active pSrc and pFAK , presumably by trafficking them to and from focal adhesions . Indeed , we found that FAK-driven invasion in the SCC cells was hugely dependent on IFITM3 , implying that its crucial role at the heart of trafficking protein networks linking Ambra1 , FAK and active Src is vital to Src/FAK-mediated invasion . One caveat of interpreting the data with the FAK P875A/P881A mutant protein is that this mutant is also impaired in binding p130Cas , although to a much lesser extent than Ambra1; importantly though , and in contrast to knockdown of Ambra1 , knockdown of p130Cas did not significantly affect invasion ( not shown ) . In summary , we have identified a novel role for Ambra1 as part of an intracellular trafficking network of proteins that control the steady-state levels , and dynamics , of active Src and FAK at focal adhesions . Ambra1 scaffolds proteins such as Dynactin 1 and IFITM3 – and here , for the first time , we link these to the spatial control of active Src and FAK , regulating their steady-state levels at focal adhesions and the autophagic targeting of active Src when FAK is absent . As a result , Ambra1 and its interacting partners control cancer cell adhesion , polarisation , proliferation in 3D and chemotactic invasion . Deregulation of these trafficking components of focal adhesion complexes inhibits Src/FAK-dependent biological processes in cancer cells . Our work demonstrates the importance of tight dynamic control of trafficking of active Src , in particular , to and from focal adhesions .
Antibodies used were as follows: anti-Paxillin ( RRID:AB_647289 ) , anti-GM130 ( RRID:AB_398141 ) , anti-p130Cas ( RRID:AB_397667 ) and anti-RACK1 ( RRID:AB_397577 ) antibodies ( BD Transduction Laboratories , New Jersey , USA ) , anti-IFITM3 ( RRID:AB_2122095 ) ( Abcam , Cambridge , UK ) , anti-CoxIV ( RRID:AB_10694213 ) , anti-FAK ( RRID:AB_10694068 ) , anti-pFAK Y397 ( RRID:AB_2173659 ) , anti-pPaxillin Y118 ( RRID:AB_2174466 ) , anti-Rab7 ( RRID:AB_1904103 ) , anti-pSrc Y416 ( RRID:AB_331697 ) , anti-Src ( clone 36D10 ) ( RRID:AB_10693939 ) , anti-LC3B ( RRID:AB_2137707 ) and anti-GAPDH ( RRID:AB_10622025 ) ( Cell Signaling Technologies , Danvers , MA , USA ) , and anti-Dctn1 ( RRID:AB_10842517 ) , anti-Ambra1 ( RRID:AB_2636939 ) and anti-pSrc Y416 ( RRID:AB_309898 ) antibodies and anti-FAK ( clone 4 . 47 ) -conjugated agarose antibody ( RRID:AB_310789 ) ( Millipore , Billerica , MA , USA ) . LC3B antibody was from MBL ( RRID:AB_1279144 ) ( MBL International , Woburn , MA , USA ) . TRITC-phalloidin was purchased from Life Technologies ( RRID:AB_2572408 ) ( Paisley , UK ) . Anti-rabbit ( RRID:AB_2099233 ) or mouse ( RRID:AB_330924 ) peroxidase-conjugated secondary antibodies were purchased from Cell Signaling Technologies . Dasatinib was obtained from Bristol Myers Squibb ( Princeton , NJ , USA ) and PF562271 from Pfizer ( Groton , CT , USA ) . FAK mutants were generated by site-directed mutagenesis using PFU Ultra Hotstart DNA polymerase ( Stratagene , Amsterdam , The Netherlands ) and the following primers ( mutated base pairs are underlined ) : P875A ( forward 5’ – GATCATGCCGCTCCAGCAAAGAAGCCCCCT – 3’ , reverse 5’ – GCGAGGGGGCTTCTTTGCTGGAGCGGCATG – 3’ ) and P881A ( forward 5’ – CAAAGAAGCCCCCTCGCGCTGGAGCCCCCC – 3’ , reverse 5’ – CAAGTGGGGGGCTCCAGCGCGAGGGGGCTT – 3’ ) . After DpnI digestion for 1 hr at 37°C , chemically competent TOP10 bacteria were transformed . FAK-deficient SCC cell lines were generated , authenticated and characterised as described previously ( Serrels et al . , 2012 , 2010 ) . SCCs were maintained in Glasgow MEM containing 10% FCS , 2 mM L-glutamine , non-essential amino acids , sodium pyruvate and MEM vitamins at 37°C , 5% CO2 . SCC FAK-WT cells were maintained in 1 mg/ml hygromycin B . Ambra1 +/+ and -/- RasV12/E1A-transformed MEFs were a generous gift from Guillermo Velasco ( Cianfanelli et al . , 2015 ) . MEFs were cultured in DMEM supplemented with 10% FCS and 2 mM L-glutamine at 37°C , 5% CO2 . Following thawing cells were used for no longer than three months . Original cells were pathogen tested using the ImpactIII test ( Idexx Bioresearch , Westbrook , ME , USA ) and were negative for all pathogens tested . All cell lines were routinely tested negative for mycoplasma contamination . Ambra1 siRNA pool ( cat . no . M-059556–01 ) , individual Ambra1 siRNAs ( cat . no . D-059556–01 , 5’ – GAAGAAUGCUGUACGAAUC – 3’; D-059556–04 , 5’ – CAACGUGCCCUCCUGCAAU – 3’ ) , Dctn1 siRNA pool ( cat . no . M-044821–01 ) , IFITM3 siRNA pool ( cat . no . M-056653–01 ) or scrambled siRNA ( cat . no . D-001206-13-20 ) were purchased from Dharmacon ( Loughborough , UK ) . FAK-WT or FAK -/- SCC cells were transiently transfected using HiPerFect ( Qiagen , Manchester , UK ) according to the manufacturer’s protocol , with a final concentration of 100 nM siRNA , respectively . Cells were analysed at 48–72 hr post transfection . The Ambra1 binding site in FAK was identified using peptide arrays as published previously ( Serrels et al . , 2010 , 2007 ) . Briefly , overlapping 25-mer peptides of FAK were spotted onto nitrocellulose and incubated with recombinant Ambra1 ( Origene , Herford , Germany ) . After extensive washes , the array was incubated with anti-Ambra1 antibody and then subjected to western blotting . For the identification of core amino acids , overlapping 25-mer peptides with one amino acid mutated at a time were used . Cells were washed twice in ice-cold PBS and then lysed in RIPA buffer ( 50 mM Tris-HCl , pH 8 . 0 , 150 mM NaCl , 1% Triton X-100 , 0 . 1% SDS and 0 . 5% sodium deoxycholate ) supplemented with PhosStop and Complete Ultra Protease Inhibitor tablets ( Roche , Welwyn Garden City , UK ) . Lysates were cleared by centrifugation at 10000 rpm for 15 min and analyzed by western blotting . Protein concentration was calculated using a BCA protein assay kit ( Thermo Scientific , Loughborough , UK ) . For immunoprecipitation , 1 mg cell lysates were incubated with 2 μg of unconjugated or 10 μl of agarose-conjugated antibodies at 4°C overnight with agitation . Unconjugated antibody samples were incubated with 10 μl of Protein A or Protein G agarose for 1 hr at 4°C . Beads were washed three times in lysis buffer and once in 0 . 6 M LiCl , resuspended in 20 μl 2x sample buffer ( 130 mM Tris-HCl , pH 6 . 8 , 20% glycerol , 5% SDS , 8% β-mercaptoethanol , bromophenol blue ) and heated for 5 min at 95°C . Samples were then subjected to SDS-PAGE analysis using the Bio-Rad TGX pre-cast gel system . Proteins were immunoblotted using the Bio-Rad Trans-blot Turbo transfer system , blocked in 5% BSA in TBST ( TBS supplemented with 1% Tween-20 ) , and incubated with primary antibody overnight at 4°C . Blots were washed three times in TBST , incubated with peroxidase-conjugated secondary antibody for 45 min at room temperature , washed as before , developed using Clarity Western ECL Substrate ( Bio-Rad , Hemel Hempstead , UK ) and imaged using a Bio-Rad ChemiDoc MP Imaging System ( Bio-Rad , Hemel Hempstead , UK ) . For mass spectrometry , SCC FAK-WT and -/- cells were lysed in RIPA buffer . For the immunoprecipitations , 2 mg cell lysates ( samples in triplicates ) were incubated with 2 μg of unconjugated antibodies ( anti-Ambra1 , anti-pSrc Y416 and rabbit-anti-IgG ) or 10 μl of agarose-conjugated antibodies ( FAK–agarose ) at 4°C overnight with agitation . Unconjugated antibody samples were incubated with 20 μl of Protein A agarose for 1 hr at 4°C . Beads were washed twice in lysis buffer and twice in PBS . Protein complexes were subjected to on-bead proteolytic digestion , followed by desalting and liquid chromatography–tandem mass spectrometry as reported previously ( Turriziani et al . , 2014 ) . The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository ( Vizcaíno et al . , 2016 ) with the dataset identifier PXD006002 . Trafficking proteins were filtered from the total datasets and the interaction network analysis was performed using Cytoscape ( RRID:SCR_003032 ) . Cells were grown on glass coverslips for 24 hr and washed once in TBS prior to fixation ( 3 . 7% formaldehyde , 100 mM PIPES , pH 6 . 8 , 10 mM EGTA , 1 mM MgCl2 and 0 . 2% Triton X-100 ) for 10 min . Cells were subsequently washed twice in TBStx ( TBS supplemented with 0 . 1% Triton X-100 ) and blocked in TBStx block ( TBStx supplemented with 3% BSA ) . Primary antibodies were incubated in TBStx block overnight at 4°C , followed by three 5 min washes in TBStx , incubated with Alexa Fluor-labelled secondary antibodies diluted 1:200 in TBStx block ( Life Technologies , Paisley , UK ) and washed as before prior to being mounted in Vectashield mounting medium containing DAPI ( Vector Labs , Peterborough , UK ) . Cells were imaged using a FV1000 confocal microscope ( Olympus , Southend-on-Sea , UK ) . For the Total Internal Reflection Fluorescence ( TIRF ) microscopy an inverted IX81 microscope ( Olympus , Southend-on-Sea , UK ) with a 150 × 1 . 45 NA UAPON TIRF objective using 491 and 561 excitation lines was used . Colocalisation was analyzed using the ImageJ plugin JaCoP ( Bolte and Cordelières , 2006 ) . The intensity of pSrc or Paxillin at focal adhesions from three independent experiments ( five cells per experiment; at least 10 focal adhesions per cells ) was measured using ImageJ ( RRID:SCR_003070 ) . Focal adhesion isolation was performed following the protocol described in ref . ( Kuo et al . , 2012 ) . Briefly , cells were rinsed with PBS and incubated with TEA buffer ( 0 . 2 M triethanolamine , pH 8 . 0 ) for 5 min . To apply hydrodynamic force , the cells were rinsed with PBS for 10 s using a Waterpik dental flosser set at 2 ( Waterpik , Reigate , UK ) . After another wash with PBS , the remaining attached focal adhesions were fixed for immunofluorescence analysis . For western blot analysis , focal adhesions were crosslinked with 3 mM dimethyl 3 , 3'-dithiobispropionimidate ( DTBP; Thermo Scientific , Loughborough , UK ) for 30 min at room temperature and then quenched with 20 mM Tris-HCl , pH 8 . 0 , for 5 min at room temperature . After another wash with 20 mM Tris-HCl , pH 8 . 0 , cells were rinsed with ice-cold PBS , and 5 ml of focal adhesion extraction buffer ( 0 . 5% Triton X-100 , 20 mM NH4OH in PBS ) was added to the cells for 5 min . Cells were rinsed with PBS for 10 s using a Waterpik dental flosser set at 2 , and the remaining attached focal adhesions were lysed in 2x sample buffer . Western blot bands were analysed by densitometry . Cell polarisation assessing the orientation of the Golgi apparatus in wounded cell monolayers was examined as described in ( Serrels et al . , 2010 ) . Briefly , 3 × 106 cells were plated on fibronectin-coated coverslips in 12-well plates for 3 hr . The cell monolayer was wounded with a pipette tip , incubated in full SCC growth medium for 1 . 5 hr and then fixed and stained with anti-GM130 antibody . Invasion was analysed as described in ( Serrels et al . , 2010 ) . Briefly , growth factor-reduced Matrigel ( BD Biosciences , Oxford , UK ) was diluted 1:1 in cold PBS and allowed to set at 37°C in transwells . 2 × 104 cells were seeded onto the underside of the transwell . After 4 hr , the transwells were washed in PBS and placed into serum-free SCC growth medium . Full growth medium containing 10% FCS was added on top of the Matrigel . After 72 hr , cell invasion was assessed by staining with 5 μM calcein ( Life Technologies , Paisley , UK ) for 1 hr . Horizontal z sections through the Matrigel were acquired at 10 μm intervals with an Olympus FV1000 confocal microscope . The images were evaluated using ImageJ software . Cells were washed twice in PBS , resuspended in PBS and rotated for 45 min at 4°C . After that , 8000 cells/well were plated in serum-free medium on fibronectin-coated 96-well plates or on plastic at 37°C . Cells were fixed after 20 min , 1 , 2 and 6 hr with 25% trichloracetic acid ( TCA ) and stored overnight at 4°C . Plates were washed five times with H2O and dried at 37°C . The fixed cells were stained with sulforhodamine B ( SRB ) ( 0 . 4% SRB in 1% acetic acid ) for 30 min , washed five times with 1% acetic acid and dried at 37°C . The SRB was dissolved in 10 mM Tris , pH 10 . 5 , for 2 hr at room temperature and the absorbance was read at 540 nm . Samples were normalised to the 6 hr time point and relative adhesion was calculated by setting the FAK-WT values to 1 . Cells were resuspended in a 1 . 4% methylcellulose solution in growth medium , plated on a layer of 0 . 9% agarose and incubated at 37°C , 5% CO2 . After nine days , images were taken from 6–10 random fields ( 10x magnification ) and colonies were counted . For measuring the intensity of pSrc Y416 staining at focal adhesions ( immunofluorescence ) , n = 5 ( 5 cells per experiment ) . For the invasion assay , the experiment was repeated with the following times: siAmbra1 , n = 5; FAK P875A/P881A mutant , n = 8; siDctn1; n = 6 and siIFITM3 , n = 6 . All other experiments shown , n = 3 . Quantification of internalised active Src was carried out by counting 100 cells and calculating the percentage . Error bars for the graphs showing the intensity of pSrc Y416 staining at focal adhesions and the invasion assays represent s . e . m . Error bars for all other experiments show s . d . Student’s t-test was carried out to calculate the statistical significance . | In animal bodies , a mesh of proteins – known the extracellular matrix – holds cells together to give the body shape and make it more stable . Cells bind to the matrix using structures called focal adhesions . However , cells do not always stay in one place: in young animals , certain cells need to move around the body to reach their final destination . Adult animals also have some cells that are able to move , for example , to close wounds . The cells move when the focal adhesions that hold these cells in place are taken apart and then rebuilt . These processes are very dynamic and happen all the time when cells move . They are normally tightly controlled to ensure that cells only migrate under appropriate conditions . However , focal adhesions are less well regulated in cancer cells , allowing the cells to migrate away from a tumour to form new tumours elsewhere in the body . Focal adhesions are large structures that contain many proteins . These proteins include FAK and Src , which are particularly important and have been well studied . In order to better understand how focal adhesions are taken apart , Schoenherr et al . wanted to discover new proteins that interact with FAK in skin cancer cells from mice . The experiments show that FAK binds to a protein called Ambra1 , which is known to control how other proteins move around inside cells . Ambra1 and FAK work together to regulate the movement of Src away from focal adhesions and into the cell . Furthermore , Ambra1 belongs to a larger network of proteins within the cancer cells that regulates the location of Src . By changing the levels of Src and FAK at focal adhesions , Ambra1 and its other binding partners can control whether the cancer cells are attached to the extracellular matrix or are free to migrate . Overall this work shows that the location and activity of Src within cells needs to be carefully controlled to stop the cells from moving at the wrong time . Further experiments will aim to understand which other proteins are involved in this network and how they contribute to the growth of cancer cells and the spread of tumours around the body . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"cancer",
"biology"
] | 2017 | Ambra1 spatially regulates Src activity and Src/FAK-mediated cancer cell invasion via trafficking networks |
The mechanisms regulating synapse numbers during development and ageing are essential for normal brain function and closely linked to brain disorders including dementias . Using Drosophila , we demonstrate roles of the microtubule-associated protein Tau in regulating synapse numbers , thus unravelling an important cellular requirement of normal Tau . In this context , we find that Tau displays a strong functional overlap with microtubule-binding spectraplakins , establishing new links between two different neurodegenerative factors . Tau and the spectraplakin Short Stop act upstream of a three-step regulatory cascade ensuring adequate delivery of synaptic proteins . This cascade involves microtubule stability as the initial trigger , JNK signalling as the central mediator , and kinesin-3 mediated axonal transport as the key effector . This cascade acts during development ( synapse formation ) and ageing ( synapse maintenance ) alike . Therefore , our findings suggest novel explanations for intellectual disability in Tau deficient individuals , as well as early synapse loss in dementias including Alzheimer’s disease .
The correct formation and subsequent maintenance of synapses is a key prerequisite for brain development , function and longevity . Precocious loss of synapses is observed in late onset neurodegenerative diseases including Alzheimer's disease ( AD ) and Frontotemporal Dementia ( FTD ) , likely contributing to the cognitive decline and neuronal decay observed in patients ( Pooler et al . , 2014; Saxena and Caroni , 2007; Serrano-Pozo et al . , 2011 ) . Therefore , the characterisation of mechanisms maintaining synapses during ageing would have major implications for our understanding of dementias . The development of synapses and their maintenance during ageing is dependent on sustained transport of synaptic proteins from the distant soma , driven by motor proteins which trail along the bundles of microtubules in axons and dendrites ( Goldstein et al . , 2008 ) . Microtubules are regulated by microtubule binding proteins which are therefore in a key position to regulate synapse formation and maintenance ( Prokop , 2013 ) . Tau is a microtubule associated protein ( MAP ) discovered in the mid-seventies ( Weingarten et al . , 1975 ) . Reduction in Tau levels has been linked to intellectual disability ( Sapir et al . , 2012 ) and a class of brain disorders termed 'dementias which lack distinctive histopathology' ( DLDH ) ( Zhukareva et al . , 2001 ) . Tau detachment from MTs is linked to prominent neurodegenerative diseases such as Alzheimer's disease , Frontotemporal Dementia and some forms of Parkinson’s disease ( Kovacs , 2015 ) . In vitro , Tau has the ability to regulate microtubule properties including stability , cross-linkage and polymerisation ( Morris et al . , 2013 ) . Through such functions , Tau would be expected to regulate multiple aspects of neuronal cell biology , but its physiological roles are still not understood and highly debated ( Morris et al . , 2013 ) . This might partly be due to experimental challenges posed by functional redundancy , where other MAPs are proposed to mask physiological roles of Tau ( Ma et al . , 2014; Takei et al . , 2000 ) . A good model in which to deal with functional redundancy is the fruit fly Drosophila melanogaster . As is ideal for studies of Tau , Drosophila neurons provide access to powerful genetics , they are readily established for research on the neuronal cytoskeleton ( Sánchez-Soriano et al . , 2010 ) , on neuronal transport ( Schwarz , 2013 ) and on synapses ( Prokop and Meinertzhagen , 2006 ) . Importantly , concepts and mechanisms gained from work in flies are often well conserved in higher organisms ( Bellen et al . , 2010; Jaiswal et al . , 2012 ) . Work in Drosophila suggested that the spectraplakin Short Stop ( Shot ) , a large actin-MT linker molecules and potent regulators of microtubules , could display potential functional overlap with Tau during microtubule stabilisation ( Alves-Silva et al . , 2012; Prokop , 2013 ) . This hypothesis is attractive because the well-conserved mammalian spectraplakin Dystonin is already linked to a neurodegenerative disease ( type VI hereditary sensory autonomic neuropathy; OMIM #614653; ) ( Ferrier et al . , 2013 ) , and its paralogue ACF7/MACF1 plays important roles during brain development ( Goryunov et al . , 2010; Ka and Kim , 2015 ) . Since ACF7 continues to be expressed in the brain , it is tempting to speculate that it might be required for neuronal maintenance ( Bernier et al . , 2000 ) . Here we use Drosophila neurons , in culture and in vivo alike , to demonstrate novel roles of Tau in regulating the formation and maintenance of synapses during ageing , by coordinating the intracellular trafficking of synaptic proteins . Thus , we show that the role of Tau in synapse regulation occurs in functional overlap with Shot . The robust shot-tau double-mutant phenotypes enabled us to study the mechanistic cascade composed of three steps: microtubule stability as the trigger , the JNK signalling pathway as the mediator and kinesin-3 mediated axonal transport of synaptic proteins as the key effector . We propose a new mechanism based on the loss of Tau function which could explain intellectual disability in MAPT ( the human tau gene ) mutant individuals and precocious synapse loss in tau-related neurodegeneration ( Saxena and Caroni , 2007; Serrano-Pozo et al . , 2011 ) .
To study synaptic roles of Drosophila Tau , we first used primary Drosophila neurons generated from tau mutant embryos . Primary fly neurons are genetically and experimentally highly amenable and provide robust cellular and subcellular readouts ( Prokop et al . , 2012 ) . These cultures are also particularly suited for the study of embryonic lethal mutations since they allow the examination of neurons beyond the embryonic lethal stage . Already 8 hr in vitro ( HIV ) , these neurons show transport of synaptic material in the growing axon ( Sánchez-Soriano et al . , 2010 ) and after 2 days in vitro ( DIV ) , they display functional presynaptic sites ( Küppers-Munther et al . , 2004; Küppers et al . , 2003 ) that can be reliably stained with antibodies against presynaptic proteins ( Figure 1—figure supplement 1 ) . They contain dense bars and synaptic vesicle accumulations which undergo excitation-dependent uptake and release ( Küppers-Munther et al . , 2004; Küppers et al . , 2003 ) . 10 . 7554/eLife . 14694 . 003Figure 1 . Tau and Shot are required for the formation of synaptic specialisations in axons . ( A ) Primary neurons at 2 DIV obtained from embryos that were wildtype ( wt ) , tau-/- , and tau-/- with elav-Gal4 driven expression of UAS-tau-GFP; neurons were stained for tubulin ( Tub , red ) and the synaptic protein Synaptotagmin ( Syt , light blue ) . ( B ) Quantification of the experiment shown in A , shown as the number of Syt puncta per neuron , normalised to wildtype ( the assessed numbers of neurons are indicated in each bar , ***PMW<0 . 001 , **PMW<0 . 01 ) . ( C ) Primary Drosophila neurons at 2DIV , obtained from embryos that were wildtype ( wt ) , tauMR22 ( tau-/- ) , shot3 ( shot-/- ) , and shot3 tauMR22 ( shot-/- tau-/- ) , co-stained with antibodies against HRP ( magenta ) and the synaptic proteins ( green ) Syt and Bruchpilot ( Brp ) ; areas emboxed with dashed lines are displayed as magnified insets showing the synaptic staining only . ( D ) Quantification of the experiments in C , displayed as number ( no . ) of Syt and Brp puncta per neuron , normalised to wildtype ( the assessed numbers of neurons are indicated in each bar , ***PMW<0 . 001; **PMW<0 . 01; *PMW<0 . 05 , ) . Scale bar: 10 µm . A statistics summary of the data shown here is available in Figure 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 00310 . 7554/eLife . 14694 . 004Figure 1—source data 1 . Summary of the statistics from Figure 1B and C . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 00410 . 7554/eLife . 14694 . 005Figure 1—figure supplement 1 . Co-localisation of presynaptic markers reveals presynaptic specialisations . Primary Drosophila neurons at 2 DIV co-stained with antibodies against the neuronal marker HRP ( red ) and different presynaptic proteins ( green and blue ) including Synaptotagmin ( Syt ) , Synapsin ( Syn ) , neuronal Synaptobrevin ( nSyb ) or Bruchpilot ( Brp ) ; whole neurons are shown , with chevrons indicating somata and the emboxed areas indicating the distal axons shown as threefold manified close-ups; presynaptic proteins show a high degree of co-localisation in the axons , 91% co-localisation of Synapsin ( Syn ) and Synaptotagmin ( Syt ) labelled spots ( sample size = 24 neurons ) , 84 . 7% of Syt and Bruchpilot ( Brp ) ( sample size = 8 neurons ) , and 81% of neuronal Synaptobrevin ( nSyb ) with Brp ( sample size = 8 neurons ) , suggesting that the majority of dots represent presynaptic specialisations . Scale bar: 10 µmDOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 00510 . 7554/eLife . 14694 . 006Figure 1—figure supplement 2 . Rescue experiments with Shot and Tau demonstrate redundant roles in synapses . ( A ) Primary neurons at 2 DIV obtained from embryos that were wildtype ( wt ) , shot-/- with elav-Gal4 driven expression of UAS-shot-GFP , shot-/- with sca-Gal4 driven expression of UAS-tau-GFP , and tau-/- with elav-Gal4 driven expression of UAS-shot-GFP; neurons were stained for tubulin ( Tub , magenta ) and Syt ( green ) . ( B ) Quantification of the experiment in A , shown as the number of Syt puncta per neuron , normalised to wildtype ( the assessed numbers of neurons are indicated in each bar , ***PMW<0 . 001; *PMW<0 . 05; ns , not significant PMW>0 . 05 ) . Scale bar: 10 µm . A statistics summary of the data shown here is available in Figure 1—figure supplement 2—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 00610 . 7554/eLife . 14694 . 007Figure 1—figure supplement 2—source data 1 . Summary of the statistics from Figure 1—figure supplement 2B . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 007 The Df ( 3R ) tauMR22 mutation ( tauMR22 ) is an embryonic lethal chromosome deletion that uncovers most of the Drosophila tau gene and is a true null allele ( Bolkan and Kretzschmar , 2014; Doerflinger et al . , 2003 ) . We found that tauMR22 mutant primary neurons at 2 DIV show a decrease in the number of puncta positive for Bruchpilot ( Brp ) and Synaptotagmin ( Syt ) ( Bruchpilot/Brp: 42%; Synaptotagmin/Syt: 59%; all compared to wildtype control neurons; Figure 1 ) . Our finding suggest that tau-deficient primary neurons contain fewer Brp and Syt positive presynaptic specialisations . In the following , we will refer to this phenotype as synapse reduction . To confirm that this reduction in synapse numbers was due to the loss of Tau , we performed rescue experiments using Gal4-induced neuronal expression of UAS-tau-GFP ( Doerflinger et al . , 2003 ) in tauMR22 mutant neurons . We found a significant improvement of the tauMR22 mutant phenotype ( Figure 1A–B ) . We concluded that absence of Tau causes a reduction in presynaptic sites . To assess potential functional overlap of Tau with the Drosophila spectraplakin Shot , we first analysed shot3 mutant primary neurons at 2 DIV . We found a reduction in synapse numbers ( Brp: 62%; Syt: 67%; Figure 1C–D ) , consistent with previous descriptions in vivo ( Löhr et al . , 2002; Prokop et al . , 1998 ) . We confirmed that this reduction in synapse numbers was due to the loss of Shot by using Gal4-induced neuronal expression of UAS-shot-GFP ( Alves-Silva et al . , 2012; Sanchez-Soriano et al . , 2009 ) which significantly rescued the synapse phenotype in shot3 mutant neurons ( Figure 1—figure supplement 2 ) , confirming the involvement of Shot . We then assessed potential functional overlap of Shot and Tau . First , we analysed primary neurons double-mutant for the shot3 and tauMR22 null alleles ( shot-tau ) which showed even lower synapse numbers ( Brp: 22%; Syt: 39%; Figure 1C–D ) than either of the single mutant neurons . Notably , these analyses were performed on clearly polarised neurons with well developed axons to exclude indirect effects caused by defective axon growth ( Figure 3—figure supplement 3 ) . Despite that , we found that the double-mutant neurons displayed reduced branch numbers ( Figure 3—figure supplement 3F ) . However , we could demonstrate that the lower number in branches is not the cause for synapse reduction by using knock-down experiments as well as rescue experiments ( explained in detail below , Figure 3—figure supplement 3 and Figure 3—figure supplement 4 , see also Discussion ) . In further support of functional overlap , also our genetic interaction studies revealed a synapse reduction phenotype in shot3/+ tauMR22/+ double heterozygous mutant neurons ( see later in Figure 5A ) . Finally , we performed cross-rescue experiments by expressing a shot transgene in tauMR22 mutant neurons and a tau transgene in shot3 mutant neurons . In both cases , Syt staining revealed a rescue of the synapse reduction phenotype ( Figure 1—figure supplement 2 ) . Taken together , our results indicate that Shot and Tau functionally overlap , rather than act hierarchically in the same pathway . Next , we investigated synaptic phenotypes in vivo . Since shot3 and tauMR22 animals are late embryonic lethal , we analysed them at late embryonic stage 16 , when Syt is already confined to nascent synaptic terminals , as can be reliably imaged at neuromuscular junctions ( NMJs; Figure 2 and Figure 2—figure supplement 1 for a schematic drawing of the embryonic NMJ ) ( Littleton et al . , 1993 ) . In shot-tau mutant embryos , Syt levels at NMJs were reduced to 48% , whereas shot mutant embryos showed a milder reduction to 71% , and tau mutant embryos no detectable effect ( Figure 2 ) . Taken together , our data suggest that Tau is required for the formation of synapses in culture and in vivo and that Tau and Shot functionally overlap in this context . 10 . 7554/eLife . 14694 . 008Figure 2 . Tau and Shot regulate the localisation of presynaptic proteins at the embryonic NMJ in vivo . ( A ) Images show the dorsal segment of inter-segmental motornerves ( Landgraf et al . , 2003 ) in stage 16 embryos that were wildtype ( wt ) , tauMR22 ( tau-/- ) , shot3 ( shot-/- ) , and shot3 tauMR22 ( shot-/- tau-/- ) , stained with antibodies against Syt ( green ) and the motorneuron-specific cell membrane protein Fasciclin II ( FasII , magenta ) . Arrowheads depict the distal end of the motoraxons where the nascent NMJs are forming; boxed areas are displayed as enlarged insets showing anti-Syt staining only . Note that cell bodies of sensory neurons contain visible levels of Syt in the mutant ( white arrows ) but not in wildtype neurons ( open arrow ) . ( B ) Quantification of the experiments in A , shown as the average intensity of Syt at the nerve ending normalised to wildtype ( the sample number of NMJs is indicated in each bar , ***PMW<0 . 001; **PMW<0 . 01; *PMW<0 . 05; ns , not significant PMW>0 . 05 ) . Scale bars: 10 µm . A statistics summary of the data shown here is available in Figure 2—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 00810 . 7554/eLife . 14694 . 009Figure 2—source data 1 . Summary of the statistics from Figure 2B . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 00910 . 7554/eLife . 14694 . 010Figure 2—figure supplement 1 . Schematic drawings of embryonic tissues analysed in this study . ( A ) Schematic horizontal view of the embryonic nervous system ( brain; OL , optic lobe; vNC , ventral nerve cord ) showing an exemplary motorneuron ( dark blue ) , muscle ( green ) and sensory neuron ( red ) . Note that cell bodies in the nervous system lie in the cortex ( Cx ) , i . e . outside the synaptic area ( Np , neuropile ) , and that sensory neurons are positioned in the periphery , often adjacent to synaptic endings of motorneurons ( NMJs , neuromuscular junctions ) . Synapses are shown as yellow dots . The stippled blue frame shows the area ( rotated 90 degrees clockwise ) shown in Figures 5C , D , 8F . ( B ) More detailed representation of the periphery with muscles , motorneuronal projections and sensory neurons ( same colours as in A ) . The stippled blue frame emboxes the area shown in Figures 2 , 4E , 5C , E , 6D , F , 7 and 8C , including an arrow head pointing at the most dorsal NMJ and arrow indicating somata of sensory neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 010 Tau and Shot remain highly expressed in mature neurons ( see later in Figure 6 ) , and we tested whether they are required also for synapse maintenance . For this , we used the GAL4-UAS system to co-express previously used and validated UAS-RNAi constructs for both genes in the same neurons ( Bolkan and Kretzschmar , 2014; Subramanian et al . , 2003 ) . This strategy takes out Tau and Shot functions with some delay , due to the late onset of GAL4 expression and the persistence of Tau and Shot proteins ( Figure 3—figure supplement 1 ) . We first used this approach in cultured primary neurons , where combined knock-down of tau and shot caused no reduction in the number of Syt-labelled presynaptic sites at 3 and 18 DIV as compared to wildtype controls ( Figure 3A–B ) , indicating normal synapse development . However , at 26 DIV , Syt puncta in knock-down neurons were reduced to 41% ( Figure 3A–B ) , which was comparable to the shot-tau double-mutant phenotype at 2 DIV ( Figure 1 ) . At all time points ( i . e . 3 , 18 and 26 DIV ) , there were no measurable changes in axonal length nor in branch number when compared to control neurons ( Figure 3—figure supplement 4 ) clearly indicating that the strong reduction in Syt positive synapses in 26 DIV knock-down neurons was not a secondary effect of morphological changes such as in number of branches and axonal length ( Figure 3A and B ) . 10 . 7554/eLife . 14694 . 011Figure 3 . Tau and Shot are required for the maintenance of synaptic markers in cultured neurons and the ageing adult fly brain . ( A ) Primary neurons at 3 DIV and 26 DIV cultured from embryos that were wildtype or jointly expressing UAS-tauRNAi and UAS-shotRNAi in all neurons driven by the pan-neuronal driver elav-Gal4 ( tauRNAi shotRNAi ) . Neurons are stained with anti-tubulin and anti-Syt; at 26 DIV , tauRNAi shotRNAi neurons display a reduction in the number of Syt puncta when compared to wildtype . ( B ) Quantification of the experiments in A , shown as the number of Syt puncta per neuron at 3 DIV , 18 DIV and 26 DIV , normalised to wildtype controls ( the number of assessed neurons is indicated in each bar; ***PMW<0 . 001; ns , not significant PMW>0 . 05 ) . ( C ) A region of Drosophila adult brains including the medulla ( delimited by dashed lines ) where Syt-GFP is expressed in dorsal cluster neurons using atonal-Gal4 , in the absence ( control ) or together with tauRNAi and shotRNAi ( tauRNAi shotRNAi ) . Brains are stained with anti-GFP at 2–5 days ( young ) and 24–29 days ( old ) after eclosion . Note that GFP-labelled synapses ( arrowheads ) are decreased in old brains upon shot and tau knock-down . ( D ) Quantification of the experiments in C , showing the normalised number of Syt-GFP-labelled puncta in old specimen per mean number of puncta in young specimens for the following phenotypes: ato-Gal4 UAS-syt-GFP alone ( control ) , co-expressing UAS-tauRNAi ( tauRNAi ) , UAS-shotRNAi ( shotRNAi ) , or both knock-down constructs ( tauRNAi shotRNAi; the number of analysed brains is indicated in each bar , ***PMW<0 . 001; ns , not significant PMW>0 . 05 ) . ( E ) Brain regions as in C , of animals expressing the membrane marker myr-tdTomato driven by ato-Gal4 revealing the morphology of the projections of dorsal cluster neurons within the medulla ; brains were from adults at 2–5 days ( young ) and 24–29 days ( old ) after eclosure , expressing myr-tdTomato either alone ( control ) or together with tauRNAi and shotRNAi ( tauRNAi shotRNAi ) . ( F ) Quantification of the experiments in E , displayed as number of branches per axon projecting into the medulla ( the number of axons analysed is indicated in each bar; ns , not significant PMW>0 . 05 ) . Scale bar: 10 µm in A and 40 µm in C and E . A statistics summary of the data shown here is available in Figure 3—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 01110 . 7554/eLife . 14694 . 012Figure 3—source data 1 . Summary of the statistics from Figure 3B , D , F . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 01210 . 7554/eLife . 14694 . 013Figure 3—figure supplement 1 . Delayed effect of RNAi mediated knock-down of Shot and Tau . ( A ) Primary neurons at 3 DIV and 25 DIV cultured from embryos that were wildtype or jointly expressing UAS-tauRNAi and UAS-shotRNAiin all neurons driven by the pan-neuronal driver elav-Gal4 ( tauRNAi shotRNAi ) . Neurons are stained with antibodies against Tau , Shot and Tubulin ( red , green and blue respectively ) ; images on the right show: a selected axon segment taken from the main image ( top ) followed by grey scale images of the separated channels for Tau ( 2nd from top ) , Shot ( 3rd from top ) and Tubulin ( bottom ) . At 25 DIV , tauRNAi shotRNAi neurons display a reduction in both Tau and Shot when compared to wildtype . ( B ) Quantification of the experiments in A , shown as mean intensity of Tau or Shot signal per neuron at 3 DIV and 25 DIV , normalised to wildtype controls ( 30–39 neurons were assessed per genotype; ***PMW<0 . 001; **PMW<0 . 01; ns , not significant PMW>0 . 05 ) . Comparative data for shot3 and tauMR22 homozygous mutant neurons are given as control , indicating low Tau background staining and incomplete knock-down of Tau at 3 DIV , but high Shot background suggesting strong or complete Shot knock-down at 25 DIV . A statistics summary of the data shown here is available in Figure 3—figure supplement 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 01310 . 7554/eLife . 14694 . 014Figure 3—figure supplement 1—source data 1 . Summary of the statistics from Figure 3—figure supplement 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 01410 . 7554/eLife . 14694 . 015Figure 3—figure supplement 2 . Schematic drawings of brain areas analysed in this study . Dorso-fronto-lateral view onto a schematic adult CNS composed of the brain and ventral nerve cord ( vNC ) . Beige areas indicate some synaptic areas , in particular the ventral nerve cord neuropile ( Np ) and the optic lobes ( OL ) composed of lamina ( 1 ) , medulla ( 2 ) , lobula ( 3 ) and lobula plate ( 4 ) . DCN neurons project to the optic lobe of the contralateral brain half where they branch out in a layered fashion . The blue stippled frame and image inset embox the area ( rotated 90 degrees counterclockwise ) shown in Figures 3E and 8E without and with synaptic markers ( yellow dots ) , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 01510 . 7554/eLife . 14694 . 016Figure 3—figure supplement 3 . Loss of function mutations in shot and tau induce morphological changes . ( A–D ) Representative examples of the shape of primary neurons at 2 DIV obtained from embryos that were wildtype ( wt , A ) , shot3 tauMR22 ( shot-/- tau-/- B ) , shot-/- tau-/- with elav-Gal4 driven expression of UAS-unc-104 ( shot-/- tau-/- UAS-unc-104 , C ) and shot3 tauMR22 wnd2 ( shot-/- tau-/- wnd-/- D ) . ( E–F ) Quantification of morphological parameters of experiment in ( A–D ) , including the length of axons ( E ) and the number of branches ( F ) , ( the assessed numbers of neurons are indicated in each bar , ***PMW<0 . 001; ns , not significant PMW>0 . 05 ) . Note that wnd2 ( wnd-/- ) restored the number of branches in shot-tau mutant neurons , suggesting that JNK not only mediates synapse regulation but also morphogenetic processes downstream of the Shot-Tau deficiency . A statistics summary of the data shown here is available in Figure 3—figure supplement 3—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 01610 . 7554/eLife . 14694 . 017Figure 3—figure supplement 4 . RNAi-mediated knock-down of Shot and Tau has no effect on axonal length and branch number . ( A–B ) Representative examples of the shape of primary neurons at 3 , 18 and 26 DIV obtained from embryos that were wildtype ( wt , A ) or jointly expressing UAS-tauRNAi and UAS-shotRNAi in all neurons driven by the pan-neuronal driver elav-Gal4 ( tauRNAi shotRNAi , B ) . ( C–D ) Quantification of morphological parameters of experiment in ( A–B ) , including the length of axons ( C ) and the number of branches ( D ) , ( the assessed numbers of neurons are indicated in each bar , ns , not significant PMW>0 . 05 ) . A statistics summary of the data shown here is available in Figure 3—figure supplement 4—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 01710 . 7554/eLife . 14694 . 018Figure 3—figure supplement 4—source data 1 . Summary of the statistics from Figure 1—figure supplement 3C and D . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 018 To assess roles in synapse maintenance also in vivo in the ageing brain , we used atonal-Gal4 ( ato-Gal4 ) to drive gene expression in dorsal cluster ( DC ) neurons of the adult brain ( Zschätzsch et al . , 2014 ) ( illustrated in Figure 3—figure supplement 2 ) . In these experiments , we expressed GFP-tagged Synaptotagmin ( Syt-GFP ) to label synapses , either alone or together with shotRNAiand/or tauRNAi . We compared young flies at 2–5 days after eclosion with old flies at 24–29 days at 29oC . We found that the number of Syt-GFP labelled synapses in DC neurons decreased to 34% in aged specimens expressing both shotRNAi and tauRNAi when normalised to young flies of the same genotype ( Figure 3C–D ) . This age-dependent decrease in synapse numbers did not occur in control flies ( Figure 3C–D ) , and single knock-down of either shot or tau only showed a non-significant tendency to lose synapses over time ( Figure 3D ) . Notably , aged double knock-down DC neurons had no reduction in the number of axonal branches ( as assessed with the myr-tdTomato membrane marker; Figure 3E–F ) , indicating that also in vivo precocious synapse decay was not due to axonal loss . From our studies in culture , in embryos and in the adult brain , we conclude that Tau and Shot are required for synapse development during early stages , and for synapse maintenance in ageing neurons , where their combined deficiency causes precocious synapse loss . Synapse formation and maintenance require that synaptic proteins synthesised in the soma are actively transported through the axon towards the distant presynaptic sites . In Drosophila primary neurons , transport of endogenous synaptic proteins already starts at 8 hr in vitro ( HIV ) when synaptic proteins appear as dotted patterns along axons and in growth cones ( Sánchez-Soriano et al . , 2010 ) ( Figure 4A–B ) . This is similar in rat hippocampal neurons ( Bonanomi et al . , 2005 ) . Already at this early stage , shot-tau double mutant neurons display a strong decrease in synaptic proteins in growth cones and axons ( Figure 4A–B ) , indicating potential intracellular transport defects . 10 . 7554/eLife . 14694 . 019Figure 4 . Intracellular transport of synaptic proteins is defective in shot-tau mutant neurons . ( A ) Primary Drosophila neurons at 8HIV , obtained from embryos that were wildtype ( wt ) and shot-tau ( shot-/- tau-/- ) stained with antibodies against pan-neuronal HRP ( magenta ) , Syt ( green ) or nSyb ( green ) ; nSyb and Syt are reduced in the growth cones ( open versus white arrow heads ) but enriched in cell bodies ( open versus white arrows ) of shot-tau mutant neurons . ( B ) Quantification of the experiments from A , given as the number of nSyb or Syt puncta in axons and growth cones; the number of analysed neurons is given in the bars ( ***PMW<0 . 001 ) . ( C ) Quantification of various transport parameters generated from live movies of axons of wildtype or shot-tau mutant neurons ( shot-/- tau-/- ) at 8 HIV with elav-Gal4 driven expression of UAS-Syt-GFP . Axonal anterograde and retrograde velocities show only subtle or no alteration in the axons of shot-tau neurons . On the contrary , the numbers of vesicles in axons of shot-tau neurons are sharply decreased and increased in the somata ( **PMW<0 . 01; *PMW<0 . 05; ns , not significant PMW>0 . 05 ) . ( D ) Magnified views of the somata from primary Drosophila neurons at 2 DIV , obtained from wildtype ( wt ) and shot-tau mutant embryos ( shot-/- tau-/- ) , co-stained with antibodies against Syt . To document the protein content within cell bodies , several z stacks per neuron were obtained and fused as maximal projection; the cell bodies show higher levels of Syt in shot-tau mutant neurons as compared to wildtype ( number of assessed cells is indicated in the bars , average staining intensity normalised to wildtype; ***PMW<0 . 001 ) . ( E ) The dorsal peripheral nervous system ( PNS ) of wildtype andshot-tau embryos at late stage 16 ( stages according to ) ( Campos-Ortega and Hartenstein , 1997 ) stained for Syt ( green ) , FasII ( red ) and the pan-neuronal nuclear marker Elav ( blue ) . The nascent NMJ at the tip of the inter-segmental motornerve ( red ) in wildtype contains high levels of Syt ( white arrow ) whereas the somata of sensory neurons ( blue and grey in insets ) contain low levels ( open arrow ) ; in shot-tau homozygous embryos the somata of sensory neurons have high levels of Syt ( arrow and inset ) , whereas there is only little staining at the nerve tip ( open arrowhead ) . Scale bars: 10 μm in A , 5 μm in D and 5 μm in E . A statistics summary of the data shown here is available in Figure 4—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 01910 . 7554/eLife . 14694 . 020Figure 4—source data 1 . Summary of the statistics from Figure 4B–D . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 020 To study intracellular transport , we analysed the dynamics of Syt-GFP using live imaging of neurons at 8 HIV . In shot-tau mutant neurons , the percentages of anterograde and retrograde displacements and retrograde velocities of Syt-GFP containing vesicles were not affected and anterograde velocities were only slightly increased . In contrast , the number of Syt vesicles within the axon showed a sharp decrease to ~40% in shot-tau mutant neurons when compared to controls ( Figure 4C ) . Notably , this decrease in axonal vesicles is accompanied by an increase in the number of somatic Syt-GFP puncta to ~159% ( Figure 4C ) . Similarly , endogenous Syt was increased in somata of shot-tau mutant neurons , both in culture and in in vivo ( Figure 4D–E ) . These phenotypes in shot-tau mutant neurons suggested aberrant intracellular trafficking of Syt-containing vesicles , potentially due to a road block in the soma . Type 3 kinesins are the predominant motors driving axonal transport of synaptic proteins ( Hirokawa et al . , 2010 ) . This is also the case for the Drosophila homologue Unc-104 ( also called Imac ) ( Pack-Chung et al . , 2007 ) . We found that unc-104170 null mutant primary neurons at 2 DIV have a vast reduction of Syt-stained synapses ( Figure 5A ) . This phenotype is strikingly similar to the one observed in shot-tau mutant neurons , and suggested that Shot-Tau might regulate Unc-104 function . 10 . 7554/eLife . 14694 . 021Figure 5 . Defects in kinesin-3 function mediate synaptic deficits in shot-tau mutant neurons . Shot and Tau interact with Unc-104 and regulate its subcellular distribution . ( A ) Primary Drosophila neurons at 2 DIV , obtained from embryos which were wildtype , homozygous for unc104imac170 ( unc104-/- ) , or triple-heterozygous for shot3 tauMR22 unc104imac170 mutations ( shot+/- tau+/-unc104+/- ) , co-stained with antibodies against tubulin ( Tub , magenta ) and Syt ( green ) . The graph shows the quantification of the data including also unc104-/+ and shot-/- tau-/- controls . ( B ) Quantification of Syt puncta in two day old neurons , obtained from embryos that were wildtype or shot-/- tau-/- with elav-Gal4 driven expression of UAS-unc-104 ( compare Figure 1D ) . ( C ) The dorsal peripheral nervous system ( PNS ) and the central nervous system ( CNS ) of wildtype and shot3 tauMR22 unc104imac170 triple heterozygous embryos at late stage 16 ( stages according to Campos-Ortega and Hartenstein , 1997 ) stained for Syt ( green ) , FasII ( red in upper panel ) and the pan-neuronal nuclear marker Elav ( blue ) ; for illustration of the imaged tissue see Figure 2-figure supplement 1 . The nascent NMJ at the tip of the inter-segmental motornerve ( red in upper panels ) in wildtype contains high levels of Syt ( arrowheads ) whereas the somata of sensory neurons ( blue; demarcated by dashed lines ) contain low levels ( open arrows ) ; in shot3 tauMR22 unc104imac170 triple heterozygous embryos the somata of sensory neurons have high levels of Syt ( arrows ) , whereas there is only little staining at the nerve tip ( open arrowhead ) . In the ventral nerve cord of wildtype ( lower panels ) , Syt is confined to the neuropile ( synapse containing CNS compartment; arrowheads ) and excluded from the cortex ( compartment with the cell bodies of inter- and motorneurons ) ; in the ventral nerve cord of shot3 tauMR22 unc104imac170 triple heterozygous embryos , there are segmental groups of cell bodies displaying higher Syt levels ( arrows ) . ( D ) Primary Drosophila neurons at 2 DIV , obtained from wildtype ( wt ) and tau-shot mutant embryos , stained with antibodies against pan-neuronal HRP ( magenta ) and Unc-104 ( green ) ; Unc-104 in distal axon segments ( emboxed and magnified in insets ) is enriched in wildtype but much weaker in shot-tau mutant neurons ( chevrons indicate neuronal somata ) . Data were quantified as average intensity of Unc-104 at the distal end of the axon divided by the average intensity at the soma . ( E ) Upper and lower panels show the same locations of late stage 16 embryos as shown in C , but taken from wildtype and shot-tau mutant embryos , stained for FasII ( magenta ) and Unc-104 ( green ) . Note the stark decrease of Unc-104 at the end of motor nerves ( open versus white arrow heads ) and the unusual accumulations of Unc-104 in the cell bodies of sensory neurons as well as in the CNS cortex in shot-tau embryos ( open versus white arrows ) . In all graphs , the number of assessed neurons is indicated in each bar; ***PMW<0 . 001; *PMW<0 . 05; ns , not significant PMW>0 . 05; scale bars: 18 μm in A , 5 μm in C/PNS , 35 μm in C/CNS , 15 μm in D and E/PNS , 35 μm in E/CNS . A statistics summary of the data shown here is available in Figure 5—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 02110 . 7554/eLife . 14694 . 022Figure 5—source data 1 . Summary of the statistics from Figure 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 02210 . 7554/eLife . 14694 . 023Figure 5—figure supplement 1 . Expression of Unc-104 rescues synaptic defects in aged adult brains . ( A ) A region of Drosophila adult brains including the medulla ( see illustration in Figure 3—figure supplement 2 ) ; UAS-nSyb-GFP is expressed in dorsal cluster neurons using atonal-Gal4 , either alone ( control ) , together with tauRNAi and shotRNAi ( tauRNAi shotRNAi ) or together with tauRNAi , shotRNAi and UAS-unc-104 ( tauRNAi shotRNAi UAS-unc-104 ) . Brains are shown at 2–6 days ( young ) and 26–30 days after eclosion ( old ) ; GFP-labelled synapses are decreased in old brains with shot-tau knock-down when compared to controls , and this effect is rescued by the expression of Unc-104 . ( B ) Quantification of experiments in A , shown as number of GFP-labelled synapses in old specimen per mean number of GFP-labelled synapses in young specimens of the respective genotype ( number of analysed brains is indicated in the bars; ***PMW<0 . 001; **PMW<0 . 01; ns , not significant PMW>0 . 05 ) . Scale bars: 5 μm in A . A statistics summary of the data shown here is available in Figure 5—figure supplement 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 02310 . 7554/eLife . 14694 . 024Figure 5—figure supplement 1—source data 1 . Summary of the statistics from Figure 5—figure supplement 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 024 To test this hypothesis , we performed genetic interaction studies . We found that primary neurons stained for Syt at 2 DIV and heterozygous for all of the three genes ( shot-/+ unc-104-/+ tau-/+ ) displayed significant reduction in the number of Syt-stained synapses when compared to heterozygous condition of the unc-104 or shot-tau mutant alleles alone ( Figure 5A ) . Also triple-heterozygous mutant embryos at late stage 16 displayed reduced Syt staining at neuromuscular terminals , but increased staining in the cell bodies of CNS and sensory neurons ( Figure 5C; see Figure 2—figure supplement 1 for a schematic drawing of the embryonic NMJ and CNS ) . Therefore , unc-104170 mutant , shot-tau mutant , and triple-heterozygous mutant neurons all show similar phenotypes , both in culture and in vivo , suggesting a functional link between these three proteins . Type 3 kinesins are anterograde motor proteins that move towards axon tips in mouse neurons ( Niwa et al . , 2013 ) , and we also find Drosophila Unc-104 to be distally enriched in the axons of primary neurons at 2 DIV and in embryonic motorneurons in vivo ( Figure 5D–E ) . In mouse , this distal localisation was shown to be suppressed when blocking kinesin-3 mobility ( Niwa et al . , 2013 ) . Also in shot-tau mutant neurons in culture and in vivo , Unc-104 localisation in distal axons is reduced whereas levels in the somata are increased ( Figure 5D–E ) , suggesting that insufficient amounts of Unc-104 move from the somata into axons . To test whether diminished Unc-104 levels in axons are the cause for the synaptic defects in shot-tau mutant conditions , we over-expressed Unc-104 , which fully restored synapse numbers in shot-tau mutant neurons at 2 DIV ( Figures 5B versus 1D ) . Notably , Unc-104 over-expression in shot-tau mutant neurons achieved this rescue of synapses in neurons with significantly less axonal branches ( Figure 3—figure supplement 3F ) , clearly demonstrating that both features are regulated independently of each other . We next examined whether Unc-104 plays comparable roles also during synapse maintenance in the ageing brain . We used the ato-Gal4 driver ( Zschätzsch et al . , 2014 ) and co-expressed Unc-104 together with shotRNAi and tauRNAi . To label synapses we expressed the presynaptic marker neuronal Synaptobrevin-GFP ( nSyb-GFP , due to technical reasons the use of nSyb-GFP was more convenient than Syt-GFP ) . Consistent with our previous findings with Syt-GFP ( Figure 3C–D ) , also nSyb-GFP revealed age-dependent synapse reduction upon shotRNAi and tauRNAi expression , clearly confirming our previous data ( Figure 5—figure supplement 1 and 8E–F ) . When Unc-104 was co-expressed , synapse reduction was clearly rescued ( Figure 5—figure supplement 1 ) . Taken together , our data are consistent with a model where Shot-Tau loss generates a road block which inhibits Unc-104 translocation from the soma into axons , causing synaptic defects at developmental stages and in ageing neurons . To address the mechanistic links from loss of Shot-Tau to aberrant transport and synaptic defects , we focussed on microtubules . Shot localises along microtubules , and shot mutant neurons treated with the microtubule-destabilising drug nocodazole display unusual gaps in their axonal microtubule bundles ( Figure 6B–C ) ( Alves-Silva et al . , 2012; Sanchez-Soriano et al . , 2009 ) . Tau also localises along microtubules ( Figure 6A , Video 1 ) , and tauMR22 mutant neurons likewise displayed axonal microtubule gaps upon nocodazole treatment which could be rescued with targeted expression of Tau ( Figure 6B–C ) . Both , shot and tau mutant neurons treated with nocodazole displayed on average one gap per axon . This number is significantly increased to ~3 gaps in shot-tau mutant neurons ( Figure 6B–C ) , demonstrating that Tau and Shot share a common function in microtubule stabilisation . 10 . 7554/eLife . 14694 . 025Figure 6 . Microtubule instability mediates aberrant JNK signalling and synaptic defects . ( A ) Live imaging of Drosophila neurons at 2 DIV , obtained from embryos carrying tau304 ( a protein trap line where the endogenous tau gene is genomically tagged with GFP ) and the microtubule binding protein Jupiter-Cherry . Endogenous Tau ( in magenta ) is observed in a pattern reminiscent of microtubules , and colocalises with Jupiter ( shown in green ) . ( B ) Axons of Drosophila neurons at 6 HIV with the following genotypes: wildtype ( wt ) , shot3 ( shot-/- ) , tauMR22 ( tau-/- ) , tau rescue ( tau-/- UAS-tau ) and shot-tau ( shot-/- tau-/- ) . Neurons were treated for 2 . 5 hrs with vehicle ( DMSO ) or 20 µM nocodazole , fixed and stained with anti-Tubulin ( Tub , magenta and white ) and phalloidin ( Pha , green ) . Only shot3 , tauMR22 , and shot-tau double mutant displayed gaps in their axonal microtubule bundles upon nocodazole treatment , but not wildtype and tau mutant embryos with Tau rescue . ( C ) Quantification of the experiments in B , indicated as the number of breaks in the microtubule staining per axon ( number of analysed neurons is indicated in bars; ***PMW<0 . 001; **PMW<0 . 01; ns , not significant PMW>0 . 05 ) . ( D ) Embryonic motoraxons of wildtype and shot-tau embryos at late stage 16 treated with vehicle ( DMSO ) or 50 nM of the microtubule stabilising drug epothilone B for 3 hr and stained with FasII ( magenta ) and Syt ( green ) ; in wildtype , the nascent NMJ at the nerve tip contains high levels of Syt ( arrowheads ) ; in shot-tau embryos there is only little Syt staining at the nerve tip ( open arrowhead ) . Treatment of shot-tau embryos with 50nM epothilone B increases the levels of Syt at the tip of motornerves ( arrowheads ) . ( E ) Quantification of the experiments shown in D , measured as the average intensity of Syt at nascent NMJs and normalised to wildtype ( number or assessed NMJ is indicated in bars; ***PMW<0 . 001; ns , not significant PMW >0 . 05 ) . ( F ) Upper ( PNS ) and lower ( CNS ) panels show the same locations of late stage 16 wildtype embryos as shown in Figure 5C , stained for FasII ( magenta ) and activated phospho-JNK ( JNK-P ) ; treatment with 100 µm nocodazole for 2 hrs induced a relocation of JNK-P from nascent NMJs ( open versus white arrow heads ) to cell bodies of sensory neurons and in the CNS cortex ( white versus open arrows ) . Scale bar: 5 μm in A , 4 μm in B , 10 μm in E , 15 μm in D/PNS and 35 μm in D/CNS . A statistics summary of the data shown here is available in Figure 6—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 02510 . 7554/eLife . 14694 . 026Figure 6—source data 1 . Summary of the statistics from Figure 6C and E . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 02610 . 7554/eLife . 14694 . 027Figure 6—figure supplement 1 . Treatment of shot-tau mutant neurons with epothilone B , increases the localisation of JNK-P at axonal tips . ( A ) Embryonic motoraxons of wildtype and shot-tau embryos at late stage 16 treated with vehicle ( DMSO ) or 50 nM of the microtubule stabilising drug epothilone B for 3 hr and stained with FasII ( magenta ) and JNK-P ( green ) ; In wildtype , JNK-P is high at nerve endings ( white arrowheads ) and below detection levels in cell bodies of sensory neurons and in the CNS cortex ( open arrows ) . This pattern is inverted in shot-tau embryos where JNK-P levels are low at nerve tips ( open arrowhead ) and high in cell bodies of sensory neurons and in the CNS cortex ( white arrows ) . Treatment of shot-tau embryos with 50 nM epothilone B increases the levels of JNK-P at the tip of motornerves ( white arrowheads ) . ( B ) Quantification of the experiments shown in A , measured as the average intensity of JNK-P at nascent NMJs and normalised to wildtype ( number of assessed NMJ is indicated in bars; ***PMW<0 . 001 ) . Scale bars: 15 μm in PNS panels and 35 μm in CNS panels . A statistics summary of the data shown here is available in Figure 6 —figure supplement 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 02710 . 7554/eLife . 14694 . 028Figure 6—figure supplement 1—source data 1 . Summary of the statistics from Figure 6—figure supplement 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 028 To test whether their roles in microtubule stabilisation and synapse regulation are linked , we treated shot-tau mutant embryos at early stage 16 for 3 hr with the microtubule-stabilising drug epothilone B ( Goodin et al . , 2004 ) . We found a significant rescue of Syt levels at motoraxonal endings , which was not observed in vehicle-treated controls ( Figure 6D–E ) . Therefore , a decrease in microtubule stability is a likely cause for defective transport of synaptic proteins in shot-tau mutant neurons . It could be argued that Shot-Tau dependent microtubule stabilisation directly regulates processive advance of kinesins in axons ( see Discussion ) , yet the rather normal transport dynamics we observed upon live imaging in shot-tau mutant neurons clearly excluded this possibility ( Figure 4C ) . Instead , we hypothesised that microtubule aberration indirectly promotes a transport roadblock in somata . As a potential mediator , we suspected the JNK signalling pathway which is known to respond to a number of cellular stresses ( see Discussion ) . To test our hypothesis , we investigated the pattern of JNK activity , using an antibody against phosphorylated JNK ( JNK-P ) ( Langen et al . , 2013 ) . In wild type embryos at stage 16 , we found high accumulations of JNK-P at motoraxon tips and low levels in the somata of CNS and sensory neurons ( Figures 6F and 7A ) , i . e . a localisation pattern similar to that of synaptic proteins and Unc-104 ( Figure 5C–E ) . This distribution was altered in single tauMR22 or shot3 mutant embryos , showing higher levels of JNK-P in neuronal somata and lower levels at the tips of motoraxons ( Figure 7A ) . This altered pattern was intensified in shot-tau double mutant neurons ( Figure 7A ) and clearly reminiscent of the redistribution patterns observed with synaptic proteins and Unc-104 in these neurons ( Figures 4E and 5E ) . Notably , these changes in the pattern of JNK activation were reproduced when inducing microtubule stress by applying nocodazole to early stage 16 wildtype embryos ( Figure 6F ) . Complementary to this finding , treatment of shot-tau mutant neurons with the microtubule stabilising drug epothilone B , increased the localisation of JNK-P at axonal tips and reduced the aberrant localisation in somata ( Figure 6—figure supplement 1A , B ) . 10 . 7554/eLife . 14694 . 029Video 1 . Live imaging of Drosophila neurons at 2 DIV , obtained from embryos carrying tau304 ( a protein trap line where the endogenous tau gene is genomically tagged with GFP ) and the microtubule binding protein Jupiter-Cherry . Endogenous Tau ( in green ) is observed in a pattern reminiscent of microtubules , and colocalises with Jupiter ( shown in red ) . The time laps were obtained every 15 s with a 3i Marianas Spinning Disk Microscope . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 02910 . 7554/eLife . 14694 . 030Figure 7 . Activated JNK correlates with the subcellular localisation of Unc-104 and Syt . Upper ( PNS ) and lower ( CNS ) panels in A-C show the same locations of late stage 16 embryos as shown in Figure 5E , but embryos are of different genotypes and stained with different antibodies , as indicated; genotypes: wildtype ( wt ) , shot3 ( shot-/- ) , tauMR22 ( tau-/- ) , shot-tau ( shot-/- tau-/- ) , elav-Gal4 driven expression of UAS-hepac ( UAS-hepac ) ; used antibodies detect FasII ( magenta ) , Syt ( green ) , Unc-104 ( green ) , activated phospho-JNK ( JNK-P ) . ( A ) In wildtype , JNK-P is high at nerve endings ( white arrow heads ) and below detection levels in cell bodies of sensory neurons and in the CNS cortex ( open arrows ) ; this pattern is inverted in tauMR22 and shot3 mutant embryos and even stronger in shot-tau embryos , i . e . Syt is reduced at nerve endings ( open arrowheads ) and upregulated in cell bodies ( white arrows ) . ( B , C ) Artificial activation of JNK with neuronal expression of Hepac suppresses high levels of Unc-104 and Syt at nascent NMJs ( open versus white arrow heads ) and increases their levels in cell bodies ( white versus open arrows ) . Scale bars: 15 μm in PNS panels and 35 μm in CNS panels . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 030 These data suggested a cascade of events where shot-tau mediated microtubule destabilisation or stress triggers abnormal JNK activation in somata which , in turn , causes a somatic block of Unc-104 mediated synaptic transport . In strong support of this hypothesis , the three key players of this cascade , JNK-P , Unc-104 and synaptic proteins , show a striking correlation by concentrating unanimously at axon tips in wildtype , but in somata in shot-tau mutant neurons ( Figures 4E , 5E and 7A ) . To prove that JNK acts downstream of shot-tau to regulate Unc-104 , we first expressed a constitutively active variant of the MAPKK Hemipterous ( HepAC ) , a known activator of the JNK pathway ( Glise et al . , 1995 ) . In late stage 16 embryos , indiscriminate JNK activation through HepAC triggered an accumulation of Unc-104 and Syt in somata and a decrease of both proteins at axon tips ( Figure 7B–C ) . Also in primary mature neurons at 2 DIV , HepAC caused a reduction in the number of synapses to 43% ( Figure 8A–B ) . Therefore , HepAC expression mimicked the defects observed in shot-tau mutant neurons , consistent with a model where aberrant JNK pathway activation upon Shot-Tau loss causes the somatic block of Unc-104-dependent synaptic transport . 10 . 7554/eLife . 14694 . 031Figure 8 . Inhibition of the JNK pathway rescues synaptic defects in shot-tau mutant neurons . ( A ) Primary Drosophila neurons at 2 DIV , obtained from embryos of the following genotypes: wildtype ( wt ) , elav-Gal4 driven expression of UAS-hepac ( UAS-hepac ) , tauMR22 ( tau-/- ) , wnd2 ( wnd-/- ) , tau-/- with elav-Gal4 driven expression of UAS-puc ( tau-/- UAS-puc ) , tauMR22 kay2 ( tau-/- kay-/- ) and shot3 tauMR22 wnd2 ( shot-/- tau-/- wnd-/- ) , all stained with antibodies against Tubulin ( tub , magenta ) and Syt ( green ) . Insets correspond to emboxed areas and show a magnified view of the Syt staining . ( B ) Quantification of experiments in A , shown as the number of Syt puncta normalised to wildtype ( number of assessed neurons is shown in the bars , ***PMW<0 . 001; **PMW<0 . 01; *PMW<0 . 05; ns , not significant PMW>0 . 05 ) . ( C ) Inter-segmental motornerves in the dorsal area of wildtype and shot3 mutant embryos at late stage 16 , stained against FasII ( magenta ) and Syt ( green ) ; insets correspond to emboxed areas and show a magnified view of the most dorsal nascent NMJs stained for Syt; note the rescue of Syt localisation if Wnd is absent in tau-shot mutant background . ( D ) Quantification of the experiments in C , measured as the average intensity of Syt normalised to wt ( number of assessed NMJs is shown in the bars; ***PMW<0 . 001; *PMW<0 . 01 ) . ( E ) A region of Drosophila adult brains including the medulla; UAS-nSyb-GFP is expressed in dorsal cluster neurons using atonal-Gal4 , either alone ( control ) , together with tauRNAi and shotRNAi ( tauRNAi shotRNAi ) or together with tauRNAi , shotRNAi and UAS-bskDN . Brains are shown at 2–6 days ( young ) and 26–30 days at 29°C after eclosion ( old ) ; GFP-labelled synapses are decreased in old brains with shot-tau knock-down when compared to controls , and this effect is rescued by the expression of BskDN . ( F ) Quantification of experiments in E , shown as number of GFP-labelled synapses in old specimen per mean number of GFP-labelled synapses in young specimens of the respective genotype ( number of analysed brains is indicated in the bars; ***PMW<0 . 001; **PMW<0 . 01 ) . Scale bars: 5 μm in A , 10 μm in C and 40 μm in E . A statistics summary of the data shown here is available in Figure 8—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 03110 . 7554/eLife . 14694 . 032Figure 8—source data 1 . Summary of the statistics from Figure 8B , D and F . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 03210 . 7554/eLife . 14694 . 033Figure 8—figure supplement 1 . Attenuation of the JNK pathway rescues aberrant Unc-104 localisation in shot-tau mutant neurons in culture . ( A ) Primary Drosophila neurons at 2 DIV , obtained from wildtype ( wt ) and shot3 tauMR22 wnd2 mutant embryos ( shot-/- tau-/- wnd-/- ) , stained with antibodies against pan-neuronal HRP ( magenta ) and Unc-104 ( green ) ; Unc-104 in distal axon segments ( emboxed and magnified in insets ) is enriched in wildtype and in shot-/- tau-/- wnd-/- mutant neurons ( chevrons indicate neuronal somata ) . This is in contrast to shot-tau mutant neurons , in which Unc-104 in distal axon segments is much weaker ( for reference see Figure 5D ) . ( B ) Data were quantified as average intensity of Unc-104 at the distal end of the axon divided by the average intensity at the soma ( **PMW<0 . 01 ) . Scale bars: 5 μm in A . A statistics summary of the data shown here is available in Figure 8—figure supplement 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 03310 . 7554/eLife . 14694 . 034Figure 8—figure supplement 1—source data 1 . Summary statistics from Figure 8—figure supplement 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 03410 . 7554/eLife . 14694 . 035Figure 8—figure supplement 2 . Attenuation of the JNK pathway rescue aberrant unc-104 localisation in shot-tau mutant embryos . ( A ) The dorsal peripheral nervous system ( PNS ) and the central nervous system ( CNS ) of wildtype wildtype , shot3 tauMR22 and shot3 tauMR22 wnd2 mutant embryos ( shot-/- tau-/- wnd-/- ) at late stage 16 stained with FasII ( red ) , Unc-104 ( green ) and elav ( blue ) for the PNS panels and FasII ( magenta ) and Unc-104 ( green ) in the CNS panels . Note the stark decrease of Unc-104 at the end of motor nerves ( open versus white arrow heads ) and the unusual accumulations of Unc-104 in the cell bodies of sensory neurons as well as in the CNS cortex in shot-tau embryos ( open versus white arrows ) . This is in contrast to shot-/- tau-/- wnd-/- mutant neurons in which Unc-104 is increased at the end of motor nerves and decreased in the cell bodies of sensory neurons as well as in the CNS cortex . Scale bars: 15 μm in PNS panels and 35 μm in CNS panels . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 035 If our model is correct , attenuation of the JNK pathway should rescue the synaptic defects in shot-tau mutant neurons . To downregulate the JNK pathway , we used loss-of-function of the JNK activating kinase Wallenda/DLK ( wnd2 ) ( Valakh et al . , 2013 ) and overexpression of the JNK inhibiting phosphatase Puckered ( Puc ) ( Martin-Blanco et al . , 1998 ) . When combined with the tauMR22 mutation , both genetic tools for JNK downregulation fully rescued the synaptic defects in primary neurons at 2 DIV ( Figure 8A–B ) . Even more , wnd2 fully rescued synapse reduction in shot-tau double mutant neurons at 2 DIV ( Figure 8A–B ) , and Syt levels at NMJs of shot-tau mutant embryos in vivo ( Figure 8C–D versus Figure 2 ) . So far , our data suggest that JNK acts downstream of shot-tau to regulate Unc-104 . In this case , attenuation of the JNK pathway should also rescue aberrant Unc-104 localisation in shot-tau mutant neurons . Accordingly , wnd2 restored correct localisation of Unc-104 in shot-tau double mutant neurons at 2 DIV ( Figure 8—figure supplement 1 ) , and in shot-tau mutant embryos in vivo ( Figure 8—figure supplement 2B versus Figure 5D , E ) . Having confirmed JNK as the essential mediator of shot-tau synaptic defects , we tested whether it acts through the canonical pathway involving the AP1 transcription factor ( Ciapponi and Bohmann , 2002 ) , or by phosphorylating other targets in the cytoplasm . For this , we used a well established LOF mutant allele of the kayak/c-fos gene ( kay2 ) which removes one constituent of the AP1 heterodimer and mimics various known JNK mutant phenotypes ( Ciapponi and Bohmann , 2002 ) . Unlike wnd2 or Puc overexpression , the kay2 mutation failed to rescue the synaptic phenotypes of tauMR22 in primary neurons ( Figure 8A–B ) . This strongly suggests that the JNK pathway inhibits synaptic transport by acting independently of AP1 dependent transcription . In conclusion , the JNK pathway is both required and sufficient to mediate between Shot-Tau loss and their downstream synaptic phenotypes in developing neurons by causing a transport roadblock , and this likely occurs through phosphorylating cytoplasmic targets in the soma . To test whether JNK plays comparable roles also during synapse maintenance in the ageing brain , we used the ato-Gal4 driver ( Zschätzsch et al . , 2014 ) and co-expressed a dominant negative variant of the Drosophila JNK homolog Basket ( bskDN ) ( Adachi-Yamada et al . , 1999 ) together with shotRNAi , tauRNAi and nSyb-GFP . We found that co-expression of bskDN was able to rescue the synapse reduction phenotype ( Figure 8E , F ) , thus confirming JNK as a mediator between the effects of shot-tau and precocious synapse decay also in ageing neurons ( summarised in Figure 9 ) . 10 . 7554/eLife . 14694 . 036Figure 9 . Schematic model of proposed function for Tau and Shot . ( A ) Neurons illustrating different phenotypes: in wildtype neurons ( wt ) , microtubules are stable ( green line ) and levels of Unc-104 ( red square ) , synaptic markers ( yellow dots ) and p-JNK ( turquoise background ) are high in axon tips; in shot-tau mutant neurons , microtubules are unstable ( stippled green lines ) , and the above listed proteins accumulate in cell bodies ( soma ) ; ubiquitous activation of JNK ( ↑JNK ) causes similar somatic accumulation of Unc-104 and synaptic markers; down-regulation of JNK ( ↓JNK ) rescues the shot-tau mutant phenotypes . ( B , C ) Schematic representation of the underlying mechanisms: In wildtype neurons ( B ) , Unc-104 is activated ( Unc* ) and mediates axonal transport of synaptic proteins ( yellow arrow ) to the axon tip , where we propose ( ? ) that active JNK inhibits Unc-104 , thus releasing its cargo for synaptic incorporation . In shot-tau mutant neurons ( C ) , unstable microtubules cause upregulation of JNK in the soma , thus inhibiting Unc-104 and trapping it as well as its cargo proteins in the soma . DOI: http://dx . doi . org/10 . 7554/eLife . 14694 . 036
The aim of our studies was to understand the role of endogenous Tau in neurons with particular attention to synapses . This effort was essentially aided by our finding that Tau and Shot are functionally redundant , and the subsequent incorporation of Shot into our studies . The robust phenotypes of shot-tau double-mutant neurons enabled us to demonstrate roles of Shot-Tau during the formation and maintenance of pre-synaptic sites in axons , and unravel the underlying mechanistic cascade which involves three major steps . Firstly , the absence of Shot-Tau causes microtubule destabilisation . Secondly , this cytoskeletal stress causes aberrant JNK activity patterns with upregulation in somata and downregulation at axon tips . Thirdly , aberrant JNK activation leads to a somatic roadblock for kinesin-3 mediated transport , thus inhibiting the delivery of synaptic proteins and eventually causing synapse loss . Depending on whether the functions of Tau and/or Shot are removed during development or ageing , either the formation or the maintenance of synapses are affected , respectively ( Figure 9 ) . Our model explaining the function of Tau and Shot in synapse establishment and maintenance by regulating intracellular transport , is supported by loss- and gain-of-function experiments , genetic interactions and cross-rescue experiments . The initial finding that shot-tau mutant neurons had reduced branch numbers , could have suggested that defects on synapse numbers is indirect . However , experiments with double knock-down in culture and in the adult brain clearly showed strong synapse reduction whilst maintaining normal branch patterns , and Unc-104 rescued synapse reduction in shot-tau mutant neurons without major increases of the branch pattern in these neurons . These results clearly demonstrate that changes in neuronal morphology are not the cause of changes in synapse number . Notably , the synaptic function of Tau described here for Drosophila might be conserved in higher animals or humans , since also aged Tau knock-out mice develop a reduction of synaptic proteins from the hippocampus ( Ma et al . , 2014 ) . Our findings provide potential new mechanistic explanations for various tau related brain disorders . For example , microdeletions in the region of MAPT ( the human tau gene ) cause intellectual disability ( Sapir et al . , 2012 ) , and Tau's synapse-promoting roles may well contribute to this pathology . Furthermore , various tauopathies are characterised by precocious pathological loss of synapses . Our data suggest that loss of tau could lead to defective synapse maintenance and eventually synapse loss . For example , a prominent group of dementias which lacks distinctive histopathology ( DLDH ) are characterised by the loss of Tau ( Zhukareva et al . , 2001 ) . Further tauopathies including Alzheimer disease , typically involve hyper-phosphorylation and aggregate formation of Tau ( Hernández and Avila , 2007; Williams , 2006 ) . In this scenario , there are two parallel , non-exclusive modalities through which Tau can cause pathology . Firstly , detached hyper-phosphorylated tau attains gain-of-function roles in the cytoplasm damaging neurons through a number of mechanisms ( Morris et al . , 2013 ) . Secondly , hyper-phosphorylation of tau causes a loss-of-function condition by depleting Tau from microtubules . However , since Tau knock-out mouse models mostly failed to show significant phenotypes and the neuronal functions of endogenous tau remain little understood , the pathological importance of Tau loss from microtubules has been marginalised ( Morris et al . , 2013 ) . Our results now re-emphasise the notion that loss of Tau from microtubules could contribute to neurodegenerative pathology and deliver mechanistic explanations . To unravel pathomechanisms caused by the loss of Tau , we mostly used combined depletion of Shot and Tau , which gave us strong phenotypes , ideal for short-term experimental approaches . However , we found similar , yet milder phenotypes if only Tau was depleted , suggesting that the mechanisms described here could well contribute to slow disease progression in tauopathies . Our discovery that spectraplakins are MAPs which functionally overlap with Tau , opens up new experimental avenues for Tau studies . So far , spectraplakins have been linked to the degeneration of sensory and autonomous neurons ( Edvardson et al . , 2012; Ferrier et al . , 2013 ) , and it remains to be elucidated whether they may have similar roles also in the brain . Our results clearly hint at this possibility . The loss of Tau and/or Shot inhibits kinesin-3 mediated transport leading to accumulation of synaptic proteins in the soma of neurons . We propose a road-block mechanism suppressing the initiation of axonal transport in somata of Shot-Tau depleted neurons , which is caused indirectly through microtubule stress and mediated by JNK ( Figure 9 ) . The involvement of microtubules in causing a transport block is supported by our experiments using microtubule stabilising and de-stabilising drugs which rescued or mimicked the shot-tau mutant phenotypes , respectively . Similarly , axonal transport defects and cognitive deficits of PS19Tg mice ( expressing the P301S mutant form of human tau ) and various other mouse and fly tauopathy models were shown to be rescued by microtubule-stabilising drugs ( Gozes , 2011; Quraishe et al . , 2013; Shemesh and Spira , 2011; Zhang et al . , 2012 ) , suggesting that the mechanisms we described may be conserved and relevant to disease . The somatic road-block is a novel mechanism through which the loss of Tau can interfere with the transport of synaptic proteins and provides potential explanations also for somatic accumulations of postsynaptic proteins such as PSD-95 , AMPA and NMDA receptors observed in mouse tauopathy models ( Hoover et al . , 2010; Shao et al . , 2011 ) . A likely mechanism causing a roadblock in intracellular transport could be the direct inactivation of Unc-104 or its associated adaptor proteins , for example through JNK or other kinases within its pathway . This mode of regulation has a clear precedent in kinesin-1 and its adaptor Jip which are directly phosphorylated by JNK leading to transport inhibition ( Stagi et al . , 2006 ) . Unfortunately , our extensive attempts to co-immunoprecipitate JNK and Kinesin-3 were unsuccessful ( data not shown ) , leaving open for now the exact molecular mechanism . We propose that aberrant JNK activation downstream of microtubule destabilisation or stress is the ultimate cause for the defective delivery of synaptic proteins in Tau and/or Shot loss of function . Also in mouse , microtubule stress leads to somatic activation of the JNK pathway , suggesting this mechanism is likely to be conserved with vertebrates ( Valakh et al . , 2015 ) . The JNK pathway is emerging as a central player in neurodegenerative diseases . Its activation is prompted by various neurodegeneration risk factors including oxidative stress , inflammation , and ageing ( Lotharius et al . , 2005; Valakh et al . , 2015 ) . Furthermore , JNK is activated in AD patients ( Coffey , 2014 ) and in several AD models where it triggers progression of the pathology ( Sclip et al . , 2014 ) . The new link between Tau/spectraplakins , JNK and synapses we propose here , is therefore likely to provide mechanistic explanations for synaptic pathology observed in AD and other tauopathies . We have delivered an important conceptual advance by revealing a new mechanistic cascade which can explain synaptic decay as the consequence of Tau loss from microtubules . Furthermore , we identified a previously unknown functional redundancy with spectraplakins as a promising new avenue for research on Tau . Our findings emphasise that Tau detachment from microtubules can be an important aspect contributing to the pathology of tauopathies in parallel to roles of hyper-phosphorylated Tau in the cytoplasm . Synaptic decay , axonal transport and alterations in the JNK pathway are emerging as central players in a wider range of adult-onset neurodegenerative diseases , and here we have aligned these factors into a concrete mechanistic cascade .
The following fly stocks were used: the Gal4 driver lines sca-Gal4 ( Sánchez-Soriano et al . , 2010 ) , elav-Gal4 ( 3rd chromosome ) ( Luo et al . , 1994 ) and ato-Gal4 ( Zschätzsch et al . , 2014 ) ; the mutant alleles Df ( 3R ) tauMR22 ( Bolkan and Kretzschmar , 2014; Doerflinger et al . , 2003 ) , shot3 ( Kolodziej et al . , 1995 ) , unc-104imac170 ( courtesy of Dr . T . Schwarz ) ( Pack-Chung et al . , 2007 ) , wnd2 ( Collins et al . , 2006 ) and kay2 ( Ciapponi and Bohmann , 2002 ) ( the latter two courtesy of S . Sweeney ) ; the UAS lines UAS-tau-GFP ( Doerflinger et al . , 2003 ) , UAS-shot-GFP ( Alves-Silva et al . , 2012; Sanchez-Soriano et al . , 2009 ) . tauGD25023 ( UAS-tauRNAi; Vienna Drosophila RNAi Center , Austria ) ( Bolkan and Kretzschmar , 2014 ) , UAS-shotRNAi ( Subramanian et al . , 2003 ) , UAS-syt-GFP ( 3rd and 2nd chromosome , Bloomington Stock Center ) , UAS-nSyb-GFP ( Bloomington Stock Center ) , UAS-tdTomato ( Zschätzsch et al . , 2014 ) , tau304 ( Bloomington Stock Center ) , Jupiter-Cherry ( Bloomington Stock Center ) , UAS-Hep-ac ( Glise et al . , 1995 ) , UAS-bskDN ( Adachi-Yamada et al . , 1999 ) and UAS-puc ( Martin-Blanco et al . , 1998 ) ( the latter five fly stocks courtesy of B . Hassan ) . Lethal fly stocks were kept over balancers carrying twist-Gal4 and UAS-GFP constructs ( Halfon et al . , 2002 ) , and combinations of mutant alleles and transgenic constructs were generated using conventional genetic crosses ( Prokop , 2013 ) . The generation of primary neuronal cell cultures was described previously ( Prokop et al . , 2012; Sánchez-Soriano et al . , 2010 ) . In brief , to generate Drosophila primary cultures , neurons were extracted from stage 11 embryos ( Campos-Ortega and Hartenstein , 1997 ) . Whole embryos were treated for 1 min with bleach to remove the chorion , sterilized for ~30 s in 70% ethanol , washed in sterile Schneider’s/FCS , and eventually homogenized with micro-pestles in 1 . 5 ml tubes containing about 21 embryos per 100 μl dispersion medium . This was followed by 4–5 min incubation in dispersion buffer containing collagenase and dispase at 37°C , followed by a wash in sterile Schneider’s/FCS and eventually resuspension in the final volume of Schneider’s medium . Cells were plated onto coverslips coated with 0 . 5 mg/ml Concanavalin A ( Sigma ) and kept as hanging drop cultures in air-tight special culture chambers ( Küppers-Munther et al . , 2004 ) usually for 8 hr , 2–3 , 18 or 26 days at 26°C . Dilutions of the MT destabilising drug nocodazole ( 20 μM; Sigma ) in Schneider’s medium were prepared from stock solutions in DMSO . For controls , equivalent concentrations of DMSO were diluted in Schneider’s medium . Stage 16 embryos were dissected flat in Dulbecco's Phosphate Buffered Saline ( Budnik et al . , 2006 ) and cultured for several hours in Schneider’s medium with or without drugs . Dilutions of the microtubule destabilising drug nocodazole ( 20 μM; Sigma ) and the microtubule stabilizer epothilone B ( 50 nM; Sigma ) in Schneider’s medium were prepared from stock solutions in DMSO . For controls , equivalent concentrations of DMSO were diluted in Schneider’s medium . Primary fly neurons were fixed in 4% paraformaldehyde ( PFA ) in 0 . 1 M phosphate buffer ( PB; pH 7–7 . 2 ) for 30 min at room temperature ( RT ) . Stage 16 embryos were dissected flat in Dulbecco's Phosphate Buffered Saline ( Budnik et al . , 2006 ) and fixed with 4% PFA for 30 min . Adult fly brains were dissected in Dulbecco's Phosphate Buffered Saline and fixed with 4% PFA for 15 min . Antibody staining and washes were performed with Phosphate Buffered Saline supplemented with 0 . 3% Triton X-100 . Staining reagents: anti-Tubulin ( clone DM1A , mouse , Sigma; alternatively , clone YL1/2 , rat , Millipore Bioscience Research Reagents ) ; anti-FasII ( clone ID4 , mouse , DSHB , RRID: AB_532376 ) ; anti-GFP ( goat , Abcam RRID: AB_305643 ) ; Cy3/FITC-conjugated anti-HRP ( goat , Jackson ImmunoResearch ) ; anti-Syn ( SYNORF1 3C11 , mouse , DSHB , RRID:AB_528479 ) ; anti-Brp ( DSHB , RRID:AB_2314867 ) ; anti-Syt ( rabbit , was a gift from Dr . S . Sweeney ) ; anti-nSyb and anti-Unc104 ( both rabbit , were a gift of Dr . T . Schwarz ) ; anti-Elav ( rat , DSHB , RRID:AB_528218 ) ; anti-pJNK ( rabbit , pTPpY , Promega , RRID:AB_430864 ) , anti-CD2 ( mouse , AbD Serotec , RRID:AB_566608 ) , anti-dTau ( Nick Lowe ) , anti-Shot ( Talila Volk ) FITC- , Cy3- or Cy5-conjugated secondary antibodies ( donkey , purified , Jackson ImmunoResearch ) . Specimens were embedded in Vectashield ( VectorLabs ) . Standard documentation was performed with AxioCam monochrome digital cameras ( Carl Zeiss Ltd . ) mounted on BX50WI or BX51 Olympus compound fluorescent microscopes . Z-stacks of embryonic CNSs were taken with a Leica DM6000 B microscope and extracted with Leica MM AF Premier software . Z-stacks of adult fly brains were taken with a Leica DM6000 B microscope or with a 3i Marianas Spinning Disk Confocal Microscope . Using custom software written in Python and NumPy , fly brain images taken with a Leica DM6000 B microscope were individually band-pass filtered ( A trous wavelet [1][2] , linear 3x3 filter , keeping scales 1–4 ) to remove stationary background . To quantify the number of synaptic densities in mature neurons in culture and the number of vesicles containing synaptic proteins in 8h neurons in culture , we used ImageJ ( RRID:SCR_003070 ) . In detail , we used thresholding to select synaptic densities from axons of single isolated cells , followed by particle analysis . For all experiments done in parallel , identical thresholds were used . For the quantification of synapses in mature neurons in culture , we selected polarised neurons with a clear distinguishable axon , the same neurons were used to study axon length and number of branches . To quantify synaptic proteins or Unc-104 in the soma of neurons , we manually selected the area of the somata using the tubulin or HRP channel and measured the signal intensity derived from the Syt or Unc-104 channel . To measure the levels of Unc-104 at the tip of axons , we selected an area of the same size at the most distal part of axons and measured the signal intensity derived from the Unc-104 channel . To quantify synaptic proteins at the tip of embryonic motorneurons in vivo . we manually selected the area occupied by the growth cones using the FasII staining and measured the signal intensity derived from the Syt channel; the background intensity was subtracted . Images used for these measurements did not contain saturated levels . Also , to measure the number of synaptic densities in DC neurons in the medulla of the adult brain , we used thresholding to select synaptic densities followed by particle analysis . The number of branches in the medulla per DC neuron was quantified manually . To quantify MT stability upon nocodazole treatment , we counted the number of breaks in the microtubule bundle per axon . Time lapse imaging of cultured primary neurons ( in Schneider's/FCS ) was performed on a Delta Vision RT ( Applied Precision ) restoration microscope using a 100x/1 . 3 Ph3 Uplan Fl objective and the Sedat filter set ( Chroma 89000 ) . The images were collected using a Coolsnap HQ ( Photometrics ) camera . The temperature control was set to 26°C . For time lapse recording , images were taken every 2 s for 2 min . To generate transport measurements , vesicles containing fluorescently tagged Syt were tracked manually using the manual tracking plugin for ImageJ . All data are shown as mean with SEM . Statistical analyses were performed in GraphPad Prism using Mann-Whitney Rank Sum Tests ( indicated as PMW ) or Chi2 ( PChi ) , with 95% confidence intervals . The exact p-values and sample size are indicated in the figure legends . For each primary neuronal cell culture experiment ( technical replication ) , approximately 30 to 40 embryos were used . Neurons obtained from those embryos were divided and cultured in 3 to 4 independent chambers ( biological replication ) . The sample size provided corresponds to the number or neurons studied . Most experiments were performed at least 2 times ( 2 technical repeats ) meaning a minimum of 60 embryos were used , and a minimum of 6 independent culture chambers were studied . These experiments are shown in Figure 1B ( tau-/- , 4 technical replications , 11 biological replications ) , Figure 1D ( Syt , shot-/- , 3 technical replications , 9 biological replications; shot-/- tau-/- , 3 technical replications , 9 biological replications ) , Figure 1D ( Brp shot-/- , 2 technical replications , 8 biological replications; tau-/- 2 technical replications , 6 biological replications ) , Figure 3B ( all genotypes at each time point 2 technical replications , 6–8 biological replications ) , Figure 4D ( all genotypes 2 technical replications , 6 biological replications ) , Figure 5B ( 2 technical replications , 6 biological replications ) , Figure 8B ( tau-/- Uas-Puc , 3 technical replications , 8 biological repeats; tau-/- wnd-/- , 2 technical replications , 6 biological repeats; shot-/- tau-/- wnd-/- , 2 technical replications , 6 biological replications ) . Other experiments were as follows: UAS-tau rescue experiment of synaptic defects ( Figure 1B ) and microtubule stability defects ( Figure 6C ) in tau-/- were performed with 1 technical replication which included at least 30 embryos distributed in 4 and 2 independent chambers or biological replications , respectively . However , for these particular experiments we used a co-culture technique in which tau-/-control neurons were cultured alongside with tau-/- UAS-Tau neurons and therefore are subject to the same environmental variations . For shot-tau mutant neurons stained with Brp ( Figure 1D ) we used 1 technical replication which included at least 30 embryos and 3 independent biological replications . For tau-/- kay-/- and kay-/- ( Figure 8B ) , we used 1 technical replication which included at least 30 embryos and 5 independent biological replications . For the measurement of Syt and nSyb synaptic puncta in 8 HIV neurons ( Figure 4B ) , we performed 1 technical replication for each synaptic protein , which included at least 30 embryos each and 2 independent biological replications . For the measurement of Unc-104 in 2 day neurons ( Figure 5D ) we performed 1 technical replication which included at least 30 embryos and 45 neurons were measured . To account for variations in the immunohistochemistry procedure , we calculated the ratio between distal axon and soma . For the quantification of axonal transport ( Figure 4C ) , we performed 2 technical replications which included at least 30 embryos each , 2 independent biological replications from which 10–14 neurons were analysed . For measurements of Syt in the nascent embryonic NMJ ( Figure 2B and 8C ) we used at least 15 embryos per genotype ( biological replication ) and performed at least 2 technical replications . Both controls and mutant embryos were dissected and stained in the same chamber and therefore subjected to equal conditions ( Figure 2B shot-/- 3 technical replications , 30 biological replications , tau-/- 2 technical replications , 15 biological replications , shot-/- tau-/- 3 technical replications , 17 biological replications; Figure 8C shot-/- tau-/-wnd-/- 2 technical replications , and 15 biological replications ) . For treatment of embryos with epothilone B ( Figure 6E ) we used at least 12 embryos per genotype and performed 2 technical replications . Both controls and shot-tau mutant embryos were present in the same treatment chambers and therefore subjected to equal conditions . For the study of synaptic phenotypes in adult brains ( Figure 3D ) , we performed at least 3 technical replications and used a minimum of 30 brains in total per genotype . For the quantification of axonal branches in adult brains we performed at least 2 technical replications and we used a minimum of 11 brains . For rescues of synaptic phenotypes in adult brains with UAS-bskDN ( Figure 8F ) we performed 3 technical replications and used a minimum of 40 brains in total . | Nerve cells form cable-like projections called axons that connect to other nerve cells to form the nervous system . Axons carry nerve impulses in the form of electrical messages , and they pass on these messages to other cells at junctions known as synapses . Specific patterns of connections between axons allow us to coordinate our movements , feel emotions and think . In Alzheimer’s disease and other neurodegenerative conditions , synapses often decay earlier than they should , which can cause important connections between nerve cells to be lost . To be able to make and maintain synapses , nerve cells transport materials from the main body of the cell along axons to the sites where synapses form . A protein called Tau and a family of proteins called the spectraplakins are linked to neurodegenerative diseases . Changes ( or mutations ) in these proteins were known to disrupt the formation and maintenance of synapses , but it was not clear how these proteins work in this context . Voelzmann et al . studied Tau and spectraplakin during synapse formation and maintenance in fruit flies . The experiments show that both proteins stabilise tube-like structures called microtubules in axons , which provide structural support to cells . The loss of Tau or spectraplakins causes the microtubules to fall apart and triggers an internal stress signalling pathway known as the JNK pathway . Activating JNK signalling blocks the transport of synaptic materials along axons , which prevents the formation of new synapses and starves existing synapses leading to their decay . The next step is to find out whether Tau and spectraplakins play similar roles in the nerve cells of mammals , which may open up new opportunities to develop therapies for Alzheimer's and other neurodegenerative diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2016 | Tau and spectraplakins promote synapse formation and maintenance through Jun kinase and neuronal trafficking |
Ran binding protein 1 ( RanBP1 ) is a cytoplasmic-enriched and nuclear-cytoplasmic shuttling protein , playing important roles in nuclear transport . Much of what we know about RanBP1 is learned from fungi . Intrigued by the long-standing paradox of harboring an extra NES in animal RanBP1 , we discovered utterly unexpected cargo dissociation and nuclear export mechanisms for animal RanBP1 . In contrast to CRM1-RanGTP sequestration mechanism of cargo dissociation in fungi , animal RanBP1 solely sequestered RanGTP from nuclear export complexes . In fungi , RanBP1 , CRM1 and RanGTP formed a 1:1:1 nuclear export complex; in contrast , animal RanBP1 , CRM1 and RanGTP formed a 1:1:2 nuclear export complex . The key feature for the two mechanistic changes from fungi to animals was the loss of affinity between RanBP1-RanGTP and CRM1 , since residues mediating their interaction in fungi were not conserved in animals . The biological significances of these different mechanisms in fungi and animals were also studied .
Eukaryotic cells each have a nucleus which segregates the nucleoplasm and cytoplasm into two isolated compartments . Exchanges between these compartments are mainly mediated through nuclear pore complex ( NPC ) , a semi-permeable channel that allows only certain classes of molecules to pass through , for example importins and exportins , which are collectively called karyopherin proteins ( Beck and Hurt , 2017 ) . Cargo entering the nucleus must possess a nuclear localization signal ( NLS ) , which binds to an importin and enters nucleus through NPC ( Rexach and Blobel , 1995 ) . In the nucleus , the GTP-bound form of Ran ( Ras-related nuclear ) protein dissociates the importin-NLS cargo , and RanGTP-Importin is exported to cytoplasm ( Izaurralde et al . , 1997; Kutay et al . , 1997 ) . For a cargo’s nuclear export , its nuclear export signal ( NES ) forms a complex with exportin in the presence of RanGTP , and together the trimeric complex translocates to the cytoplasm through NPC ( Ullman et al . , 1997 ) . In the cytoplasm , RanGTP complexes ( with either importin or exportin-NES ) are hydrolyzed to RanGDP by GTPase-activating protein RanGAP , dissembling the complexes and recycling karyopherins and Ran for further rounds of nuclear transport ( Bischoff et al . , 1995a ) . However , importin or exportin-NES displays extremely high affinity ( in the nM range ) for RanGTP and inhibits RanGAP-facilitated RanGTP hydrolysis ( Bischoff and Görlich , 1997; Askjaer et al . , 1999 ) . Efficient hydrolysis requires Ran binding proteins containing one or more Ran binding domains ( RBDs , around 150 residues each ) to dissociate RanGTP from karyopherin prior to hydrolysis ( Beddow et al . , 1995; Bischoff et al . , 1995b; Vetter et al . , 1999 ) . In human , there are two such proteins , namely RanBP1 and RanBP2 . While the predominantly cytoplasmic RanBP1 contains one RBD , the cytoplasmic rim-attached RanBP2 has four RBDs ( Bischoff et al . , 1995b ) . These RBDs are the tightest binders of RanGTP , Kd being around 1 nM , whereas it binds to RanGDP at only approximately 10 µM affinity ( Görlich et al . , 1996; Kuhlmann et al . , 1997; Delphin et al . , 1997 ) . In addition , RanBP1 in the cytoplasm functions in the disassembly of nuclear export complexes ( also before RanGTP is hydrolyzed ) , through dissociating NES containing cargoes from the complexes ( Floer and Blobel , 1999 ) . CRM1 ( Chromosomal Region Maintenance 1 , also known as Exportin-1 ) is a major nuclear export factor that is responsible for nuclear export of a plethora of proteins containing NES sequence ( s ) ( Stade et al . , 1997 ) . In yeast ( S . cerevisiae ) , after binding to RanGTP and CRM1 , RanBP1 allosterically closes the groove , releasing CRM1 cargoes into the cytoplasm ( Koyama and Matsuura , 2010 ) . Sequence analysis indicates that in contrast with fungi RanBP1 ( which includes yeast RanBP1 ) , animal RanBP1 contains NES sequence C-terminal to its RBD ( Zolotukhin and Felber , 1997 ) ( Figure 1—figure supplement 1 ) . It is reported that the NES of human RanBP1 ( hRanBP1 ) is responsible for its cytoplasmic accumulation ( RanBP1 is a shuttling protein ) ( Zolotukhin and Felber , 1997; Künzler et al . , 2000 ) . If animal RanBP1 binds to RanGTP-CRM1 similarly as yeast RanBP1 ( yRanBP1 ) , an apparent paradox then exists: its NES binding to CRM1 is inhibited by its own RBD . Specifically , if animal RanBP1 binds to CRM1 through NES in the nucleus to prepare for nuclear export , its RBD might immediately interact with RanGTP on CRM1 ( because of proximity and high affinity ) and dissociate its own NES before animal RanBP1 is exported . One may argue that NES may play a role in recruiting animal RanBP1 to CRM1 . However , RanBP1-RanGTP-CRM1 complex displayed much higher affinity than NES-CRM1-RanGTP complex in yeast ( Maurer et al . , 2001 ) ; thus , the recruiting purpose seems unnecessary and unlikely . It should be noted that there are about ten residues between RBD and NES , which are insufficient to cover the distance ( about 70 Å ) between RBD and NES in space . Therefore in theory , the NES and RBD of animal RanBP1 would not bind to CRM1-RanGTP simultaneously . It is fascinating as to why animal RanBP1 requires an extra NES while fungi RanBP1 does not; what factor ( s ) prevents animal RBD from dissociating its own NES during its nuclear export; whether the NES of animal RanBP1 functions in cargo dissociation by direct competition with NES of cargo; and how animal RBD and NES binding to CRM1-RanGTP is regulated in time and space in cells . Intrigued by these long-standing questions , we performed biochemical , biophysical , and cellular studies on RanBP1 and related proteins . Our work not only solved those puzzles , but also discovered unexpected animal RanBP1 nuclear export and cargo dissociation mechanisms distinctive from those in the yeast .
In yeast , RanBP1 , RanGTP and CRM1 form a tight complex whereby RanBP1 forces H9 loop of CRM1 to allosterically close NES binding groove and dissociate NES . In order to visualize the mode of animal RanBP1’s binding to RanGTP-CRM1 , we first attempted to solve the crystal structure of complex formed by mouse RanBP1 ( mRanBP1 ) , human RanGTP ( hRanGTP ) and human CRM1 ( hCRM1 ) . However , although the yeast complex formed readily as expected , the equivalent complex with animal proteins hardly formed under similar conditions ( Figure 1A ) . The human and yeast Ran protein shares 83% sequence identity and are often used interchangeably ( Koyama and Matsuura , 2010; Sun et al . , 2013 ) . When RanBP1 , RanGTP and CRM1 were mixed at 5:3:1 molar ratio and passed through a size exclusion chromatography column , the yeast proteins formed stable complex and were co-eluted , but not the case for the animal proteins ( Figure 1—figure supplement 2 ) . Interestingly , a greater amount of animal complex was formed when concentration of RanGTP was increased ( Figure 1B ) . In contrast , the yeast complex was not affected by RanGTP concentration ( Figure 1B ) . It should be noted that the mammalian complex was dependent upon high RanGTP but not RanGDP concentration ( Figure 1—figure supplement 3 ) . These results suggest that mRanBP1 is somewhat different from yRanBP1 , in forming complex with RanGTP-CRM1 . To identify which region of mRanBP1 mediated the interaction with hCRM1 , we cloned GST-tagged mRanBP1 mutants with NES mutation ( termed as NESmut ) , the C terminus deletion ( ΔC ) or only the NES ( NESmRanBP1 ) . Surprisingly , both NESmut and ΔC lost binding to hCRM1 completely , in either low or high RanGTP concentration ( Figure 1C ) . In contrast , NESmRanBP1 bound to hCRM1 in the presence of RanGTP , and was outcompeted by supraphysiological NES ( Engelsma et al . , 2008 ) from the Minute Virus of Mice ( Figure 1D ) , suggesting that NESmRanBP1 is a regular NES that binds to NES groove on hCRM1 . Further , the NES of mRanBP1 is weaker than that of regular NES from PKI ( Protein kinase inhibitor ) ( Figure 1E ) . Previously , it was reported that the NES of mRanBP1 was essential for its nuclear export ( Plafker and Macara , 2000 ) . Indeed , while mRanBP1 was localized to the cytoplasm , NESmut ( or mRanBP1 in the presence of CRM1 inhibitor KPT-330 ) was significantly re-localized to the nucleus ( Figure 1F ) . Unexpectedly , yRanBP1 was exclusively localized in the nucleus ( Figure 1F , right column ) , though RanBP1 proteins are known to be cytoplasmic localized proteins ( Künzler et al . , 2000; Plafker and Macara , 2000 ) . The reason for this counter-intuitive observation will be explained in later sections . Altogether , these results conclude that the NES of mRanBP1 is necessary and sufficient for its interaction with hCRM1 in the presence of RanGTP , and this interaction is crucial for nuclear export of mRanBP1 . Previously we showed that ΔC or NESmut did not bind to hCRM1 in the presence of RanGTP ( Figure 1C ) . Relating to RanBP1’s function of CRM1 cargo dissociation , we wondered whether either ΔC or NESmut possesses cargo dissociation ability , and whether NESmRanBP1 is involved in dissociating cargo , possibly through direct competition with cargo’s NES . Pull down assay showed that ΔC or NESmut dissociated cargoes as potently as WT mRanBP1 ( Figure 2A , Figure 2—figure supplement 1 ) , suggesting that NES of mRanBP1 is dispensable for cargo dissociation . Further , if direct competition plays a role , the NES of mRanBP1 should be stronger than that of the cargo . However , our earlier result showed that NESmRanBP1 was weaker than cargo PKI’s NES ( Figure 1E ) . Taken together , we conclude that mRanBP1 does not dissociate cargo through direct competition of cargo’s NES . Since ΔC-RanGTP did not bind to hCRM1 but still dissociated cargo ( Figures 1C and 2A ) , we therefore speculated that ΔC dissociates cargo through sequestering RanGTP from the export complex , which would result in a transient , low-affinity binary complex of hCRM1 and NES cargo that automatically dissociates . Indeed , at high concentration of RanGTP , cargo dissociation activity of ΔC was fully inhibited ( Figure 2B ) , because excess of RanGTP not only saturated ( and inhibited ) mRanBP1 , but also allowed CRM1 to form complex with GST-PKI . Further , purified 1:1 complex of ΔC-RanGTP , where RBD was already saturated with RanGTP , was unable to dissociate cargo ( Figure 2C ) . These results illustrate that the Ran binding domain , but not NESmRanBP1 , is required for CRM1 cargo dissociation , through stripping RanGTP out of the nuclear export complexes . We also performed the above cargo dissociation experiments using yeast protein in parallel . In contrast to mRanBP1 , cargo dissociation ability of yRanBP1 was not inhibited by excess of Ran , and yRanBP1-yRanGTP remained active to dissociate CRM1 cargo ( Figure 2B and C ) . Further , when RanGTP is in large molar excess than CRM1 , yRanBP1 dissociated cargo as long as its concentration was higher than CRM1 concentration; in contrast , mRanBP1 dissociated cargo only at concentration that was higher than that of RanGTP ( Figure 2D ) . It is evident that yeast RanBP1 dissociates cargo through sequestering CRM1-RanGTP , distinctive from animal RanBP1’s cargo dissociation mechanism ( Figure 3A ) . Previously , it was reported that CRM1 or more generally karyopherins protect bound RanGTP from GAP mediated hydrolysis ( Bischoff and Görlich , 1997; Lounsbury and Macara , 1997 ) . We showed that indeed RanGTP hydrolysis is inhibited when complexed to CRM1 and NES ( in the absence of RanBP1 ) , both in yeast and in human . Since human proteins do not form tight trimeric CRM1-RanGTP-RanBP1 complex , it is predicted that hCRM1 would not protect RanGTP-RanBP1 from GAP hydrolysis like the yeast proteins . As expected , the rate of hydrolysis in yeast was partially inhibited with addition of yCRM1 , while there was no change of hydrolysis rate with addition of hCRM1 ( Figure 3B ) . Using semi-permeabilized cells , we further showed that when CRM1 is the rate limiting factor , nuclear export of cargo is indeed faster in the presence of human CRM1/Ran/RanBP1 than yeast proteins ( Figure 3C , D ) , possibly due to faster rate of CRM1 recycling ( by faster rate of GAP hydrolysis ) , although it does not rule out other possibilities ( will be discussed later ) . These experiments are consistent with earlier reports that RanBP1 is required for RanGAP activity , and consistent with our model that animal CRM1/Ran/RanBP1 does not form high-affinity trimeric complex as in yeast . Previously , we showed that mRanBP1 bound to hCRM1 through its NES and that ΔC did not dissociate cargo in excess of RanGTP . Since ΔC forms an extremely tight complex with RanGTP ( Figure 1C ) , we hypothesized that when RanGTP is excessive , mRanBP1 binds to one RanGTP through its RBD and binds to hCRM1 through its NES , while hCRM1 simultaneously binds to another RanGTP , forming a tetrameric complex ( RanGTP-mRanBP1 ) -hCRM1-RanGTP ( Figure 4A ) . In this model , unlike the yRanBP1 that contacts both yCRM1 and yRanGTP , animal RanBP1’s RBD only binds to RanGTP ( but not CRM1 ) . Further , in contrast to H9 loop stabilized closure of NES groove in yeast ( Koyama and Matsuura , 2010 ) , animal CRM1’s H9 loop would be shifted away from the NES groove ( as in pdb 3NC0 ) ( Fung and Chook , 2014 ) , opening its NES groove for interaction with NES of RanBP1 ( Figure 4A ) . This model is consistent with all previous results , including that the binding between mRanBP1 and hCRM1 requires excessive RanGTP ( Figure 1A , B ) , that NES is the only interacting site between mRanBP1 and hCRM1 ( Figure 1C , D , F ) , and that ΔC-RanGTP does not dissociate CRM1 cargo ( Figure 2 ) . To further validate this model , we analyzed the size exclusion peaks of yeast and animal RanBP1-RanGTP-CRM1 complexes mixed at a 2:5:1 molar ratio . The calculated RanGTP to RanBP1 molar ratio in the peak of the animal complex was approximately twice ( 2 . 05 fold ) over that of the yeast complex ( Figure 4—figure supplement 1 ) , suggesting that animal complex contains two RanGTP proteins . To differentiate the two RanGTP proteins in the tetramer , we performed a pull down assay using GST-RanGTP and a truncated form of Ran , Ran1-179 , which binds to CRM1 but does not bind to mRanBP1 due to lack of C-terminus . Clearly , Ran1-179 and hCRM1 bound to GST-RanGTP beads in the presence of mRanBP1 but not ΔC or buffer ( Figure 4B ) . The bound Ran1-179 can be explained by the formation of the proposed tetrameric complex , where GST-RanGTP binds to mRanBP1 , which further binds to hCRM1-Ran1-179 through the NES of mRanBP1 . Though ΔC binds to GST-RanGTP , the lack of NES rendered it unable to bind CRM1; therefore there is a lack of Ran1-179 band in the bound fraction . By size exclusion chromatography , hCRM1 , RanGTP , mRanBP1 and Ran1-179 formed a reasonable 1:1:1:1 complex , whereas Ran1-179 did not co-elute with CRM1 , RanGTP and RanBP1 in yeast ( Figure 4C and D ) . Further , RanGTP titration into hCRM1 and mRanBP1 by ITC produced an exothermic phase and an endothermic phase , representing the heat change from forming RanGTP-mRanBP1 and the tetrameric complex , respectively ( Figure 4E ) . In contrast , the titration of RanGTP into yCRM1 and yRanBP1 displayed merely an exothermic phase representing the formation of the trimeric complex ( Figure 4F ) . Taken together , we conclude that RanBP1 forms a tetrameric complex with CRM1 and two RanGTP proteins in animals ( when RanGTP excessive ) , where RanBP1-RanGTP binds to CRM1-RanGTP through the NES of RanBP1 . Since the RanBP1 tetrameric complex forms only in excess of RanGTP , and since the nucleus is enriched with RanGTP ( Izaurralde et al . , 1997 ) , we tried to detect the presence of nuclear tetramer using the bimolecular fluorescence complementation ( BiFC ) approach . We fused one Ran with CYFP ( C-terminal fragment of YFP ) and another Ran with NYFP ( N-terminal fragment of YFP ) , and co-transfected the two plasmids into 293 T cells . The formation of the tetrameric complex would bring the fluorescent fragments within proximity , promoting the assembly of an intact YFP ( Figure 5A ) . Please note that this assay would not differentiate which Ran fusion binds to RanBP1 ( or CRM1 ) in the tetramer , and possibly that both configurations could elicit YFP signals . GST-mRanBP1 pull down showed that these fusion Ran proteins are still functional ( Figure 5—figure supplement 1 ) . When the fusion plasmids were co-transfected , weak YFP signals were observed ( Figure 5B yellow arrows ) , suggesting possible tetramer formation . In fact , not all double fusion transfected cells emitted weak signals , possibly those cells did not have a detectable level of nuclear RanBP1 ( Figure 5B white arrows ) . To generate an appropriate negative control for this experiment , we mutated C-terminus of Ran ( A181W , P184H and L209H , named RanCmut , single mutation is insufficient to disrupt its tight binding to RanBP1 ) , so that it is still active ( able to promote CRM1-NES binding ) but no longer binds to RanBP1 , therefore unable to form RanBP1 tetramer ( Figure 5C ) . When the two RanCmut fusion constructs were co-transfected , no YFP signal was observed ( Figure 5B ) , suggesting that the weak signals that we observed earlier using RanWT are true signals . The weak signals were probably due to low level of RanBP1 in the nucleus ( Richards et al . , 1996 ) and competition by endogenous RanGTP . In order to enhance the signals that we observed , we further transfected NLSSV40-fused hRanBP1 or hNESmut ( which is the human version of NESmut ) to artificially increase the nuclear pool of RanBP1 along with the two RanWT fusion plasmids . In contrast to NLS-hRanBP1 which significantly promoted nuclear YFP formation , NLS-hNESmut inhibited the weak YFP signals observed in the absence of NLS-hNESmut ( possibly inhibiting tetramer formation by sequestering Ran fusions ) ( Figure 5B ) . This further demonstrates that NES is the only interface between mRanBP1 and CRM1 , without which the formation of tetramer is inhibited . Consistent with the fact that RanGTP is almost absent in cytoplasm ( Izaurralde et al . , 1997 ) , strong cytoplasmic YFP signals were not observed during image collection , validating that the complex forms only in excess of RanGTP . As an alternative approach , we measured the Fluorescence Resonance Energy Transfer ( FRET ) efficiency between CFP-Ran and Alexa Fluor 546 immuno-labelled Myc-Ran , while co-transfecting either RanBP1Δlinker-NLS ( forms tetramer; linker refers to residues between RBD and NES; linker was deleted to reduce distances between two Rans and enhance FRET efficiency ) or RanBP1ΔNES-NLS ( does not form tetramer ) . In agreement with the BiFC experiment , the nuclear FRET efficiency is significantly ( p<0 . 0001 ) higher when transfected with RanBP1Δlinker-NLS ( Figure 5—figure supplement 2 ) . In summary , we detected the RanBP1 tetrameric complex in ( and only in ) the nucleus of human cells . The key nuclear export and cargo dissociation mechanistic change from yeast to animals is the loss of binding between RanBP1-RanGTP and CRM1 . We then asked why animal RanBP1-RanGTP did not bind to CRM1 as the yeast proteins . Through a pull down assay to test inter-species CRM1-RanGTP-RanBP1 complex formation , we found that in contrast to yRanBP1-hRanGTP-yCRM1 complex , neither yRanBP1-hRanGTP-hCRM1 nor ΔC-hRanGTP-yCRM1 complex was formed ( Figure 6A ) . Therefore , neither hCRM1 nor ΔC is compatible for trimeric complex formation . We previously showed that yRanBP1 was unexpectedly localized in the nucleus of HeLa cells ( Figure 1F ) . This could be explained by Figure 6A: yRanBP1 does not form trimeric or tetrameric complex with human CRM1/RanGTP , thus is not exported to the cytoplasm of HeLa cells . It should be noted that human and yeast Ran proteins are not critical for trimeric complex formation . In this study , we discovered a small difference between human and yeast Ran , that is hRan-yRanBP1-yCRM1 complex has lower affinity than yRan-yRanBP1-yCRM1 ( Figure 4F; refer also to Maurer et al . , 2001 ) . This reduction in affinity agrees with pull down that hRan-yRanBP1-yCRM1 but not yRan-yRanBP1-yCRM1 could be dissembled by NES ( which is the reason for using yRan but not hRan for yeast complex in cargo dissociation experiments in Figure 2 ) . Though the affinity of Ran-yRanBP1-yCRM1 is significantly reduced with hRan , we showed that hRan-yRanBP1-yCRM1 complex formed readily by pull down ( Figure 6A ) , and this complex is competent in cellular nuclear export ( will be presented in a later section ) . The above results suggested that essential residues for trimeric complex formation in yeast CRM1/RanBP1 were not conserved in animals . When crystal structures of hRanBP1-hRanGTP ( pdb:1K5G ) and hCRM1 ( pdb:3GB8 ) were aligned onto yCRM1-hRan-yRanBP1 ( pdb:4HAT ) , several differences between RanBP1 and CRM1 of the two species were observed ( Figure 6B , C ) . First , unlike yRanBP1 ( 17 ) , hRanBP1 residues preceding its RBD domain are ordered and inserted between the closely-packed RanGTP and CRM1 ( Figure 6B ) , possibly hindering trimeric complex formation by steric clash . Second , V150 that is in close proximity with yCRM1 in yRanBP1 corresponds to A114 in hRanBP1 ( Figure 6C ) . V150 is conserved to be valine or isoleucine in fungi ( Figure 6—figure supplement 1 ) , but not conserved in hRanBP2 and RanBP1 from human , mouse , zebrafish and frog ( Figure 1—figure supplement 4 ) . The residues discussed above were mutated in NESmut and yRanBP1 to test whether they impact binding towards yCRM1 in the presence of RanGTP . Pull down results showed that either A114V or deletion of N terminus ( Del 1–37 ) on NESmut could partially rescue binding to yCRM1 ( Figure 6D , lane 2 and 4 ) . In agreement , V150A mutation in yRanBP1 partially inhibited binding of yCRM1 ( Figure 6D , lane 7 and 8 ) . In contrast , mutation of more distant residue ( A119 in NESmut mutated to yeast-equivalent S ) did not significantly affect binding . Triple mutant ( A114V , A119S and Del1-37 , labelled as TM ) rescued yCRM1 binding to a comparable level as yRanBP1 ( Figure 6D , lane 6 , 7 ) . Taken together , both the N-terminus and CRM1 interacting region of animal RanBP1 are incompatible for trimeric complex formation . Similarly , several regions of CRM1 were mutated to identify critical sequence changes that abolished the formation of trimeric complex in animals . In the yRanBP1 binding interface , side chain of the strictly conserved yCRM1 T753 is in close contact with yRanBP1 V150 ( Figure 6C ) . The corresponding residue in human is Q742 , being strictly conserved in animals . Pull down results showed that Q742T mutation partially rescued binding between hCRM1 and yRanBP1-RanGTP ( Figure 6E , lane 3 ) . Similarly , T753Q or T753G mutation in yCRM1 also reduced its binding to yRanBP1-RanGTP ( Figure 6E , lane 5–8 ) . V739A mutation in hCRM1 ( V739 equivalent residues in yCRM1 is A ) , however , did not rescue binding ( Figure 6E , lane 2 ) . Besides the yRanBP1 interaction surface discussed above , yRanBP1 binds to yCRM1 indirectly through RanGTP . Human RanGTP forms 26 hydrogen bonds with yCRM1 in yRanBP1-hRan-yCRM1 structure ( Figure 6—figure supplement 2 ) , and those residues that form hydrogen bonds are strictly conserved in yeast Ran ( Figure 6—figure supplement 3 ) . However , six hydrogen bonds are predicted to be lost if yCRM1 is replaced with hCRM1 ( Figure 6F , Figure 6—figure supplement 4 ) . Five hydrogen bonds are contributed by N-terminal residues ( E172 , Q42 , N116 ) in yCRM1 , and one by a C terminal residue ( D947 ) ( Figure 6F , Figure 6—figure supplement 2 ) . Instead of making single mutations individually , the N-terminal 200 residues of hCRM1 were replaced with yeast equivalent residues 1–188 , and the chimera protein ( hCRM1N_swap ) was able to weakly bind to yRanBP1-RanGTP ( Figure 6E , lane 4 ) . Similarly , when the N-terminal 1–188 region of yCRM1 was replaced by human equivalent N-terminal region 1–200 ( yCRM1N_swap ) , binding to yRanBP1-RanGTP was significantly reduced ( Figure 6E , lane 9 ) . Pull down assay did not produce a significant difference by further mutating D947 ( data not shown ) . Therefore , loss of binding between hCRM1 and yRanBP1-RanGTP is due to sequence divergence of hCRM1 from yCRM1 , including the RanBP1 contact surface and Ran contact surfaces . Previously , we showed that yeast RanBP1 is localized in the nucleus in HeLa cells . When yCRM1 was co-transfected with yRanBP1 , yRanBP1 was significantly ( p<0 . 0001 ) relocalized to the cytoplasm ( Figure 7A , B ) , suggesting that yCRM1 but not endogenous hCRM1 formed a nuclear export complex with yRanBP1 . In contrast , co-transfected yCRM1T753Q , the mutant that displayed reduced binding to yRanBP1-RanGTP by pull down , barely promoted nuclear export of yRanBP1 ( p<0 . 0001 ) . Similarly , the predominantly nucleus-localized tipple mutant of NESmut ( NESmut ) was significantly ( p<0 . 0001 ) exported to the cytoplasm when co-transfected with yCRM1 , but not yCRM1T753Q ( Figure 7C , D ) . These cellular studies fully recapitulated what was observed by pull down ( Figure 6 ) , supporting the molecular mechanism for loss of trimeric complex formation in animals . Endogenous human RanBP1 knock-down and further rescue with mCherry-tagged yRanBP1 , yRanBP1V150A , mRanBP1 , ΔC or NESmut did not change the localization of endogenous CRM1 , the nuclear export cargo NFκB and the transfected nuclear import cargo GFP-NLS ( Figure 7—figure supplement 1 ) . This is probably due to the presence of functionally redundant RanBP2 in mammals ( Villa Braslavsky et al . , 2000 ) , though RanBP1 is reported to be an essential gene in yeast ( Petersen et al . , 2000 ) . Cargo’s proper localization also suggests that the nucleus concentration of mis-localized RanBP1 is probably too low to disrupt the cellular RanGTP gradient ( which is critical for nuclear transport ) ( Izaurralde et al . , 1997 ) . On the other hand , cytoplasmic localization of endogenous RanBP1 and NFκB was inhibited by siCRM1 , which could be rescued by the expression of either hCRM1 and yCRM1T753Q , but not yCRM1 and hCRM1Q742T ( Figure 7E–G ) . Though hCRM1Q742T is much more like hCRM1 than yCRM1T753Q in sequence , function-wise the opposite is true , suggesting that Q742 ( T753 in yeast ) is an important function-defining residue . Moreover , hCRM1Q742T displayed strong nuclear rim staining in all acquired images ( Figure 7E ) , which might be due to the formation of weakly GAP-resistant CRM1-RanGTP-RanBP2 trimeric complex . Negative control using transfected mCherry-NLS displayed no localization changes under siCRM1 and different rescue conditions ( Figure 7H ) . We previously showed that when CRM1 is limited , yeast proteins ( that form tight trimeric complex ) are less efficient than human proteins in nuclear export of cargoes . However , one might argue that yeast proteins are innately less efficient when used in human cells . We therefore performed a similar experiment using all yeast proteins , that is WT proteins that form trimer , and mutant yeast proteins that do not ( Figure 8A ) . The mutations include yCRM1T753Q and yRanBP1V150A , which inhibits trimer formation ( Figure 6D , E ) , and fusion of NESmRanBP1 to C-terminus of yRanBP1V150A to prevent nuclear accumulation of yRanBP1 . Statistical analysis showed a faster nuclear export rate for mutant than WT proteins ( Figure 8B ) , reconfirming that the nuclear export rate is higher when not forming tight trimer . By doubling the energy expenditure , animal cells may be able to do 'more work' and could export against a higher nucleo-cytoplasmic concentration difference . To test this hypothesis , we analyzed the subcellular localization of transfected GFP-NLS and mCherry-NES in human ( HeLa ) and yeast cells . Astonishingly , while mCherry-NES showed predominant cytoplasmic localization ( 22% nuclear ) in human cells , the same protein in yeast cells showed much higher nucleus localization level ( 43% nuclear ) ( Figure 8C , D ) , though mCherry-NES bound to yeast and human CRM1 transportors with similar strength ( Figure 8—figure supplement 1 ) . Similarly , GFP-NLS , which displayed similar binding strength to human and yeast transportors ( Figure 8—figure supplement 1 ) , was much less nucleus localized in yeast cells ( 59% nuclear ) compared to in human cells ( 82% nuclear ) ( Figure 8C , D ) . In addition , mCherry-NES were expressed at very low level in yeast cells compared to mammalian cells ( Figure 8—figure supplement 2A ) , while the Ran and CRM1 concentrations in both cells are comparable ( Figure 8—figure supplement 2B , C ) , suggesting that the lower export efficiency in yeast should not be due to saturation of export capacity . In summary , these results advocate the notion that nuclear transport efficiency is higher in human cells than in yeast cells .
Nuclear export of RanBP1 is essential for all eukaryotic cells since excess of RanBP1 in the nucleus sequesters nuclear RanGTP , inhibits nuclear transport and is toxic to cells ( Izaurralde et al . , 1997; Richards et al . , 1996 ) . In both yeast and human , RanBP1 is exported by nuclear export factor CRM1 ( 19 ) . Though yeast and mammalian proteins are used in this work , the mechanisms revealed in this study should be applicable to fungi and animals . Due to the lack of NES , fungi RanBP1 should form trimeric complex through its RBD interacting with both RanGTP and CRM1 . This explains the observation that intact RBD is required for its nuclear export ( Künzler et al . , 2000 ) . Obviously , NES is not needed for its nuclear export because fungi RBD is sufficient for its nuclear export . In animals , NES but not RBD is required for its export because animal RBD-RanGTP does not bind to CRM1 at sufficient high affinity to be exported . We show that this is due to degeneration of the interface residues essential for the formation of trimeric complex . Since nucleus is enriched with RanGTP , a tetrameric animal export complex of ( RanBP1-RanGTP ) -CRM1-RanGTP would form in animals , whereby RanBP1-RanGTP binds to the NES groove of CRM1 ( through the NES of RanBP1 ) as an NES cargo . It should be noted that animal tetrameric complex probably exists only in the nucleus , whereas the fungi trimeric complex could exist both in nucleus and cytoplasm . In the nucleus , animal RanBP1 cargo dissociation activity is inhibited with excessive RanGTP , and it would not dissociate any CRM1 cargo or its own NES before reaching the cytoplasm . When animal RanBP1 tetrameric complex enters the cytoplasm , where RanBP1/RanGAP is present at high concentration and RanGTP at low concentration , the complex would probably be quickly dissembled ( shutting off cytoplasmic YFP signals in the BiFC experiment ) . In the cytoplasm , fungi and animal RanBP1 also use different mechanisms to displace CRM1 cargoes . Both fungi and animals use RBD domains for cargo dissociation , and we showed that NES of animal RanBP1 is dispensable for cargo dissociation . Fungi RanBP1 binds tightly to RanGTP-CRM1 and closes the NES groove to actively release bound NES . This ends up sequestering available CRM1 , before RanGTP is hydrolyzed with the help of RanGAP . In animals , excessive RanBP1 ( and RanBP2 ) in the cytoplasm sequesters RanGTP from the NES-CRM1-RanGTP complex , resulting in a transient low-affinity binary CRM1-NES complex , which dissembles automatically . However , we do not exclude the possibility that a transient complex of animal RanBP1-RanGTP-CRM1 forms like the fungi complex , which actively dissociates NES from CRM1 , immediately followed by dissociation of RanBP1-RanGTP-CRM1 . In fact , active displacement of cargo was observed previously with the yCRM1 mutant ( P754D ) , which is defective in the stable yRanBP1-yRanGTP-yCRM1 ternary complex formation ( Koyama and Matsuura , 2010 ) . Consistent with the proposed cargo dissociation mechanisms , we show that in the presence of RanGAP and RanBP1 , yeast but not human CRM1 partially inhibits RanGTP hydrolysis . Besides dissociating cargo from CRM1 , RanBP1 is essential for the disassembly of tightly-bound Ran-karyopherin complexes . In fact , the Ran-sequestering mechanism seems to be more prevalent among RanBP1 mediated RanGTP dissociation from different karyopherins . Only human and yeast Importin β1 ( Lounsbury and Macara , 1997; Deane et al . , 1997; Floer et al . , 1997; Bischoff et al . , 1995a ) , and yeast CRM1 ( Künzler et al . , 2000; Maurer et al . , 2001 ) , form very tight complexes with RanBP1-RanGTP , which are partially GAP resistant . For human importin β2 ( Lounsbury and Macara , 1997 ) , importin β3 ( Deane et al . , 1997 ) , yeast importin 4 ( Schlenstedt et al . , 1997 ) , human importin 7 ( Görlich et al . , 1997 ) , human importin 8 ( Görlich et al . , 1997 ) , human Exportin-1 ( Paraskeva et al . , 1999; Askjaer et al . , 1999 ) , human Exportin-2 ( Kutay et al . , 1997; Bischoff et al . , 1995a ) and human exportin-t ( Kutay et al . , 1998 ) , their affinities with RanGTP-RanBP1 are relatively weak and are effectively dissembled in the presence of RanGAP . In the case of human and yeast KPNB1 , NLS is required for efficient release of Ran-RanBP1 for GAP hydrolysis ( Bischoff and Görlich , 1997; Floer et al . , 1997 ) . For yCRM1 , there is no corresponding RanBP1-RanGTP release factor reported thus far . Apparently , one round nuclear export of fungi RanBP1 costs energy of hydrolysing one GTP , while one round nuclear export of animal RanBP1 costs energy of two GTPs . Since RanBP1 is a constantly shuttling protein , animals may spend significantly more energy than fungi in the long term . We believe that at the same time of suffering this disadvantage in energy consumption , forming tetrameric complex may also confer at least two advantages for animals . First , as mentioned above , forming tetrameric complex does not inhibit GAP mediated hydrolysis , which means faster recycling rate of nuclear export machinery . This possibly further implies less congestion of nuclear pore and faster rate of nuclear transport , or translated into a higher nucleo-cytoplasmic concentration difference of cargoes . Indeed , we found that both nuclear import and export of cargoes were substantially more efficient in human cells than in yeast cells ( Figure 8C ) . Though we showed that this is not due to differences in binding strength to transport receptors nor cargo expression levels in two types of cells , our results do not eliminate other possibilities such as a leakier NPC in yeast; further experiments are needed to fully understand how much the yeast RanBP1 system ( which partially inhibits GAP hydrolysis ) contributes to the observed lower concentration gradient of cargoes . Second , the interface region and several other regions in fungi have to be conserved in order to maintain trimeric complex formation and export nuclear RanBP1 . In animals , the existence of specialized NES in RanBP1 has eliminated that requirement , allowing multiple regions in both RanBP1 and CRM1 to diverge and participate in other yet-to-be discovered cellular functions . Compared with fungi , animals are readily mobile and more complex in cellular/tissue organization . Whether the appearance of RanBP1 NES in animals is involved in higher order cellular organization , tissue development or independent mobility , and whether the tetrameric complex plays roles in other processes warrant further studies . Further , the different mechanisms observed in this study could aid in development of inhibitors against fungi by targeting the critical binding interfaces discussed above to inhibit the formation of trimeric RanBP1 nuclear export complex . Anti-fungal medicines developed by this strategy are potentially safe for human beings , since our RanBP1 nuclear export system does not rely on the formation of the trimeric complex .
The human , mouse or yeast RanBP1 ( or their mutants ) was cloned into a pGEX-4T1 based expression vector incorporating a TEV-cleavable N-terminal GST-tag fusion . The plasmid was transformed into Escherichia coli BL-21 ( DE3 ) and grown in LB Broth medium . Expression of protein was induced by the addition of 0 . 5 mM isopropyl β-D-1-thiogalactopyranoside ( IPTG ) , and the culture was grown overnight at 18 ˚C . Cells were harvested and sonicated in lysis buffer ( 50 mM Tris pH 8 . 0 , 200 mM NaCl , 10% glycerol , 2 mM DTT , 1 mM EDTA and 1 mM PMSF ) . RanBP1 was purified on a GST column and eluted after TEV cleavage in a buffer containing 20 mM Tris pH 8 . 0 , 200 mM NaCl , 10% glycerol , 1 mM EDTA and 2 mM DTT . This was followed by a Superdex 200 increase gel filtration column on the ÄktaPure ( GE Healthcare ) using the gel filtration buffer ( 20 mM Tris pH 8 . 0 , 200 mM NaCl , 10% glycerol , 2 mM DTT ) . Eluted proteins were frozen at −80°C at 5–10 mg/ml . Alternatively , GST-RanBP1 was eluted with 20 mM Tris pH 8 . 0 , 200 mM NaCl , 1 mM EDTA , 2 mM DTT , 10 mM reduced glutathione , and purified by Superdex 200 increase column . Both mouse and human RanBP1 have been used in this study and they are highly similar ( 95% identity , Figure 1—figure supplement 4 ) . ΔC construct contains deletion of C terminal residues’ PGKNDNAEKVAEKLEALSVREAREEAEEKSEEKQ’ . His-tagged proteins ( human and yeast CRM1 and related mutants ) were expressed in E . coli grown in TB Broth medium . The proteins were induced in the presence of 0 . 5 mM IPTG overnight at 25 ˚C , and purified by Nickel beads . 6 × His tagged proteins were eluted with 300 mM Imidazole pH 7 . 5 , 300 mM NaCl , 10% glycerol , and 2 mM BME . All proteins used were purified by S200 prior to pull down . To assess different interactions , GST-tagged proteins were immobilized on GSH beads . For GST-RanBP1 , a wash step was performed immediately after immobilization to remove unbound GST-RanBP1 . Soluble proteins at indicated concentrations were incubated with the immobilized proteins in a total volume of 1 ml for two hours at 4°C . After two washing steps , bound proteins were separated by SDS/PAGE and visualized by Coomassie Blue staining . Each experiment was repeated at least twice and checked for consistency . The pull down buffer contains 20 mM Tris pH 8 . 0 , 200 mM NaCl , 10% glycerol , 5 mM MgCl2 , 0 . 005% Triton-X100 , and 5 mM DTT if not specified . RanWT contains about 95% RanGDP and is used as RanGDP ( Zhang et al . , 2018 ) . Human RanL182A and yeast RanL184A used in this paper are C-terminus destabilized , therefore loaded with more than 80% of GTP , and are used as RanGTP ( Zhang et al . , 2018 ) . ITC experiments were conducted at 20°C using ITC200 ( Microcal ) in a buffer containing 20 mM Tris pH 8 . 0 , 200 mM NaCl and 5 mM MgCl2 . Q69L and L182A double mutant of Ran were used in this assay because this Ran mutant is 100% GTP bound ( Zhang et al . , 2018 ) . For animal complex , 250 µM Ran was titrated into the sample cell containing 20 µM hCRM1 and 20 µM mRanBP1 . For yeast complex , 150 µM Ran was titrated into the sample cell containing 15 µM yCRM1 and 15 µM yRanBP1 . Each experiment was repeated at least twice . Data were processed by NITPIC ( Scheuermann and Brautigam , 2015 ) and fitted by SEDPHAT ( Houtman et al . , 2007 ) . pEGFP-C1 and pmCherry-C1 plasmids were used to express EGFP or mCherry-tagged CRM1 or RanBP1 constructs . HeLa and 293 T cells were obtained from the Cell Bank of Chinese Academic of Sciences ( Shanghai , China ) . HeLa cells were maintained in Dulbecco’s modified Eagles medium ( Hyclone ) supplemented with 10% ( vol/vol ) fetal bovine serum ( Biological Industries ) . For experiments without siRNA , cells were transfected with TurboFect transfection reagent ( ThermoFisher ) and fixed after 24 hr of transfection . For siRNA work , cells were first transfected with siRanBP1 ( GGGCAAAACUGUUCCGAUUUG ) or siCRM1 ( CUCAGAAUAUGAAUACGAATT ) using Lipo2000 ( ThermoFisher ) ( Fan et al . , 2011 ) . After 24 hr , cells were transfected with different plasmids in the presence of polyethylenimine ( PEI ) transfection reagent . Cells were visualized 48 hr after the second transfection . Antibodies against RanBP1 ( ProteinTech , 1:400 ) , CRM1 ( absin , 1:400 ) , NFκB ( SAB , 1:250 ) , HA ( CST , 1:1000 ) , and Myc ( ProteinTech , 1:1000 ) were used . Images were acquired by Olympus FV-1000 confocal microscope , and were analyzed using NIH ImageJ and Graphpad software . NYFP-RanWT-HA , Myc-RanWT-CYFP , NYFP-RanCmut-HA and Myc-RanCmut-CYFP were cloned into pCDNA3 . 1 ( + ) expression vectors . 293 T cells were seeded in 24-well plates containing circular coverslips slides in Dulbecco’s modified Eagles medium ( Hyclone ) supplemented with 10% ( vol/vol ) fetal bovine serum ( Biological Industries ) , 100 U/mL penicillin and 100 μg/mL streptomycin in a 5% CO2 atmosphere at 37°C . Twenty-four hours later , cells were transfected with plasmids in 1 mg/ml PEI ( Polyethylenimine , Polysciences ) transfection reagent . After another 24 hr , cells were treated with 25 µM biliverdin to boost the YFP signals . Two hours later , cells were fixed and incubated with the primary antibody anti-HA ( CST , rabbit ) and anti-Myc ( ProteinTech , mouse ) , then with secondary antibody anti-mouse ( Beyotime , Dylight 405 ) and anti-rabbit ( ThermoFisher , Cy5 ) . Images were acquired by Olympus FV-1000 confocal microscope , and analyzed using NIH ImageJ software . Human ( 0 . 4 µM ) or yeast wide type Ran ( 0 . 3 µM ) protein was first mixed with hRanBP1 ( 1 µM ) or yRanBP1 ( 1 µM ) , respectively . Then samples were briefly incubated with or without 1 µM CRM1 from respective species . Since Ran has a very slow intrinsic nucleotide exchange rate , 0 . 5 µM RCC1 and 100 µM of GTP was added to reload Ran with GTP after each cycle of GAP-mediated hydrolysis in all samples . The samples were then incubated with 0 . 3 µM yeast RanGAP for 0 , 15 , 45 and 75 min . Control samples were done with 1 µM CRM1 , 1 µM GST-PKI and 0 . 3 µM of Ran for both species . The amount of free phosphate generated was measured using GTPase assay kit ( Bioassay ) and Multiskan FC microplate reader ( ThermoFisher ) . Each reaction was repeated four times in parallel . We performed in vitro nuclear export experiment with slight modifications ( Cassany and Gerace , 2009 ) . First , 1 µM GST-NESPKI-NLSIBB ( GST-NES-NLS ) , 1 µM KPNB1 , 1 µM NTF2 , 1 × energy regeneration system ( Cassany and Gerace , 2009 ) , 0 . 01% Triton-X100 , and 2 µM of hRan were added to semi-permeabilized HeLa cells and incubated at room temperature for 60 mins to accumulate nuclear cargoes . The very low concentration of Triton-X prevents non-specific cytoplasmic binding , but does not permeate nuclear envelopes ( Figure 3—figure supplement 1 ) . After washing , the cells were incubated with either human or yeast proteins ( 0 . 1 µM CRM1 , 1 µM Ran and 1 µM RanBP1 ) , energy regeneration system , 0 . 01% Triton-X , for different time points at room temperature with gentle shaking . After reaction , the cells were washed , fixed and visualized by immunostaining with GST antibody ( Santa Cruz , 1:400 ) . Statistics were based on measurements from at least 30 cells for each sample , and statistical significance was calculated by one-way ANOVA test in Graphpad software . In summary , we have shown that animal RanBP1 binds to CRM1 only in the nucleus where there is excessive RanGTP , and forms a complex with CRM1 and two RanGTP proteins through its NES but not RBD . This complex is stable and is exported to the cytoplasm where it is dissembled by RBD containing proteins and RanGAP . In contrast to the CRM1-RanGTP sequestering cargo dissociation mechanism in fungi , animal RanBP1 dissociates cargo through stripping RanGTP from nuclear export complexes . The key difference between fungi and animals is that animal RanBP1-RanGTP does not bind to CRM1 as it does in fungi , due to degeneration of interface residues on both CRM1 and RanBP1 , though either change is sufficient to reduce binding . The different mechanisms of RanBP1 nuclear export highlight how animals increase catalysis rate on the expense of more energy consumption . It is unexpected and interesting that while essential functions of orthologous proteins are conserved , profound differences exist in the underlining mechanisms . | Plant , animal and fungal cells all store their DNA inside the cell’s nucleus . Small molecules can freely cross the membrane that surrounds the nucleus , but pores in the membrane control when larger molecules enter or leave . This transport process is an essential part of healthy cell behavior . To leave the nucleus , large molecules need to carry a coded sequence called a nuclear export signal . In yeast cells , which are often used to study cell biology , this sequence allows cargo to bind to a groove in so-called molecular cargo vehicles , such as a protein called CRM1 . A protein called RanGTP binds to CRM1 to supply the energy needed to transport molecules across the membrane . Outside of the nucleus , another protein called RanBP1 closes up the groove in the CRM1 protein to help remove the cargo by interacting with RanGTP and CRM1 to form a ‘complex’ . The version of RanBP1 found in animal cells has its own nuclear export signal , which led researchers to question whether it works in the same way as yeast RanBP1 . To find out , Li et al . compared yeast RanBP1 with mouse RanBP1 . This revealed that mouse RanBP1 lacks the amino acids that allow it to interact with CRM1 in the fashion of yeast RanBP1 . When unloading cargo from CRM1 , mouse RanBP1 does not form a complex with Ran and CRM1; instead , it works entirely through removing RanGTP from CRM1 . This process is more efficient than the one used by yeast cells , but it uses twice as much energy . The results presented by Li et al . demonstrate that even processes that are essential to cells can be optimized to fit the needs of different species . Future work could potentially exploit the differences in the export processes used by fungi and animal cells to develop new anti-fungal treatments . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2019 | Distinct RanBP1 nuclear export and cargo dissociation mechanisms between fungi and animals |
Combining clonal analysis with a computational agent based model , we investigate how tissue-specific stem cells for neural retina ( NR ) and retinal pigmented epithelium ( RPE ) of the teleost medaka ( Oryzias latipes ) coordinate their growth rates . NR cell division timing is less variable , consistent with an upstream role as growth inducer . RPE cells divide with greater variability , consistent with a downstream role responding to inductive signals . Strikingly , the arrangement of the retinal ciliary marginal zone niche results in a spatially biased random lineage loss , where stem- and progenitor cell domains emerge spontaneously . Further , our data indicate that NR cells orient division axes to regulate organ shape and retinal topology . We highlight an unappreciated mechanism for growth coordination , where one tissue integrates cues to synchronize growth of nearby tissues . This strategy may enable evolution to modulate cell proliferation parameters in one tissue to adapt whole-organ morphogenesis in a complex vertebrate organ .
To maintain proper proportions , growth must be regulated at the level of the whole body , the size of each organ , and the size of tissues within an organ ( Roselló-Díez and Joyner , 2015 ) . Some regulatory mechanisms are shared , while others are specific to each level or to particular organs ( Lui and Baron , 2011; Roselló-Díez and Joyner , 2015 ) . Systemic signals couple nutrition to growth to coordinate growth of all organs at the organismal level ( Buchmann et al . , 2014; Droujinine and Perrimon , 2016 ) . In addition to extrinsic systemic factors , transplantation experiments showed that many organs , including the eye , grow autonomously according to intrinsic factors ( Wallman and Winawer , 2004; Roselló-Díez and Joyner , 2015 ) . Growth coordination mechanisms have been studied at the level of the whole organism and inter-organ communication ( Buchmann et al . , 2014; Droujinine and Perrimon , 2016 ) , but feedback mechanisms between constituent tissues of an organ remain largely unexplored both experimentally and at a conceptual level ( Buchmann et al . , 2014 ) . Teleost fish grow throughout their lives , increasing massively in size ( Johns and Easter , 1977 ) . The teleost medaka ( Oryzias latipes ) grows roughly ten-fold from hatching to sexual maturity within 2–3 months ( Figure 1—figure supplement 1A ) . Unlike embryonic morphogenesis , during post-embryonic growth all organs must scale with the increasing body size while fully functioning . In the eye , continuous growth must be additionally balanced with continuous shape-keeping: Proper optics , and thus vision , requires a precise 3D shape . Highly visual shallow water fish such as medaka have near-perfect hemispherical eyes ( Fernald , 1990; Nishiwaki et al . , 1997; Beck et al . , 2004 ) . The growth rates of all eye tissues must perfectly match , otherwise the organ would deform , akin to a bimetallic strip . Thus , the eye of fish provides an excellent system to explore how anatomically and functionally distinct tissues coordinate to grow and maintain the shape of an organ in functional homeostasis ( Johns and Easter , 1977; Centanin et al . , 2014 ) . The vertebrate eye consists of multiple concentric tissues , including the neural retina ( NR ) and the retinal pigmented epithelium ( RPE ) ( Figure 1A; Table 1 ) . In fish and amphibians , these tissues grow from a ring-shaped stem cell niche in the retinal periphery: the ciliary marginal zone ( CMZ ) ( Johns , 1977; Harris and Perron , 1998; Amato et al . , 2004 ) . The CMZ can be subdivided into a peripheral stem- and a central progenitor cell domain; stem cells are believed to have the potential for indefinitely many cell divisions while progenitor cells divide only a handful of times ( Raymond et al . , 2006; Centanin et al . , 2014; Wan et al . , 2016; Shi et al . , 2017 ) . At the very periphery of the CMZ , about 5 rows of cells express the stem cell marker retina-specific homeobox gene 2 ( Rx2 ) ( Reinhardt et al . , 2015; Wan et al . , 2016; Tang et al . , 2017 ) . The CMZ is a bi-partite niche , with tissue-specific stem cells for NR and RPE ( Shi et al . , 2017 ) . In medaka , stem cells for NR and RPE are strictly separate , as demonstrated by transplantations at blastula stage and genetic recombination after hatching ( Centanin et al . , 2011; Centanin et al . , 2014 ) . Thus , medaka NR and RPE are independently growing tissues with identical topology . As a population , CMZ cells appositionally add new cells in concentric rings as shown by label incorporation with thymidine analogues ( Johns , 1977; Centanin et al . , 2011 ) . Individual stem cells labelled by genetic markers form clonal progeny in so-called Arched Continuous Stripes ( ArCoS; Figure 1B ) ( Centanin et al . , 2011; Centanin et al . , 2014 ) . Medaka NR stem cells produce the full complement of neuronal cells in apico-basal clonal columns ( Figure 1—figure supplement 2A’–B ) ( Centanin et al . , 2011; Centanin et al . , 2014; Lust and Wittbrodt , 2018 ) . These differentiated retinal cells grow little in size ( Johns , 1977 ) , retain their relative position over time ( Johns , 1977; Centanin et al . , 2011 ) , and have negligible death rates ( Johns and Easter , 1977; Stenkamp , 2007 ) . Thus , the only parameter available to NR and RPE to coordinate their growth rates is the proliferation of the tissue-specific CMZ stem cells . Stem cells have long been defined by an unlimited self-renewal capacity ( Watt and Hogan , 2000; Clevers and Watt , 2018 ) . Two general strategies underlie long-term maintenance of stem cells: 1 ) a deterministic model where every single division produces a stem- and a progenitor daughter cell ( ‘invariant asymmetry’ ) ; and 2 ) a stochastic model where cells divide symmetrically , and the daughter cells have a probability to stay as stem cells or commit to a progenitor fate ( ‘neutral drift’ ) ( Watt and Hogan , 2000; Clevers and Watt , 2018 ) . One tenet of this model is neutral competition: Stem cells randomly displace one another , resulting in the ‘loss’ of lineages where all progeny commit to a progenitor fate until the entire niche is occupied by a single clone ( Colom and Jones , 2016; Clevers and Watt , 2018 ) . Strikingly , the medaka retina diverges from the neutral drift model . The CMZ maintains a polyclonal stem cell population for both the NR and the RPE , and in particular NR stem cells undergo asymmetric self-renewing divisions throughout the life of the animal ( Centanin et al . , 2011; Centanin et al . , 2014 ) . It remains unclear whether stem cell proliferation in the CMZ follows a purely deterministic model , or whether it follows a strategy in-between invariant asymmetry and neutral drift . In this work we combine in vivo and in silico clonal analysis in the NR and RPE of medaka to address how these tissues coordinate their growth rates . We find that RPE stem cells have highly variable cell division timing consistent with a downstream role in the control hierarchy , whereas NR stem cells display less variability consistent with an upstream role in inducing growth in nearby tissues . Our simulation predicts that the spatial segregation of stem and progenitor CMZ domains is an emergent property , as the topology of the retinal niche preconditions the retina to a spatially biased neutral drift . NR stem cells deviate from a purely random drift model by preferential division axis orientation and differential modulation of division parameters along the CMZ circumference . We propose that during post-embryonic growth of the teleost eye , the NR CMZ forms a hub for integrating external and internal stimuli that affect cell division parameters , which ultimately direct the growth and shape of the entire eye .
Retinal cells follow an exquisite spatiotemporal order ( Figure 1B–C , Figure 1—figure supplement 1B ) . Thus , clones derived from stem cells are a frozen record of past cell divisions ( Centanin et al . , 2011; Centanin et al . , 2014 ) , offering a window of opportunity to study stem cell properties in the NR and RPE . We experimentally generated NR ArCoS by randomly labelling individual NR stem cells using the Rx2::ERT2Cre , Gaudí2 . 1 line in hatchling medaka , and analyzing the eyes in adult fish as previously described ( Centanin et al . , 2014; Reinhardt et al . , 2015 ) . The Rx2 promoter drives the inducible Cre recombinase in stem cells at the very periphery of the CMZ ( Reinhardt et al . , 2015 ) . A recombined stem cell generates a stripe of GFP-positive progeny in an otherwise GFP-negative retina ( Centanin et al . , 2014 ) . In proximal view , NR ArCoS emanated as rays from the central embryonic retina , the part of the eye that was already differentiated at the timepoint of Cre-mediated recombination ( Figure 1D ) . We visualized RPE ArCoS by mosaic knockout of pigmentation using CRISPR/Cas9 targeted to the gene oculo-cutaneous albinism 2 ( Oca2 ) , which is required for melanosome maturation ( Fukamachi et al . , 2004; Lischik et al . , 2019 ) . RPE stem cells with a bi-allelic mutation in Oca2 generate unpigmented stripes , analogous to RPE ArCoS obtained by transplantation ( Centanin et al . , 2011 ) . RPE ArCoS frequently branched , forming irregular stripes variable in size and shape ( Figure 1E ) . These qualitative differences in clonal pattern suggested that despite their identical topology , the division behavior of NR and RPE stem cells differed . Clonal data generates a distribution of outcomes that is challenging to analyse and easy to misinterpret ( Klein et al . , 2007 ) . The curved retinal surface and spatial extent of the niche pose a further challenge . We overcome these challenges by comparing experimental clonal data with simulated clonal data from a 3D agent based cell-center overlapping spheres model built in the platform EPISIM ( Sütterlin et al . , 2013; Sütterlin et al . , 2017; Sütterlin , 2019 ) . This modelling technique represents cells as discrete objects ( e . g . spheres ) that physically interact through forces acting on the cell centers; the spheres are allowed to slightly overlap to simulate cell deformability and allow a tight cell packing ( Sütterlin et al . , 2013; Sütterlin et al . , 2017 ) . This level of abstraction is ideally suited to the tightly packed pseudocrystalline mosaic of retinal cells ( Johns , 1981; Nishiwaki et al . , 1997; Pérez Saturnino et al . , 2018 ) , and has been used previously to model clonal data in skin and gut epithelia ( Osborne et al . , 2010; Buske et al . , 2011; Li et al . , 2013 ) . Our retinal tissue model consists of a layer of spheres ( representing either NR or RPE cells ) on a hemisphere ( representing the rest of the organ that is not explicitly modelled; Figure 2A ) . The RPE is a monolayer , thus each model cell corresponds to one RPE cell . In the NR , CMZ stem cells form a monolayer , and their differentiated progeny arrange in multiple neuronal layers ( Johns , 1977; Raymond et al . , 2006 ) . We observed that clonal progeny of CMZ stem cells retained close proximity with little spread tangential to the retinal surface , forming clonally related ‘columns’ ( Figure 1—figure supplement 2A’-B ) ( Centanin et al . , 2011; Centanin et al . , 2014; Lust and Wittbrodt , 2018 ) . We took advantage of this fact to abstract each differentiated clonal column as a single cell in the simulation . In vivo , the spatial extent of the CMZ stem cell domain is believed to be defined by cues such as nearby blood vessels ( Wan et al . , 2016; Tang et al . , 2017 ) . Therefore , we defined the virtual stem cell domain with a fixed size of 25 µm , that is 5 rows of cells , reflecting the endogenous scale of the Rx2-expressing CMZ domain ( Reinhardt et al . , 2015; Wan et al . , 2016; Tang et al . , 2017 ) . In vivo , NR stem cells divide predominantly asymmetrically , but also undergo symmetric divisions ( Centanin et al . , 2014 ) . The rates of asymmetric and symmetric divisions are unknown; likewise , it is unknown whether these rates are deterministically defined or an emergent property of an underlying stochastic system . Since stochastic cell divisions successfully describe the proliferation of committed retinal progenitor cells in larval zebrafish ( Wan et al . , 2016 ) , we used a simple stochastic mechanism for our initial model . Virtual stem cells commit to divide with a fixed probability pdivision=126h−1 and intervals between subsequent cell divisions must fulfill a minimum cell cycle length tcellCycle=24h . These values lie within a biologically plausible range estimated from experimentally measured growth rates and a parameter scan of the simulation ( Appendix 1 section 3 . 3 ) . All divisions are symmetric , resulting in two stem cells; cells differentiate and stop cycling when they exit the virtual CMZ after being pushed out by cellular crowding . To prevent physically implausible cell crowding , cell-center based models include a density-dependent inhibition of cell division ( Pathmanathan et al . , 2009; Osborne et al . , 2017; Sütterlin et al . , 2017 ) . In our model , inhibition occurs in cells whose average overlap with all neighbors exceeds a fraction of the cell’s diameter given by the model parameter δol_threshold ( Figure 2—figure supplement 1; Appendix 1 , section 2 . 4 ) . Based on in vivo observations ( Lyall , 1957; Johns , 1977; Ohki and Aoki , 1985 ) , the growing virtual eye gradually moves cells apart as it expands , thus decreasing cell density ( Figure 2—figure supplement 2; Appendix 1 section 2 . 2 ) . Continuous proliferation in the CMZ counteracts this decrease in vivo ( Johns , 1977; Johns and Easter , 1977 ) ; likewise , the ever-increasing virtual cell population optimally fills the hemisphere at all times ( Video 1; Video 2 ) . Our model distills the complexity of the system and replicates the exquisite spatiotemporal growth order observed in vivo ( Figure 2B’ , B’’ ) . Conceptually , we reasoned that feedback between tissues in an organ can be wired in two fundamental ways: Either the tissue of interest acts upstream to induce growth of other tissues ( Figure 2C’; ‘inducer growth mode’ ) , or , vice versa , the tissue of interest lies downstream of growth cues from another tissue in the organ ( Figure 2D’; ‘responder growth mode’ ) . Possible biological mechanisms for these growth modes could be mechanical , biochemical , or a combination of both . For example , in the inducer growth mode cells could instruct organ growth by modifying the extracellular matrix or by paracrine signalling ( Buchmann et al . , 2014; Droujinine and Perrimon , 2016 ) . These stimuli instruct tissues with the responder growth mode to grow , for example by alleviating contact inhibition or by providing permissive proliferation signals ( Buchmann et al . , 2014; Droujinine and Perrimon , 2016 ) . In an organ composed of multiple tissues , one tissue may be the driver for growth , while the rest follows . We examined how these two conceptual growth modes affected stem cell dynamics in the simulation . In our implementation of the inducer growth mode , an increase in cell number induces growth of the virtual eye’s radius ( Appendix 1—equation ( 5 ) ) . Implicit in this growth mode is the assumption that cell division is not inhibited by the degree of cell crowding normally present in the tissue ( otherwise the organ would never grow ) . Therefore , we set the tolerated overlap threshold δol\_threshold=0 . 4 , a value which we determined by parameter scan to minimize cell division inhibition while preventing physically implausible crowding ( Appendix 1 , section 3 . 2 ) . In the responder growth mode , we let the radius grow linearly over time ( Appendix 1—equation ( 6 ) ) . In this growth mode , cells must stop dividing until they receive an external stimulus . We take advantage of the pre-existing local density sensing to implement a physical stimulus akin to contact inhibition . Thus , we set the tolerated overlap threshold δol\_threshold=0 . 2 to maximize cell division inhibition at homeostatic density ( Appendix 1 , section 3 . 2 ) . As growth of the hemisphere decreases cell density , cells dynamically respond to growth of the eye by resuming divisions . In short , the growth modes in our simulation differ only in: 1 ) the growth equation for the radius of the hemisphere , 2 ) the value of the threshold parameter δol\_threshold where local cell density inhibits cell divisions ( for details , the reader is referred to Appendix 1 , sections 2 . 3; 2 . 4; and 3 . 2 ) . We obtained virtual ArCoS regardless of growth mode ( Figure 2C’’ , D’’ ) . The growth mode strongly impacted on the shape of ArCoS . Clones in the inducer growth mode formed well-confined stripes with low variation in shape ( Figure 2C’’’ ) . In the responder growth mode , the virtual clones frequently intermingled and broke up into smaller clusters ( Figure 2D’’’ ) . Specifically , the growth modes impacted on variation in cell division timing ( Figure 2C’’’’ , D’’’’ ) . In the responder growth mode , local competition for space increased cell division intervals , particularly among cells exceeding the tolerated overlap threshold δol\_threshold=0 . 2 ( Figure 2D’’’’ ) . Thus , the model predicted distinct levels of variation in cell division timing in retinal tissues following the inducer or responder growth modes . Since the position of cells in the retina reflects their birth order ( Centanin et al . , 2011; Centanin et al . , 2014 ) , we reasoned that in the extreme case of no variation in cell division timing , each clone forms a continuous , unbranching stripe ( Figure 3B , left ) . In the opposite highly variable case , clones frequently branch or merge into polyclones , as well as fragment into several small patches ( Figure 3B , right ) . Thus , with increasing variation in cell division timing , we expect an increasing variation in clone width , and an increasing incidence of clone branching and fragmentation . To quantitatively underpin our previous observations , we compared simulated clones of the inducer and responder growth modes to clones in the NR and RPE ( Figure 3A’ , A’’ ) . We circumvented biases associated with fusion and fragmentation of clones by analyzing ‘patches’ , that is contiguous domains of segmented pixels . A patch may entail a ( sub- ) clone , or multiple clones ( i . e . a polyclone ) ( Figure 3—figure supplement 1; Video 3 ) . To assay our experimental and simulated data , we unrolled the retina with a coordinate transform ( Figure 3—figure supplement 2C ) and quantified three different metrics: patch width variance , branching , and fragmentation . To assay patch width variance , we aligned and superimposed all patches ( Figure 3C’ , C’’ ) , and quantified the distribution of maximum patch width ( Figure 3—figure supplement 2A; Figure 3—figure supplement 2—source data 3 . ) . Confirming our previous qualitative observations , NR patches formed a narrow stripe , while the width of RPE patches showed much greater variation ( Figure 3C’; Figure 3—figure supplement 2A ) . The variance of NR and RPE patches was significantly different at the 0 . 05 level ( p=3 . 50∙10−12 , F-test of equality of variance ) . In striking agreement to the experimental data , simulated patches in the inducer growth mode had low variation in width , while patches in the responder growth mode spread widely ( Figure 3C’’; Figure 3—figure supplement 2A ) . The variances in the simulated conditions were significantly different at the 0 . 05 level ( p=5 . 84∙10−7 , F-test of equality of variance ) , but highly similar between NR and inducer ( p=0 . 56 , F-test of equality of variance ) ; and RPE and responder ( p=0 . 21 , F-test of equality of variance ) . To measure branching we skeletonized the patches , and quantified the distribution of nodes per patch and condition ( Figure 3D; Figure 3—source data 5 ) . Patches in the NR and in the inducer growth mode were overwhelmingly stripe-like with no branch points ( Figure 3D; inset I ) , with similar node distribution ( p=0 . 64 , Wilcoxon rank sum test ) . In contrast , both NR and inducer differed significantly at the 0 . 05 level from the distribution in the RPE and responder growth mode ( NR-RPE: p=3 . 93∙10−6; NR-responder: p=3 . 26∙10−4; inducer-RPE: p=6 . 24∙10−7; inducer-responder: p=7 . 00∙10−5 , Wilcoxon rank sum test ) . Patches in the RPE and in the responder growth mode frequently bifurcated or merged , creating branching shapes with inclusions indicative of clone intermingling ( Figure 3D; inset III ) . RPE and responder growth mode were highly similar in this metric ( p=0 . 38 , Wilcoxon rank sum test ) . Not all patches were contiguous with the embryonic retina . Such ‘late arising patches’ result if a cell divided intermittently with periods of dormancy , leaving clone fragments behind ( Figure 3B , highly variable scenario ) . We quantified fragmentation by plotting the occurrence of late arising patches along the normalized post-embryonic retinal radius ( Figure 3E; Figure 3—source data 6 ) . In the NR late patches clustered in the central post-embryonic retina and waned thereafter . Thus clone fragments were not equally distributed , consistent with lower levels of cell division variability and a majority of continuous stripe-like clones . In contrast , the RPE displayed an even distribution indicative of frequent fragmentation throughout the life of the animal as predicted for the highly variable scenario ( NR-RPE: p=1 . 74∙10−3 , Wilcoxon rank sum test ) . The simulated data showed the same tendency , to a lesser degree , as the central peak in late patches was higher in the inducer growth mode and peripheral late patches occurred more frequently in the responder growth mode ( Figure 3E; inducer-responder: p=0 . 10 , Wilcoxon rank sum test ) . In this metric , the RPE stood out from the NR and both simulated conditions ( RPE-inducer: p=6 . 94∙10−5; RPE-responder: p=0 . 04 , Wilcoxon rank sum test ) , indicating a high degree of fragmentation and thus cell division variability . Together , these data show that NR and RPE have different degrees of variability in cell division timing . The NR displayed lower variability consistent with the simulated inducer growth mode , while the RPE showed higher levels of variability that even exceeded what we modelled with the responder growth mode . Thus , our data support a model where NR and RPE concertedly expand relying on different growth modes , which manifest in differently shaped ArCoS . Both the NR and simulations displayed a cluster of late patches in the central post-embryonic retina ( Figure 3E ) . Additionally , when discounting late patches , the distribution of patch length showed clear bimodality ( Figure 3—figure supplement 2B ) , suggesting that beyond fragmentation an additional stochastic process took place after clonal labelling . The region at the border to the embryonic retina , the ‘induction ring’ , marks the original position of the CMZ at the timepoint of Cre-mediated recombination ( Figure 4E ) . To investigate the stem cell dynamics in the induction ring we turned to the simulation . Surprisingly , the virtual induction ring contained many few-cell clones unrelated to any ArCoS ( Figure 4A’ , encircled by pink dashed lines ) . In these clones , all stem cells left the niche and thus differentiated ( ‘terminated clones’ ) . Nested inductions showed that sister stem cells within one clone segregated into subclones ( Figure 4A’–A’’ , highlighted ArCoS ) . However , only some of these subclones generated virtual ArCoS . Again , terminated clones clustered in the virtual induction ring ( Figure 4A’’ , encircled by black dashed lines ) , demonstrating that the pattern repeated itself regardless of the timepoint of virtual induction . Therefore , since central positions were occupied by short terminated clones , many stripe-like patches necessarily began in more peripheral positions , explaining the peak in late arising patches . In our model , all proliferative cells were equipotent stem cells . Nevertheless , a subset of these virtual stem cells proliferated only a few times before terminally differentiating , resulting in a bimodal distribution of patch lengths ( Figure 3—figure supplement 2B ) . Notably , the overwhelming majority of virtual ArCoS emerged from the periphery of the induction ring ( Figure 4A’–A’’; Video 4 ) , as confirmed by tracing back the position of the founder stem cells at simulation step 0 , while centrally located cells formed exclusively terminated clones ( Figure 4B ) . This behavior is highly reminiscent of retinal progenitor cells in vivo , which are believed to reside in the central CMZ ( Raymond et al . , 2006; Shi et al . , 2017 ) . Strikingly , only a minority of virtual stem cells formed ArCoS , while the vast majority formed terminated clones ( Figure 4B ) . Together , these data show that the virtual stem cell population subdivided into two functional domains that mirror the current model of the retinal niche with a peripheral stem- and a central progenitor domain ( Raymond et al . , 2006; Shi et al . , 2017 ) . Importantly , this subdivision was not imposed onto the simulation , but emerged dynamically . The central-most cells were poised to differentiate by being pushed out of the niche by divisions of their more peripheral neighbors . This neutral competition occurred continuously , as demonstrated by nested virtual inductions ( Figure 4A’–A’’ ) . Thus , the spatial segregation of stem- and progenitor domains is an emergent property of the system . Our simulations uncovered a role of stochastic drift in the niche , and lead us to the following two predictions: First , a large proportion of stem cells is lost by neutral competition and forms terminated clones . Thus , ArCoS should be a minority among labelled clones . Second , there is a spatial bias in this drift: The majority of ArCoS will derive from peripheral cells but some will derive from more central positions . Similarly , the majority of terminated clones will derive from central positions , but some will derive from peripheral positions . To address these predictions experimentally , we again labelled NR stem cells in hatchlings using the Rx2::ERT2Cre , GaudíRSG line ( Centanin et al . , 2014; Reinhardt et al . , 2015 ) , which when recombined results in a nuclear GFP signal , and analysed the eyes at adult stage . Few-cell clusters in the induction ring vastly outnumbered ArCoS , showing that terminated clones were the most common type of clone ( n = 1129 terminated clones in 20 retinae; Figure 4C’–C’’ , Figure 4—figure supplement 1A–B ) . A small fraction of terminated clones extended into the post-embryonic retina ( Figure 4C’–C’’ , yellow arrowheads ) . ArCoS , which by definition always reach the retinal margin , were less frequent ( Figure 4C’–C’’ , pink arrowheads; n = 36 ArCoS in 20 retinae ) . Thus , Rx2-expressing cells in the CMZ included cells that proliferated indefinitely as well as cells that proliferated only a few times before differentiating . The preponderance of terminated clones shows that ArCoS-forming cells are a minority , in line with our first prediction . To address the spatially biased stochastic drift , we examined at which position in the induction ring clones contained their central-most pixels in experiment and simulation ( Figure 4C’–C’’ , D’–D’’ , F ) . Among terminated clones , the majority started in central positions ( experiment: 77 . 3%; simulation: 61 . 0% ) , while a minority were exclusively located in the peripheral induction ring or in the post-embryonic retina ( experiment: 22 . 7%; simulation: 39 . 0% ) . The difference in proportions between experiment and simulation may indicate that the simulation underestimates the number of terminated clones . Nevertheless , a sizeable subset of experimental terminated clones derived from the periphery of the stem cell domain of the CMZ , indicating that some stem cells drifted into a progenitor-like state . Among experimental ArCoS , the vast majority ( 86 . 1% ) started in the periphery , but 13 . 9% derived from a central position , showing that some cells located in the central progenitor domain of the CMZ drifted into a lifelong stem cell fate . Strikingly , the ratios for peripheral and central ArCoS in the simulation are nearly identical ( p=1 . 00 , 2-sample test for equality of proportions ) , showing that the simulation captures ArCoS dynamics extremely well . Together , these data support a model of stochastic drift with a peripheral-stem and central-progenitor bias that is conditioned by the physical topology of the niche . NR ArCoS formed stripes that appeared slightly narrower than in the simulation ( Figure 3A’–A’’ , C’–C’’ ) . In simulations , the division axis was not oriented ( ‘random division axis’ ) . The thin clonal stripes suggested that NR stem cells had a preferential axis of division along the radial ( central-peripheral ) coordinate , while circumferential divisions occurred with lower frequency than expected for a random division axis orientation . We wondered whether NR stem cell division orientation could relate to shaping the organ . An inducer growth mode does not necessarily imply regulation of organ shape . To use an analogy , a mass of dough grows from within ( similar to the inducer growth mode ) , but its shape can be imposed externally by a mold ( i . e . the dough does not affect shape regulation ) . In the NR , the shape could plausibly be imposed externally by any of the surrounding tissues , and in this case , it would have no role in organ shape regulation ( Figure 5A ) . As the space available for cells is imposed externally , any orientation of division axes is theoretically possible; after division cells will locally shift to optimally fill space . In an alternative scenario , organ shape could be regulated by oriented cell divisions of CMZ stem cells ( Figure 5B’ ) . In this scenario , a precise orientation of division axes is necessary . We calculated the ideal proportion of circumferential and radial divisions required to maintain hemispherical geometry . We assumed two principal axes of division , and that each new cell contributed either to the area of the CMZ or to the rest of the eye ( Figure 5B’’ ) . Circumferential divisions ( two daughter cells stay in the CMZ ) must be balanced by radial divisions ( one daughter cell is poised to leave the niche and differentiate ) . A hemispherical eye of radius R has the area ( 1 ) Aeye=2πR2 , while the CMZ forms a band of width w at the base of the eye with area ( 2 ) ACMZ=2πRw . Thus , we obtain an ideal ratio of circumferential to radial divisions of1:Aeye−ACMZACMZ , ( 3 ) 1:R−ww , that is for every one circumferential division , there must be R−ww radial divisions . Since R≫w , radial divisions must be more frequent than circumferential divisions , and the frequency of radial divisions increases as the retinal radius grows . To quantify circumferential stem cell divisions in experimental and simulated data , we took advantage of the exquisite temporal order of NR growth to measure ArCoS width – a proxy for circumferential stem cell divisions . To this end , we developed a pipeline that unrolled the retina as described before , and measured the number of pixels along each radial position normalized by the total circumference – effectively the angle enclosed by two rays traversing the center of the embryonic retina and the clone boundaries at every radial position ( Figure 5D’’ ) . To only include lifelong stem cells , we focused our analysis on the post-embryonic retina and excluded the central portion including the induction ring . As expected , with increasing probability to divide along the circumferential axis , average clone width increases in the simulation ( Figure 5—figure supplement 1A’–B ) . When division axes perfectly match the ratio in Equation 3 , the simulation becomes the limiting case of shape regulation where the hemispherical shape is always maintained . Thus , we modelled how the ‘ideal division axis’ ratio given by Equation 3 affected simulated ArCoS in the inducer growth mode and compared this to experimental data as well as simulations with random division axis ( Figure 5C’–C’’’ ) . Experimental ArCoS width averaged to 4 . 87° ( Figure 5D’ black graph; n = 99 ArCoS across seven retinae ) . In contrast to experimental data , ArCoS width in simulations with random division axis averaged to 7 . 28° ( Figure 5D’ blue graph; n = 102 clones from five simulation runs; compared to experimental data: p=1 . 94∙10−7 , Welch two-sample t-test ) . In simulations with ideal division axis , ArCoS width closely matched experimental data , averaging at 4 . 54° ( Figure 5D’ , red graph; n = 133 clones from five simulation runs; compared to experimental data: p=0 . 37 , Welch two-sample t-test ) . These data show that NR stem cell divisions were not randomly oriented , but instead were preferentially oriented along the central-peripheral axis . Moreover NR stem cells underwent radial and circumferential divisions at a rate consistent with a role in organ shape regulation . We observed that in the retina of the surface-dwelling medaka , the position of the embryonic retina was not centered , but instead was shifted ventrally ( Figure 6A’ ) . As a result , the post-embryonic retina was longer dorsally than ventrally ( ratio dorsal to ventral length: mean = 1 . 42; standard deviation = 0 . 29; n = 10 retinae ) . The embryonic retina covered the entire retinal surface at induction ( Figure 6A’’ ) . Equal growth around the circumference should maintain the embryonic retina in the center . The ventral-ward shift indicated that along the CMZ circumference , ventral stem cells had different division parameters . We probed the feasibility of different scenarios in generating a ventral shift in an in silico screen . First , we discerned two ways for stem cells in the ventral domain ( defined as a 90° sector; Figure 6—figure supplement 2 ) to select a different division behavior: Either a lineage-bound intrinsic signal ( e . g . epigenetic imprinting ) , or a lineage-independent extrinsic signal ( e . g . a local diffusible molecule ) . Second , we altered two cell division parameters: The probability of division , which we varied between half ( pdiv_ventral=0 . 5⋅ pdiv_non-ventral ) or equal to the value in the non-ventral sector ( pdiv_ventral= pdiv_non-ventral ) , and the preferential axis of cell division , which we varied between circumferentially-biased ( pcirc_ventral=1 ) and radially-biased ( pcirc_ventral=0 ) . In control simulations where all cells behaved equally , the embryonic retina stayed centered ( Figure 6B’ , C’ ) . For a lineage-bound intrinsic signal , a circumferential bias lead to massive enlargement of ventral lineages at the expense of adjacent clones without affecting the embryonic retina ( Figure 6B’’ ) . Reducing proliferation probability resulted in termination of ventral lineages , as adjacent clones displaced them from the virtual niche ( Figure 6B’’’ ) . An intrinsic signal resulted in a ventral shift only if circumferential bias was combined with lower proliferation probability ( Figure 6B’’’’ – condition I ) . In these simulations , circumferential divisions allowed ventral lineages to physically occupy niche positions ( preventing their displacement ) while lower proliferation reduced pressure on cells of the embryonic retina , allowing a ventral shift . In the scenario of a lineage-independent extrinsic signal , two conditions resulted in a ventral shift of the embryonic retina: Both lower division probability ( Figure 6C’’’ – condition II ) and the combination of lower division probability with circumferential division axis bias ( Figure 6C’’’’ – condition III ) . To identify which scenario was most plausible , we analysed patches in the ventral and non-ventral sectors . Both in experiments and all three simulated conditions , patch shape in the non-ventral sector was similar ( Figure 6D’–D’’’’ ) . Although there was a tendency for ventral clones to terminate more often , the width distribution of experimental NR patches did not differ substantially between non-ventral and ventral sectors ( Figure 6D’ , E’ , Figure 6—figure supplement 1D’; p=0 . 84 , Wilcoxon rank sum test ) . In contrast , this latter criterion was violated by two of the three simulated scenarios ( Figure 6D’’–D’’’’ and E’’–E’’’’ , Figure 6—figure supplement 1D’’-D’’’’ ) . In condition I , ventral ArCoS started narrow but then broadened ( Figure 6E’’ ) and interdigitated circumferentially ( Figure 6—figure supplement 1A , black arrowheads ) , unlike the very uniform stripes in the experimental data . The broader ventral ArCoS lead to a more dispersed distribution compared to the non-ventral sector ( Figure 6—figure supplement 1D’’; p=4 . 31∙10−14 , Wilcoxon rank sum test ) . In condition II , the majority of ventral ArCoS formed very narrow stripes , but at the border to the non-ventral sector ArCoS were broad and curved ( Figure 6—figure supplement 1B , black arrowheads ) . Again , this resulted in more shape variation ( Figure 6E’’’ ) . Nevertheless , these outliers were outweighed by a high density of narrow clones , such that the overall distribution was similar between ventral and non-ventral sectors ( Figure 6—figure supplement 1D’’’; p=0 . 12 , Wilcoxon rank sum test ) . Clones in the ventral and non-ventral sectors were qualitatively similar in condition III ( Figure 6E’’’’ , Figure 6—figure supplement 1C ) . Ventral clones however tended to be broader , resulting in a more dispersed distribution compared to the non-ventral sector ( Figure 6—figure supplement 1D’’’’; p=7 . 29∙10-7 , Wilcoxon rank sum test ) . In conclusion , ventral NR stem cells have a different behavior than elsewhere along the circumference , leading to a ventral-ward shift of the embryonic retina . The simulations suggest that this different behavior consists of modulation of proliferation parameters by an extrinsic signal in the ventral CMZ .
The coordinated growth of multiple independent tissues is a ubiquitous process in biology . In this work , we used the post-embryonic growth of NR and RPE in the eye of medaka as a model system of coordination in an organ where both growth and shape must be precisely regulated . Eye size in fish scales to the body size ( Lyall , 1957; Johns and Easter , 1977 ) . Body size , and thus eye growth rates greatly vary among individuals and depend on environmental factors ( Johns , 1981 ) . This natural malleability implies that feedback coupling plays a dominant role rather than the precise parametrization of each tissue growth and cell proliferation rate . Our simulations showed that inducer and responder growth modes impacted on variability in cell division timing , ultimately resulting in distinct clonal patterns that reproduced the experimentally observed differences between NR and RPE . RPE cells divided with high variability , indicative of periods of long quiescence where they waited for proliferative cues . NR cells displayed lower variability , supporting an upstream role in regulating growth ( Figure 7A ) . Although our implementation of the responder growth mode used a mechanical stimulus ( local cell density ) , a biochemical stimulus could equally well represent the system . Our model highlights an underappreciated mechanism whereby tissues coordinate by inducer and responder roles . Such division of labor among tissues might apply more generally to multiple organ systems , for example hair follicle cells in mouse induce the growth of underlying adipose tissue through hedgehog signalling ( Zhang et al . , 2016 ) . Intriguingly , hedgehog signalling also regulates the NR/RPE boundary in the CMZ of medaka ( Reinhardt et al . , 2015 ) , suggesting that signals mediating coordination of proliferative cell populations might be conserved . The topology of the retinal niche lead to a spatially biased neutral drift where stem and progenitor compartments spontaneously emerged . All virtual cells had equal potency , yet only a fraction realized their full stem cell potential . Peripheral cells had a high chance to become canalized in a stem cell fate , while central cells were more likely to act as progenitor cells with limited proliferation potential ( Figure 7B ) . Our experimental data support a spatially biased neutral drift . Fusion of clones may have lead us to overestimate ArCoS deriving from the central domain , which represent progenitors reverting to a stem cell fate . Nevertheless , terminated clones arising from the very periphery of the niche unambiguously demonstrate that some stem cells failed to self-renew throughout the life of the animal . Moreover , our finding that only cells in the first two rows of the CMZ have stem cell potential is consistent with in vivo time-lapse data ( Wan et al . , 2016; Tang et al . , 2017 ) . Interestingly , retrograde movement of row 2 cells into row 1 of the CMZ occurs in vivo ( Wan et al . , 2016 ) , which we also observed in our simulations . CMZ progenitor cells can be subdivided into two populations ( Harris and Perron , 1998; Raymond et al . , 2006 ) : First , peripheral multipotent progenitors ( i . e . able to generate all retinal neurons and glia ) which differ from stem cells only in their proliferative potential . Second , central progenitors that are restricted both in proliferative and differentiation potential , which likely act as a transit-amplifying zone , both increasing the proliferative output and cross-regulating to produce a full neuronal complement with the correct proportions of cell types ( Pérez Saturnino et al . , 2018 ) . Our data support an alternative model that identifies peripheral multipotent progenitors as stem cells that have been outcompeted . All terminated clones we examined were multipotent and spanned all retinal layers ( Figure 1—figure supplement 2 ) . Thus , as in many other systems ( Clevers and Watt , 2018 ) , our work highlights the limitation of strictly defining stem cells as infinitely self-renewing , or a posteriori based on their ArCoS-forming capacity . Importantly , although stochastic competition is most apparent in the early phase after clonal induction , it occurs continuously as demonstrated by late arising patches ( Figure 3E ) and nested inductions ( Figure 4A’–A’’ ) . The shift from an ‘early stochastic’ to ‘late polyclonal’ growth observed in other systems ( Nguyen et al . , 2017 ) may simply result from clonal growth masking the underlying stochasticity . Due to this stochasticity , it is impossible to tell at any moment with absolute certainty if a given cell will perpetually function as a stem cell . Neutral drift in a finite-sized environment such as adult mammalian tissues must ultimately result in a monoclonal niche ( Snippert et al . , 2010; Colom and Jones , 2016; Clevers and Watt , 2018 ) . In fish , homeostatic growth expands niches , and thus the number of stem cells increases ( Centanin et al . , 2011 ) . In principle , niche expansion reduces the impact of competition on clonal loss , but does not completely abolish it . Indeed , neutral drift leads to gradual loss of polyclonality in the intestine and muscle of fish ( Aghaallaei et al . , 2016; Nguyen et al . , 2017 ) . Organs may limit monoclonal drift by physically isolating niches ( Aghaallaei et al . , 2016 ) . In the intestine of both mammals and fish , physical isolation of multiple niches results in a polyclonal organ built up of monoclonal units ( Snippert et al . , 2010; Aghaallaei et al . , 2016 ) . In contrast , the CMZ is a physically contiguous niche that nevertheless maintains polyclonality lifelong both in the NR and the RPE ( Centanin et al . , 2011; Centanin et al . , 2014 ) . As shown in this work , the retina is not devoid of stochastic competition . Then how does it conserve its polyclonality ? Conceptually , the clonal growth of the retina resembles a population expanding into a new habitat , as studied in the context of evolutionary theory ( Hallatschek and Nelson , 2010 ) . Specifically for a radially expanding population , it has been mathematically proven that ( assuming pure neutral genetic drift ) no single clone will ever take over and clonal sectors perpetually coexist ( Hallatschek and Nelson , 2010; Korolev et al . , 2012 ) . Growth of the perimeter is faster than circumferential expansion of clones , thus preserving population diversity ( Hallatschek and Nelson , 2010 ) . Interestingly , in the NR , the biased division axis further reduces competition ( Figure 5 ) , thus increasing niche polyclonality . In summary , the geometry of the CMZ niche prohibits the total loss of polyclonality . Our analysis of NR cell divisions implies that cells sense the radius of the eye to regulate organ shape . Across vertebrates , the retina integrates visual input to adapt organ shape to optimize optics , a process called ‘emmetropization’ ( Wallman and Winawer , 2004 ) . In chicken , emmetropization is regulated by specialized neurons distributed across the retina that send their axons to the CMZ , implicating the CMZ in regulation of eye shape ( Fischer et al . , 2008 ) . Visual cues also guide emmetropization in fish ( Kröger and Wagner , 1996; Shen et al . , 2005; Shen and Sivak , 2007 ) . Eye growth in young fish predominantly occurs by cell addition , while in older fish CMZ proliferation decreases ( Johns , 1981 ) coincident with a decrease in emmetropization plasticity ( Shen and Sivak , 2007 ) . Thus , in fish , emmetropization correlates with CMZ proliferation . Experiments in chicken and zebrafish support the existence of two principal axes of stem cell division , that is circumferential and central-peripheral ( Fischer et al . , 2008; Ritchey et al . , 2012; Wan et al . , 2016 ) . Notably , the predominance of central-peripheral divisions and decreasing frequency over time of circumferential divisions in CMZ stem cells that is predicted by Equation 3 is supported by in vivo imaging data ( Wan et al . , 2016 ) and previous long-term clonal analyses ( Centanin et al . , 2014 ) . Altogether , the data support a model where the NR perceives the retinal radius through visual cues , and that cell divisions in the NR contribute to shaping the eye . The retinae of many fishes grow asymmetrically , perhaps to maintain the relative positions of receptive fields of neurons ( Johns , 1977; Johns , 1981; Easter , 1992 ) . Ecology dictates a distribution of subdomains enriched in specialized neuronal circuits and retinal cell subtypes ( Zimmermann et al . , 2018 ) . Interestingly , in green sunfish , the area that grows slowest displays highest visual acuity ( Cameron , 1995 ) . Medaka predominantly gaze upwards in their native shallow rice paddies , and a higher ventral acuity has been presumed based on photoreceptor densities ( Nishiwaki et al . , 1997 ) . Thus , slower ventral growth may have evolved to match ecological requirements for medaka vision . Our in silico screen identified three scenarios consistent with asymmetric ventral growth . Based on clonal patterns , an extrinsic signal driving lower proliferation ( and potentially also circumferential divisions ) appears most plausible . Experimental eye re-orientation in vivo implied an eye-internal mechanism independent on body axes or visual cues in regulating asymmetric retinal growth ( Cameron , 1996 ) . The origin of this signal and how it scales with the growing eye to always affect a similarly-sized retinal sector remains to be elucidated . The retina integrates global systemic cues such as nutrition to scale with body size ( Johns and Easter , 1977 ) , local eye-internal cues to generate an asymmetric retinal topology ( Cameron , 1996 ) , and external visual cues to adapt the shape of the organ ( Kröger and Wagner , 1996; Shen and Sivak , 2007 ) . In chicken and goldfish , visual cues and nutrients feed into the CMZ through growth factor signalling ( Boucher and Hitchcock , 1998; Fischer et al . , 2008; Ritchey et al . , 2012 ) . We propose that NR cells in the CMZ act as a hub to coordinate organ growth; in the eye of fish , this happens at the level of cell proliferation parameters , which affect eye growth , eye shape , and retinal topology ( Figure 7C ) . Indeterminate , lifelong growth is a widespread evolutionary strategy ( Karkach , 2006 ) . Given the geometrical constraints of the eye with respect to optics , a peripheral proliferative domain is the most parsimonious architecture to ensure that the differentiated neuronal cell mosaic is not disturbed by constant proliferation . Fishes are the largest vertebrate clade , with a huge diversity of eye shapes , such as cylindrical eyes in deep-sea fish ( Fernald , 1990 ) . By modulating CMZ proliferation parameters , evolution can adapt whole-organ morphogenesis to perfectly fit to the species’ ecological niche . | By the time babies reach adulthood , they have grown many times larger than they were at birth . This development is driven by an increase in the number and size of cells in the body . In particular , special types of cells , called stem cells , act as a reservoir for tissues: they divide to create new cells that will mature into various specialized structures . The retina is the light-sensitive part of the eye . It consists of the neural retina , a tissue that contains light-detecting cells , which is supported by the retinal pigment epithelium or RPE . In fish , the RPE and neural retina are replenished by distinct groups of stem cells that do not mix , despite the tissues being close together . Unlike humans , fish grow throughout adulthood , and their eyes must then keep pace with the body . This means that the different tissues in the retina must somehow coordinate to expand at the same rate: otherwise , the retina would get wrinkled and not work properly . Tsingos et al . therefore wanted to determine how stem cells in the neural retina and RPE co-operated to produce the right number of new cells at the right time . First , stem cells in the eyes of newly hatched fish were labelled with a visible marker so that their divisions could be tracked over time to build cell family trees . This showed that stem cells behaved differently in the neural retina and the RPE . Computer simulations of the growing retina explained this behavior: stem cells in the neural retina were telling the RPE stem cells when it was time to divide . Combining results from the simulations with data from the experiments revealed that a stem cell decided to keep up dividing partly because of its position in the tissue , and partly because of random chance . To be healthy , the body needs to fine-tune the number of cells it produces: creating too few cells may make it difficult to heal after injury , but making too many could lead to diseases such as cancer . Understanding how tissues normally agree to grow together could therefore open new avenues of treatment for these conditions . | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"developmental",
"biology",
"computational",
"and",
"systems",
"biology"
] | 2019 | Retinal stem cells modulate proliferative parameters to coordinate post-embryonic morphogenesis in the eye of fish |
Molecular mimicry is an evolutionary strategy adopted by viruses to exploit the host cellular machinery . We report that SARS-CoV-2 has evolved a unique S1/S2 cleavage site , absent in any previous coronavirus sequenced , resulting in the striking mimicry of an identical FURIN-cleavable peptide on the human epithelial sodium channel α-subunit ( ENaC-α ) . Genetic alteration of ENaC-α causes aldosterone dysregulation in patients , highlighting that the FURIN site is critical for activation of ENaC . Single cell RNA-seq from 66 studies shows significant overlap between expression of ENaC-α and the viral receptor ACE2 in cell types linked to the cardiovascular-renal-pulmonary pathophysiology of COVID-19 . Triangulating this cellular characterization with cleavage signatures of 178 proteases highlights proteolytic degeneracy wired into the SARS-CoV-2 lifecycle . Evolution of SARS-CoV-2 into a global pandemic may be driven in part by its targeted mimicry of ENaC-α , a protein critical for the homeostasis of airway surface liquid , whose misregulation is associated with respiratory conditions .
The surface of SARS-CoV-2 virions is coated with the spike ( S ) glycoprotein , whose proteolysis is key to the infection lifecycle . After the initial interaction of the S-protein with the ACE2 receptor ( Walls et al . , 2020 ) , host cell entry is mediated by two key proteolytic steps . The S1 subunit of the S-protein engages ACE2 , and viral entry into the host cell is facilitated by proteases that catalyze S1/S2 cleavage ( Belouzard et al . , 2012; Belouzard et al . , 2009 ) at Arginine-667/Serine-668 ( Figure 1a ) . This is followed by S2’ site cleavage that is required for fusion of viral-host cell membranes ( Hoffmann et al . , 2020; Walls et al . , 2020 ) .
We hypothesized that the virus may mimic host substrates to achieve proteolysis . Comparing human-infecting SARS-CoV-2 with SARS-CoV strains , as well as with candidates of zoonotic origin ( Pangolin-CoV and Bat-CoV RaTG13 ) , shows that SARS-CoV-2 has evolved a unique sequence insertion at the S1/S2 site ( Zhang et al . , 2020; Figure 1a ) . Although the S protein of SARS-CoV-2 shares high sequence identity with the S proteins of Pangolin-CoV ( 92% ) and Bat-CoV RaTG13 ( 97% ) , the furin insertion site seems to be uniquely acquired by SARS-CoV-2 . The resulting tribasic 8-mer peptide ( RRARSVAS ) on the SARS-CoV-2 S1/S2 site is conserved among 10 , 956 of 10 , 967 circulating strains deposited at GISAID ( https://www . gisaid . org/ ) ( Elbe and Buckland-Merrett , 2017 ) , as of April 28 , 2020 ( Supplementary file 1a ) . This peptide is also absent in over 13 , 000 non-COVID-19 coronavirus S-proteins from the VIPR database ( Carrillo-Tripp et al . , 2009 ) . Strikingly , examining over 10 million peptides ( 8-mers ) of 20 , 350 canonical human proteins from UniProtKB shows that the peptide of interest ( RRARSVAS ) is present exclusively in human ENaC-ɑ , also known as SCNN1A ( p-value=4E-4 ) ( see Materials and methods ) . The location of this SARS-CoV-2 mimicked peptide in the ENaC-ɑ structure is in the extracellular domain ( Noreng et al . , 2018; Figure 1b ) . This suggests that the SARS-CoV-2 may have specifically evolved to mimic a human protease substrate . ENaC regulates sodium ion ( Na+ ) and water homeostasis , and ENaC’s expression levels are controlled by aldosterone and the associated Renin-Angiotensin-Aldosterone System ( RAAS ) 6 . In distal lung airways , ENaC is known to play a key role in controlling fluid reabsorption at the air–liquid interface ( Rossier and Stutts , 2009 ) , and similar to SARS-CoV2 , ENaC-ɑ also needs to be proteolytically activated for its function ( Vallet et al . , 1997 ) . FURIN cleaves the equivalent peptide on mouse ENaC-ɑ between the Arginine and Serine residues in the 4th and 5th positions respectively ( RSAR|SASS ) ( Hughey et al . , 2004a; Hughey et al . , 2004b ) , akin to the recent report establishing FURIN cleavage at the S1/S2 site of SARS-CoV-2 ( Walls et al . , 2020; Figure 1b ) . It is conceivable that human ENaC activation may be compromised in SARS-CoV-2 infected cells , for instance by SARS-CoV-2 exploiting host FURIN for its own activation . The likely consequence would be low ENaC activity on the surface of the airways leading to compromised fluid reabsorption ( Planès et al . , 2010; Yurdakök , 2010 ) , an important lung pathology in COVID-19 patients with acute respiratory distress syndrome ( ARDS ) . Indeed , the exact mechanism of SARS-CoV-2’s potential impact of ENaC activation needs to be investigated . Although the furin-like cleavage motifs can be found in other viruses ( Coutard et al . , 2020 ) , the exact mimicry of human ENaC-ɑ cleavage site raises the specter that SARS-CoV-2 may be hijacking the protease network of ENaC-ɑ for viral activation . We asked whether there is an overlap between putative SARS-CoV-2 infecting cells and ENaC-ɑ expressing cells . Systematic single cell expression profiling of the ACE2 receptor and ENaC-ɑ was performed across human and mouse samples comprising ~1 . 3 million cells ( Venkatakrishnan et al . , 2020; Figure 1c ) . Interestingly , ENaC-ɑ is expressed in the nasal epithelial cells , type II alveolar cells of the lungs , tongue keratinocytes , and colon enterocytes ( Figure 1c and Figure 2—figure supplements 1–6 ) , which are all implicated in COVID-19 pathophysiology ( Shweta et al . , 2020; Venkatakrishnan et al . , 2020 ) . Further , ACE2 and ENaC-ɑ are known to be expressed generally in the apical membranes of polarized epithelial cells ( Butterworth , 2010; Musante et al . , 2019 ) . The overlap of the cell-types expressing ACE2 and ENaC-ɑ , and similar spatial distributions at the apical surfaces , suggest that SARS-CoV-2 may be leveraging the protease network responsible for ENaC cleavage . Beyond FURIN , which cleaves the S1/S2 site ( Walls et al . , 2020 ) , we were intrigued by the possibility of other host proteases also being exploited by SARS-CoV-2 . We created a 160-dimensional vector space ( 20 amino acids x eight positions on the peptide ) for assessment of cleavage similarities between the 178 human proteases with biochemical validation from the MEROPS database ( see Materials and methods; 0 < protease similarity metric <1 ) ( Rawlings et al . , 2018 ) . This shows that FURIN ( PCSK3 ) has overall proteolytic similarity to select PCSK family members , specifically PCSK5 ( 0 . 99 ) , PCSK7 ( 0 . 99 ) , PCSK6 ( 0 . 99 ) , PCSK4 ( 0 . 98 ) , and PCSK2 ( 0 . 94 ) ( Supplementary file 1b ) . It is also known that the protease PLG cleaves the ɣ-subunit of ENaC ( ENaC-ɣ ) ( Passero et al . , 2008 ) . In order to extrapolate the tissue tropism of SARS-CoV-2 from the lens of the host proteolytic network , we assessed the co-expression of these proteases concomitant with the viral receptor ACE2 and ENaC-ɑ ( Figure 2 ) . This analysis shows that FURIN is expressed with ACE2 and ENaC-ɑ in the colon ( immature enterocytes , transit amplifying cells ) and pancreas ( ductal cells , acinar cells ) of human tissues , as well as tongue ( keratinocytes ) of mouse tissues . PCSK5 and PCSK7 are broadly expressed across multiple cell types with ACE2 and ENaC-ɑ , making it a plausible broad-spectrum protease that may cleave the S1/S2 site . In humans , concomitant with ACE2 and ENaC-ɑ , PCSK6 appears to be expressed in cells from the intestines , pancreas , and lungs , whereas PCSK2 is noted to be co-expressed in the pancreas ( Figure 2 ) . It is worth noting that the extracellular proteases need not necessarily be expressed in the same cells as ACE2 and ENaC-ɑ . Among the PCSK family members with the potential to cleave the mimicked 8-mer peptide , it is intriguing that the same tissue can house multiple proteases and also that multiple tissues do share the same set of proteases .
Our findings emphasize that redundancy may be wired into the mechanisms of host proteolytic activation of SARS-CoV-2 . This study should stimulate the design of experiments that confirm the working hypothesis generated by our unbiased and systematic computational analysis . The mimicry of a cleavable host peptide central to pulmonary , renal , and cardiovascular function provides a new perspective to the evolution of SARS-CoV-2 in causing a global coronavirus pandemic .
The complete S-protein sequence for SARS-CoV ( Uniprot ID: P59594 ) and SARS-CoV-2 was obtained from uniprot ( ftp://ftp . uniprot . org/pub/databases/uniprot/pre_release/ ) . The sequences of Pangolin-CoV and Bat-CoV RaTG13 were obtained from the VIPR database ( https://www . viprbrc . org/ ) . Sequence alignments using Clustal-W , and comparison of SARS-CoV-2 versus other coronavirus strains were performed using JalView17 . We enumerated 10 , 257 , 893 ( 10 . 26M ) 8-mers from 20 , 350 reviewed uniprot reference sequences from human proteome ( Proteome ID: UP000005640 , as accessed on May 4th 2020 ) . The previously identified SARS-CoV-2 8-mer ‘RRARSVAS’ was in fact found in ENaC-ɑ protein ( Uniprot ID: P37088; p-value ≈ 10 . 26M/208 = 4E-4; chance of finding that particular 8-mer anywhere in the reference sequences ) . The position frequency matrix ( PFM ) of the individual proteases obtained from the MEROPS database ( Rawlings et al . , 2018 ) was converted to a probability weight matrix ( PWM ) ( normalized and scaled ) ( Supplementary file 1b ) . Out of 178 proteases , there were 146 proteases that had specificity information available on the eight mer peptide spanning the cleavage site ( ±4 ) . The 20 ( amino acids ) x 8 ( position ) matrix defined for each of the proteases were flattened into a single vector with 160 elements . We performed a cosine similarity calculation between all pairs ( X , Y ) of protease specificity vector . The similarity was derived as the normalized dot product of X and Y: K ( X , Y ) = <X , Y> / ( ||X||*||Y|| ) ) . We performed a systematic expression profiling of the ACE2 and ENaC-ɑ across 65 published human and mouse single-cell studies comprising ~1 . 3 million cells using nferX Single Cell platform ( Supplementary file 1c , https://academia . nferx . com/ ) ( Venkatakrishnan et al . , 2020 ) . The ACE2 expression could be detected in 66 studies ( 59 studies of human samples and 7 studies of mouse samples ) spanning across ~50 tissues , over 450 cell-types and ~1 . 05 million cells . In order to call a given cell-type to be positive for both ACE2 and a protease we applied a cutoff of 1% of the cells in the total cell-type cluster population to have a non-zero count associated with both ACE2 and the respective protease . The mean expression of the proteases , ENaC-ɑ and ACE2 was derived for individual cell population within each of the studies . The cell-type information was obtained from the author annotations provided for each of the studies . The analysis was performed separately on the mouse and human datasets . For each protease , the mean expression of given cell-population ( mean log[cp10k +1] counts ) was Z-score normalized ( to ensure the sd = 1 and mean ~0 for all the genes ) to obtain relative expression profiles across all the samples . The same normalization was applied to ACE2 and ENaC-ɑ and both human and mouse datasets were analyzed independently by generating heatmaps . The cell types having zero-expression values of ACE2 were also included as negative control to probe the expression of various proteases . We performed an analysis to identify the cell types with significant overlap of ACE2 and ENaC-ɑ expression . To this end , we shortlisted cell types in which ENaC-ɑ is expressed in a significantly higher proportion of ACE2-expressing cells than in the overall population of cells of that sub-type . We computed the ratios of these proportions , and used a corresponding Fisher exact test to compute significance . | Viruses hijack the cellular machinery of humans to infect their cells and multiply . The virus causing the global COVID-19 pandemic , SARS-CoV-2 , is no exception . Identifying which proteins in human cells the virus co-opts is crucial for developing new ways to diagnose , prevent and treat COVID-19 infections . SARS-CoV-2 is covered in spike-shaped proteins , which the virus uses to gain entry into cells . First , the spikes bind to a protein called ACE2 , which is found on the cells that line the respiratory tract and lungs . SARS-CoV-2 then exploits enzymes called proteases to cut , or cleave , its spikes at a specific site which allows the virus to infiltrate the host cell . Proteases identify which proteins to target based on the sequence of amino acids – the building blocks of proteins – at the cleavage site . However , it remained unclear which human proteases SARS-CoV-2 co-opts and whether its cut site is similar to human proteins . Now , Anand et al . show that the spike proteins on SARS-CoV-2 may have the same sequence of amino acids at its cut site as a human epithelial channel protein called ENaC-α . This channel is important for maintaining the balance of salt and water in many organs including the lungs . Further analyses showed that ENaC-α is often found in the same types of human lung and respiratory tract cells as ACE2 . This suggests that SARS-CoV-2 may use the same proteases that cut ENaC-α to get inside human respiratory cells . It is possible that by hijacking the cutting mechanism for ENaC-α , SARS-CoV-2 interferes with the balance of salt and water in the lungs of COVID-19 patients . This may help explain why the virus causes severe respiratory symptoms . However , more studies are needed to confirm that the proteases that cut ENaC-α also cut the spike proteins on SARS-CoV-2 , and how this affects the respiratory health of COVID-19 patients . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"medicine",
"computational",
"and",
"systems",
"biology"
] | 2020 | SARS-CoV-2 strategically mimics proteolytic activation of human ENaC |
Oxidative protein folding in the endoplasmic reticulum ( ER ) has emerged as a potentially significant source of cellular reactive oxygen species ( ROS ) . Recent studies suggest that levels of ROS generated as a byproduct of oxidative folding rival those produced by mitochondrial respiration . Mechanisms that protect cells against oxidant accumulation within the ER have begun to be elucidated yet many questions still remain regarding how cells prevent oxidant-induced damage from ER folding events . Here we report a new role for a central well-characterized player in ER homeostasis as a direct sensor of ER redox imbalance . Specifically we show that a conserved cysteine in the lumenal chaperone BiP is susceptible to oxidation by peroxide , and we demonstrate that oxidation of this conserved cysteine disrupts BiP's ATPase cycle . We propose that alteration of BiP activity upon oxidation helps cells cope with disruption to oxidative folding within the ER during oxidative stress .
A largely unrecognized yet significant source of intracellular reactive oxygen species ( ROS ) is oxidative folding within the lumen of the endoplasmic reticulum ( ER ) . Ero1 is a primary enzymatic catalyst of biosynthetic disulfide bonds ( Araki and Inaba , 2012; Ramming and Appenzeller-Herzog , 2012 ) . A byproduct of Ero1 activity is peroxide , a ROS . Studies with recombinant Ero1 demonstrate that for every disulfide bond generated by Ero1 a peroxide molecule is formed ( Tu and Weissman , 2002; Gross et al . , 2006 ) . Extrapolating from these in vitro results , it is estimated that Ero1 activity in living cells could generate up to 25% of the cellular ROS produced during protein synthesis ( Tu and Weissman , 2004 ) . Fluorescent probes that measure ROS in living cells indicate that the amount of ROS in the ER lumen exceeds the quantity of ROS within the mitochondria , a substantial and well-characterized source of intracellular ROS ( Malinouski et al . , 2011 ) . Recently , several systems that function to consume ROS in the ER lumen have been uncovered in mammalian cells . A peroxiredoxin ( PRDX4 ) and two glutathione peroxidases ( GPx7/8 ) have been demonstrated to reduce peroxide to water ( Tavender and Bulleid , 2010; Zito et al . , 2010; Nguyen et al . , 2011 ) . These enzymes enhance the efficiency of oxidative protein folding by coupling peroxide reduction with oxidation of the disulfide-bond forming enzyme PDI , ultimately converting peroxide generated as a byproduct of disulfide bond formation into additional nascent chain disulfides ( Tavender et al . , 2010; Zito et al . , 2010; Nguyen et al . , 2011 ) . We expect these identified detoxification pathways operate within the ER alongside additional and more general systems conserved across eukaryotes that remain to be elucidated . Lower eukaryotes maintain robust Ero1-dependent protein oxidation pathways that produce ROS , however these organisms do not contain homologs of either mammalian PDRX4 or GPx7/8 . To discover systems that respond to redox imbalance within the ER , we capitalized on our prior identification of a key homeostatic feedback system that modulates the activity of Ero1 in disulfide bond formation ( Sevier et al . , 2007 ) . Regulation of Ero1 activity normally serves to maintain redox homeostasis in the ER lumen and helps to moderate the amount of ROS produced by Ero1 . Disruption of the Ero1 feedback loop ( de-regulation of Ero1 ) is detrimental to the ER redox environment , resulting in severe redox imbalance in the ER and oxidative stress ( Sevier et al . , 2007 ) . We took advantage of the ROS generated by overproduction of a de-regulated , constitutively active mutant of Ero1 ( Ero1-C150A-C295A; hereafter referred to as Ero1* ) to identify cellular systems activated in response to an overabundance of ER ROS . Here we report that oxidative stress in the ER of yeast , created by Ero1* overproduction , results in direct modification by peroxide of a conserved cysteine in the ATP-binding site of the molecular chaperone BiP . We show that modification of BiP decouples the ATPase and peptide-binding activities of BiP , disrupting the normal allostery between these domains . We propose that oxidation of BiP during conditions of increased ER ROS both enhances the ability of BiP to bind polypeptides and prevents consumption of cellular ATP by hydrolysis . We suggest that this conserved ROS-sensing mechanism has evolved to protect against oxidative stress in the ER by minimizing protein aggregation and maintaining ER folding homeostasis during conditions of excess ROS .
A search of ER-localized protein sequences for potentially redox-responsive amino acid ( s ) drew our attention to a cysteine residue highly conserved among BiP orthologs . This cysteine is localized within the ATP-binding pocket ( Figure 1 ) and is the sole cysteine in yeast BiP ( Kar2 Cys63 ) . To probe the role of this conserved cysteine , we replaced the cysteine in yeast BiP with alanine and found that the cysteine is not required for the essential cellular function of BiP . Specifically , we observed that a kar2-C63A allele can complement an otherwise inviable KAR2 chromosomal deletion strain ( kar2Δ ) ( Figure 1C ) . Viability of the cysteine-less Kar2 strain was not a consequence of upregulation of the unfolded protein response ( UPR ) , which has the potential to compensate for decreased Kar2 function through a transcription-mediated increase in Kar2 levels ( Beh and Rose , 1995 ) . As evidence , an UPR-promoter element ( UPRE ) -lacZ reporter showed a similar basal level of UPR induction for both wild-type and kar2-C63A strains ( Figure 1D ) . BiP is not only a target of the UPR but also modulates UPR induction ( Pincus et al . , 2010 ) . Replacement of the BiP cysteine with alanine did not perturb the UPR response; treatment with the reductant dithiothreitol ( DTT ) , which disrupts disulfide bond formation and causes unfolded proteins to accumulate in the ER lumen , resulted in a comparable UPR induction in both wild-type and kar2-C63A strains ( Figure 1D ) . 10 . 7554/eLife . 03496 . 003Figure 1 . BiP contains a conserved cysteine that is dispensable for yeast viability . ( A ) Ribbon diagram of human BiP nucleotide binding domain in complex with calcium-ADP ( PDB entry 3IUC ) ( Wisniewska et al . , 2010 ) . ADP and the conserved BiP cysteine are shown as colored sticks . Calcium atoms are represented as blue spheres . ( B ) Magnified representation of the conserved cysteine and CaADP from the BiP structure in panel A . Additional amino acid side chains are shown as lines . ( C ) CSY214 containing the plasmids pCS681 , pCS685 , or empty vector were spotted onto SMM plates with or without 5-fluoroorotic acid ( 5-FOA ) and incubated for 2 d at 30°C . ( D ) CSY5 or CSY275 containing a UPRE-lacZ reporter plasmid ( pJC8 ) were cultured in SMM-ura at 30°C , treated with 0 or 2 mM dithiothreitol ( DTT ) for 2 hr , and assayed for beta–galactosidase activity . Three independent transformants of each strain were grown and assayed in duplicate . Data represent the mean of averaged values for the three transformants ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 03496 . 003 Although a kar2-C63A mutant showed no obvious defects in characteristic BiP activities , kar2-C63A cells were highly sensitive to hyper-oxidation of the ER lumen by overexpression of Ero1* . kar2-C63A strains containing either integrated ( Figure 2A ) or plasmid-borne ( Figure 2B ) ERO1* alleles showed an increased sensitivity to hyper-oxidation of the ER by Ero1* overexpression relative to a wild-type strain . In Figure 2A , ERO1* was integrated into the yeast genome ( at the CAN1 locus ) , allowing for stable and uniform expression of Ero1* . In Figure 2B , Ero1* expression was induced from a plasmid as described previously ( Sevier et al . , 2007 ) . Importantly , overexpression of a catalytically inactive version of ERO1* did not affect growth of the kar2-C63A strain relative to the wild-type strain ( Figure 2B ) , confirming that the loss of viability upon Ero1* expression in the kar2-C63A mutant is a consequence of the hyper-oxidizing activity of Ero1* . As noted above , the kar2-C63A mutant alone ( without Ero1* overexpression ) grew at a rate similar to wild-type ( Figure 2A , B ) . 10 . 7554/eLife . 03496 . 004Figure 2 . A cysteine-less BiP strain is sensitive to increased ER oxidation . ( A ) CSY170 and CSY278 strains containing an integrated galactose-inducible ERO1* were spotted onto SMM or SMM Gal plates , and plates were incubated for 2 d ( glucose ) or 3 d ( galactose ) at 30°C . ( B ) CSY5 or CSY275 strains transformed with plasmids pCS452 , pCS504 , or empty vector were spotted onto SMM-ura or SMM Gal-ura plates and incubated for 2 d ( glucose ) or 3 d ( galactose ) at 30°C . ( C and D ) CSY5 or CSY275 strains were spotted onto SMM plates containing 0–2 mM diamide ( C ) or DTT ( D ) , and plates were incubated at 30°C for 2 d . DOI: http://dx . doi . org/10 . 7554/eLife . 03496 . 004 Furthermore , the kar2-C63A mutant showed greater sensitivity than a wild-type strain to the small molecule oxidant diamide ( Figure 2C ) , indicating a general role for the BiP cysteine in giving resistance to cellular oxidative stress . The kar2-C63A and wild-type strains showed similar growth in the presence of the reductant DTT ( Figure 2D ) , suggesting that the conserved BiP cysteine has a selective role in managing oxidative but not reductive stress . We speculated that oxidation of the BiP cysteine may trigger the ability of wild-type BiP to promote cell growth during hyper-oxidizing ER conditions . Accordingly , we used a biotin-switch assay to monitor the redox state of the BiP cysteine during oxidative stress ( Jaffrey and Snyder , 2001; Figure 3A ) . Cells grown with or without Ero1* induction were lysed under acidic conditions to protonate free thiols , which decreases thiol reactivity and limits post-lysis oxidation . Lysates were subsequently brought to neutral pH in the presence of N-ethylmaleimide ( NEM ) , to block free thiols , followed by beta-mercaptoethanol ( BME ) treatment to reduce cysteines originally oxidized in the cell lysates . Cysteines originally oxidized in the cell , and subsequently reduced by BME , were then modified with maleimide-biotin . FLAG-tagged Kar2 was isolated from the lysates with anti-FLAG beads , separated by SDS-PAGE , and biotinylated Kar2 was detected by western blotting ( Figure 3B ) . 10 . 7554/eLife . 03496 . 005Figure 3 . BiP's cysteine is oxidized upon hyperoxidation of the ER by Ero1* . ( A ) Schematic for the biotin-switch assay used in panels B and C . ( B ) The biotin-switch assay was used on lysates prepared from strains CSY316 and CSY319 containing either pCS452 or an empty vector . Strains were grown in galactose medium to induce Ero1* , and Kar2 was immunoprecipitated from lysates postbiotin-maleimide treatment . The relative proportion of Kar2 with an oxidized cysteine in the cell lysates under stressed ( ERO1* ) and non-stressed conditions was determined by comparing the intensity of the Kar2-biotin signal relative to the total level of Kar2 . As a control ( lane 4 ) , no reductant was added postNEM treatment and prior to biotin-maleimide addition . ( C ) Lysates were prepared from CSY316 containing pCS452 grown in galactose medium . The biotin-switch assay was performed with BME ( as in panel B ) or sodium arsenite ( NaAsO2 ) as the reductant . Kar2 was immunoprecipitated from lysates postbiotin-maleimide treatment . ( D ) Schematic for the use of DAz-2 to detect sulfenic acid in panel E and F . ( E ) Strains were grown as in panel B and treated with DAz-2 . Kar2 was immunoprecipitated from lysates , and Staudinger ligation was performed with phosphine-biotin . DAz-2 addition was detected with an avidin probe . Note , a higher molecular weight Kar2 band corresponding to untranslocated Kar2 was observed in the Ero1*-overexpression strain ( lane 2 ) . Overlay of the two images confirmed that the avidin signal corresponds to the lower molecular weight mature Kar2 band ( data not shown ) . ( F ) Strains were grown in glucose medium , and cells were exposed to 5 mM cumene hydroperoxide ( CHP ) for 30 min prior to harvest . Treatment with DAz-2 and sample processing were as described in panel E . DOI: http://dx . doi . org/10 . 7554/eLife . 03496 . 005 Cells overexpressing Ero1* exhibited an approximately twofold increase in the fraction of biotinylated Kar2 ( Figure 3B , lanes 1 and 2 ) , demonstrating a BME-reducible modification on Kar2 under hyper-oxidizing ER conditions . No Kar2-biotin signal was detected upon Ero1* induction in a strain containing a cysteine-less variant of Kar2 ( Figure 3B , lane 3 ) , indicating biotinylation was limited to the Kar2 thiol . When lysates were not treated with BME prior to maleimide-biotin addition , a negligible biotin signal was detected for Kar2 ( Figure 3B , lane 4 ) , confirming that Kar2 biotinylation was not a consequence of residual Kar2 thiols available due to incomplete alkylation by NEM . There are many known thiol-oxidation outcomes that are reversible by BME , including disulfide bonds , nitrosothiols , sulfenic acids , and glutathiolated cysteines ( Chung et al . , 2013 ) . As a primary role of Ero1 is to facilitate disulfide bond formation in the ER , we suspected that Kar2 might form a disulfide bond with itself or another protein upon Ero1* overproduction . However , we did not observe a mobility shift on a non-reducing SDS-polyacrylamide gel for Kar2 isolated from cells expressing Ero1* , which would be expected if Kar2 was disulfide bonded to another protein ( data not shown ) . Consequently , we focused on anticipated changes in small molecular oxidants in the ER that could modify Kar2 . Specifically , as peroxide is a byproduct of Ero1* activity , we considered oxidation of the BiP cysteine thiol ( −SH ) to sulfenic acid ( −SOH ) . To determine if peroxide oxidizes the Kar2 cysteine , we replaced the reductant BME in the biotin-switch assay with sodium arsenite , a reductant selective for sulfenic acid ( Torchinsky , 1991 ) . Treatment with sodium arsenite resulted in Kar2 biotinylation ( Figure 3C , lane 1 ) , demonstrating that Kar2 undergoes direct modification by peroxide . These data were confirmed using the dimedone-based DAz-2 probe , which binds directly to sulfenic acid-modified proteins ( Figure 3D; Leonard et al . , 2009 ) . Kar2 proteins were isolated from cell lysates treated with DAz-2 , Staudinger ligation was performed to add a biotin-moiety to any Kar2-DAz-2 species , and Kar2 modified by DAz-2-biotin was detected by western blotting with an avidin probe . An increase in biotinylated Kar2 was observed in DAz-2-treated cells overexpressing Ero1* relative to cells grown in the absence of stressor ( Figure 3E ) , corroborating sulfenic acid formation at the site of the BiP cysteine during conditions of excess ER ROS . Oxidation of the Kar2 cysteine to sulfenic acid was detected also in cells exposed to exogenous peroxide . Addition of 5 mM cumene hydroperoxide ( CHP ) to cells for 30 min resulted in the recovery of biotinylated wild-type Kar2 from DAz-2 treated lysates , indicative of sulfenic acid formation on the Kar2 cysteine ( Figure 3F , lane 3 ) . Recovery of biotinylated-Kar2 was dependent on both the presence of the Kar2 cysteine and peroxide treatment ( Figure 3F ) . Sensitivity of the kar2-C63A strain to Ero1* overexpression implies that BiP cysteine oxidation normally protects wild-type cells against oxidative cellular damage and a corresponding loss of viability during Ero1* overproduction . To provide more direct evidence for BiP oxidation in protection against oxidative ER stress , we created BiP alleles designed to mimic the modified form of BiP . Alleles were generated that contained a negatively charged or bulky amino acid in place of the BiP cysteine . We reasoned that the aromatic groups may recapitulate a structural perturbation caused by conversion of cysteine to sulfenic acid . ( We will colloquially refer to these as bulky substitution alleles ) . Due to its charge , we anticipate aspartic acid may be a close mimetic of sulfenic acid , although it is likely an even better mimetic of sulfinic acid ( −SO2H ) , which is formed when sulfenic acid is further oxidized by peroxide . When expressed ectopically , each of these alleles protected cells against a loss of viability during conditions of increased ER ROS , effectively mimicking the phenotype anticipated for oxidation of the BiP cysteine . Specifically , we found that addition of a plasmid coding for Kar2-C63D , Kar2-C63F , Kar2-C63Y , or Kar2-C63W suppressed the growth defect of a kar2-C63A strain overexpressing Ero1* ( Figure 4 ) . The growth of these strains could not be attributed solely to an excess of Kar2; addition of a Kar2-C63A expressing plasmid did not allow for growth of cells overexpressing Ero1* ( Figure 4 ) . The differences in growth for strains carrying the kar2-C63D/F/Y/W plasmids vs a kar2-C63A plasmid were most striking when cells were exposed to a combination of Ero1* and heat ( 37°C ) ( Figure 4 ) ; heat may induce an additional folding burden on the ER , which may serve to accentuate the growth differences between strains . Notably , cells with kar2-C63F/Y/W plasmids grew better than strains containing a wild-type KAR2 plasmid ( Figure 4 ) ; these data may indicate an optimal ratio of unmodified-to-modified Kar2 that is closely matched by the combination of the genomic kar2-C63A and plasmid-borne kar2-C63F/Y/W alleles . The bulky alleles ( Kar2-C63F/Y/W ) were more protective than Kar2-C63D ( Figure 4 ) , suggesting that the Kar2-C63F/Y/W may be a more effective mimetic of the modified form than Kar2-C63D . 10 . 7554/eLife . 03496 . 006Figure 4 . Substitution of the BiP cysteine with an amino acid containing a negatively charged or large side chain enables protection against hyper-oxidation of the ER lumen . ( A and B ) CSY278 containing ( A ) plasmids pCS681 , pCS685 , pCS802 , or empty vector and ( B ) plasmids pCS681 , pCS685 , pCS687 , pCS688 , pCS750 , or empty vector were spotted on SMM-leu or SMM Gal-leu plates , and plates were incubated at 30°C and 37°C for 2 d ( glucose ) or 3 d ( galactose ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03496 . 006 Remarkably , although the negatively charged and bulky amino acid substitutions of Cys63 ( Kar2-C63D/F/Y/W ) protect against hyper-oxidizing conditions , these alleles lack essential BiP activity . KAR2 is an essential gene in yeast . A Kar2-C63W allele could not support growth of a chromosomal deletion of KAR2 ( Figure 5A ) ; Kar2-C63D , Kar2-C63F and Kar2-C63Y mutants were viable but temperature sensitive for growth ( Figure 5A , B ) . BiP normally facilitates nascent chain translocation into the ER lumen . Decreased polypeptide translocation into the ER was observed for the Kar2-C63D , Kar2-C63F , and Kar2-C63Y strains after a 90 min shift to 37°C , further corroborating a loss of vital BiP activity for these cysteine substitution mutants ( Figure 5C ) , Untranslocated polypeptides were readily detected by the accumulation of precursor protein forms ( pre-Kar2 , pre-PDI , Gas1 precursor ) that lack the characteristic size shifts associated with post-translational processing that occurs after translocation into the ER lumen , including signal sequence processing , glycosylation ( PDI , Gas1 ) , and GPI anchor addition ( Gas1 ) ( Figure 5C ) . In keeping with a loss of vital BiP activity and disruption of the ER folding environment , an approximately fivefold induction of the UPR was observed in the Kar2-C63D/F/Y strains at both permissive ( 24°C ) and restrictive ( 37°C ) temperatures relative to wild-type cells ( Figure 5D ) . UPR levels in the mutant strains at 37°C were comparable to those observed in cells treated with DTT , an established robust inducer of the UPR ( Figure 5D ) . Significantly , the phenotypes observed for Kar2-C63D , Kar2-C63F , and Kar2-C63Y alleles contrast markedly the Kar2-C63A mutant , which showed wild-type activity with respect to growth , protein translocation , and UPR activity ( Figure 5 ) and afforded no protection against oxidative stress ( Figure 4 ) . Note the ability of the loss-of-function Kar2-C63D/F/Y/W alleles to rescue the cell death phenotype from Ero1* overexpression ( Figure 4 ) was observed at 37°C , the temperature at which none of these alleles are viable as the sole copy of cellular BiP . 10 . 7554/eLife . 03496 . 007Figure 5 . Replacement of the BiP cysteine with aspartic acid , phenylalanine , tyrosine , or tryptophan results in decreased BiP function . ( A ) CSY214 containing the plasmids pCS681 , pCS802 , pCS687 , pCS688 , pCS750 or empty vector were spotted onto SMM plates with or without 5-fluoroorotic acid ( 5-FOA ) and incubated for 2 d at 30°C . ( B ) CSY289 , 290 , 368 , 292 , and 293 were spotted onto YPD plates and incubated at 24° , 30° , and 37°C . ( C ) Strains from B were cultured at 24°C to log-phase in YPD and shifted to 37°C for 90 min prior to harvest . Accumulation of unprocessed untranslocated forms of the proteins Kar2 , PDI , and Gas1 were detected by western blotting . ( D ) Strains from B containing an UPRE-lacZ reporter plasmid ( pJC8 ) were cultured in SMM-ura at 24°C to log-phase and shifted to 37°C ( with or without 2 mM DTT ) for 90 min prior to harvest . Samples were assayed for beta–galactosidase activity . Three independent transformants of each strain were grown and assayed in duplicate . Data represent the mean of averaged values for the three transformants ± SD . ( E–G ) ATP hydrolysis was assessed by determining the fraction of [alpha-32P]ATP converted to [alpha-32P]ADP as described in the 'Materials and methods' . Panels F and G show CHP and NEM-treated samples , respectively . Samples without chemical additions in panels F and G were mock treated to match the CHP or NEM-treatment . Data represent the means ± SD of three independent assays . DOI: http://dx . doi . org/10 . 7554/eLife . 03496 . 007 The conserved BiP cysteine is located in the ATPase domain ( Figure 1 ) , and we speculated that the decrease in cellular BiP function observed upon introduction of a charged or bulky side chain at the cysteine position was a byproduct of altered BiP ATPase activity . To monitor ATPase activity , we expressed wild-type and BiP cysteine mutants in Escherichia coli and assayed the purified recombinant proteins in vitro . As anticipated , Kar2-C63D , Kar2-C63F , Kar2-C63Y , and Kar2-C63W strains all exhibited negligible hydrolysis of ATP ( Figure 5E ) , whereas Kar2-C63A showed ATP hydrolysis activity equivalent to that observed for wild-type Kar2 ( Figure 5E ) . To correlate the bulky substitution BiP mutants with adduct formation at the BiP cysteine , we followed the ATPase activity of BiP after oxidation with peroxide ( CHP; Figure 5F ) or alkylation with NEM ( Figure 5G ) . Oxidized or alkylated BiP exhibited a loss of ATPase activity similar to that observed for the negative or bulky amino acid cysteine-substitution mutants ( Figure 5E–G ) . Treatment of Kar2-C63A with peroxide or NEM did not decrease ATP hydrolysis , confirming that the effect of peroxide and NEM was linked to cysteine modification ( Figure 5F , G ) . The block in polypeptide translocation into the ER observed in cells expressing the BiP oxidation mimetic alleles raised the interesting possibility that oxidation of the BiP cysteine may trigger a block in polypeptide movement into the ER lumen . To assess whether oxidation of BiP impedes polypeptide translocation , we checked if cells overexpressing Ero1* accumulate unprocessed untranslocated precursor proteins , which show characteristic size shifts on SDS-PAGE . We observed that Ero1* overproduction results in an inhibition of protein translocation , which was most striking when Ero1* was overproduced in a sensitized strain background ( an ire1Δ strain ) that is unable to induce the UPR ( Figure 6 ) . A robust translocation block was not as reproducible in wild-type ( IRE1+ ) cells , yet we did observe precursor protein accumulation in some assays using a wild-type strain background , which was dependent both on Ero1* overexpression and the presence of the Kar2 cysteine ( e . g . , Figure 3E , pre-Kar2 accumulation ) . We speculate the ire1Δ background facilitates a robust translocation block because there is no transcriptional upregulation of KAR2 upon Ero1*-induction ( due to the loss of a UPR response ) ( Sevier et al . , 2007 ) . Lower levels of BiP in the ire1Δ strain may result in a higher fraction of modified BiP upon Ero1* expression , which may lead to a more dramatic impact on ER translocation . 10 . 7554/eLife . 03496 . 008Figure 6 . Overexpression of Ero1* causes an accumulation of untranslocated polypeptides . ( A ) CSY44 and ( B ) CSY172 containing plasmids pAF112 ( ERO1; E ) , pCS452 ( ERO1*; E* ) , or empty vector were cultured in galactose medium for 5 hr to induce Ero1 and Suc2 expression . Accumulation of unprocessed untranslocated forms of the proteins CPY , Kar2 , PDI , Gas1 , and Suc2 were detected by western blotting . DOI: http://dx . doi . org/10 . 7554/eLife . 03496 . 008 As noted above , it is intriguing that the translocation block correlated with conditions ( Ero1* overproduction ) that result in BiP modification ( Figure 3 ) . It is also notable that the sporadically observed translocation block in IRE1+ cells expressing Ero1* is dependent on the presence of the Kar2 cysteine ( e . g . , Figure 3E , pre-Kar2 accumulation ) . Together these data suggest BiP modification may account for the observed translocation attenuation; however , at present we are unable to definitively link BiP oxidation with a block in polypeptide translocation . Attempts to facilitate a robust translocation block in wild-type cells with exogenous oxidant ( diamide or CHP ) , or a combination of oxidant ( Ero1* , diamide , CHP ) and heat ( 37°C ) , have been unsuccessful; none of these conditions stimulate a robust reproducible translocation attenuation in wild-type cells . Experiments to study the role of the BiP cysteine in an ire1Δ background , where we do observe a strong translocation attenuation , have also been uninformative . To test the role of the BiP cysteine , we constructed a kar2-C63A ire1Δ double mutant overexpressing Ero1* , which would be expected to show a less pronounced translocation block if BiP oxidation triggers translocation attenuation . Although viable , a kar2-C63A ire1Δ double mutant exhibits heterogeneous colony size and slow growth , and experiments to study translocation attenuation during oxidative stress with this strain have produced inconsistent results . Given the correlation between alleles of BiP that show a loss of ATPase activity and protection against oxidative stress , we asked whether ectopic addition of any BiP ATPase mutant could confer protection against over-oxidation of the ER . Mutation of a conserved threonine in the BiP ATPase domain to glycine has been established in both yeast ( T249 ) and hamster BiP ( T229 ) to prevent ATP hydrolysis ( Wei et al . , 1995; Steel et al . , 2004 ) . We expressed and purified recombinant Kar2-T249G and confirmed this mutant exhibits a severe defect in ATP hydrolysis relative to wild-type Kar2 ( Figure 7A ) . Strikingly , ectopic expression of a Kar2-T249G mutant was not able to suppress the growth defect of a kar2-C63A strain overexpressing Ero1* ( Figure 7B ) . These data suggest that decreased ATP hydrolysis by BiP , which we observed upon BiP oxidation , is not sufficient to allow for protection against redox imbalance in the ER lumen . These data imply that Kar2-C63D/F/Y/W mutants alter BiP activity in a manner distinct from Kar2-T249G . 10 . 7554/eLife . 03496 . 009Figure 7 . A BiP ATPase mutant is not sufficient to protect cells during oxidative stress . ( A ) ATP hydrolysis was assessed by determining the fraction of [alpha-32P]ATP converted to [alpha-32P]ADP as described in the 'Materials and methods' . Data represent the means ± SD of three independent assays . ( B ) CSY278 containing plasmids pCS681 , pCS774 , pCS687 , or empty vector were spotted on SMM-leu or SMM Gal-leu plates , and plates were incubated at 37°C for 3 d . ( C ) CSY275 containing a UPRE-lacZ reporter ( pCS852 ) and plasmids pCS681 , pCS802 , pCS774 , pCS687 , pCS688 , or pCS750 were cultured in SMM-ura-leu at 24°C to log-phase and shifted to 37°C ( with or without 2 mM DTT ) for 90 min prior to harvest . Three independent transformants of each strain were grown and assayed for beta–galactosidase activity in duplicate . Data represent the mean of averaged values for the three transformants ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 03496 . 009 The robust UPR induction we observed in cells containing Kar2-C63D/F/Y alleles as the sole copy of BiP ( Figure 5D ) raised the possibility that these ATPase deficient BiP alleles might also elicit a UPR response when introduced into the kar2-C63A background . We reasoned UPR induction by the ATPase deficient BiP alleles could facilitate the protection conferred by ectopic expression of these Kar2 mutants during oxidative stress ( Figure 4 ) . Notably , introduction of a plasmid-borne copy of a Kar2-C63D/F/Y/W allele to kar2-C63A cells did result in an enhanced UPR relative to cells expressing wild-type Kar2 ( or a Kar2-C63A mutant ) , primarily observed at high temperature ( 37°C ) ( Figure 7C ) . These data suggest UPR induction may contribute to the enhanced growth observed for these same strains under oxidative stress ( Figure 4 ) , and these data imply BiP oxidation during stress may trigger a UPR response . However , although UPR induction may augment growth during oxidative stress , UPR induction does not appear sufficient to manage the disruption to the ER environment in Ero1* overexpressing cells . Ectopic expression of the Kar2-T249G mutant in the kar2-C63A background ( which does not facilitate growth during oxidative stress ) resulted in a similar UPR as observed for the Kar2-C63D/F/Y/W mutants ( Figure 7C ) . Recent data show that it is possible to decouple the normal allostery observed between Hsp70 ATPase and peptide binding activities . The small molecule YM-08 ( an analog of the Hsp70 inhibitor MKT-077 ) has been shown to enhance Hsp70 binding to misfolded proteins and simultaneously inhibit Hsp70 ATPase activity ( Miyata et al . , 2013 ) . Similarly , several mutations in the bacterial Hsp70 DnaK have been reported that allow DnaK to prevent protein aggregation in the absence of robust ATPase activity ( Chang et al . , 2010 ) . These data raised the interesting possibility that oxidation of BiP could be a physiological means to allosterically decouple BiP activity , allowing BiP to prevent protein aggregation during excess ER ROS despite minimal ATPase activity . To test whether modified BiP maintains the capacity to lessen protein aggregation , we employed an assay based on the in vitro aggregation of rhodanese . When a solution of reduced and chemically denatured rhodanese is rapidly diluted out of denaturant , rhodanese forms large aggregates that can be detected by light scattering ( Langer et al . , 1992; Figure 8 ) . Significantly , we observed that the presence of the ATPase-deficient Kar2 cysteine-substitution alleles ( Kar2-C63D/F/Y/W ) during dilution not only lessened aggregation ( minimized light scattering ) of denatured rhodanese but also was more effective than wild-type BiP at minimizing rhodanese aggregation ( Figure 8A ) . An enhanced capacity to prevent aggregation relative to wild-type BiP was observed as well for oxidized BiP ( peroxide treated; Kar2-SOH ) and alkylated BiP ( Kar2-NEM ) ( Figure 8B , C ) . Treatment of a Kar2-C63A mutant with peroxide or NEM did not enhance its ability to lessen aggregation of rhodanese , demonstrating that the enhanced capacity to limit aggregation correlates with modification of the BiP cysteine ( Figure 8B , C ) . Notably , the ATPase-deficient Kar2-T249G mutant ( which did not enhance growth of cells during oxidative stress; Figure 7 ) did not show an enhanced ability to prevent aggregation of rhodanese ( Figure 8A ) , supporting a correlation between a greater holdase capacity of BiP and enhanced viability during oxidative stress . It is worth pointing out that at the concentrations used in our assay , addition of wild-type , Kar2-C63A , and Kar2-T249G mutants did not allow for significant aggregation protection; the presence of any of these three proteins at the time of dilution resulted in light scattering comparable to the light scattering observed by rhodanese diluted into a solution of bovine serum albumin ( BSA; Figure 8A ) . Assays with a second model substrate , IgY , showed similar trends . Kar2-C63F/Y and alkylated BiP ( Kar2-NEM ) showed a greater ability to curtail IgY aggregation than wild-type Kar2 ( Figure 8D , E ) . Kar2-C63D and Kar2-C63W were prone to aggregation at high temperature ( required for IgY aggregation ) , and we were unable to assess their activity in the IgY assay ( data not shown ) . 10 . 7554/eLife . 03496 . 010Figure 8 . BiP cysteine mutants that protect cells during oxidative stress are more effective than wild-type BiP in suppressing polypeptide aggregation . ( A–C ) Denatured rhodanese was diluted to a final concentration of 1 µM in the presence of 4 µM BSA or wild-type , mutant , peroxide-treated , or alkylated BiP . Samples in panels B and C were mock treated to match the CHP or NEM-treatment . Rhodanese aggregation was followed by monitoring light scattering at 320 nm over a period of 5 min ( D and E ) After denaturation in 6 M guanidine and 40 mM DTT , IgY was diluted to final concentration of 0 . 7 µM at 45°C in the presence or absence of 0 . 7 µM wild-type , mutant , or alkylated BiP . IgY aggregation was followed by monitoring light scattering at 360 nm over a period of 60 min . Data in panels A–E are representative traces from at least three trials . ( F ) CSY278 containing plasmids pCS681 , pCS802 , pJW5 , pCS687 , pCS844 , pCS688 , pCS845 , pCS750 , pCS846 , or empty vector were spotted on SMM-leu or SMM Gal-leu plates , and plates were incubated at 37°C for 2 d ( glucose ) or 2-3 d ( galactose ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03496 . 010 These in vitro data suggest that the ability of modified BiP to bind polypeptides could account for the BiP-mediated protection against the deleterious effects of excess ROS in the ER . Consistent with this interpretation , we observed that Kar2 peptide-binding activity contributes to the protective activity of the Kar2-C63D/F/Y/W mutants in cells . A Kar2-C63D/F/Y/W mutant with an additional mutation in the peptide-binding domain , G445D , ( analogous to DnaK-G400D; Burkholder et al . , 1996 ) showed decreased cell growth under conditions of increased ER ROS ( relative to the Kar2-C63D/F/Y/W mutant alone ) ( Figure 8F ) , indicating that the ability to bind polypeptides is essential for BiP-mediated protection of cells during oxidative stress .
We propose that BiP oxidation is part of a redox signaling circuit within the ER , wherein hyper-oxidizing conditions in the ER are offset through modulation of BiP activity . A primary outcome of oxidative stress is oxidative damage to proteins , including the mispairing of cysteine residues to form non-native disulfide bonds as well as the oxidation of cysteine thiols with small molecules , which prevents disulfide bond formation . These oxidized proteins are unlikely to be able to fold properly until ER conditions are returned to a less oxidized state . We suggest that activation of a redox switch in BiP upon ROS accumulation in the ER converts BiP from a chaperone driven by ATP hydrolysis to a high avidity polypeptide holdase . An enhanced capacity for oxidized BiP to bind polypeptides may serve to maintain proteins in a folding-competent state until the ER returns to a more reduced status . Upon restoration of a more reduced ER environment , reduction of BiP would reestablish BiP activity as an ATP-driven chaperone to facilitate protein folding through its binding and release of polypeptides . Inactivation of ATPase activity for a pool of BiP upon oxidation may additionally benefit cells by limiting ATP turnover during oxidative stress . In the absence of modification , BiP has the potential to consume ATP while carrying out futile chaperone cycles of binding and release from polypeptides locked into non-native states by covalent bonds; inactivation of BiP ATPase activity would limit hydrolysis of cellular ATP as a byproduct of futile folding cycles . Finally , inhibition of the translocation activity of BiP under hyper-oxidizing conditions could decrease the flux of nascent polypeptides into the ER further decreasing the burden on the ER folding machinery ( Figure 6 ) . We are unable to definitively link translocation attenuation with BiP modification due to the lack of a measurable translocation defect in wild-type cells under ER oxidative stress . Yet these data do not preclude a more modest decrease in polypeptide flux during oxidative stress; note we do not anticipate all BiP becomes modified during oxidative stress , which is consistent with the lack of a complete translocation block during stress . Data from our BiP cysteine-mutant alleles suggest that a single population of all reduced or all oxidized BiP is insufficient to maintain cell viability throughout a range of redox conditions . Alleles of BiP that mimic the phenotypic outcomes of BiP oxidation ( Kar2-C63D/F/Y/W ) are unable to function effectively as the only copy of BiP in cells during standard growth conditions ( Figure 5 ) . Conversely , a mimetic of reduced BiP ( Kar2-C63A ) is unable to support robust growth of cells as the only form of cellular BiP during oxidative ER stress ( Figure 2 ) . We anticipate the robust growth we observed in the presence of both Kar2-C63A plus Kar2-C63F/Y/W is due to the genetic recapitulation of the normally reduced and oxidized pools of ER BiP . Intriguingly , multiple means may exist to post-translationally control BiP activity to benefit cells under oxidative stress . Recently mammalian BiP was established as a main player in an oxidative stress response pathway initiated by activation ( oxidation ) of GPx7 ( NPGPx ) ( Wei et al . , 2012 ) . Wei et al . reported that GPx7 acts as a direct sensor of cellular ROS , and they propose that oxidized GPx7 catalyzes the formation of an intramolecular BiP Cys41-Cys420 disulfide . Similar to what we observed upon BiP oxidation , Wei et al . show that formation of an intramolecular BiP disulfide enhances the ability of BiP to bind denatured luciferase in vitro . Notably , GPx7 and the disulfide-bonded cysteines in BiP are found only in mammals and not in other eukaryotes ( yeast BiP has one cysteine ) , suggesting the GPx7-BiP signaling pathway is likely unique to mammalian cells . Yet it is significant that in their study Wei et al . observed in vitro treatment of recombinant mammalian BiP with peroxide facilitated not only the Cys41-Cys420 disulfide but also a sulfonic acid ( −SO3H ) at Cys41 , which corresponds to Kar2 Cys63 ( Wei et al . , 2012 ) . These data demonstrate that mammalian BiP Cys41 is prone to direct oxidation by peroxide like yeast BiP , solidifying the capacity for BiP to serve also as a direct sensor of ROS in mammalian cells . Emerging data suggest that decoupling of ATPase and chaperone functions through cysteine oxidation could be a property shared by several Hsp70s . For example , in vitro alkylation of the three cysteines in the cytosolic yeast Hsp70 Ssa1 triggers a loss of ATPase activity and augments the capacity of Ssa1 to prevent aggregation of denatured luciferase ( Liu et al . , 1996; Hermawan and Chirico , 1999 ) . Cys15 in Ssa1 is equivalent to the conserved BiP cysteine , and changes in chaperone activity for alkylated Ssa1 could occur through a mechanism similar to what we describe for BiP . Alternatively , the mechanism may be distinct from what we describe for BiP and make use of the other two Ssa1 cysteine residues; indeed , precedent exists for a redox-sensing role for Ssa1 Cys264 and Cys303 ( but not Cys15 ) in the derepression of Hsf1 in response to a variety of thiol-reactive compounds ( Wang et al . , 2012 ) . Irreversible inhibition of the Hsp70 ATPase activity of mammalian Hsp70 ( HSPA1A ) has been shown to occur upon oxidation of the same two cysteines with methylene blue ( Miyata et al . , 2012 ) . Intriguingly , a separate study reported that the in vitro acquisition of peptides by cytosolic Hsp70 is enhanced in the presence of hydrogen peroxide , raising the possibility that ATPase and chaperone activities of Hsp70 may also be decoupled by oxidation ( Callahan et al . , 2002 ) . To the best of our knowledge our study shows the first example of an Hsp70 modified by endogenous ROS; given the reported susceptibility of numerous Hsp70s to oxidation by exogenous oxidants it will be exciting to determine if and what endogenous oxidants are sensed by the cytoplasmic and mitochondrial Hsp70s . Notably , redox modification of Hsp70 cysteines is not restricted to eukaryotes but also has been shown in bacteria . Oxidation of Cys15 in the bacterial Hsp70 DnaK ( equivalent to Kar2 Cy63 ) has been demonstrated to occur when bacteria are exposed to exogenous oxidants at elevated temperatures ( Winter et al . , 2005 ) . Markedly distinct from what we observe for BiP , oxidant exposure inhibits DnaK's ability to prevent luciferase aggregation ( Winter et al . , 2005 ) . Although modification of the DnaK cysteine is detected in cells exposed to oxidant , a DnaK-C15A mutant shows the same oxidant-induced loss of activity as wild-type DnaK , suggesting that cysteine oxidation may contribute to but does not account for the inactivation of DnaK chaperone activity ( Winter et al . , 2005 ) . Interestingly , the inactivation of DnaK during oxidative stress occurs concurrently with oxidant-induced activation of holdase activity for the general chaperone Hsp33 ( Winter et al . , 2005 ) . Thus oxidative stress appears to evoke similar functional outcomes in both bacteria and eukaryotes ( loss of Hsp70 ATP-dependent chaperone activity and enhanced holdase capacity ) , but in eukaryotes this can be achieved through a switch in BiP activity alone whereas in bacteria this outcome is a product of altered activities for a coupled DnaK-Hsp33 system . The position of the BiP cysteine within the ATP binding pocket ( less than 10 Å from the nucleotide ) ( Wisniewska et al . , 2010; Yan et al . , 2011 ) suggests that oxidation of the cysteine thiol may prevent the correct positioning of the nucleotide within the ATP-binding pocket . Sulfenic acid at the site of the cysteine is unlikely itself to physically block ATP entry into the cleft; it seems more likely that cysteine oxidation will perturb the surrounding residues , which in turn may alter residues necessary for ATP positioning or hydrolysis , leading to decreased ATPase activity . It is well established that the nucleotide and peptide binding domains of BiP are allosterically coupled; nucleotide binding and hydrolysis modulates both the structure of the peptide binding domain and its affinity for polypeptides ( Zuiderweg et al . , 2013 ) . We propose that structural changes within the nucleotide-binding pocket upon cysteine oxidation are similarly propagated to the peptide-binding domain to alter peptide affinity , allowing for enhanced peptide binding by BiP without ATP hydrolysis . Sulfenic acids are generally considered reactive species prone to further oxidation ( Jacob et al . , 2006; Reddie and Carroll , 2008; Roos and Messens , 2011 ) , which raises the possibility that direct modification of BiP with peroxide primes BiP for further oxidative modification ( s ) . Our data indicate that addition of sulfenic acid accounts for at least some portion of the oxidized BiP cysteine that occurs upon oxidative stress . However , our data do not rule out the possibility of additional BiP cysteine modifications . Indeed , the sensitivity of the kar2-C63A strain to the thiol-oxidant diamide ( Figure 2 ) , thought to primarily impact the redox balance of cells through oxidation of the cellular anti-oxidant glutathione ( Kosower et al . , 1969 ) , likely indicates the potential for a non-peroxide adduct on the BiP cysteine . Our experiments detect only a small fraction of BiP isolated from cells in an oxidized state; we estimate recovery of less than 1% of the total isolated BiP in a modified form . Low yields of oxidized cellular BiP species have limited our ability to use mass spectrometry for conclusive identification of any additional modifications . It is worth noting that the limited recovery of oxidized BiP from cells likely reflects the difficulty in preserving oxidized thiols , which are prone to reduction during lysate preparation . We speculate that a pool of modified BiP , greater than the recovered 1% , exists in cells during oxidative stress; indeed , it is difficult to envision that a 1% pool of oxidized BiP in wild-type cells during oxidative stress could account for the striking growth difference between the wild-type and kar2-C63A strains during oxidative ER stress ( Figure 2 ) . If BiP undergoes oxidation by more than one type of molecule , it is possible that distinct biochemical activities for BiP may be observed dependent on the specific redox modification . Sulfenic acid is known to prime proteins for modification by glutathione ( Gallogly and Mieyal , 2007 ) , and the abundance of reduced glutathione in the ER could allow for formation of a reversible BiP protein-glutathione adduct . Consistent with the proposed susceptibility of BiP to glutathione modification , a proteomics approach previously identified mammalian BiP ( and 22 additional proteins ) as a substrate for glutathiolation during diamide-induced oxidative stress in endothelial cells ( Lind et al . , 2002 ) . The efficacy of the BiP-C63F/Y/W alleles may reflect the ability of these larger amino acid side chains to mimic BiP glutathiolation and protection . Sulfenic acid can also be further oxidized by peroxide to form sulfinic ( −SO2H ) and sulfonic ( −SO3H ) acid species , which are generally considered irreversible modifications . Modification of BiP by excess peroxide could denote irreparable cellular damage , initiating a pathway for self-destruction ( e . g . , like apoptotic death induced by unresolved signaling through the unfolded protein response ) ( Walter and Ron , 2011 ) . Conversely , glutathiolation of BiP could be an important means to protect BiP from irreversible oxidation with excess peroxide , mediating a reversible signaling system for stress protection . We uncovered BiP oxidation upon genetic manipulation of cells to create hyper-oxidizing ER conditions; the ability to alter the redox environment of the ER was key to reveal BiP oxidation and has been similarly vital for identification of other homeostatic pathways in the ER such as the UPR ( Kozutsumi et al . , 1988; Dorner et al . , 1990; Gething and Sambrook , 1992 ) . Although the UPR was originally designated as a stress response system based on its importance during acute stress , it is now appreciated that the UPR plays crucial roles in maintaining ER plasticity and function under normal growth conditions ( Moore and Hollien , 2012 ) . Similarly , we propose that BiP oxidation likely influences cell growth and development in the absence of extreme redox imbalance . Notably , modified BiP was detectable even in the absence of Ero1* overproduction ( Figure 3 ) , consistent with a role for BiP oxidation not only during stress but also during non-stressed growth conditions . Increased ROS content within the ER as a consequence of enhanced secretory capacity could increase the pool of oxidized BiP to help maintain secretory protein dynamics . Assays similar to those described herein should enable future studies of yeast and mammalian BiP oxidation under non-stress conditions .
Plasmids used in this study are listed in Table 1 . All yeast expression plasmids derive from the pRS vector series ( Sikorski and Hieter , 1989 ) . pCS623 and pCS739 contain a XhoI–XhoI fragment of genomic KAR2 . Additional yeast Kar2-encoding plasmids contain KAR2 with ∼1 kb each of 5′ and 3′ untranslated sequences flanked with engineered BamHI and SacI restriction sites . Plasmids pCS623 and pCS681 show identical phenotypes in yeast . To create pCS757 and pCS760 , a FLAG epitope ( DNA sequence: GGAGATTATAAGGATGACGACGATAAGGGT ) was inserted by two-step fusion PCR immediately prior to the DNA sequence in KAR2 encoding for the HDEL retrieval signal . pCS584 was made by placing a ∼2 . 4 kb CAN1 fragment into pRS316 , and then replacing an EcoRI-SpeI piece within CAN1 with an EcoRI-SpeI fragment from pCS452 . A pET21b ( EMD Millipore , Billerica , MA ) Kar2 plasmid series was made that expresses a C-terminal his6-tagged Kar2 residues 40–668 . A pET28a ( EMD Millipore ) Kar2 plasmid series codes for a N-terminally tagged Kar2 amino acids 42–682 . To make pCS675 , sequence coding for the Sec63J domain ( residues 121–221 ) was cloned into pGEX-4T-1 ( GE Healthcare , UK ) . The UPRE-lacZ reporter pCS852 was generated by destroying the LEU2 marker in pJC8 ( Cuozzo and Kaiser , 1999 ) by MfeI digestion followed by ligation of the cut plasmid backbone . Amino acid substitutions were made by QuikChange site-directed mutagenesis ( Stratagene , Santa Clara , CA ) . All mutations were confirmed by sequencing . 10 . 7554/eLife . 03496 . 011Table 1 . PlasmidsDOI: http://dx . doi . org/10 . 7554/eLife . 03496 . 011NameDescriptionMarkersSourcepJC8UPRE-LacZ reporterCEN URA3 LEU2Cuozzo and Kaiser , 1999pCS852UPRE-LacZ reporterCEN URA3This studypAF112PGAL1-ERO1-mycCEN URA3Sevier et al . , 2007pCS452PGAL1-ERO1*-mycCEN URA3Sevier et al . , 2007pCS504PGAL1-ero1*-C100A-C105A-mycCEN URA3Sevier et al . , 2007pCS584can1::PGAL1-ERO1*-mycCEN URA3This studypCS739kar2-C63AURA3This studypCS623KAR2CEN URA3This studypCS681KAR2CEN LEU2This studypCS685kar2-C63ACEN LEU2This studypCS802kar2-C63DCEN LEU2This studypCS687kar2-C63FCEN LEU2This studypCS688kar2-C63YCEN LEU2This studypCS750kar2-C63WCEN LEU2This studypCS774kar2-T249GCEN LEU2This studypJW5kar2-C63D-G445DCEN LEU2This studypCS844kar2-C63F-G445DCEN LEU2This studypCS845kar2-C63Y-G445DCEN LEU2This studypCS846kar2-C63W-G445DCEN LEU2This studypCS757KAR2-FLAGCEN LEU2This studypCS760kar2-C63A-FLAGCEN LEU2This studypCS630kar2- ( 40-668 ) -His6AMPThis studypCS631kar2- ( 40-668 ) -C63A-His6AMPThis studypJW4kar2- ( 40-668 ) -C63D-His6AMPThis studypCS658kar2- ( 40-668 ) -C63F-His6AMPThis studypCS643kar2- ( 40-668 ) -C63Y-His6AMPThis studypCS644kar2- ( 40-668 ) -C63W-His6AMPThis studypCS639kar2- ( 40-668 ) -T249G-His6AMPThis studypCS675GST-sec63J- ( 121-221 ) AMPThis studypCS817His6-kar2- ( 42-682 ) KANThis studypCS818His6-kar2- ( 42-682 ) -C63AKANThis studypCS822His6-kar2- ( 42-682 ) -C63DKANThis studypCS819His6-kar2- ( 42-682 ) -C63FKANThis studypCS820His6-kar2- ( 42-682 ) -C63YKANThis studypCS821His6-kar2- ( 42-682 ) -C63WKANThis studypKP113His6-kar2- ( 42-682 ) -T249GKANThis study Saccharomyces cerevisiae strains were grown and genetically manipulated using standard techniques ( Adams et al . , 1998 ) . YPD is rich medium with 2% glucose . SMM is synthetic minimal medium supplemented with 2% glucose or a specified carbon source: 2% raffinose ( SMM Raf ) or 2% galactose ( SMM Gal ) . SCAA is minimal medium containing 0 . 67% yeast nitrogen base , 2% casamino acids , an amino acid supplement ( 0 . 004% adenine , histidine , and methionine , 0 . 006% leucine and lysine , 0 . 002% tryptophan ) , and a specified carbon source: 2% raffinose ( SCAA Raf ) or 2% galactose ( SCAA Gal ) . Uracil or leucine medium supplements were removed to select for plasmids as needed . Strains used in this study are listed in Table 2 . CKY1026 is described in ( Sevier et al . , 2007 ) . The KanMX marker in CKY1026 was swapped for NatMX using homologous recombination to make CSY158 ( Goldstein and McCusker , 1999 ) . For CSY172 , a KanMX-GAL promoter module was inserted before the SUC2 open reading frame using PCR-mediated gene-modification ( Longtine et al . , 1998 ) . To create CSY170 , a disrupted CAN1 fragment containing PGAL1-ERO1*-myc was released from pCS584 , and transformed into CKY263 . Stable integrants were selected on plates lacking arginine and containing 60 µg/ml canavanine . A kar2Δ strain ( CSY214 ) was made by replacing the KAR2 coding sequence with KanMX in a homozygous GAL2 ura3-52 leu2-3 , 112 diploid . The resultant heterozygous diploid was transformed with pCS623 and a viable MATa Ura+ KanMX+ segregant was recovered after sporulation . CSY289 , 290 , 368 , 292 and 293 were generated by transformation of CSY214 with pCS681 , pCS685 , pCS802 , pCS687 , or pCS688 , followed by selection against pCS623 by plating on SMM with 5-FOA . CSY308 was made by one-step gene replacement of PEP4 with NatMX in CSY214 ( Kozutsumi et al . , 1988 ) . CSY316 and CSY319 were generated by transformation of CSY308 with pCS757 or pCS760 , followed by counter-selection of pCS623 on SMM with 5-FOA . CSY275 was created by replacement of KAR2 with kar2-C63A using a two-step method with linearized pCS739 as described in Rothstein ( 1991 ) . Double mutant combinations of the kar2-C63A , ire1Δ , and can1::PGAL1-ERO1*-myc mutants were created using standard genetic techniques of mating , sporulation , and scoring . 10 . 7554/eLife . 03496 . 012Table 2 . Yeast strainsDOI: http://dx . doi . org/10 . 7554/eLife . 03496 . 012StrainGenotypeSourceCKY263/CSY5MATa GAL2 ura3-52 leu2-3 , 112Lab collectionsCKY264/CSY6MATα GAL2 ura3-52 leu2-3 , 112Lab collectionsCKY1026/CSY44MATa GAL2 ura3-52 leu2-3 , 112 ire1Δ::KanMXLab collectionsCSY158MATa GAL2 ura3-52 leu2-3 , 112 ire1Δ::NatMXThis studyCSY172MATa GAL2 ura3-52 leu2-3 , 112 ire1Δ::NatMX KanMX:PGAL1-SUC2This studyCSY275MATa GAL2 ura3-52 leu2-3 , 112 kar2-C63AThis studyCSY277MATa GAL2 ura3-52 leu2-3 , 112 kar2-C63A ire1Δ::NatMXThis studyCSY170MATa GAL2 ura3-52 leu2-3 , 112 can1::PGAL1-ERO1*-mycThis studyCSY278MATa GAL2 ura3-52 leu2-3 , 112 kar2-C63A can1::PGAL1-ERO1*-mycThis studyCSY214MATa GAL2 ura3-52 leu2-3 , 112 kar2Δ::KanMX [pCS623]This studyCSY289MATa GAL2 ura3-52 leu2-3 , 112 kar2Δ::KanMX [pCS681]This studyCSY290MATa GAL2 ura3-52 leu2-3 , 112 kar2Δ::KanMX [pCS685]This studyCSY368MATa GAL2 ura3-52 leu2-3 , 112 kar2Δ::KanMX [pCS802]This studyCSY292MATa GAL2 ura3-52 leu2-3 , 112 kar2Δ::KanMX [pCS687]This studyCSY293MATa GAL2 ura3-52 leu2-3 , 112 kar2Δ::KanMX [pCS688]This studyCSY308MATa GAL2 ura3-52 leu2-3 , 112 kar2Δ::KanMX pep4Δ::NatMX [pCS623]This studyCSY316MATa GAL2 ura3-52 leu2-3 , 112 kar2Δ::KanMX pep4Δ::NatMX [pCS757]This studyCSY319MATa GAL2 ura3-52 leu2-3 , 112 kar2Δ::KanMX pep4Δ::NatMX [pCS760]This study BL21 ( DE3 ) pLysS cells containing pET-derived plasmids were grown overnight at 37°C to saturation in Luria–Bertani ( LB ) medium containing 34 µg/ml chloramphenicol and 100 µg/ml ampicillin or 15 µg/ml kanamycin . Cells were diluted 1:20 in LB with antibiotics , grown for 2 hr at 37°C , moved to 24°C , and protein expression was induced with 0 . 4 mM isopropyl-β-D-thiogalactopyranoside ( IPTG ) . Cells were harvested 2–6 hr post-IPTG addition , and cell pellets were frozen at −80°C . Pellets were thawed and solubilized for 30 min on ice with 20 ml sonication buffer ( 50 mM HEPES , pH 7 . 4 , 0 . 3 M NaCl , 10 mM imidazole ) plus 1 mM PMSF , 1 µM pepstatin A , and 5 mM BME per 1 l cell culture . Cells were lysed by treatment with lysozyme followed by sonication , and insoluble material was removed by centrifugation at 16 , 000×g for 20 min . Supernatant was incubated for 30 min at 4°C with a slurry of 50% Ni-NTA agarose resin ( Qiagen , Germany ) and loaded into a column , or lysate was loaded directly onto a HiTrap chelating column ( GE Healthcare ) charged with nickel . C-terminally tagged Kar2 preps were washed extensively to remove contaminating ATPase activity similar to ( McClellan et al . , 1998 ) . Briefly , resin was washed with 10 column volumes ( cv ) sonication buffer , 10 cv sonication buffer with 5% glycerol , 1% Triton-X-100 , 10 cv sonication buffer with 5% glycerol , 1 M NaCl , 10 cv sonication buffer with 5% glycerol , 5 mM ATP , 10 mM MgCl2 , 10 cv sonication buffer with 5% glycerol , 0 . 5 M Tris–HCl , pH 7 . 4 , and 10 cv sonication buffer with 5% glycerol , 25 mM imidazole . Purified protein was eluted with 3 cv sonication buffer with 5% glycerol , 0 . 25 M imidazole . Protein was exchanged into 40 mM Tris–HCl , pH 7 . 4 , 80 mM NaCl , 10% glycerol using a PD-10 column . For N-terminally tagged Kar2 , resin was washed with 5 cv wash buffer ( 20 mM HEPES , pH 7 . 5 , 0 . 5 M NaCl , 10% glycerol , 10 mM imidazole ) and 15 cv wash buffer with 25 mM imidazole . Purified protein was eluted with 3 cv wash buffer with 0 . 2 M imidazole . Protein was exchanged into 10 mM Tris–HCl , pH 7 . 4 , 50 mM NaCl , 10% glycerol using a NAP-5 column . All rhodanese and IgY assays were performed with N-terminally tagged Kar2; the C-terminally tagged Kar2 is not full length and was not active in the protein aggregation assays . GST-Sec63J protein ( pCS675 ) was expressed in EN2 cells ( dnaKΔ ) , which were generously provided by Nadia Benaroudj ( Institut Pasteur , Paris , France ) and are described in Ratelade et al . ( 2009 ) . Bacteria containing pCS675 were grown overnight to saturation at 30°C in LB medium containing 100 µg/ml ampicillin , diluted 1:20 in LB with ampicillin , grown for 2 . 5 hr at 30°C , and protein expression was induced with 0 . 2 mM IPTG . Cells were harvested 2 hr post-IPTG addition and cell pellets were frozen at −80°C . Pellets were thawed and solubilized for 30 min on ice with 20 ml PBS plus 2 mM EDTA , a complete protease inhibitor cocktail ( Roche , Switzerland ) , 1 mM PMSF , and 5 mM BME per 1 l cell culture . Cells were lysed by treatment with lysozyme followed by sonication . After addition of benzonase and 0 . 1% Triton-X-100 , insoluble material was removed by centrifugation at 20 , 000×g for 20 min . Supernatant was incubated for 1 hr at 4°C with a slurry of 50% glutathione-sepharose ( GE Healthcare ) . Resin was loaded into a column and washed with 20 cv PBS with 2 mM EDTA , 20 cv PBS with 2 mM EDTA , 1 M KCl , 0 . 1% Triton-X-100 , and 10 cv PBS . Protein was eluted with 10 cv 50 mM Tris–HCl , pH 8 , 10 mM reduced glutathione , 10% glycerol . A vivaspin-15 column ( GE Healthcare ) was used for glutathione removal and buffer exchange into 20 mM HEPES , pH 6 . 8 , 75 mM KOAc , 0 . 25 M sorbitol , 5 mM MgOAc2 , 10% glycerol . All purified proteins were flash frozen in liquid nitrogen and stored at −80°C . Protein concentrations were determined by BCA protein assay ( Thermo Fisher Scientific , Waltham , MA ) using bovine serum albumin as a standard . To prepare alkylated Kar2 , 1 nmol Kar2 was incubated with 20 nmol TCEP ( diluted from a 0 . 1 M TCEP stock prepared in 1 . 5 M Tris–HCl , pH 8 . 8 ) for 1 hr at room temperature in a 40 µl total volume of buffer suitable for downstream assays . Kar2 was then incubated with 40 nmol NEM for 1 hr at room temperature , and the reaction was quenched with 1 µmol DTT . To prepare oxidized Kar2 , 1 nmol Kar2 was incubated with 70 nmol CHP for 30 min at 30°C in 70 µl total volume of 10 mM Tris-HCl , pH 7 . 4 , 50 mM NaCl . CHP was subsequently removed using a Bio-spin P6 column ( Bio-Rad , Hercules , CA ) equilibrated with 10 mM Tris-HCl , pH 7 . 4 , 50 mM NaCl . Unmodified Kar2 control proteins were prepared identically except buffer was substituted for NEM or CHP . To measure ATPase activity , 1 µM C-terminally tagged Kar2 and 2 . 5 µM GST-Sec63J were incubated with 0 . 1 mM of cold ATP and 0 . 45 µCi of [alpha-32P]ATP ( Perkin–Elmer , Waltham , MA ) in ATPase buffer ( 50 mM Tris–HCl , pH 7 . 4 , 50 mM KCl , 5 mM MgCl2 , 1 mM DTT ) in a total volume of 45 µl at room temperature . For peroxide-treated Kar2 samples , 3 µM Kar2 and 5 µM GST-Sec63J were incubated in assay buffer lacking DTT ( to prevent Kar2 reduction ) . At various time points , 5 µl aliquots were removed and activity was quenched with the addition of an equal volume of 2X stop solution ( 0 . 14% SDS , 32 mM EDTA , 0 . 2 M NaCl ) . Samples ( 1–2 µl ) were spotted on polyethyleneimine cellulose TLC plates ( Sigma-Aldrich , St . Louis , MO ) , and plates were developed in 1 M formic acid and 0 . 5 M LiCl . Conversion of ATP to ADP was imaged with a phosphorimager and quantified using ImageQuant ( GE Healthcare ) . Strains transformed with the UPRE-lacZ reporter plasmid pJC8 ( Cuozzo and Kaiser , 1999 ) or pCS852 were grown in SMM medium and treated with 0 or 2 mM DTT for 1 . 5 hr ( 37°C ) or 2 hr ( 30°C ) prior to harvesting of exponential phase cells . Cells were permeabilized and beta–galactosidase activity measured as described ( Guarente , 1983 ) . Three transformants were assayed in duplicate per strain . CSY289 , 290 , 368 , 292 , and 293 were cultured in YPD at 24°C until mid-log phase , at which time samples were divided and half of the cultures were moved to a 37°C water bath incubator . After 90 min , 5 OD600 units were harvested for each culture . CSY44 and CSY172 containing pAF112 , pCS452 , or empty vector were grown overnight at 30°C in SMM Raf or SMM with 2% lactate and 2% gycerol , respectively . Cells were subcultured into SMM Gal and 5 OD600 equivalents were harvested after 5 hr of growth at 30°C . Lysates were prepared as described ( Kushnirov , 2000 ) . Samples were suspended in 100 µl of sample buffer ( 62 . 5 mM Tris–HCl , pH 6 . 8 , 10% glycerol , 2% SDS , 0 . 01% bromophenol blue ) containing 2% BME and 10 µg/ml pepstatin A and boiled for 3 min at 100°C . Proteins were separated by SDS-PAGE and detected with appropriate antiserum after transfer to nitrocellulose . CSY316 and CSY319 containing pRS316 or pCS452 were cultured overnight at 30°C in SMM Raf or SCAA Raf , subcultured into SMM Gal or SCAA Gal the next morning , and grown for 6–8 hr at 30°C . Cells ( 10 OD600 equivalents ) were harvested by centrifugation , and pellets were flash frozen in liquid nitrogen and stored at −80°C . Cells were suspended in 40 µl of 10% TCA . Zirconium beads were added , and cells were lysed in a FastPrep 24 instrument ( MP Biomedical , Santa Ana , CA ) with two 1 min pulses at speed 6 separated by a 5 min rest on ice . Samples were diluted with 1 ml of 10% TCA , and liquid was transferred to a new tube . Proteins were precipitated by centrifugation at 21 , 000×g for 10 min at 4°C , and pellets were washed once with ice-cold 5% TCA and once with ice-cold ethanol . Pellets were suspended in 500 µl of urea-containing cysteine modification buffer ( CMBU ) ( 0 . 1 M HEPES-NaOH , pH 7 . 4 , 1% SDS , 10 mM DTPA , 6 M urea ) with a complete ultra protease inhibitor cocktail ( Roche ) and 0 . 1 M NEM . Samples were rotated for 30 min at room temperature , 500 µl of CMBU with or without 10% BME was added , and samples were rotated for another 15 min at room temperature . Proteins were separated from small molecules by centrifugation after a 5 min incubation on ice with 10% TCA as above . Pellets were washed once with 5% TCA , twice with ethanol , and suspended in 300 µl of CMBU with 0 . 1 mM maleimide-biotin ( Sigma-Aldrich ) . For sodium arsenite treatment , pellets previously untreated with BME were suspended in 300 µl of CMBU with 20 mM sodium arsenite ( Sigma-Aldrich ) and 0 . 1 mM maleimide-biotin . After a 30 min rotation at room temperature , unreacted maleimide was quenched with 4% BME for 5 min , and small molecules and proteins were separated by addition of TCA as above . Pellets were solubilized in 100 µl CMBU , diluted with 1 ml IP buffer ( 50 mM Tris–HCl , pH 7 . 4 , 0 . 15 M NaCl , 1% Triton-X-100 ) , and incubated for 10 min at room temperature . Insoluble material was removed by centrifugation at 21 , 000×g for 5 min at 4°C prior to addition of 30 µl of 50% anti-FLAG M2 bead slurry ( Sigma-Aldrich ) . Samples were rotated at 4°C for 1 hr , beads were washed three times with IP buffer , and proteins were eluted for 5 min at 100°C with 30 µl 2X sample buffer containing 10% BME . Proteins were separated by SDS-PAGE , transferred to nitrocellulose , and probed with an avidin-Alexa488 conjugate or Kar2 antiserum and an Alexa546-conjugated anti-rabbit IgG . Fluorescent signal was detected using a ChemiDoc MP System ( Bio-Rad ) . For Ero1* experiments , CSY316 and CSY319 containing pRS316 or pCS452 were grown overnight at 30°C in SMM Raf , subcultured the following morning in SMM Gal , and returned to 30°C for 4 hr . For CHP treated samples , cells were grown in SMM at 30°C to mid-log at which time cells were treated with 5 mM CHP for 30 min . Cells ( 20 OD600 equivalents ) were harvested by filtration or centrifugation and suspended in 100 µl of lysis buffer ( 0 . 1 M HEPES , pH 7 . 4 , 0 . 3 M NaCl , 10% glycerol , 0 . 1 mM DTPA ) with 1 mM PMSF , 200 U/ml catalase , 2 mM DAz-2 ( Cayman Chemicals , Ann Arbor , MI ) , and 2X complete protease inhibitor cocktail . Zirconium beads were added , cells were lysed in a FastPrep 24 instrument with three 1 min pulses at speed 4 separated by 5 min rests on ice , and 100 µl of lysis buffer with 1 mM DAz-2 and 10% Triton-X-100 was added prior to sample rotation for 20–40 min at room temperature . Lysis buffer ( 800 µl ) was added , samples were incubated for 20 min at room temperature , and insoluble material was removed by centrifugation at 14 , 000×g for 5 min . Anti-FLAG M2 beads ( 50 µl of a 50% slurry ) were added to the soluble material and incubated for 1 hr at 4°C . Beads were washed twice with lysis buffer containing 1% Triton-X-100 , and once with lysis buffer . Beads were incubated with 25 µl of lysis buffer without DTPA and with 100 µM phosphine-biotin ( Cayman Chemicals ) for 2 hr at 37°C , and the reaction was quenched with 25 µl of 2X sample buffer . Proteins were separated by SDS-PAGE , transferred to nitrocellulose , and probed with an avidin-Alexa488 conjugate or Kar2 antiserum and a fluorescent or HRP-conjugated secondary antibody . Fluorescent and chemiluminescent signals were detected using a ChemiDoc MP System . Denatured rhodanese was prepared by incubating 50 µM bovine rhodanese ( Sigma-Aldrich ) for 1 hr at room temperature in 20 mM HEPES , pH 7 . 4 , 6 M guanidine-HCl , 0 . 1 M NaCl , 5 mM DTT . Denatured protein aliquots were flash frozen and stored at −80°C . Rhodanese aggregation was performed by diluting thawed denatured rhodanese to 1 µM in 0 . 2 ml of 20 mM HEPES , pH 7 . 4 , 50 mM KCl containing 5 mM MgCl2 , 1 mM ATP , and 4 µM Kar2 or 4 µM BSA . Rhodanese aggregation was followed by monitoring the light scattering at 320 nm for 5 min . IgY was purified using the Chicken IgY Purification kit ( Thermo Fisher Scientific ) and suspended in 0 . 1 M Tris–HCl , pH 8 , at a final concentration of 50 mg/ml . IgY aggregation was performed according to the methods of Stronge et al . ( 2001 ) with 0 . 7 µM final of each Kar2 and IgY . IgY was denatured for a minimum of 2 hr , and used within 3 hr of preparation . | The endoplasmic reticulum is the cellular compartment where approximately one third of the cell's proteins are made . Inside , chaperone molecules bind to newly made protein chains and help them to fold into the three-dimensional structure required for the protein to work correctly . A chaperone called Ero1 helps to facilitate this folding process by catalyzing a reaction that forms strong chemical bonds , which help stabilize the final protein structures . However , this help from Ero1 comes at a cost: forming a stabilizing bond this way also produces a peroxide molecule as a byproduct . Peroxide is a ‘reactive oxygen species’: a chemical that can oxidize and damage proteins and DNA , which can potentially kill the cell . Three other enzymes in the endoplasmic reticulum can convert peroxide into water , to protect the cells from reactive oxygen species build-up . However , not all cells that use Ero1 have these other enzymes , suggesting that other pathways must exist to manage reactive oxygen species . Wang et al . took advantage of yeast cells containing a hyperactive mutant version of the Ero1 enzyme to look for alternative detoxifying mechanisms that occur when the cell is stressed by an excess of reactive oxygen species . In these cells , Wang et al . observed that the high levels of reactive oxygen species caused part of a chaperone molecule called BiP to oxidize . This modification of BiP acts like a switch that the reactive oxygen species flip on . When activated by the reactive oxygen species , BiP enhances its activity as a folding molecular chaperone , keeping proteins apart . This is thought to allow BiP to minimize the protein misfolding that may otherwise occur in the wake of the damage caused by the building levels of peroxide . Wang et al . created a mutant BiP chaperone that mimics the oxidized form , and found that it also protects cells from the damage inflicted by the excess of reactive oxygen species . Wang et al . propose that the BiP chaperone may be an important sensor of reactive oxygen species that changes its activity when these harmful chemicals are present and helps to protect the cell from damage . The success in mimicking the protective effects of oxidized BiP with a mutant BiP suggest that in the future one may be able to design small molecule drugs that bind to BiP to produce the activity of the modified form . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2014 | Redox signaling via the molecular chaperone BiP protects cells against endoplasmic reticulum-derived oxidative stress |
Aberrant signaling through the Raf/MEK/ERK ( ERK/MAPK ) pathway causes pathology in a family of neurodevelopmental disorders known as 'RASopathies' and is implicated in autism pathogenesis . Here , we have determined the functions of ERK/MAPK signaling in developing neocortical excitatory neurons . Our data reveal a critical requirement for ERK/MAPK signaling in the morphological development and survival of large Ctip2+ neurons in layer 5 . Loss of Map2k1/2 ( Mek1/2 ) led to deficits in corticospinal tract formation and subsequent corticospinal neuron apoptosis . ERK/MAPK hyperactivation also led to reduced corticospinal axon elongation , but was associated with enhanced arborization . ERK/MAPK signaling was dispensable for axonal outgrowth of layer 2/3 callosal neurons . However , Map2k1/2 deletion led to reduced expression of Arc and enhanced intrinsic excitability in both layers 2/3 and 5 , in addition to imbalanced synaptic excitation and inhibition . These data demonstrate selective requirements for ERK/MAPK signaling in layer 5 circuit development and general effects on cortical pyramidal neuron excitability .
The canonical Ras/Raf/MEK/ERK ( ERK/MAPK ) signaling pathway is a key intracellular signaling cascade downstream of cell surface receptors critical for brain development ( Samuels et al . , 2009 ) . In the developing cortex , the ERK/MAPK pathway is thought to be particularly important for neuronal responses to neurotransmitters and receptor tyrosine kinase ( RTK ) ligands , such as FGFs and neurotrophins . In the mature brain , it is well established that ERK/MAPK plays a central role in the activity-dependent plasticity of neural circuits ( Shilyansky et al . , 2010; Thomas and Huganir , 2004 ) . Importantly , a number of human neurodevelopmental syndromes have been linked to aberrant ERK/MAPK activity . This related group of human syndromes , increasingly referred to as RASopathies , are caused by genetic mutations in core components or regulators of the ERK/MAPK signaling cascade ( Rauen , 2013 ) . Macrocephaly , neurodevelopmental delay , cognitive impairment , and epilepsy are frequently observed in RASopathy patients with clinical manifestations being dependent on the precise causative mutation ( Rauen , 2013 ) . RASopathies are most often associated with hyperactive ERK/MAPK signaling ( eg . Neurofibromatosis type 1 ( NF1 ) , Noonan , Costello , and Cardiofaciocutaneous ( CFC ) syndromes ) ( Rauen , 2013 ) . However , mutations that lead to diminished ERK/MAPK activation have been identified in a subset of LEOPARD and CFC syndrome patients ( Kontaridis et al . , 2006; Nowaczyk et al . , 2014 ) . Abnormal Ras/MAPK signaling has also been observed in models of other monogenic neurodevelopmental disorders including Fragile X syndrome and Tuberous Sclerosis ( Chévere-Torres et al . , 2012; Faridar et al . , 2014; Osterweil et al . , 2013; Zhang et al . , 2014 ) . Recent exciting work has shown that pharmacological normalization of pathological Ras/MAPK activity is sufficient for correcting select cellular and behavioral abnormalities in Fragile X , Tuberous Sclerosis , NF1 , Noonan , and Costello syndrome mutant mice ( Cui et al . , 2008; Lee et al . , 2014; Li et al . , 2005; Osterweil et al . , 2013; Wang et al . , 2012; Zhang et al . , 2014 ) . However , our understanding of these disorders remains rudimentary as there is limited information on brain cell type-specific consequences of either loss- or gain-of-function through this pathway . Accumulating evidence also suggests that pathological ERK/MAPK signaling contributes to certain forms of autism . Altered Ras/MAPK signaling has been identified as a common downstream mediator of divergent genetic mutations linked to autism , and MAPK3/ERK1 is present in a region of 16p11 . 2 mutated in ~1% of cases of autism ( Eichler and Zimmerman , 2008; Gilman et al . , 2012; Gilman et al . , 2011; Kumar et al . , 2008; Pinto et al . , 2010; Pucilowska et al . , 2015; Weiss et al . , 2008 ) . Little is known about how ERK/MAPK signaling might relate to the pathogenesis of autism . An important current research theme is that the behavioral manifestations of autism spectrum disorders ( ASDs ) may be linked to both functional hypo- and hyper-connectivity between distinct brain regions ( Geschwind and Levitt , 2007; Just et al . , 2007; Keown et al . , 2013; Supekar et al . , 2013 ) . Furthermore , recent work in postmortem brains of autistic patients suggests that local patches of disorganization , in which cortical layers 4–5 are particularly affected , play an important role in disease pathogenesis ( Stoner et al . , 2014 ) . In one study , co-expression network analyses of autism-linked genetic mutations suggested that layer 5 in prefrontal and sensorimotor cortex is a key site of convergence for pathogenesis ( Willsey et al . , 2013 ) . Whether aberrant ERK/MAPK signaling might result in cortical layer disorganization and defective long-range connectivity is unknown . To address questions of cell type specificity and consequences for circuit formation , we have defined the effects of ERK/MAPK loss- and gain-of-function on the development of cortical pyramidal neurons . Pyramidal neuron-specific functions of ERK/MAPK signaling were assessed by deleting the upstream kinases Map2k1/Mek1 and Map2k2/Mek2 ( hereafter referred to as Map2k1/2 ) or overexpressing hyperactive Map2k1 . Conditional deletion of Map2k1/2 led to major disruption of layer 5 with noticeably fewer CTIP2-expressing large neurons compared to controls . Further , long range axon extension of layer 5 corticospinal projection neurons during early development was markedly impaired . Subsequent to delayed entry of axons into the cervical spinal cord , many layer 5 projection neurons in sensorimotor cortices underwent apoptosis . Gain-of-function ERK/MAPK signaling also affected layer 5 CST neurons with a resultant decrease in axon elongation and associated increase in axon branching . The morphological requirement for ERK/MAPK signaling was specific for layer 5 , as layer 2/3 was not disrupted and callosal projection neurons in upper cortical layers do not exhibit overt changes in axon extension or targeting following Map2k1/2 deletion . In contrast to the layer-specific functions of ERK/MAPK on axonal development , we found that ERK/MAPK was required for the expression of ARC and other plasticity-associated genes across all cortical lamina . Further , loss of ERK/MAPK signaling in pyramidal neurons disrupted excitatory and inhibitory neurotransmission and altered intrinsic excitability in both layers 2/3 and 5 . Our data reveal unexpectedly specific requirements for ERK/MAPK signaling in layer 5 circuit development and general effects on the excitability of cortical pyramidal neurons in multiple layers .
Previous work has shown that MAPK1/3 ( aka ERK1/2 ) is activated in embryonic cortical neurons , albeit at much lower levels than in the ventricular zone ( Faedo et al . , 2010; Li et al . , 2014; Pucilowska et al . , 2012; Toyoda et al . , 2010 ) . In western blots of sensorimotor cortical lysates from P1 , 2 , 7 , 14 , and 21 day old mice , we found that the levels of pan-MAPK1/3 ( ERK1/2 ) and pan-MAP2K1/2 show a steady but evident increase from a relatively lower level at birth ( Figure 1—figure supplement 1A ) . Phosphorylated-MAPK1/3 ( ERK1/2 ) and phosphorylated-MAP2K1/2 levels were also relatively low at birth but increased noticeably by P7 and peaked at P14 ( Figure 1—figure supplement 1A ) ( Oliveira et al . , 2008 ) . The pattern of phosphorylated-MAPK1/3 ( ERK1/2 ) expression in P3 histological sections of cortex did not exhibit any clear-cut laminar specificity ( Figure 1—figure supplement 1B ) . These findings indicate that ERK/MAPK signaling is activated in the developing cortex , peaking during the second postnatal week . We generated a mouse model to test the direct , neuron-autonomous role of ERK/MAPK signaling by inactivating Map2k1/2 specifically in immature mouse cortical excitatory neurons . We conditionally deleted Map2k1/2 with a Cre-dependent Map2k1 allele , a germ-line Map2k2 deletion allele , and Cre-recombinase under the control of the NeuroD6/Nex promoter ( Map2k1loxp/loxp;Map2k2-/-;Neurod6-Cre , referred to hereafter as Map2k1/2;Neurod6-Cre ) ( Goebbels et al . , 2006 ) . As expected , Neurod6-Cre activated reporter-gene expression in the Cre-dependent EYFP mouse line , Ai3 , in excitatory , but not inhibitory , neurons in the neocortex by mid-embryogenesis ( Figure 1—figure supplement 1C–E ) ( Madisen et al . , 2010 ) . Western blotting of neocortical lysates and immunolabeling show that Map2k1/2;Neurod6-Cre mice exhibit significantly reduced MAP2K1 levels by birth and reduced phosphorylation of ERK/MAPK substrates , RSK , and MSK ( Figure 1—figure supplement 1F–H ) . Complete loss of MAP2K1 protein would not be expected in whole cortical lysates due to MAP2K1 expression in inhibitory interneurons and non-neuronal cell types ( Figure 1—figure supplement 1F ) . These data show that the Neurod6-Cre-mediated genetic targeting strategy is effective at inducing loss of ERK/MAPK signaling in developing cortical excitatory neurons . Map2k1/2;Neurod6-Cre pups were born at normal Mendelian ratios without overt differences from littermate controls . However , a delay in overall growth could be detected by the end of the first postnatal week and lethality was invariably observed in the third to fourth postnatal week . Behaviorally , P14 Map2k1/2;Neurod6-Cre mice exhibited spontaneous and persistent hindlimb clasping when lifted by the tail , an indicator of neurological impairment . A qualitatively similar effect on growth , neurological function , and viability was also observed in Mapk1loxp/loxp;Mapk3-/-;Neurod6-Cre mice ( data not shown ) , demonstrating that phenotypes are conserved following deletion of different core components of the ERK/MAPK cascade . Map2k1/2;Neurod6-Cre mice exhibited an average reduction in body weight of 37 . 1 ± 11 . 3% at P14 and neocortical volume was reduced by 23 . 2 ± 6 . 4% ( mean ± SEM , n=7 , <0 . 0001 ) . A significant increase in neuron ( NEUN/RBFOX3+ cell ) density was observed in Map2k1/2;Neurod6-Cre sensory cortices from 1636 ± 185 neurons/mm2 to 2168 ± 231 neurons/mm2 ( mean ± SEM , n=5 , p=0 . 002 ) ( Figure 1A–B ) . Given the significant difference in cortical size , we extrapolated a relative estimate of neuronal number by multiplying the density measurement by the relative change in cortical volume between individual littermate control and mutant pairs . This estimate showed no significant difference in relative global neuron number between mutant and control cortices ( 98 . 0 ± 3 . 5% , mean ± SEM , n=5 , p=0 . 61 ) . We did , however , detect a significant reduction in the size of NEUN+ soma in P14 Map2k1/2;Neurod6-Cre sensory cortices in all cortical layers with the most pronounced effect on layer 5 neurons ( Figure 1C ) . In contrast , no decrease in the size of parvalbumin-expressing , GABAergic neurons in the cortex was observed ( data not shown ) . These data show that ERK/MAPK signaling is necessary for neurological function and the somal growth of excitatory neurons , but do not provide strong support for a widespread role in regulating global excitatory neuron number . 10 . 7554/eLife . 11123 . 003Figure 1 . Loss of ERK/MAPK signaling leads to a reduction in the number of CTIP+ layer 5 neurons and reduced neuronal somal size . ( A–B ) . Immunostaining of P14 control ( A ) and Map2k1/2;Neurod6-Cre ( B ) sagittal forebrain sections for all neurons , callosal projection neurons , and subcortical projection neurons with NEUN , SATB2 , and CTIP2 , respectively , revealed an aberrant pattern of CTIP2 expression in layer 5 of mutant sensorimotor cortices ( yellow arrows ) ( n=6 , scale bar=100 µm ) . ( C ) Quantification of somal size was performed by measuring the cross-sectional area of randomly selected , NEUN labeled soma in distinct cortical layers . A reduction in the size of NEUN labeled soma in P14 mutant sensory cortices was detected in all cortical layers , but was particularly pronounced in layer 5 ( n= >300 total neurons per layer derived from five independent mice per condition , mean ± SEM , *p<0 . 001 ) . ( D–I ) Representative confocal images of CTIP2 immunolabeling in radial columns of primary motor ( D–E ) , sensory ( F–G ) , and visual ( H–I ) cortex from P14 control ( D , F , H ) and Map2k1/2;Neurod6-Cre ( E , G , I ) brains ( scale bar=30 µm ) . ( J ) Quantification of the relative number of layer 5 CTIP2+ neurons as a proportion of the total number of NEUN+ neurons in a cortical column revealed a substantial decrease in motor and sensory , but not visual , cortices in P14 mutant mice ( n=4 , mean ± SEM , *p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 00310 . 7554/eLife . 11123 . 004Figure 1—figure supplement 1 . Developmental changes in ERK/MAPK activity and mouse models for loss of ERK/MAPK signaling in cortical excitatory neurons . ( A ) Whole cortical lysates show significant changes in the expression and phosphorylation of ERK/MAPK components , MAP2K1/2 and MAPK1/3 ( ERK1/2 ) , during the first three weeks of postnatal development ( n=3 ) . ( B ) Immunofluorescent staining reveals the ubiquitous distribution of activated , phospho-MAPK1/3 ( ERK1/2 ) in all cortical layers in P2-3 mouse sensorimotor cortices . ( n=3 , scale bar=200 µm ) ( C–E ) Neurod6-Cre induces recombination of the Cre dependent EYFP reporter , Ai3 , in postmitotic excitatory neurons throughout the cortex as shown here in P14 ( C ) sagittal brain sections . Parvalbumin expressing GABAergic interneurons ( D , arrows ) do not exhibit EYFP expression ( E ) , further demonstrating the specificity of recombination in excitatory neurons . ( F ) P1 and P14 Map2k1/2;Neurod6-Cre cortices show pronounced loss of MAP2K1 expression and reduced phosphorylation of ERK/MAPK pathway substrates , RSK and MSK ( n=5 ) . ( G–H ) Representative confocal images of P14 control mouse brains show that cortical excitatory neurons in the cortex and hippocampus exhibit a high level of MAP2K1 immunolabeling ( G ) . MAP2K1 levels are profoundly reduced in the cortex ( yellow arrows ) and hippocampal subfields CA1-CA3 ( blue arrow ) in P14 Map2k1/2;Neurod6-Cre mice ( H ) ( n=6 , scale bar=2 mm ) . ( I–J ) MAP2K1 expression is absent in the cortex of Map2k1/2;Emx1-Cre mice ( J ) when compared to P21 littermate control cortices ( I ) ( n=3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 004 Recombination in Neurod6-Cre mice is particularly strong in the cerebral cortex , however , sparse recombination has been detected outside of the cortex ( Figure 1—figure supplement 1C ) ( Goebbels et al . , 2006 ) . To confirm the deficits in Map2k1/2;Neurod6-Cre mice are specifically due to cortical malfunction , we deleted Map2k1/2 using an independent Cre line , Emx1-Cre . Emx1-Cre induces recombination in dorsal telencephalic progenitors by E9 . 5 that generate excitatory neurons specifically within the cortex and hippocampus ( Gorski et al . , 2002 ) . Map2k1loxp/loxp;Map2k2-/-;Emx1-Cre ( Map2k1/2;Emx1-Cre ) exhibit cortex specific MAP2K1 deletion and a pattern of growth delay , clasping behavior , and lethality similar to that observed in Map2k1/2;Neurod6-Cre mice ( Figure 1—figure supplement 1I–J ) . We conclude that ERK/MAPK signaling in cortical excitatory neurons is necessary for appropriate nervous system function and mouse viability . The profound decrease in layer 5 neuron soma size caused by Map2k1/2 deletion suggests that ERK/MAPK signaling is particularly important for the development of this layer . Comprehensive analyses of layer 5 neurons have identified specific gene expression patterns , morphologies , and electrophysiological characteristics for this neuronal subtype ( Greig et al . , 2013; Kwan et al . , 2012; Leone et al . , 2008 ) . We analyzed the expression pattern of a critical transcription factor for layer 5 development , CTIP2/BCL11B , in control and mutant cortices ( Figure 1A–B ) ( Arlotta et al . , 2005 ) . The appearance of co-labeled CTIP2+/NEUN+ neurons in layer 5 appeared disrupted in rostral P14 Map2k1/2;Neurod6-Cre cortices compared to controls ( Figure 1B , yellow arrows ) . CTIP2+ layer 5 neurons are a small percentage of the total cortical neuron population . In the sensory cortex of control mice , we found that CTIP2+ layer 5 neurons represent only 9 . 61 ± 0 . 84% ( mean ± SEM , n=5 ) of all NEUN+ neurons within a radial column . Thus , our previous global NEUN estimates across an entire cortical column were not sensitive enough for detecting changes confined to sparse neuronal subtypes . To quantify the number of CTIP2+ neurons , we determined the relative proportion of cells in layer 5 that express CTIP2+ as a percentage of NEUN+ cells across all lamina in a radial cortical column . This measurement was performed in motor , sensory , and visual cortices . Indeed , we found that P14 Map2k1/2;Neurod6-Cre primary motor and sensory cortices exhibit a clear decrease in the relative proportion of CTIP2+ neurons in layer 5 ( Figure 1D–G , J ) . Qualitatively similar results were observed in P14 Map2k1/2;Emx1-Cre cortices ( Figure 2—figure supplement 1A–B ) . The relative proportion of CTIP2+ neurons in visual cortices ( Figure 1H–I , J ) was not significantly diminished in Map2k1/2;Neurod6-Cre mutants , suggesting the effect in sensory and motor cortex was due to loss of the cell type and not a result of ERK/MAPK regulation of global CTIP2 expression levels . Layer 5 neurons are morphologically heterogeneous with distinct subpopulations that can be differentiated by cortico-cortical or subcortical axonal projections . We examined layer 5 projections in the hindbrain corticobulbar tract ( CBT ) and spinal cord corticospinal tract ( CST ) of P14 Map2k1/2 mutant mice . Immunostaining for a well-established marker of corticospinal projections , PKCγ , and genetic labeling with Ai3 revealed a profound decrease in the size of the CBT in P14 Map2k1/2;Neurod6-Cre hindbrains ( Figure 2A–B , Figure 2—figure supplement 1E–F ) . CST labeling was also strikingly reduced in the cervical spinal cord and essentially absent in the lumbar spinal cord in both Map2k1/2;Neurod6-Cre ( Figure 2C–D ) and Map2k1/2;Emx1-Cre mutants ( Figure 2—figure supplement 1C–D ) . Analysis of rare Map2k1/2 mutants that survived as late as P24 revealed that no CST axons were present in lumbar spinal cords ( data not shown ) . These data provide further evidence of an overt loss of corticospinal neurons by the second postnatal week . 10 . 7554/eLife . 11123 . 005Figure 2 . Corticospinal tract defects in Map2k1/2;Neurod6-Cre mice . ( A–D ) To further evaluate the loss of layer 5 projection neurons , we examined the expression of a well-established corticospinal tract marker , PKCγ . Compared to control hindbrains ( A ) and spinal cords ( C ) , a profound decrease in corticospinal tract labeling was observed in the Map2k1/2;Neurod6-Cre hindbrain ( B-yellow arrows , scale bar=200 µm ) and spinal cord ( D ) , consistent with the reduced number of CTIP2+ layer 5 neurons ( df=dorsal funiculus , n=3 , scale bar=50 µm ) . ( E–F ) The Thy1-YFP reporter line , YFP16 , labels a small fraction of layer 5 neurons in sensorimotor cortex . Many layer 5 neurons labeled in this line have large complex apical dendritic arbors ( yellow arrows ) that are consistent with the known morphology of subcerebral projection neurons , including corticospinal neurons . EYFP expressing neurons heavily branch in layer 1 as shown in representative confocal images of sagittal sections from P18 control mice ( E , yellow arrows ) . In Map2k1/2;Neurod6-Cre cortices , we noted a significant reduction in the number of EYFP labeled neurons in layer 5 with complex apical dendritic arbors ( F ) ( n=3 , scale bar=100 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 00510 . 7554/eLife . 11123 . 006Figure 2—source data 1 . Reduced expression of layer 5 neuron markers following loss of Map2k1/2 . ( A ) Gene expression profiling of RNA extracts from the cortex of control and Map2k1/2;Neurod6-Cre mice collected at P9 ( n=2 ) and P14 ( n=3 ) revealed significantly diminished expression ( fold change>1 . 5 and p<0 . 05 ) of many genes that are known to be expressed in layer 5 neurons ( red filled boxes ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 00610 . 7554/eLife . 11123 . 007Figure 2—figure supplement 1 . Reduced number of corticospinal neurons following loss of Map2k1/2 . ( A–B ) Similar to results with Neurod6-Cre mice , deletion of Map2k1/2 with Emx1-Cre also led to a disruption of CTIP2 labeling in layer 5 sensorimotor cortices in Map2k1/2;Emx1-Cre mice ( B ) when compared to controls ( A ) ( n=3 , scale bar=200 µm ) . ( C–D ) Visualization of the CST in spinal cord cross-sections of P14 Map2k1/2;Emx1-Cre;Ai3 mice revealed a near complete absence of corticospinal axons in the lumbar segment of the spinal cord compared to control ( E ) ( n=3 , scale bar=50 µm ) . ( E–F ) The size of the EYFP labeled CBT was reduced in P14 Map2k1/2;Neurod6-Cre;Ai3 hindbrains ( F ) when compared to controls ( E ) ( n=3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 007 Corticospinal neurons represent a subset of the entire CTIP2+ population in sensorimotor layer 5 ( Arlotta et al . , 2005 ) . The lack of layer 5 neuron loss in the Map2k1/2;Neurod6-Cre visual cortex suggests that ERK/MAPK signaling is dispensable for the development of projection neurons targeting structures other than the spinal cord . Past work has shown that callosally projecting layer 5 neurons have significantly shorter and less complex apical dendritic arbors than subcortical projection neurons ( Larsen et al . , 2007; Molnár and Cheung , 2006 ) . To determine which of these classes was affected by Map2k1/2 deletion , a Thy1-based reporter , YFP-16 , that fluorescently labels a small proportion of layer 5 neurons in sensorimotor cortex , was bred with Map2k1/2;Neurod6-Cre mice ( Feng et al . , 2000 ) . These mutants clearly show that the fluorescently labeled layer 5 neurons in P14-P21 Map2k1/2;Neurod6-Cre;YFP-16 sensorimotor cortices have substantially shorter apical dendrites than in controls ( Figure 2E–F ) . The reduction in large , tufted neurons in layer 5 of mutant mice provide further evidence for a deficit in the development of corticospinal projection neurons . To gain further insight into the cellular and molecular mechanisms utilized by ERK/MAPK signaling to regulate cortical development , we performed microarray profiling of P14 whole cortical lysates from control and Map2k1/2;Neurod6-Cre mutants . The validity of the screen was verified by a highly significant -2 . 76 ± 0 . 13 ( p-val<0 . 001 ) -fold reduction in the expression of Map2k1-exon 3 at the probe-level in P14 mutant samples . We observed a pronounced decrease in a large number of genes that are known to be highly expressed in layer 5 projection neurons at P9 or P14 , including Etv1/Er81 , Adcyap1 , Ldb2 , Dkk3 , Lmo4 , and Opn3 ( fold-change>1 . 5 p<0 . 05 ) ( Arlotta et al . , 2005; Chen et al . , 2005; Chen et al . , 2005; Molyneaux et al . , 2007 ) ( Figure 2—source data 1 ) . Moreover , Fezf2 , a well-known master transcription factor important for corticospinal neuron development , showed a strong trend toward diminished expression in Map2k1/2;Neurod6-Cre mutants ( Chen et al . , 2005; Chen et al . , 2005; Molyneaux et al . , 2005 ) . The gene profiling results provide further support for a specific loss of layer 5 neurons possibly due to cell death or altered fate specification in Map2k1/2 mutant mice . The loss of layer 5 neurons following deletion of Map2k1/2 could be due to an early disruption in the initial specification of this neuronal subtype . During layer 5 neuron development , Neurod6-Cre is not expressed until neurons are post-mitotic ( Goebbels et al . , 2006; Wu et al . , 2005 ) . However , it remained possible that post-mitotic stages of embryonic layer 5 neuron specification were altered during embryogenesis . In newborn mutant pups , we found that the expression and number of CTIP2+ neurons in presumptive layer 5 was not diminished following Map2k1/2 deletion ( Figure 3A–C ) . Thus , ERK/MAPK signaling is not required for the initial establishment of the correct numbers of the CTIP2+ deep-layer neuron population . 10 . 7554/eLife . 11123 . 008Figure 3 . Layer 5 neuron corticospinal axon outgrowth requires ERK/MAPK signaling in vivo . ( A–C ) Expression of a well-known master transcription factor for layer 5 neurons , CTIP2 , was intact in cortical layer 5 as shown in representative confocal images of newborn control ( A ) and Map2k1/2;Neurod6-Cre ( B ) sensorimotor cortices ( scale bar=50 µm ) . Quantitation of the number of CTIP2-expressing nuclei in cortical layer 5 ( C ) did not reveal a significant difference in CTIP2+ neuron density between control and mutant neonates ( n=3 , mean ± SEM , p=0 . 89 ) . ( D–E ) In vivo DiI injections into the sensorimotor cortex of P0 . 5 control and Map2k1/2;Neurod6-Cre neonates were performed and mice were collected three days after injection . The extent of anterograde DiI labeling was analyzed in coronal sections through the spinal cord ( D ) . We observed a significant decrease in the extent of corticospinal ( cst ) elongation in mutant mice , especially in the lower cervical/thoracic spinal cord segments ( yellow arrows in D ) ( n=3 , scale bar=100 µm ) . Quantitation of CST length relative to the medullary decussation revealed a significant decrease in corticospinal axon growth in mutant spinal cords ( E ) ( n=3 , mean ± SEM , *p=0 . 02 ) . ( F–I ) The initial stages of corticospinal elongation in the dorsal spinal cord can be visualized in sagittal sections of the caudal medulla and rostral spinal cord from P2 Emx1-Cre;Ai3 mice . Immunoenhancement of the Ai3 reporter with an EGFP antibody clearly shows that the caudal extension of corticospinal axons ( yellow arrowheads ) is profoundly reduced in P2 Map2k1loxp/loxp;Map2k2-/-;Emx1-Cre;Ai3 mutant mice ( H ) when compared to Map2k1loxp/wt;Map2k2-/-;Emx1-Cre;Ai3 controls ( F ) ( n=3 , scale bar=200 µm ) . Whole mount visualization of the CST coursing through the ventral medulla in control ( G ) and mutant ( I ) hindbrains did not reveal an overt difference in corticospinal growth ( n=3 ) . ( J–P ) Compared to control mice ( J–L ) , Neurod6-Cre mediated deletion of Igf1r did not alter the relative area of PKCγ labeling in the cervical ( K , N , P ) or lumbar ( L , O , P ) CST compared to the medullary CBT ( J , M ) in mutant mice ( M–O ) at P21 ( n=3 , scale bar=50 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 00810 . 7554/eLife . 11123 . 009Figure 3—figure supplement 1 . The rostrocaudal elongation of layer 5 axons in the spinal cord requires ERK/MAPK , but not IGF1R , signaling . ( A–B ) In vivo DiI injections into the sensorimotor cortex were performed in P0 . 5 control and Map2k1/2;Neurod6-Cre neonates to label subcortical projection neuron afferents ( A ) . Mice were collected three days after injection and the extent of DiI labeling in multiple subcortical targets was analyzed in coronal sections through the brain ( B ) . A substantial difference in the innervation of the thalamus ( th ) , cerebral peduncle ( cp ) , superior colliculus ( sc ) , or pons ( po ) was not detected in mutants ( B ) ( n=3 , scale bar=100 µm ) . C-D . Cross sections through the medulla were labeled with EYFP and PKCγ in P3 control ( C ) and Map2k1/2;Neurod6-Cre;Ai3 mutant ( D ) hindbrains . No overt change in CST size was detected in mutant mice , providing evidence that cortical layer 5 axon outgrowth to the hindbrain does not require ERK/MAPK signaling ( n=3 , scale bar=100 µm ) . ( E–F ) At P3 , IGF1 expression is high in the inferior olive adjacent to the location of the developing CST in control neonates ( E ) . Map2k1/2;Neurod6-Cre mutants did not exhibit a qualitative difference in IGF1 expression ( F ) ( n=3 , scale bar=100 µm ) . ( G–K ) A substantial loss of IGF1Rβ protein expression ( yellow arrows ) was observed in the developing cortex of Igf1rloxp/loxp;Neurod6-Cre mutant mice ( H ) when compared to controls ( G ) ( scale bar=200 µm ) . The loss of IGF1R signaling led to a relatively reduced brain size when compared to P7 littermates ( I ) , but had no effect on the expression of CTIP2 or CUX1 in the sensorimotor cortex of Igf1rloxp/loxp;Neurod6-Cre mutants ( K ) relative to controls ( J ) ( n=3 ) ( scale bar=100 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 009 By P3 in wild-type neonates , corticospinal axons have projected through the ventral hindbrain , crossed the midline at the medullary/spinal cord boundary , and are extending into cervical spinal cord through the dorsal funiculus ( Schreyer and Jones , 1982 ) . We asked whether ERK/MAPK signaling was required for the initial outgrowth of corticospinal projections in vivo . In vivo DiI injections into the motor cortex of P0 . 5 neonates were performed to assess subcortical axon growth , especially into the spinal cord ( Figure 3—figure supplement 1A ) . Strikingly , analysis of anterogradely labeled axonal projections at P3-4 revealed a highly significant decrease in the extension of corticospinal axons into the lower cervical/upper thoracic segments of spinal cord ( Figure 3D–E ) . Moreover , a profound decrease in the caudal extension of descending corticospinal axons into the spinal cord of P2 Map2k1/2;Emx1-Cre;Ai3 mice was also observed ( Figure 3F , H ) . DiI , PKCγ , and Ai3 labeling of subcortical projections in Map2k1/2;Neurod6-Cre and Map2k1/2;Emx1-Cre mice revealed that layer 5 neuron growth into the tectum and pons/medulla was not significantly decreased ( Figure 3G , I; Figure 3—figure supplement 1A–D ) . These data demonstrate that the initial growth of corticospinal axons to the level of the hindbrain is not significantly disrupted in Map2k1/2 mutants , however , ERK/MAPK signaling is clearly necessary for corticospinal axon elongation into the spinal cord . In vitro studies have demonstrated that IGF1 signaling acting via PI3K and ERK/MAPK promotes corticospinal axon outgrowth in vitro ( Özdinler and Macklis , 2006 ) . IGF1 is also focally and intensely expressed in the medulla in the presumptive inferior olivary nucleus during the precise developmental time frame ( P1 – P2 ) that CST axons normally enter the spinal cord ( Figure 3—figure supplement 1E ) . IGF1 expression is temporally regulated , being present by E18 and much diminished by P14 ( data not shown ) . Map2k1/2 deletion in CST axons does not affect IGF1 expression by these cells ( Figure 3—figure supplement 1E–F ) . Thus , IGF1 may act via ERK/MAPK within corticospinal neurons to regulate corticospinal outgrowth . We functionally tested whether cortical neuron specific loss of IGF1R would disrupt the CST development by generating Igf1rloxp/loxp;Neurod6-Cre mutants ( Liu et al . , 2009 ) . Total body weight and brain size were significantly reduced in Igf1rloxp/loxp;Neurod6-Cre mutants ( Figure 3—figure supplement 1I ) and a clear loss of IGF1R immunoreactivity was observed in P7 Igf1rloxp/loxp;Neurod6-Cre cortices ( Figure 3—figure supplement 1G–H ) . However , the expression of CTIP2 in layer 5 was not significantly altered ( Figure 3—figure supplement 1J–K ) . Further , PKCγ labeling of the CST in the cervical and lumbar spinal cord did not reveal a significant difference in rostrocaudal extension in Igf1rloxp/loxp;Neurod6-Cre mutants ( Figure 3J–P ) . These data suggest that direct regulation of ERK/MAPK in layer 5 neurons by IGF1R signaling is not a significant regulator of neuronal survival and CST development in vivo . By P3 we noted a dramatic increase in the number of activated caspase-3 labeled neurons in layer 5 in Map2k1/2;Neurod6-Cre mutants when compared to controls ( Figure 4A–C ) . Modest caspase-3 activation was also observed in layer 6 , but not in upper layers 1–4 ( Figure 4C ) . In line with the elevation of caspase-3 activity , we observed IBA1+ microglia in P3 Map2k1/2;Neurod6-Cre cortices with a ramified morphology that appeared to be engulfing CTIP2+ neurons ( Figure 4D–E ) . To further confirm the absence of corticospinal neurons , we retrogradely labeled corticospinal projection neurons by injecting DiI into the cervical spinal cord in Map2k1/2;Neurod6-Cre neonates at P3 , the time point when dying cells could initially be detected in the cortex . Analysis of retrogradely labeled corticospinal neurons in the primary motor cortex at P7 revealed a significantly reduced number of DiI labeled cells consistent with the death of these neurons during this time period ( Figure 4F–G ) . Interestingly , retrograde DiI labeling of P5 layer 5 neurons that project into the contralateral hemisphere did not reveal a substantial decrease in P9 mutants relative to controls ( Figure 4H–I ) . These data indicate that loss of Map2k1/2 results in overt caspase-3 activation and corticospinal neuron death that can be first detected at P3-4 . 10 . 7554/eLife . 11123 . 010Figure 4 . Initiation of layer 5 neuron death by P3 in Map2k1/2;Neurod6-Cre mutants . ( A–B ) Representative confocal images of immunolabeling for cleaved activated caspase-3 , a well-known marker of neuronal apoptosis in P3 control ( A ) and Map2k1/2;Neurod6-Cre ( B ) sensorimotor cortices . Note the extensive increase in the number of activated caspase-3+ cells co-labeled with CTIP2 in layer 5 of mutant cortices ( B , yellow arrows ) ( n=4 , scale bar=50 µm ) . ( C ) Quantification of activated caspase-3+ cells in upper layers ( layer 1–4 ) , layer 5 ( CTIP2+ ) and layer 6 revealed a pronounced elevation in the number of apoptotic cells in layer 5 in P3 Map2k1/2;Neurod6-Cre mice relative to controls . The numbers of activated caspase-3+ cells are comparable in upper layers and doubled in layer 6 relative to controls . ( n=4 , mean ± SEM , * p<0 . 05 ) D-E . Relative to controls ( D ) , many microglia ( IBA1+ ) were observed with processes surrounding CTIP2 labeled neurons in layer 5 of mutant cortices ( E ) ( n=3 , scale bar=5 µm ) . F-G . Retrograde labeling of corticospinal neurons at P3 via DiI injection into the cervical spinal cord revealed a substantial reduction in the number of DiI labeled neuronal bodies in mutant ( G ) sensorimotor cortices at P7 when compared to controls ( F ) ( n=3 , scale bar=20 µm ) . ( H–I ) Retrograde labeling of layer 5 neurons that project into the contralateral hemisphere by injection of DiI in contralateral cortices , did not reveal an obvious loss in mutant ( I ) mice when compared to controls ( H ) ( n=3 , scale bar=100 um ) . ( J ) Quantification of the relative number of layer 5 SATB2+/CTIP2- neurons as a proportion of the total number of NEUN+ neurons in a cortical column revealed a substantial decrease in motor , but not sensory , cortices in mutant mice ( n=3 , mean ± SEM , *p=0 . 003 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 010 We examined the possibility that transformation of CTIP2+ corticospinal neurons into an alternative phenotype during the neonatal period contributes to the reduction in CTIP2+ layer 5 neuron number observed at P14 . We tested whether CTIP2+ layer 5 neurons might be transforming into a callosal phenotype by quantifying the proportion of cortical neurons that express SATB2 , an important transcription factor for the differentiation of callosal neurons ( Alcamo et al . , 2008; Britanova et al . , 2008 ) . Since SATB2 is coexpressed by a minority of CTIP2+ neurons and SATB2 plays a transient role in corticospinal differentiation , we only assayed SATB2+ layer 5 neurons that were CTIP2- ( Leone et al . , 2015; McKenna et al . , 2015 ) . A significant decrease in the proportion of SATB2+/CTIP2- neurons in layer 5 could be detected in motor cortex , but not sensory cortex , in P14 Map2k1/2;Neurod6-Cre mice when compared to controls ( Figure 4J ) . These findings demonstrate that the reduced number of CTIP2+ neurons in layer 5 does not coincide with a compensatory increase in the proportion of SATB2+/CTIP2- , presumably callosal , layer 5 neurons . Many neurodevelopmental syndromes that involve mutations in canonical members of the ERK/MAPK cascade exhibit enhanced ERK/MAPK activity ( Rauen , 2013 ) . Thus , we assessed effects of hyper-activation of the ERK/MAPK pathway on developing cortical excitatory circuits . A Cre-dependent constitutively-active Map2k1S217/222Q ( ca-Map2k1 ) overexpressing line was crossed with Neurod6-Cre and Emx1-Cre mice to induce gain-of-function ERK/MAPK signaling ( Krenz et al . , 2008 ) . Immunolabeling for MAP2K1 confirmed overexpression of the ca-Map2k1 allele in the ca-Map2k1;Neurod6-Cre cortex ( Figure 5—figure supplement 1 ) . In contrast to the loss-of-function mutants , ca-Map2k1;Neurod6-Cre mice are viable and able to breed , but are reduced in weight . Adult male control mice weighed 41 . 01 ± 1 . 17g while mutants were 27 . 68 ± 0 . 67g ( mean ± SEM , n=14 controls , 11 mutants ) . Ca-Map2k1;Emx1-Cre are viable and grossly normal , but exhibit lethality between 6 and 10 weeks of age . We first asked whether ERK/MAPK hyperactivation led to defects in lamination , particularly in layer 5 . CTIP2 labeling of mature sensorimotor cortices showed no overt defects in the specification or number of layer 5 neurons in ca-Map2k1;Neurod6-Cre mice ( Figure 5A–C ) . Based on our previous results , we hypothesized that hyperactivation of ERK/MAPK would lead to enhanced corticospinal axon growth into the spinal cord . In vivo DiI injections of P0 . 5 motor cortices were performed to assess the extent and pattern of layer 5 subcortical axon outgrowth . Innervation of the thalamus and medulla appeared normal in P3 . 5 gain of function mutants ( Figure 5D ) . Surprisingly , we detected a marked decrease in the initial extension of DiI labeled corticospinal afferents into the spinal cord in P3-4 ca-Map2k1;Neurod6-Cre mutants ( Figure 5D–E ) . Further , genetic labeling of corticospinal projections revealed a significantly diminished CST size in the spinal cord dorsal funiculus of both ca-Map2k1;Neurod6-Cre and ca-Map2k1;Emx1-Cre mutants ( Figure 6—figure supplement 1A–C ) . These data show that enhanced ERK/MAPK signaling results in a decrease in the elongation of corticospinal axons into the spinal cord that persists into adulthood . 10 . 7554/eLife . 11123 . 011Figure 5 . Gain of function ERK/MAPK signaling in ca-Map2k1;Neurod6-Cre mice decreases corticospinal extension into the spinal cord . ( A–C ) Representative confocal images of sensory cortices show that the expression and distribution of the callosal projection neuron marker , SATB2 , and subcerebral projection neuron marker , CTIP2 , in ca-Map2k1;Neurod6-Cre forebrains ( B ) appears normal when compared to littermate controls ( A ) ( n=4 , scale bar=100 µm ) . The relative proportion of CTIP2+ layer 5 neurons as a percentage of NEUN+ neurons within a radial unit did not show a significant difference between adult mutant and control motor cortices ( C ) ( n=3 , mean ± SEM , p=0 . 2 ) . ( D–E ) In vivo DiI injections into P0 neonates were performed to label corticospinal axons during initial stages of elongation into the spinal cord ( D ) . Analysis of DiI labeling in the spinal cord dorsal funiculus at P3 revealed a significant decrease in the extent of axonal elongation in mutant spinal cords relative to controls ( E ) ( n=3 , mean ± SEM , *p=0 . 033 , scale bar=100 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 01110 . 7554/eLife . 11123 . 012Figure 5—figure supplement 1 . Mouse model for excitatory neuron specific gain-of-ERK/MAPK signaling in the cortex . ( A–B ) Representative confocal images of forebrain sections from controls ( A ) and ca-Map2k1;Neurod6-Cre mutants ( B ) demonstrate that the expression of MAP2K1 is substantially higher in excitatory neurons across all layers of the cortex , but not in the striatum ( n=4 , scale bar=100 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 012 We tested whether the final pattern of axonal elongation and arborization was altered in ca-Map2k1;Neurod6-Cre mutants using a Cre-dependent , tdTomato-expressing viral vector ( AAV5-CAG-FLEX-tdTomato ) . Following unilateral injection of AAV into primary motor cortex at P1 , the extent of axonal labeling was assessed in the white matter tract and grey matter in hindbrain ( Figure 6A–F ) and spinal cord ( Figure 6G–J ) sections . To determine the amount of axonal elongation that occurred between the hindbrain and the spinal cord , high resolution confocal imaging and measurement of axonal RFP labeling in the white matter of the hindbrain corticobulbar tract ( CBT , Figure 6C–D ) and cervical spinal cord CST ( Figure 6I–J ) was performed . We then calculated the amount of RFP labeling in the spinal cord white matter tract relative to the amount of labeling in the hindbrain white matter tract within individual mice to provide a measure of axonal elongation . While control mice showed little reduction in the extent of labeling in the spinal cord white matter tract relative to the hindbrain , ca-Map2k1;Neurod6-Cre mutant mice exhibited a significant 60 . 14 ± 8 . 3% reduction in labeling in the spinal cord CST relative to the hindbrain ( Figure 6M ) . These findings provide further support for a substantial and persistent decrease in the elongation of corticospinal axons following hyperactivation of ERK/MAPK . 10 . 7554/eLife . 11123 . 013Figure 6 . Hyperactivation of ERK/MAPK enhances axonal branching in the hindbrain and spinal cord . ( A–J ) AAV5-FLEX-tdTomato injections into control ( Neurod6-Cre ) and ca-Map2k1;Neurod6-Cre motor cortices at P1 ( inset in A–B ) results in labeling of subcerebral axon projections in P30 hindbrains ( A–F ) and spinal cords ( G–L ) . To measure axonal elongation from the hindbrain to spinal cord , high resolution confocal images of the extent of axonal labeling in sections of the hindbrain corticobulbar tract ( cbt , C–D ) and brachial spinal cord corticospinal tract ( cst , I–J ) were collected and compared . The ratio of corticospinal to corticobulbar tract axonal labeling was significantly decreased in ca-Map2k1;Neurod6-Cre mutants ( M ) , providing further evidence for a reduction in corticospinal axon elongation at P30 ( n=3 , mean ± SEM , *p<0 . 01 , scale bar=500 µm ) . The extent of axonal branching was measured by comparing the amount of axonal labeling in the hindbrain ( A–B , zoom in E–F ) or spinal cord grey matter ( G–H , zoom in K–L ) , to the amount of axonal labeling in the corresponding white matter tract , the CBT ( C–D ) or CST ( I–J ) , respectively . We observed a significant increase in the relative ratio of grey/white matter labeling in ca-Map2k1;Neurod6-Cre mutant hindbrains ( n=4 , mean ± SEM , *p<0 . 05 ) and spinal cords ( n=3 , mean ± SEM , *p<0 . 05 ) when compared to controls ( N ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 01310 . 7554/eLife . 11123 . 014Figure 6—figure supplement 1 . ca-Map2k1;Emx1-Cre mice exhibit alterations in the pattern of corticospinal outgrowth in the spinal cord similar to ca-Map2k1;Neurod6-Cre mutants . ( A–D ) The entire pattern of corticospinal branching can be visualized in Emx1-Cre;Ai3 mice . P30 ca-Map2k1;Emx1-Cre;Ai3 mice showed a significant reduction in the area of corticospinal labeling in the white matter of lumbar spinal cord segments ( B and C ) when compared to Emx1-Cre;Ai3 controls ( A and C ) ( n=4 , mean ± SEM , *p=0 . 01 , scale bar=100 µm ) . When the amount of fluorescent labeling in the grey matter relative to the white matter was assessed ( D ) , mutant mice exhibited an increased extent of axonal branching per axon ( n=4 , mean ± SEM , *p=0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 014 In striking contrast to the decreased axonal elongation into the spinal cord , we noted a significant increase in the density of axonal labeling in the hindbrain grey matter of ca-Map2k1;Neurod6-Cre mutants ( Figure 6A–B , E–F ) . The hindbrain reticular nucleus is known to receive input from the primary motor cortex ( Esposito et al . , 2014 ) . The ratio of axonal labeling in the hindbrain grey matter relative to labeling in the hindbrain white matter ( CBT ) provided quantitative evidence of enhanced axonal arborization in the ca-Map2k1;Neurod6-Cre mutants ( Figure 6N ) . We also tested whether increased arborization could be detected in the spinal cord . The absolute level of axonal labeling density in the spinal cord grey matter was reduced in mutant mice ( Figure 6I-L ) . However , a comparison of the ratio of axonal labeling in the spinal cord grey matter relative to axonal labeling in the spinal cord white matter tract suggests that CST axon arborization is increased per axon ( Figure 6N ) . A similar result was also observed in the spinal cord of ca-Map2k1;Emx1-Cre;Ai3 mutants ( Figure 6—figure supplement 1A–B , D ) . In sum , our findings show that ERK/MAPK hyperactivation reduces the number of axons that extend longitudinally down the spinal cord , but the extent of arborizing axonal outgrowth into the hindbrain and spinal cord grey matter is enhanced . To definitively evaluate the requirement for ERK/MAPK signaling during callosal projection neuron development , we tested the effect of Map2k1/2 deletion on cortical layer 2/3 neurons . Excitatory neurons in layer 2/3 primarily project to intracortical targets ipsilaterally and contralaterally through the corpus callosum . We first asked whether ERK/MAPK signaling regulates upper layer neuron number and differentiation by assessing the expression of CUX1 and SATB2 , two critical transcriptional regulators of callosal projection neuron development , in P14 Map2k1/2;Neurod6-Cre cortices ( Alcamo et al . , 2008; Britanova et al . , 2008 ) . We detected a modest , but significant increase in the proportion of total NEUN+ neurons that express CUX1 in layer 2–4 in Map2k1/2;Neurod6-Cre sensory cortices ( Figure 7A–B , E ) . An increase in the proportion of CUX1 labeled neurons in mutant cortices might be an expected ratiometric result of the loss of layer 5 neurons . However , other mechanisms might contribute , such as effects on intermediate progenitors ( Li et al . , 2012 ) . SATB2 expression in upper cortical layers showed no major changes at either P1 ( Figure 3A–B ) or P14 ( Figure 1A–B; Figure 7C–D ) . Lastly , our microarray analysis did not show decreases in a number of genes known to be expressed in upper layer cortical neurons , in fact some genes were modestly increased ( Figure 2—source data 1 ) . We conclude that ERK/MAPK signaling is dispensable for the survival and specification of upper layer callosal projection neurons during early excitatory circuit development . 10 . 7554/eLife . 11123 . 015Figure 7 . The differentiation and morphology of callosal projecting neurons in layer 2/3 does not require ERK/MAPK signaling during development . ( A–E ) Analysis of intracortical neuron markers , CUX1 ( A–B ) and SATB2 ( C–D ) , from confocal images of P14 control ( A , C ) and Map2k1/2;Neurod6-Cre ( B , D ) sensory cortices did not reveal significant differences in overall expression . ( E ) Quantitation of the proportion of CUX1 labeled neurons in layer 2–4 as a percentage of NEUN labeled neurons in a radial unit was not decreased in mutant mice when compared to controls ( n=3 , mean ± SEM , *p=0 . 033 , scale bar = 100 µm ) . ( F–L ) Cell autonomous deletion of Map2k1/2 and labeling of a subset of layer 2/3 cortical neurons through unilateral in utero electroporation ( IUEP ) of pNeurod1-Cre and ploxp-STOP-loxpEGFP into the E14 . 5 ventricular zone . Following IUEP , coronal forebrain sections from electroporated P14 Map2k1wt/loxp;Map2k2-/-control ( F–G ) and Map2k1loxp/loxp;Map2k2-/- mutant ( H–I ) mice were immunostained for EGFP . Representative two-dimensional projections of confocal Z-stack images of EGFP-labeled layer 2/3 neuron bodies and dendrites in the electroporated hemisphere ( F , H ) and associated callosally projecting axons in the contralateral hemisphere ( G , I ) are shown ( scale bar=50 µm ) . ( J ) The extent of axonal innervation was assessed by quantifying the number of labeled pixels within contralateral radial units . A significant difference in axonal innervation was not detected following loss of ERK/MAPK signaling ( n=3 , mean ± SEM , p=0 . 97 ) . K . High resolution confocal Z-stacks of EGFP expressing layer 2/3 neurons were collected from control and mutant sections . Reconstruction and analysis of randomly selected layer 2/3 neurons show that the length of apical dendritic arbors was modestly reduced in Map2k1/2 deleted layer 2/3 neurons ( K ) ( n=10 control and 11 mutant neurons , mean ± SEM , p=0 . 04 ) . No significant effect on basal dendrite length was detected in mutant neurons ( mean ± SEM , p=0 . 625 ) . ( L ) Somal size assessment revealed a substantial decrease in the cross sectional size of EGFP labeled layer 2/3 neuron cell bodies following IUEP mediated Map2k1/2 deletion ( n>100 neurons per condition , mean ± SEM , *p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 01510 . 7554/eLife . 11123 . 016Figure 7—figure supplement 1 . ERK/MAPK signaling is dispensable for the differentiation and axonal morphology of callosal projection neurons in layer 2/3 . ( A–B ) Co-in utero electroporation ( IUEP ) of pNeurod1-Cre and ploxp-STOP-loxp-EGFP plasmids into E14 . 5 control ( Map2k1loxp/wt;Map2k2-/- ) and mutant ( Map2k1loxp/loxp;Map2k2-/- ) embryonic ventricles results in layer 2/3 neuron autonomous Map2k1/2 deletion and fluorescent labeling of excitatory neuron dendrites and axons . A . Immunohistochemical detection of MAP2K1 and EGFP expression in cortical layer 2/3 neurons of P14 Map2k1loxp/loxp;Map2k2-/- mice that were electroporated at E14 . 5 with pNeurod1-Cre and ploxp-STOP-loxpEGFP plasmids . Note the substantial loss of MAP2K1 expression in EGFP labeled mutant pyramidal neurons ( yellow arrows ) when compared to unlabeled control excitatory neurons ( white arrowheads ) . ( B ) Representative neurolucida reconstructions of layer 2/3 neurons from P14 control ( Map2k1loxp/wt;Map2k2-/- ) and mutant ( Map2k1loxp/loxp;Map2k2-/- ) cortex following electroporation with pNeuroD-Cre and ploxp-STOP-loxpEGFP at E14 . 5 ( scale bar=100 µm ) . ( C–I ) IUEP of pCAG-dsRed2 into E14 . 5 control and Map2k1/2;Neurod6-Cre mutant embryos leads to fluorescent labeling of layer 2/3 cortical neurons and callosal axons . P15 forebrain sections from control ( C ) and Map2k1/2;Neurod6-Cre mutant ( D , scale bar=2 mm ) mice were immunostained for NEUN and dsRED2 . High resolution confocal images show that upper layer neurons were electroporated at similar efficiency in control ( E ) and mutant ( G ) forebrains and contralateral axon innervation is not different ( F , H , scale bar = 50 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 016 We tested whether ERK/MAPK signaling is required for the morphological differentiation of upper layer neurons . Plasmids expressing Neurod1-Cre and a Cre-dependent EGFP construct were injected into the lateral ventricle of E14 . 5 embryos and unilaterally electroporated into developing radial progenitors . In utero electroporation ( IUEP ) of Map2k1loxp/loxp;Map2k2-/-embryos at this stage of development resulted in concurrent layer 2/3 neuron-autonomous deletion of Map2k1/2 and fluorescent labeling of a fraction of layer 2/3 dendritic and axonal arbors ( Figure 7F–I , Figure 7—figure supplement 1B ) . This approach resulted in clear loss of MAP2K1 immunolabeling in P14 Map2k1loxp/loxp;Map2k2-/-layer 2/3 neurons that had been electroporated with Neurod1-Cre and Cre-dependent EGFP when compared to surrounding non-electroporated neurons ( Figure 7—figure supplement 1A ) . The somal size of labeled neurons was significantly decreased in Map2k1/2 deleted layer 2/3 neurons that were electroporated with Neurod1-Cre , providing further evidence that ERK/MAPK signaling acts as a direct regulator of neuron cell body size ( Figure 7L ) . Morphological reconstruction of dendritic arbors revealed a modest statistically significant reduction in the length of apical , but not basal , dendrites following layer 2/3-autonomous loss of ERK/MAPK signaling when compared to controls ( Figure 7F , H , K; Figure 7—figure supplement 1B ) . Analyses of biocytin filled layer 2/3 neurons in P14-17 Map2k1/2;Neurod6-Cre visual cortices also revealed a modest but significant decrease in dendritic outgrowth without significant alterations in dendritic branching ( Figure 8—figure supplement 1D ) . IUEP resulted in clear labeling of callosal axons in contralateral , homotypic regions of sensory cortex ( Figure 7G , I ) . In contrast to corticospinal axons , no significant effect of Map2k1/2 deletion on layer 2/3 callosal axon innervation was detected ( Figure 7G , I , J ) . In a complementary approach , electroporation of dsRed2-expressing plasmids into E14 . 5 Map2k1/2;Neurod6-Cre mice also revealed no significant difference in the innervation of the contralateral cortex when compared to controls ( Figure 7—figure supplement 1C–H ) . These data demonstrate that in callosally projecting layer 2/3 neurons , ERK/MAPK signaling cell-autonomously regulates somal size and subtle aspects of dendrite outgrowth , but not axonal morphology . Many studies have shown that ERK/MAPK signaling regulates neuronal activity and synaptic plasticity ( Di Cristo et al . , 2001; Kushner et al . , 2005; Thomas and Huganir , 2004 ) . Our data show that ERK/MAPK signaling played a limited role in the morphological development of callosal layer 2/3 neurons , therefore , we tested whether the electrophysiological development of these neurons was altered in Map2k1/2;Neurod6-Cre mice . Indeed , gene expression profiling of P14 Map2k1/2;Neurod6-Cre whole cortical samples revealed diminished levels of many well-established plasticity-associated immediate early genes ( IEGs ) , including Arc , Fos , Npas2 , 4 , and Egr1 , 2 , 4 ( Figure 8A ) ( West and Greenberg , 2011 ) . Western blot and immunohistochemical screening confirmed a significant decrease in ARC protein expression in mutant cortices in essentially all cortical lamina when compared to littermate controls ( Figure 8B–D ) . In our P14 gene expression profiling , we also observed significantly decreased expression of many ion channels and neurotransmitter receptors that regulate neuronal excitability , including GABA/glycine receptor subunits ( Gabr-a5 , -b1 , -a1 , -b2 , Glra2 ) , sodium channels ( Scn3b , Scn2a1 ) , potassium channels ( Kcn-g1 , -v1 , -g3 ) , Hcn1 , Accn1 , Cacna2d3 , and Gria3 . Perhaps surprisingly , gene expression profiling of P21 gain of function ca-Map2k1;Neurod6-Cre cortices showed fewer significantly altered transcripts and did not reveal significant changes in plasticity- or activity-associated genes when compared to the loss-of-function cortices ( Figure 8A ) . This finding supports past studies showing modest effects of MAP2K1/2 hyperactivation on neuronal and glial transcript levels ( Nateri et al . , 2007; Sheean et al . , 2014 ) . Possible explanations are that negative feedback loops associated with the cascade and cortical network homeostatic effects reduce the consequences of gain of function . 10 . 7554/eLife . 11123 . 017Figure 8 . ERK/MAPK signaling promotes ARC expression and reduces neuron excitability in layer 2/3 and 5 pyramidal neurons during development . ( A ) Microarray profiling of whole cortical lysates detected a significant decrease in the expression of activity-dependent genes in P14 Map2k1/2;Neurod6-Cre mice compared to controls ( n=3 ) . ( B ) Western blotting of P14 control and Map2k1/2;Neurod6-Cre whole cortical lysates confirmed the decrease in ARC expression in mutant cortices ( B ) ( n=3 ) . ( C-D ) Immunostaining with an ARC antibody shows reduced expression across all cortical layers in P14 Map2k1/2;Neurod6-Cre mutants ( D ) when compared to controls ( C ) ( n=3 , scale bar=100 µm ) . ( E–F ) Average action potential frequency-current ( F/I ) curves recorded from layer 2/3 ( E ) and layer 5 ( F ) pyramidal neurons in acute slices of control and Map2k1/2;Neurod6-Cre sensory cortices . Recordings were performed in the presence of DNQX , APV , and picrotoxin to block all synaptic activity . Both layer 5 ( n=16 control , 14 mutant neurons , mean ± SEM; F/I Slope t-test , p<0 . 0001 ) and layer 2/3 ( n=14 control , 14 mutant neurons , mean ± SEM; F/I Slope t-test , p<0 . 0001 ) pyramidal neurons in Map2k1/2;Neurod6-Cre sensory cortices exhibited a marked increase in action potential firing frequency in response to increasing current injections . ( G–H ) Recordings of miniature excitatory postsynaptic currents in layer 2/3 pyramidal neurons from the visual cortex revealed a significant increase in the amplitude , but not frequency , of mEPSCs in neurons lacking Map2k1/2 ( n=12 neurons of each genotype from three independent litters , mean ± SEM , t-test , p=0 . 0012 ) ( G ) . In contrast , a substantial decrease in the frequency , but not amplitude , of mIPSCs from Map2k1/2;Neurod6-Cre neurons was observed ( H ) ( n=12 neurons of each genotype , mean ± SEM , t-test , p=0 . 0002 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 01710 . 7554/eLife . 11123 . 018Figure 8—figure supplement 1 . Loss of ERK/MAPK signaling leads to imbalanced excitatory and inhibitory synaptic drive in layer 2/3 neurons . ( A ) Intrinsic membrane properties and action potential properties of layer 5 ( n=16 control , 14 mutant neurons , mean ± SEM ) and layer 2/3 ( n=14 control , 14 mutant neurons , mean ± SEM ) neurons measured in current clamp . ( B–C ) Average miniature postsynaptic current waveforms recorded from layer 2/3 pyramidal neurons in control ( black ) and mutant ( red ) visual cortices . ( D ) Morphological reconstructions of confocal Z-stack images of biocytin backfilled P14 control and Map2k1/2;Neurod6-Cre layer 2/3 neurons following electrophysiological recordings . Quantification of dendritic length shows a significant decrease in apical and total dendritic length in Map2k1/2;Neurod6-Cre layer 2/3 neurons , but no significant effect on basal dendritic length or branch points could be detected ( n=7 neurons per condition from three independent litters , mean ± SEM , scale bar=50 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11123 . 018 The observed differences in activity-dependent IEGs and ion channels indicated that electrophysiological parameters might be disrupted specifically in loss-of-function mutants . We directly tested whether ERK/MAPK was required for the normal development of intrinsic excitability using whole-cell recordings from P14 Map2k1/2;Neurod6-Cre sensory cortices . We observed a significant decrease in the average membrane capacitance and the membrane time constant , tau , of mutant P14 pyramidal neurons ( Figure 8—figure supplement 1A ) . This result is an expected outcome from the smaller somal size of neurons in Map2k1/2;Neurod6-Cre cortices . No significant difference in the resting membrane potential or action potential amplitude/width was observed in mutant neurons ( Figure 8—figure supplement 1A ) . However , layer 2/3 neurons fired significantly more action potentials in response to increasing amounts of injected current , demonstrating that loss of Map2k1/2 increases layer 2/3 neuronal excitability . In mutant neurons , the action potential threshold was significantly reduced and the slope of the action potential output in response to current injection was increased by 42 . 6% relative to controls ( Figure 8E , Figure 8—figure supplement 1A ) . Interestingly , a similar increase in intrinsic hyperexcitability was noted when we recorded from the surviving , presumably callosal , layer 5 neurons in P14 Map2k1/2;Neurod6-Cre sensory cortices ( Figure 8F , Figure 8—figure supplement 1A ) . These data show that ERK/MAPK signaling normally reduces the intrinsic excitability of both cortical layer 2/3 and layer 5 neurons during early postnatal development . We sought to further understand how ERK/MAPK signaling regulates cortical excitability by measuring excitatory and inhibitory neurotransmission in neurons lacking Map2k1/2 . To further minimize indirect effects due to overt loss of layer 5 neurons in sensorimotor cortices , we examined basal synaptic properties in visual cortex layer 2/3 neurons from Map2k1/2;Neurod6-Cre mice , where neuron number is relatively intact . We tested whether loss of ERK/MAPK signaling influences synaptic transmission by recording miniature excitatory and inhibitory postsynaptic currents ( mEPSCs , mIPSCs ) from layer 2/3 neurons in control and Map2k1/2;Neurod6-Cre cortices . Interestingly , visual cortex layer 2/3 neurons exhibited a significant increase in mEPSC amplitude , while mEPSC frequency was unaffected ( Figure 8G ) . These findings suggest that loss of ERK/MAPK signaling increases the strength of glutamatergic synapses , likely through increases in postsynaptic AMPA receptor number or function . We next tested whether GABAergic neurotransmission was disrupted in Map2k1/2;Neurod6-Cre cortices . In contrast to spontaneous excitatory transmission , mutant layer 2/3 pyramidal neurons had a substantial decrease in mIPSC frequency , but not mIPSC amplitude ( Figure 8H ) . Overall , these findings demonstrate that loss of Map2k1/2 significantly disrupts the ratio of excitation to inhibition through increases in glutamatergic neurotransmission and decreases in GABAergic neurotransmission .
In vitro analyses have suggested that ERK/MAPK signaling acts as an important regulator of cortical neuron dendritic and axonal morphogenesis ( Dijkhuizen and Ghosh , 2005; Kumar et al . , 2005; Wu et al . , 2001 ) . However , our in vivo data reveal a remarkably specific requirement for the ERK/MAPK pathway in the morphological development of layer 5 long range projection neurons . ERK/MAPK signaling is dispensable for the early phases of CST neuronal differentiation , axon growth all the way to the medulla , and even responsiveness to midline guidance cues . However , a drastic delay in corticospinal axon growth was observed in Map2k1/2 mutants between P2-4 when CST axons in control mice start to invade the spinal cord . By P14 in controls , CST axons have reached the lumbar spinal cord . In P14 mutant mice , only a few corticospinal axons were present in the cervical spinal cord and no axons reached lumbar segments . Trophic cues linked to ERK/MAPK activation , specifically , BDNF , IGF1 , IGF2 , GDNF , and pleiotrophin , have been shown to regulate developing corticospinal neuron growth ( Dugas et al . , 2008; Giehl et al . , 1998; Özdinler and Macklis , 2006; Ueno et al . , 2013 ) . IGF1 has been described as a particularly potent growth factor for CST development both in vitro and in vivo ( Özdinler and Macklis , 2006 ) . IGF1 protein levels are higher in the spinal cord than the brain in the early postnatal stage ( Rotwein et al . , 1988 ) and we identified a focal pattern of IGF1 protein expression in the ventral medulla near the location where developing corticospinal axons enter the cervical spinal cord . However , we quite surprisingly found that deletion of Igf1r in long-range projection neurons did not lead to the predicted failure of CST development . Our results suggest that CST axon growth in the spinal cord may be orchestrated by multiple growth factors or via non-neuronal actions of IGF1 . For example , vascular endothelial cells also express IGF1R and endothelial cell-derived trophic cues are potent regulators of corticospinal neuron survival and outgrowth in vitro ( Dugas et al . , 2008 ) . Overall our data support the view that multiple cues converge upon ERK/MAPK to regulate transcriptional and cell biological processes required for extension of long axons . Following the delayed entry of axons into the spinal cord in Map2k1/2 mutants , we detected layer 5 neurons undergoing apoptosis in a restricted time frame during the first postnatal week . Counts revealed loss of roughly half of the CTIP2-expressing layer 5 neurons , which reflects a substantial proportion of the CST population . The reduced number of CST neurons in ERK/MAPK mutants is reminiscent of the loss of CST neurons after lesions to developing corticospinal projections in the neonatal period ( Merline and Kalil , 1990; Tolbert and Der , 1987 ) . Loss of trophic support from the spinal cord could plausibly account for these findings; an effect that might be exacerbated by known consequences of ERK/MAPK disruption on retrograde transport of growth factor signaling components including signaling endosomes ( Mitchell et al . , 2012 ) . We cannot completely exclude that the reduced number of CTIP2+ neurons in layer 5 results , in part , from conversion into an alternative neuronal type . For example , loss of Fezf2 results in conversion of corticospinal neurons into a callosal fate ( Chen et al . , 2005; Chen et al . , 2005; Molyneaux et al . , 2007 ) . Our assessment of SATB2+/CTIP2- layer 5 neurons did not reveal a coincident increase in the number of presumed callosally fated layer 5 neurons in Map2k1/2;Neurod6-Cre mice . Interestingly we did not find similar layer 5 apoptosis occurring in conditional Igf1r mutants , placing our results at odds with a recent study suggesting that microglia-derived IGF1 is required for the survival of layer 5 long range projection neurons ( Ueno et al . , 2013 ) . Whatever the mechanism , our data show that layer 5 corticospinal neurons are remarkably vulnerable to loss of ERK/MAPK signaling during the neonatal period and our results may be relevant to motor system dysfunction in RASopathy patients and layer 5 disorganization observed in ASDs . Perhaps surprisingly , rostrocaudal corticospinal axon extension was also highly diminished in the setting of enhanced ERK/MAPK signaling . At every stage examined from P3 onward , the number of CST axons in the dorsal funiculus in response to gain of ERK/MAPK signaling was reduced . A distinct feature of axon growth in gain of function mutants was enhanced axonal branching in the hindbrain . Normally , over the first two postnatal weeks , ERK/MAPK activation is upregulated , coincident with increased BDNF/TrkB expression ( Maisonpierre et al . , 1990; Timmusk et al . , 1994 ) . BDNF/TrkB signaling has been shown to promote corticospinal axon branching in vitro ( Özdinler and Macklis , 2006 ) . We suggest that hyperactivation of ERK/MAPK triggers mechanisms normally associated with BDNF induced corticospinal branching , possibly resulting in premature closure of the period of rostrocaudal axon elongation . Reduced axon extension of layer 5 projection neurons in the setting ERK/MAPK hyperactivation would likely have major implications for cortical circuit development in the human brain where distances are vastly longer than in rodents . In addition to effects on axonal connectivity , dysregulated ERK/MAPK signaling is likely to disrupt glutamatergic signaling ( Di Cristo et al . , 2001; Thomas and Huganir , 2004 ) . Studies have demonstrated that mature RASopathy mouse models exhibit reduced hippocampal LTP and spatial memory impairment ( Costa et al . , 2002; Cui et al . , 2008; Lee et al . , 2014 ) . Our findings provide genetic confirmation for prior work using pharmacological inhibitors demonstrating ERK/MAPK regulation of plasticity-associated genes induced by excitatory activity , including Arc , Egr2 , and Fos ( Majdan and Shatz , 2006; Panja et al . , 2009; Tropea et al . , 2006; Waltereit et al . , 2001 ) . Further , we establish a link between ERK/MAPK and expression of Npas4 , Nrn1/Cpg16 , and in vivo ( Coba et al . , 2008 ) . Unexpectedly , hyper-activation of ERK/MAPK signaling with the ca-Map2k1S217/221Q mutant had little effect on the expression of these same plasticity associated genes . These findings are in line with a past study of hippocampal gene expression in a different ca-MAP2K1 mouse mutant ( Nateri et al . , 2007 ) . The severity of effects on expression of plasticity associated genes is likely correlated with the magnitude of ERK/MAPK hypo- or hyper-activity in response to specific mutations . Since ERK/MAPK dependent changes in plasticity associated gene expression were widespread , we asked whether ERK/MAPK has general or layer-specific effects on developing projection neuron excitability and synaptic transmission . Our work demonstrates that complete loss of ERK/MAPK signaling leads to increased intrinsic excitability of both layer 2/3 and layer 5 pyramidal neurons . Further , we show that Map2k1/2 deleted layer 2/3 neurons exhibited increased excitatory synaptic strength . We conclude that ERK/MAPK signaling contributes to excitatory/inhibitory balance during the early postnatal period by reducing the excitability of cortical excitatory neurons . Changes in excitatory drive in a subset of layer 2/3 neurons have been shown to trigger a compensatory homeostatic increase in inhibitory drive , thus maintaining a stable synaptic excitatory/inhibitory ratio ( Xue et al . , 2014 ) . Remarkably , layer 2/3 pyramidal neurons lacking Map2k1/2 exhibit reduced inhibitory synaptic input in the absence of a fully compensatory change in excitation . Loss of ERK/MAPK activity in cortical pyramidal neurons may disrupt the homeostatic balance between synaptic excitation and inhibition , a mechanism hypothesized to form the neurological basis of autism ( Rubenstein and Merzenich , 2003 ) . ERK/MAPK may also regulate the release of BDNF from excitatory neurons , a well-known regulator of inhibitory synapse formation ( Huang et al . , 1999; Kohara et al . , 2007; Porcher et al . , 2011 ) . During postnatal cortical pyramidal neuron development , input resistance decreases and the frequency of GABAergic and amplitude of glutamatergic spontaneous neurotransmission increase ( Desai et al . , 2002; Morales et al . , 2002 ) . Our findings are consistent with pyramidal neurons lacking Map2k1/2 failing to undergo these developmental alterations and further suggest that loss of ERK/MAPK signaling may arrest normal physiological development . Importantly , the alterations in action potential threshold in neurons lacking Map2k1/2 may secondarily result from increased sodium channel density due to the smaller neuronal size or the observed increase in membrane resistance , which may also influence quantal amplitude measurements . Our observations that disruption of ERK/MAPK signaling alters excitatory/inhibitory balance and the expression of select plasticity associated genes , coupled with known effects on Hebbian forms of synaptic plasticity ( Thomas and Huganir , 2004 ) , suggest that alterations in ERK/MAPK signaling are likely to dramatically disrupt network activity and cortical re-wiring . Hypotonia , muscle weakness , and delay in motor milestones are often observed in RASopathy patients ( Dileone et al . , 2010; Mejias et al . , 2011; Oberman et al . , 2012; Stevenson et al . , 2012; Tidyman et al . , 2011 ) . Although some of these symptoms may be due to alterations in muscle development , we have shown here that upper motor neuron and corticospinal tract development are especially sensitive to both gain and loss of ERK/MAPK activity . Thus , deficits in motor function or motor learning in Ras/MAPK Syndromes may be explained , in part , by altered corticospinal connectivity . Importantly , our data point to distinct effects in gain vs loss of function mutants . Our results therefore argue for a mutation-specific approach to correct neurological dysfunction within the RASopathy spectrum . Aberrant long-range circuit development has also been proposed as a defining feature of ASD pathogenesis and that layer 5 may be a focal point ( Stoner et al . , 2014; Willsey et al . , 2013 ) . An interesting possibility is that altered activation of ERK/MAPK at early developmental stages in response to certain ASD linked mutations ( Fmr1 , Mecp2 , Tsc1/2 ) and environmental insults ( hypoxia , inflammation , etc . ) might contribute to select cortical circuit abnormalities , especially for neurons projecting over long distances .
Animal experiments were performed in accordance with established protocols approved by the Institutional Animal Care and Use Committee at the University of North Carolina–Chapel Hill and Arizona State University and NIH guidelines for the use and care of laboratory animals . All mice were housed in standard conditions with food and water provided ad libitum and maintained on a 12 hr dark/light cycle . Experiments were replicated a minimum of three times with mice derived from independent litters . Neurod6-Cre or Emx1-Cre expression alone did not have a detectable effect on the phenotypes described in this manuscript . Thus , Cre-expressing or Cre-negative littermates were utilized as controls unless indicated otherwise . Map2k1loxp/loxpmice possess a loxp flanked exon 3 while Map2k2-/- mice contain a neo insertion in exons 4–6 , which encodes the kinase domain ( Belanger et al . , 2003; Bissonauth et al . , 2006 ) . Map2k2-/- mice are viable and breed normally . Loxp-STOP-loxp-caMAP2K1 mice were kindly provided by Dr . Maike Krenz and Dr . Jeffrey Robbins ( Krenz et al . , 2008 ) ; the Igf1rloxp/loxp were kindly provided by Dr . Ping Ye ( Liu et al . , 2009 ) ; the Neurod6-Cre mice were kindly provided by Dr . Klaus Nave and Dr . Sandra Goebbels ( Goebbels et al . , 2006 ) ; and the Emx1-Cre mice were kindly provided by Dr . Franck Polleux ( Gorski et al . , 2002 ) . Ai3 mice were purchased from Jackson laboratories ( Madisen et al . , 2010 ) . All mice in this study were of mixed genetic background . Genomic DNA extracted from tail or toe samples was utilized for mouse genotyping by PCR using standard techniques . Primers for gene amplification are as follows ( listed 5’-3’ ) : Cre - TTCGCAAGAACCTGATGGAC and CATTGCTGTCACTTGGTCGT amplify a 266 bp Cre allele; Map2k1 –CAGAAGTTCCCACGACACTA , CTGAAGAGGAGTTTACGTCC , and GTCTGTCACTTGTCTTCTGG amplifies a 372 bp wild type and a 682 bp floxed allele; Map2k2 – CTGACCTTCCTGTAGGTG , ACTCACGGACATGTAGGA , and AGTCATAGCCGAATAGCCTC amplify a 293 bp wild-type allele and a 450 bp knockout allele; caMAP2K1 -GTACCAGCTCGGCGGAGACCAA and TTGATCACAGCAATGCTAACTTTC amplify a 600 bp mutant allele; Ai3 – AAGGGAGCTGCAGTGGAGTA , CCGAAAATCTGTGGGAAGTC , ACATGGTCCTGCTGGAGTTC , and GGCATTAAAGCAGCGTATCC amplify a 297 bp wild-type allele and a 212 bp Ai3 allele; Igf1r-CTTCCCAGCTTGCTACTCTAGG and CAGGCTTGCAATGAGACATGGG amply a 124 and a 220 bp band for wild-type and floxed alleles . P0-P5 litters were removed as a group , cryo-anesthetized on wet-ice for 3–5 min , and immediately injected with 50-500nl of solution using a 5 uL Hamilton syringe fitted with a 32 gauge beveled needle mounted to a stereotaxic arm . For viral labeling , the AAV5-CAG-FLEX-tdTomato vector was prepared by the UNC Viral Vector Core and diluted in sterile PBS , 5% sorbitol , and 0 . 1% Fast Green to allow for visualization prior to injection . For DiI tracing , a 10% DiI solution ( Life Technologies ) was prepared in DMSO and injected into the primary motor cortex or the cervical spinal cord . Upon completion of the injection , pups recovered on a heating pad and were returned as a group to the home cage . In utero electroporations were performed as previously described ( Li et al . , 2012 ) with a few modifications . In brief , E14 . 5 timed-pregnant females were anesthetized with isoflurane and the uterine horn was accessed through a cesarean section procedure . The lateral ventricles of the embryos were injected with 1–2 µg of plasmid prepared using an EndoFree plasmid purification kit ( Qiagen ) diluted in 1 x PBS with 0 . 1% Fast Green dye for visualization . Five electrical pulses were delivered at 30V ( 50 ms duration ) with a 950 ms interval using 5 mm paddle electrodes . The uterine horns were then gently reinserted into the abdominal cavity and the abdomen wall and skin was sutured . Electroporated mice were sacrificed at the appropriate time point and processed for analysis . Mice of the appropriate age were anesthetized and perfused transcardially with 4% paraformaldehyde/PBS . For cryoprotection , subdissected samples were incubated in a graded series of 10% , 20% , and 30% sucrose/PBS at 4°C before embedding in O . C . T . compound and freezing . Cryostat sections were collected on Fisherbrand Superfrost/Plus slides ( Fisher Scientific ) and air-dried prior to staining . For some experiments , brains were dissected , postfixed , and mounted in agarose prior to vibratome sectioning . For immunofluorescent staining , sections were rinsed in PBS and blocked with 5% normal serum/0 . 1% Triton X-100/PBS at room temperature . Primary antibodies were diluted in blocking solution and incubated 1–2 days at 4°C with gentle agitation . The antibodies utilized were; rabbit anti-Parvalbumin ( Swant ) , chicken anti-EGFP ( Aves Labs ) , rat anti-CTIP2 ( Abcam ) , rabbit anti-SATB2 ( Abcam ) , rabbit anti-CUX1 ( Santa Cruz ) , mouse anti-NEUN ( Chemicon ) , rabbit anti-Cleaved Caspase-3 ( Cell Signaling Technology ) , rabbit anti-IBA1 ( Wako ) , goat anti-IGF1 ( R&D Research ) , rabbit anti-PKCγ ( Santa Cruz ) , rabbit anti-MAP2K1/2 ( MEK1/2 ) ( Abcam ) , rabbit anti-P-MAPK1/3 ( ERK1/2 ) ( Cell Signaling Technology ) and rabbit anti-IGF1Rβ ( Cell signaling Technology ) . After rinsing in PBS/T , the secondary antibody was diluted in blocking solution and added overnight at 4°C . Secondary antibodies included Alexa Fluor 488 , 546 or 568 , and 647 conjugated anti-rabbit , anti-mouse , anti-rat , or anti-goat IgG ( Invitrogen ) . For some experiments slides were then incubated in Hoechst or DAPI for nuclear labeling , rinsed , and mounted . Images were collected with a Zeiss LSM 710 , 780 , or Leica SP5 laser scanning confocal microscope . Confocal images of regions of interest were collected from individual brain sections for each animal . For assessment of relative neocortical volume , cortical area was measured in five anatomically matched coronal sections along the rosto-caudal axis , averaged , and normalized against the control brain . For assessment of CTIP2 , CUX1 , SATB2 , and NEUN expressing cells , regions of primary sensory , motor , and visual cortex from at least three anatomically matched sections of mutant and control cortices were defined using morphological and anatomical features . Radial columns were outlined within the cortical region of interest and measured . The total sampled area from a specific region of cortex of a single brain ranged from 1–3 mm2 . Individual layer boundaries were determined by the changes in density and appearance of NEUN labeling . Images were then transferred into ImageJ , a pixel intensity threshold was set manually by an observer blind to the genotype , and watershed segmentation was performed . The binary image mask was then analyzed using the particle analysis tool to count the number of events with a min-max size cutoff of 50–600 µm2 for NEUN labeling and 30–300 µm2 for CTIP2+ , CUX1+ , or SATB2+ labeled nuclei . For NEUN density determination , the total number of NEUN+ cells within a radial column was divided by the area of the column and averaged across at least three separate columns within the cortical region of interest . For determination of the relative proportion of CTIP2+ or SATB2+ cells in layer 5 or CUX1+ cells in layers 2–4 , the number of each labeled cell within specific laminar boundaries was divided by the total NEUN count within the entire radial column to determine the proportion of each neuronal subtype per radial column . Results from this analysis were averaged across three individual radial columns per cortical region and normalized against the littermate control analyzed in parallel for each mutant . These images were also utilized for determination of the cross sectional area of neuronal soma . Randomly selected , well-labeled NEUN expressing neurons that included a DAPI labeled nucleus were outlined manually in Photoshop and measured . For analysis of axonal innervation , a modification of the cell counting procedure described above was utilized where images of the region of axonal innervation from at least three anatomically matched sections were collected , manually thresholded in ImageJ , and the number of labeled pixels was measured . For measurement of ca-Map2k1;Emx1-Cre spinal cords background signal was determined by measuring a non-innervated area in the same section and subtracted from the mean intensity of an anatomical region of interest . Representative images have been cropped and adjusted for brightness and contrast in Photoshop for presentation . Student's t-test was used for statistical analysis . Sensorimotor cortices were dissected from both mutant and litter mate control mice and lysed inRIPA buffer ( 0 . 05M Tris-HCl , pH 7 . 4 , 0 . 5M NaCl , 0 . 25% deoxycholic acid , 1% NP-40 , and 1 mM EDTA , Millipore ) supplemented with 0 . 1% SDS , protease inhibitor cocktail ( Sigma ) and phosphatase inhibitor cocktail II and III ( Sigma ) . Lysates werecleared by centrifugation and protein concentration was determinedusing the Bio-Rad protein assay ( Bio-Rad ) using BSA as a standard . Equal amounts of protein were denatured in reducing sample buffer , separated by SDS-PAGE gels , and blotted to PVDF membranes ( Bio-Rad ) . Blots were blocked with 5% BSA in TBS containing 0 . 5% Tween 20 ( TBS-T ) for 1 hr at roomtemperature , then incubated overnight at 4°C with primaryantibodies . The primary antibodies used were rabbit anti-phospho MAPK1/3 ( ERK1/2 ) ( Thr202/Tyr204 ) ( Cell Signaling Technology , Inc ) , rabbit anti-MAPK1/3 ( ERK1/2 ) ( CST ) , rabbit anti-phospho-p90RSK ( Thr573 ) ( Cell Signaling Technology , Inc . ) , rabbit anti-RSK ( Cell Signaling Technology , Inc . ) , rabbit anti-MSK1 ( Ser360 ) ( Abcam ) , rabbit anti-MSK1 ( Cell Signaling Technology , Inc . ) , rabbit anti-MAP2K1/2 ( MEK1/2 ) ( Cell Signaling Technology , Inc . ) , rabbit anti-ARC ( Synaptic System ) and anti-GAPDH ( Cell Signaling Technology , Inc . ) . After washing with TBS-T , membraneswere incubated with HRP-conjugated secondary antibodies in 5% milk in TBS-T for 2 hr at room temperature . Blots were washed with TBS-T and detection wasperformed with SuperSignal West Pico chemiluminescent substrate ( Thermo Scientific ) . Dorsal cortices from control and mutant embryos derived from two independent litters at P9 and three independent litters at P14 were dissected and total RNA was extracted with Trizol ( Invitrogen ) , followed by mRNA extraction with a Qiagen RNeasy Mini kit per manufacturer’s instructions . RNA was assayed for quality and quantity with an Agilent 2100 Bioanalyzer and a Nanodrop spectrophotometer . Total RNA was amplified , labeled , and hybridized on Affymetrix arrays in the UNC Functional Genomics Core . Slides were processed and scanned according to the manufacturer protocol . Data was further processed using the RMA algorithm for background adjustment , quantile normalization , and median polish probeset summarization . P14 and P21 cortical samples were log2 transformed and differential expression and p-values were calculated using a mixed-model ANOVA in Partek Genomics Suite . Transcripts demonstrating an average change in expression that is >1 . 5 fold ( up-regulated ) or < –1 . 5 fold ( down-regulated ) and a p-value < 0 . 05 were considered differentially expressed . To examine Map2k1 exon 3 expression specifically , individual probe ID#s 561826 ( TGGTTCCGGATTGCGGGTTTGATCT ) , 1056599 ( TCCGGATTGCGGGTTTGATCTCCAG ) , and 1079432 ( CCGGATTGCGGGTTTGATCTCCAGG ) were processed as described without probeset summarization . All microarray data from this study have been deposited into the NCBI GEO Database under accession number GSE75129 . For cortical slice preparation , mice were anesthetized with pentobarbital sodium and decapitated after disappearance of corneal reflexes . Brains were rapidly removed and 350 μm coronal sections were cut using a vibrating microtome ( Leica VT1200S ) in ice-cold dissection buffer containing ( in mM ) : 87 NaCl , 2 . 5 KCl , 1 . 25 NaH2PO4 , 25 NaHCO3 , 75 sucrose , 10 D- ( + ) -glucose , 1 . 3 ascorbic acid , 7 MgCl2 and 0 . 5 CaCl2 , bubbled with 95% O2 and 5% CO2 . Following dissection , slices recovered for 20 min at 35°C in artificial cerebrospinal fluid ( ACSF ) containing ( in mM ) : 124 NaCl , 3 KCl , 1 . 25 Na2PO4 , 26 NaHCO3 , 1 MgCl2 , 2 CaCl2 and 20 D- ( + ) -glucose , saturated with 95% O2 , 5% CO2 and were then kept at 21–24°C for at least 40 min . During this period and before transferring sections to the recording chamber , slices were maintained in ACSF supplemented with 1 . 25 mM ascorbic acid . Recordings were made in a submersion chamber at 32°C +/- 1°C in ACSF , except where noted . Patch pipettes were pulled from thick-walled borosilicate glass with open tip resistances of 2–7 MΩ . Neurons were recorded in either voltage- or current-clamp configuration with a patch clamp amplifier ( Multiclamp 700A ) , and data were acquired and analyzed using pCLAMP 10 software ( Molecular Devices ) . L2/3 or L5 neurons were visually identified with infrared differential interference contrast optics . All recordings were performed from excitatory neurons as verified by the presence of dendritic spines and/or an apical dendrite oriented towards the pial surface . To minimize variability , recordings were performed from a mutant and control littermate with identical , same day conditions . Voltage-clamp recordings were performed from visual cortical layer 2/3 neurons , as described previously ( Larsen et al . , 2011 ) . Events with a rapid rise time and exponential decay were identified as mEPSCs or mIPSCS using an automatic detection template in pCLAMP 10 . To isolate mEPSCs , slices were placed in a submersion chamber , maintained at 32°C and perfused at 2 mL/min with oxygenated ACSF containing 200 nM tetrodotoxin and 50 μM picrotoxin . Pipettes were filled with internal solution containing ( in mM ) : 100 K-Gluconate , 20 KCl , 0 . 2 EGTA , 4 Mg-ATP , 0 . 3 Na-GTP , 10 HEPES , 10 Na-phosphocreatine , and 0 . 5% ( w/v ) neurobiotin with pH adjusted to 7 . 25 with 1M KOH and osmolarity adjusted to ~300 mOsm by addition of sucrose . To isolate mIPSCs , ACSF included 20 μM 6 , 7-dinitroquinoxaline-2 , 3-dione ( DNQX ) , 100 μM D , L-2-amino-5-phosphonopentanoic acid ( D , L-AP5 ) , and 200 nM tetrodotoxin . Pipettes were filled with a high internal chloride concentration to increase the chloride driving force at -70 mV and contained ( in mM ) : 134 KCl , 2 NaCl , 10 HEPES , 0 . 2 EGTA , 10 sucrose , 4 Mg-ATP , 0 . 3 Na-GTP , 14 Naphosphocreatine , 0 . 5% ( w/v ) neurobiotin , with pH adjusted to 7 . 2 with KOH and osmolarity adjusted to ~300 mOsm by addition of sucrose . All voltage-clamp recordings were sampled at 10 kHz and Bessel filtered at 2 kHz . Series and input resistances were monitored throughout the experiments by measuring the response to a −5 mV step during each sweep . Series resistance was calculated using the capacitive transient at the onset of the step and input resistance was calculated from the steady-state current during the step . No series resistance compensation was applied . Intrinsic excitability recordings were performed in L2/3 and L5 somatosensory cortex with AMPA/Kainate , NMDA , and GABA ( A ) receptor-mediated synaptic transmission blocked by adding 20 μM DNQX , 100 μM D , L-AP5 , and 50 μM picrotoxin to the ACSF . Throughout the experiment , current was injected to maintain a −70 mV resting potential . Action potential threshold was defined as the voltage at which dv/dt=20 V/s . Adaptation ratio was measured as the ratio of the third interspike interval to the last interspike interval at currents which evoked 6–7 action potentials to minimize confounds in adaptation ratio due to differences in action potential number between genotypes . Action potential amplitude was measured relative to the membrane potential prior to current injection ( -70 mV ) . Action potential slope was analyzed by comparing threshold currents for generating action potentials to responses at two times the threshold current , which was divided by the difference in current levels . Input resistance was calculated by measuring the steady-state response to a 50 pA , 300 ms step at each current injection and was averaged across all current injections . Resting membrane potential was measured at the onset of the recording . Current clamp recordings were sampled at 20 kHz and Bessel filtered at 10 kHz . Membrane capacitance and time constant values ( Figure 8—figure supplement 1 ) were calculated in voltage-clamp at the onset of the recording . Capacitance values were measured from the averaged integrated transient capacitive current produced during a 10 mV step sampled at 100 kHz in membrane test mode . Membrane time constant values were measured from the transient portion of the current response to the voltage step and were fit by a single exponential . | In the nervous system , cells called neurons form networks that relay information in the form of electrical signals around the brain and the rest of the body . Typically , an electrical signal travels from branch-like structures at one end of the cell , through the cell body and then along a long fiber called an axon to reach junctions with another neurons . The connections between neurons start to form as the nervous system develops in the embryo , and any errors or delays in this process can cause severe neurological disorders and intellectual disabilities . For example , genetic mutations affecting a communication system within cells known as the ERK/MAPK pathway can lead to a family of syndromes called the “RASopathies” . Abnormalities in this pathway may also contribute to certain types of autism . However , it is not clear how alterations to the ERK/MAPK pathway cause these conditions . Xing et al . investigated whether ERK/MAPK signaling regulates the formation of connections between neurons and the activity of neurons in mouse brains . The experiments showed that the growth of axons that extend from an area of the brain called the cerebral cortex towards the spinal cord are particularly sensitive to changes in the level of signaling through the ERK/MAPK pathway . On the other hand , inhibiting the pathway has relatively little effect on the growth of axons within the cerebral cortex . Further experiments showed that many neurons in the cerebral cortex require the ERK/MAPK pathway to activate genes that alter neuronal activity and the strength of the connections between neurons . Xing et al . ’s findings suggest that defects in the connections between the cerebral cortex and different regions of the nervous system may contribute to the symptoms observed in patients with conditions linked to alterations in ERK/MAPK activity . Future studies will focus on understanding the molecular mechanisms by which ERK/MAPK pathway influences the organization and activity of neuron circuits during the development of the nervous system . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2016 | Layer specific and general requirements for ERK/MAPK signaling in the developing neocortex |
In the unfolded protein response ( UPR ) , stress in the endoplasmic reticulum ( ER ) activates a large transcriptional program to increase ER folding capacity . During the budding yeast UPR , Ire1 excises an intron from the HAC1 mRNA and the exon products of cleavage are ligated , and the translated protein induces hundreds of stress-response genes . Using cells with mutations in RNA repair and decay enzymes , we show that phosphorylation of two different HAC1 splicing intermediates is required for their degradation by the 5′→3′ exonuclease Xrn1 to enact opposing effects on the UPR . We also found that ligated but 2′-phosphorylated HAC1 mRNA is cleaved , yielding a decay intermediate with both 5′- and 2′-phosphates at its 5′-end that inhibit 5′→3′ decay and suggesting that Ire1 degrades incompletely processed HAC1 . These decay events expand the scope of RNA-based regulation in the budding yeast UPR and have implications for the control of the metazoan UPR .
During the unfolded protein response ( UPR ) , protein folding stress in the lumen of the endoplasmic reticulum leads to oligomerization of the transmembrane kinase/endoribonuclease Ire1 and the processing of a cytoplasmic mRNA to yield splicing intermediates with 2′ , 3′-cyclic phosphate ( PO4 ) and 5′-hydroxyl ( OH ) termini ( Gonzalez et al . , 1999 ) . In budding yeast , excision of an intron from the HAC1u mRNA ( ‘u’ denoting the unspliced mRNA ) by Ire1 is followed by exon ligation by the multifunctional Trl1 RNA ligase ( Sidrauski et al . , 1996 ) involving 5′-phosphorylation of the 5′-OH product , adenylylation of the 5′-PO4 , and resolution of the 2′ , 3′-cyclic PO4 to a 2′-PO4/3′-OH . The newly produced 3′-OH serves as the nucleophile to attack the 5′-adenylate intermediate , yielding a ligated mRNA with an internal 2′-PO4 . The 2′-PO4 is assumed to be removed in a separate reaction by the 2′-phosphotransferase , Tpt1 , in a NAD+-dependent reaction ( Culver et al . , 1997 ) . The spliced mRNA , called HAC1s mRNA ( ‘s’ denoting spliced mRNA ) ( Li et al . , 2018 ) , is translated into a transcription factor that activates the expression of dozens of stress-response genes to mitigate protein-folding stress ( Ron and Walter , 2007 ) . In addition , Hac1 activates its own promoter in a positive feedback loop that generates more HAC1u and permits sustained UPR activation ( Ogawa and Mori , 2004 ) ( Figure 1A ) . Control of this positive feedback loop ensures UPR suppression during normal growth and rapid activation upon stress exposure . To facilitate the control of UPR activation , HAC1u contains cis-regulatory elements that suppress unintended translation and promote rapid processing . A long-range base-pairing interaction between the 5′-UTR and intron prevents ribosome initiation to suppress translation of HAC1u mRNA ( Chapman and Walter , 1997; Di Santo et al . , 2016 ) . If a ribosome initiates on HAC1u , translation through the 5′-exon/intron junction yields a truncated protein with an intron-encoded C-terminal peptide ‘degron’ that targets it for ubiquitylation and degradation ( Di Santo et al . , 2016 ) . A stem-loop ( the ‘3′-BE’ ) in the 3′-untranslated region of HAC1 tethers the mRNA to the ER membrane , ensuring rapid Ire1-mediated cleavage following ER stress ( Aragón et al . , 2009 ) . Previous work found unexpected roles for RNA decay and repair enzymes acting on HAC1 mRNA in the budding yeast unfolded protein response . Ire1 is a metal-ion-independent endonuclease that produces RNA cleavage products with 5′-OH termini ( Gonzalez et al . , 1999 ) . In cells lacking the cytoplasmic 5′→3′ exonuclease Xrn1 , HAC1 splicing intermediates accumulate with 5′-PO4 termini , indicating that a RNA 5′-kinase phosphorylates HAC1 processing intermediates and that not all HAC1 splicing intermediates are productively ligated ( Harigaya and Parker , 2012; Peach et al . , 2015 ) . In addition to its role in HAC1 exon ligation , Trl1 is required to relieve translational attenuation of HAC1s by an unknown mechanism ( Mori et al . , 2010 ) . In cells expressing the T4 bacteriophage RNA repair enzymes PNK and RNL1 in lieu of TRL1 , ligated HAC1 molecules contained single nucleotide deletions from the 3′-terminus of the 5′-exon , indicating that a 3′→5′ exonucleolytic activity acts on the cleaved 5′-exon ( Schwer et al . , 2004 ) and nuclear 3′→5′ decay of HAC1u liberates the 3′-BE , tuning the activation potential of the UPR ( Sarkar et al . , 2018 ) . Recent studies showed that RNA decay also plays a role in the UPR in other organisms . During UPR activation in the fission yeast , Ire1 incises specific mRNAs to promote their stabilization or degradation ( Guydosh et al . , 2017; Kimmig et al . , 2012 ) . This mode of Ire1 cleavage is similar to the metazoan Regulated Ire1-Dependent Decay ( RIDD ) pathway wherein Ire1 incises some ER-localized mRNAs and the cleavage products are degraded by Xrn1 and the cytoplasmic exosome ( Hollien and Weissman , 2006 ) . Here , we used budding yeast with mutations in RNA repair and decay enzymes to show that HAC1 splicing intermediates are processed at multiple steps prior to ligation , limiting the impact of spurious Ire1 activation and unintentional HAC1 cleavage . Our studies also show that incompletely spliced HAC1s mRNA is targeted for degradation , which may be used to attenuate the UPR .
We recently showed that the functions of the essential RNA repair enzymes Trl1 and Tpt1 in budding yeast can be genetically bypassed by the expression of intronless tRNAs , which are able to support translation in trl1∆ and tpt1∆ cells ( Cherry et al . , 2018 ) . Because Trl1 is required for HAC1s ligation and subsequent UPR activation ( Sidrauski et al . , 1996 ) , trl1∆ cells are unable to grow on media containing tunicamycin ( Figure 1B ) . In contrast , the general growth defect of tpt1∆ cells is unaffected by tunicamycin ( Cherry et al . , 2018 ) ( Figure 1B ) , indicating that these cells can activate the UPR . Combination of trl1∆ and tpt1∆ with mutations in 5′→3′ and 3′→5′ decay factors xrn1∆ or ski2∆ led to more pronounced growth defects than the single deletions , but removal of these decay factors did not affect the growth deficit of trl1∆ or tpt1∆ cells on tunicamycin ( Figure 1B ) . Given the multiple enzymatic roles of Trl1 and Tpt1 during RNA repair , we sought to understand how the loss of these enzymes affected HAC1 mRNA splicing . We visualized HAC1 splicing intermediates by northern blotting with probes for the HAC1 3′-exon and intron ( Figure 1C , D ) and found cleavage and ligation of HAC1u in wild-type cells in the presence of tunicamycin , leading to high levels of HAC1s . As expected , trl1∆ cells lacking RNA ligase activity did not produce HAC1s upon tunicamycin treatment . However , cleaved 3′-exon and intron accumulated upon tunicamycin treatment in trl1∆ cells ( Figure 1C , D ) , indicating a defect in 3′-exon decay . Cleaved HAC1 3′-exon often appears as a smear of products between ~450 nt and ~575 nt ( Figure 1C ) ; we attribute this size heterogeneity to differences in poly ( A ) tail presence or length , as the 5′-ends of these products occur uniformly at one site ( Figure 5A ) . The 2′-phosphotransferase Tpt1 is essential in budding yeast to remove 2′-phosphate groups from ligated tRNAs ( Culver et al . , 1997 ) , but its role in HAC1 mRNA processing during the UPR has not been defined . Although the growth of tpt1∆ cells is unaffected by tunicamycin ( Figure 1B ) , specific perturbations of HAC1 processing in tpt1∆ cells indicate that residual 2′-phosphate groups on HAC1 mRNA cause defects in cleavage and ligation . Whereas tunicamycin treatment led to cleavage of HAC1u and production of HAC1s in tpt1∆ cells , the levels of HAC1s ( Figure 1C ) and excised intron ( Figure 1D ) are lower than in wild-type cells . In addition , despite the fact that tpt1∆ cells have functional RNA ligase , cleaved 3′-exon accumulated to high levels upon tunicamycin treatment ( Figure 1C , lane 6 ) . To further investigate the 3′-exon decay defect , we examined splicing of HAC1 in xrn1∆ cells . We found that HAC1s accumulated in the absence of tunicamycin ( Figures 1C , 2A , D , E and F ) . This promiscuous processing was surprising given that HAC1s is undetectable in wild-type cells under normal growth conditions , and it suggested that Xrn1 somehow limits production of HAC1s . In xrn1∆ cells , 3′-exon accumulated to modest levels in both the absence and presence of tunicamycin ( Figure 2A ) , whereas in trl1∆ cells , HAC1 3′-exon accumulated to higher levels ( Figure 2A and B ) . Moreover , the abundance of 3′-exon was similar in trl1∆ and trl1∆ xrn1∆ cells ( Figure 2A ) , indicating that Xrn1 requires Trl1 for 3′-exon degradation . Previous work showed that Trl1 5′-kinase activity is required for the Xrn1-mediated degradation of excised tRNA introns in budding yeast ( Wu and Hopper , 2014 ) , and we considered whether this pathway also degraded HAC1 3′-exon . Indeed , expression of a kinase-inactive version of Trl1 ( Trl1-D425N ) ( Wang et al . , 2006 ) did not restore Xrn1-mediated decay of the 3′-exon ( Figure 2B ) , affirming that the 5′-kinase activity of Trl1 ligase is required for Xrn1-mediated suppression of HAC1 splicing . We also tested whether the ligase activity of Trl1 affected HAC1 3′-exon abundance using an adenylyl-transferase/ligase defective allele ( Trl1-K114A ) ( Sawaya et al . , 2003 ) and found that additional 3′-exon accumulates compared to wild-type ( Figure 2C ) , indicating that ligation also contributes to processing of free 3′-exon . Furthermore , we examined the accumulation of 3′-exon in cells lacking Dxo1 , a distributive , 5′-phosphate-dependent 5′→3′ exonuclease ( Chang et al . , 2012 ) , and found that HAC1 3′-exon accumulation was unaffected in dxo1∆ cells . In addition , the levels of 3′-exon were similar in xrn1∆ and dxo1∆ xrn1∆ cells ( Figure 2D ) , indicating that Xrn1 is the primary factor responsible for 5′→3′ decay of the 3′-exon . Together these data indicate that ligation and Xrn1-mediated 5′→3′ decay compete for the 5′-phosphorylated 3′-exon splicing intermediate ( Figure 2G , top ) . Examination of HAC1 splicing in trl1∆ cells expressing the E . coli RtcB RNA ligase ( Tanaka et al . , 2011 ) provided additional evidence of a competition between ligation and decay . RtcB catalyzes ligation of 2′ , 3′-cyclic PO4 and 5′-OH RNA termini via a unique mechanism involving nucleophilic attack of the 5′-OH on a 3′-guanylate intermediate; accordingly , RtcB does not have 5′-kinase activity ( Chakravarty et al . , 2012 ) . We found that upon tunicamycin treatment , HAC1s was produced in trl1∆ ( RtcB ) cells ( Figure 2E ) , as shown previously ( Tanaka et al . , 2011 ) . However , under normal growth conditions , trl1∆ ( RtcB ) cells also promiscuously spliced HAC1s at levels similar to xrn1∆ cells ( Figure 2E and F ) . We propose that because HAC1 ligation by RtcB does not involve a 5′-phosphate intermediate , Xrn1 is unable to degrade the 5′-hydroxyl exon product of Ire1 cleavage , tipping the balance toward ligation and producing HAC1s under normal growth conditions ( Figure 2G , bottom ) . Thus Xrn1-mediated decay of HAC1 3′-exon appears to counteract a low rate of background Ire1 cleavage to ensure the UPR is only activated when legitimately stressed . In several instances , cells with mutations in repair and decay factors can splice HAC1 but fail to grow on media containing tunicamycin ( Figure 3A ) , indicating that HAC1s production is not sufficient to activate the UPR . We assayed expression of KAR2 , an ER chaperone and direct target of the Hac1 transcription factor ( Kohno et al . , 1993 ) , and found that all repair and decay mutants express significantly less KAR2 mRNA upon tunicamycin treatment ( Figure 3B ) , consistent with another layer of UPR regulation downstream of HAC1s production . Excised HAC1 intron accumulates upon tunicamycin treatment in trl1∆ , xrn1∆ , and trl1∆ xrn1∆ cells ( Figure 3C ) . The intron decay products that accumulate in these cells are indicative of kinase-mediated decay: in xrn1∆ cells that lack Xrn1 but have 5′-kinase activity , intron products appear with a few distinct , smaller products below the full-length 252 nt excised intron ( Figure 3C , lanes 3 and 4 ) . In contrast , excised intron accumulated as a uniform , ~250 nt product in trl1∆ cells that lack 5′-kinase activity ( Figure 3C , lanes 5–8 ) , independent of XRN1 status . Moreover , production of shorter decay products ( like those present in xrn1∆ cells ) was dependent on Trl1 5′-kinase catalytic activity ( Figure 3D ) . Interestingly , expression of the adenylyl-transferase-dead/ligase-dead allele , trl1-K114A , also led to accumulation of some free HAC1 intron ( Figure 3I ) , potentially indicating a role for ligation in the processing of liberated HAC1 intron . Together , these data show that excised HAC1 intron is a substrate for kinase-mediated decay with strict dependence on a 5′-phosphorylation step to promote 5′→3′ decay . A single deletion of Dxo1 had no effect on HAC1 intron degradation ( Figure 3E ) ; however , the sizes of smaller decay products in xrn1∆ dxo1∆ cells were subtly different than in xrn1∆ cells ( Figure 3E ) , indicating that Dxo1 and other exonucleases partially degrade excised HAC1 intron , but only when it accumulates in xrn1∆ cells . Consistent with this notion , we found that the cytoplasmic exosome also contributes to HAC1 intron turnover ( Figure 3H ) , but this mode of decay is unlikely to regulate the UPR as the growth of ski2∆ cells is unaffected by tunicamycin ( Figure 1B ) . In trl1∆ ( RtcB ) cells , excised HAC1 intron accumulates as a circle , evinced by its altered mobility and resistance to Xrn1-mediated decay in vivo ( Figure 3F ) , and its resistance to RNase R degradation in vitro ( Figure 3G ) . Circularization of the HAC1 intron by RtcB in trl1∆ cells is facilitated by 5′-OH and 2′ , 3′-cyclic PO4 termini created by Ire1 cleavage and the absence of Trl1 end modification activities that could otherwise produce termini incompatible with RtcB ligation ( 5′-PO4 or adenylylate; and 2′-PO4/3′-OH ) . It is noteworthy that circularized intron accumulates to high levels in the absence of tunicamycin ( Figure 3F and G ) , indicating that Ire1 catalyzes a low level of intron excision ( and 3′-exon excision ( Figure 1C ) ) from HAC1u during normal growth , leading to the accumulation of stable , circularized introns in the presence of RtcB . The HAC1 intron and 5′-UTR form an extensive base-pairing interaction that inhibits ribosome initiation ( Chapman and Walter , 1997; Di Santo et al . , 2016 ) . Thus , together these data evoke a model in which kinase-mediated decay of the excised HAC1 intron is required for HAC1s translation , and a failure to degrade HAC1 intron—even when HAC1s is produced—prevents HAC1s translation and subsequent expression of stress-responsive genes . Despite their ability to make HAC1s , xrn1∆ and trl1∆ ( RtcB ) cells have growth defects on media containing tunicamycin ( Figure 3A ) and the relative severity of their defects parallels the accumulation of excised intron in these cells . Cells lacking Xrn1 have a modest growth defect and accumulate linear intron , which can be degraded by other exonucleases ( Figure 3 ) . In contrast , trl1∆ ( RtcB ) cells have a severe growth defect and accumulate high levels of a stable , circularized intron that is immune to exonucleolytic decay ( Figure 3F and G ) . We believe these findings resolve the mystery of the previously identified ‘second function’ of Trl1 required for UPR activation ( Mori et al . , 2010 ) , namely that—in addition to its ligase activity—Trl1 initiates kinase-mediated decay of the excised HAC1 intron , relieving its repressive effect on HAC1s and activating translation . In addition to the canonical 5′-exon product of 5′-splice site cleavage , a second product uniquely accumulates in tpt1∆ and tpt1∆ ski2∆ cells that is ~50 nt shorter than full-length 5′-exon ( Figure 4A ) . Expression of a catalytically inactive form of Tpt1 ( Tpt1-R138A , ( Sawaya et al . , 2005 ) ) in tpt1∆ cells failed to rescue this defect ( Figure 4B ) , affirming that the catalytic activity of Tpt1 is required to prevent accumulation of the shorter 5′-exon fragment . A corresponding elongated 3′-exon fragment accumulates in tpt1∆ , and more intensely in tpt1∆ xrn1∆ cells , indicating it is degraded by Xrn1 ( Figure 4C–E ) . The elongated 3′-exon is specifically detected using a northern probe with a sequence complementary to the distal 3′-end of the 5′-exon ( Figure 4C , D ) , indicating that a portion of the 5′-exon is responsible for the increased size of this decay intermediate . Moreover , a fragment of similar size hybridizes to a probe for the 3′-exon , suggesting that the sequence derived from 5′-exon is linked to the 3′-exon ( Figure 4E ) ; together these data indicate that HAC1s is cleaved upstream of the 2′-phosphorylated ligation junction in tpt1∆ cells , and these products are degraded by both Xrn1 and the cytoplasmic exosome . To determine whether HAC1s sequence is sufficient for cleavage , we expressed plasmid-encoded HAC1u and HAC1s in hac1∆ cells to identify HAC1 splicing intermediates . We performed the analysis on trl1∆ cells , reasoning that if HAC1s sequence were sufficient to cause cleavage , we would expect an accumulation of 3′-exon in cells unable to carry out ligation or degrade the products by kinase-mediated decay . Tunicamycin treatment of trl1∆ hac1∆ cells expressing HAC1u caused Ire1-mediated cleavage and accumulation of cleaved 3′-exon ( Figure 4F ) . However , we found that full length HAC1s was the only product produced in the trl1∆ hac1∆ cells in the presence and absence of tunicamycin ( Figure 4F ) . Furthermore , expression of HAC1s in a tpt1∆ background failed to produce additional fragments ( Figure 4G ) , whereas expression of HAC1u is sufficient in the tpt1∆ background to produce free HAC1 3′-exon , consistent with its secondary cleavage . Together , these results ( Figure 4F and G ) indicate that the HAC1s transcript is not sufficient to recapitulate the secondary cleavage , and that HAC1u must be cleaved and ligated to produce the 2′-phosphorylated HAC1s secondary cleavage substrate . To further characterize this secondary cleavage product , we determined the 5′ end of the elongated 3′-exon . We analyzed the 5′-ends of cleaved 3′-exon by primer extension and found that cleaved 3′-exon was barely detectable in wild-type cells upon tunicamycin addition , whereas 3′-exon accumulated in trl1∆ cells due to a lack of kinase-mediated decay ( Figure 5A ) . In tpt1∆ cells treated with tunicamycin , two products accumulated upon tunicamycin addition: a product consistent with canonical length 3′-exon and a small amount of elongated 3′-exon ( Figure 5A ) . The elongated product accumulated in tpt1∆ xrn1∆ cells , again indicating it is degraded by Xrn1 ( Figure 5A ) . To test this prediction ( summarized in Figure 5B ) , we measured the susceptibility of 3′-exon fragments to treatment in vitro with recombinant Xrn1 ( rXrn1/TEX ) . As expected , fragments from xrn1∆ cells were degraded by rXrn1 ( Figure 5C ) , establishing that they have 5′-PO4 termini , whereas 3′-exon fragments from trl1∆ cells were resistant to rXrn1 degradation ( Figure 5C ) , indicating that they have 5′-OH termini . The instructive findings came from examining 3′-exon accumulation in tpt1∆ cells . The 3′-exon fragment of canonical length that accumulates in tpt1∆ cells was resistant to rXrn1 treatment ( Figure 5C ) , while the elongated fragment of 3′-exon is susceptible to rXrn1 treatment ( Figure 5C , lanes 15 and 16 ) . These observations indicate that 3′ product of secondary cleavage of HAC1s is a substrate of Xrn1 in vivo and in vitro , raising the possibility that it may also be a substrate of kinase-mediated decay , depending on the chemistry of the endoribonuclease that generates the secondary cleavage . We propose these products are created via two steps: ( i ) HAC1s is cleaved ~50 nt upstream of the 2′-phosphorylated ligation junction; ( ii ) Xrn1 partially degrades the intermediate fragment to the site of 2′-phosphorylation , which inhibits further degradation ( Figure 5D ) . Under this model , the 3′-exon fragment that accumulates in tpt1∆ cells has both 5′-PO4 and 2′-PO4 moieties at its first position , which inhibits Xrn1-mediated decay in vivo and in vitro ( Figures 1C , 4D , 5A and C ) . The accumulation of HAC1 decay intermediates in tpt1∆ cells over time further supports a model of HAC1s cleavage by Ire1 . In tpt1∆ cells , the accumulation of secondary cleavage product coincides with the increase in production of HAC1s at 20 min ( Figure 6 ) . In tpt1∆ cells , cleaved HAC1 3′-exon is present at low levels at steady state and accumulates over the course of two hours upon tunicamycin treatment ( Figure 6A ) . HAC1s is also generated in tpt1∆ cells , but at significantly reduced levels compared to wild-type ( Figure 6A ) . It is also notable that tpt1∆ xrn1∆ cells and tpt1∆ ski2∆ cells accumulate more spliced HAC1s than tpt1∆ cells ( Figure 6E ) , suggesting that some HAC1s molecules or splicing intermediates in tpt1∆ cells are degraded , possibly because they contain 2′-PO4 moieties . At all time points , tpt1∆ cells contain more free 3′-exon than HAC1s , a ratio opposite to wild-type cells ( Figure 6A ) , indicating that that 3′-exon cleaved from HAC1u accumulates as a result of partial decay of 5′- and 2′-phosphorylated HAC1s . The augmented accumulation of 3′-exon in tpt1∆ cells is also observed in the primer extension analysis ( Figure 5A ) . The in vitro ability of Xrn1 to only partially degrade elongated HAC1 3′-exon from tpt1∆ xrn1∆ cells , and inability to degrade canonical 3′-exon from tpt1∆ cells , indicates that the accumulation of free HAC1 3′-exon is likely caused primarily by blocked 5′→3′ degradation .
Many different regulatory events impinge on HAC1 mRNA to control its localization and processing . It has been assumed that cleavage of HAC1u by Ire1 is the rate-limiting step for UPR activation . Counter to this view , we found that decay of HAC1 splicing intermediates is required for both UPR activation and suppression . We found several examples wherein ‘kinase-mediated decay’ ( KMD ) degrades HAC1 splicing intermediates containing 5′-OH termini by sequential 5′-phosphorylation and 5′-phosphate-dependent 5′→3′ exonucleolytic degradation ( Figure 7 ) . We propose that after 3′-splice site cleavage by Ire1 , the Trl1 5′-kinase domain associates with and phosphorylates the 5′-OH of the 3′-exon product ( Figure 7 ) . Dissociation of the Trl1 kinase active site from the 5′-PO4 product then enables a competition between reassociation of Trl1 ( now its adenylyltransferase/ligase domain ) to catalyze ligation—or Xrn1 to catalyze degradation . In some circumstances , this balance is tipped to favor ligation even in the absence of overt UPR stimulation: a lack of decay in xrn1∆ cells favors ligation , whereas the lack of 5′-kinase activity in trl1∆ ( RtcB ) cells renders Xrn1 decay irrelevant ( Figure 2E and F ) . Xrn1 is abundant in budding yeast ( Ghaemmaghami et al . , 2003 ) , which may efficiently suppress the UPR under normal conditions by degrading spuriously cleaved HAC1 3′-exon intermediates . Regulation of the ligation step of HAC1s splicing makes intuitive sense because it is the last opportunity to act during splicing; once ligated , HAC1s is again protected by a 7-methylguanosine cap and a poly- ( A ) tail . Spurious HAC1 splicing was previously reported in trl1∆ cells expressing the tRNA ligase from Arabidopsis thaliana ( Mori et al . , 2010 ) . It is noteworthy that there are mechanistic differences between the tRNA ligases of A . thaliana and another yeast ( K . lactis ) ( Remus and Shuman , 2014 ) , suggesting that these differences could impact the balance between kinase-mediated decay and ligation in budding yeast expressing plant RNA ligase . We also found that excised HAC1 intron is a substrate of kinase-mediated decay ( Figure 3 ) . Indeed , we believe that phosphorylation of HAC1 intron by Trl1 to promote kinase-mediated decay is the previously proposed ‘second role’ of Trl1 ligase in activating HAC1 translation independent of ligation ( Mori et al . , 2010 ) . In this previous study , excised and circularized HAC1 intron was found to remain associated with HAC1s , inhibiting translation . Kinase-mediated decay of excised intron therefore likely relieves the long-range base-pairing interaction that prevents HAC1s translation ( Chapman and Walter , 1997; Di Santo et al . , 2016; Rüegsegger et al . , 2001 ) , explaining how HAC1s can accumulate without concomitant UPR activation . This second layer of control over HAC1s translation by KMD adds another failsafe mechanism to prevent its translation and unintentional UPR activation . Previous examples of kinase-mediated decay of bacterial mRNAs ( Durand et al . , 2012 ) , eukaryal tRNA introns ( Wu and Hopper , 2014 ) , ribosomal RNA processing intermediates ( Gasse et al . , 2015 ) , and no-go mRNA decay cleavage products ( Navickas et al . , 2018 ) suggest that this mode of decay may be widespread . Coupling of RNA 5′-kinase and 5′→3′ exonucleolytic decay activities in the context of kinase-mediated decay may regulate the UPR in other organisms . Splicing of Xbp-1 mRNA in metazoans ( the functional homolog of HAC1 ) is catalyzed by Ire1-mediated removal of an intron and ligation by RtcB RNA ligase ( Kosmaczewski et al . , 2014; Lu et al . , 2014 ) . Fundamental differences in the chemistry of RNA ligation between fungal and metazoan ligases suggest that Xbp-1 may be subject to a biochemically distinct mode of regulation . Because RtcB depends on 5′-OH and 3′-PO4 termini for catalysis ( Chakravarty et al . , 2012 ) , activities that remodel 5′-OH RNA termini could divert Ire1-generated 5′-OH splicing intermediates from productive ligation . To that point , cyclic nucleotide phosphodiesterase ( CNP ) and RtcA ( a 2′ , 3′-RNA cyclase ) were shown to ‘tune’ the UPR in metazoans by competing with RtcB/HSCP117 for ligation substrates ( Unlu et al . , 2018 ) . Specifically , CNP hydrolyzes the 2′ , 3′-cyclic phosphate compatible RtcB to a 2′-PO4 incompatible with RtcB , decreasing ligation of Xbp1 . Conversely , RtcA , which converts 2′-PO4 RNA to 2′ , 3′-cyclic phosphate , makes the terminus compatible with RtcB , thus enhancing Xbp1 splicing . Additionally , an RNA 5′-kinase ( e . g . , Clp1 ) may phosphorylate the 3′-exon product of Xbp-1 cleavage , simultaneously inhibiting ligation by RtcB and promoting its degradation by a 5′-phosphate-dependent exoribonuclease to limit UPR activation . In this vein , it is noteworthy that kinase-inactivating mutations in Clp1 cause neurodevelopmental defects and neuronal dysfunction in humans , mice , and zebrafish ( Hanada et al . , 2013; Karaca et al . , 2014; Schaffer et al . , 2014 ) , possibly due to chronic UPR activation in neural tissues ( Clayton and Popko , 2016 ) . Xrn1 degrades mRNA fragments generated during metazoan Regulated Ire1-dependent Degradation ( RIDD ) ( Hollien and Weissman , 2006 ) ; however , it is not known how these 5′-OH cleavage products of Ire1 are phosphorylated for Xrn1-mediated decay . The RNA 5′-kinase Clp1 ( Weitzer and Martinez , 2007 ) and polynucleotide kinase Nol9 ( Heindl and Martinez , 2010 ) are candidates for this activity , though neither are known to phosphorylate mRNA decay intermediates . While 5′→3′ decay plays a major role in UPR regulation , we found little evidence for UPR regulation by 3′→5′ decay activity . As shown previously ( Schwer et al . , 2004 ) , the exposed 3′-end of cleaved HAC1 5′-exon is a substrate for 3′→5′ decay ( Figure 4 ) . But while excised HAC1 intron is stabilized in ski2∆ cells lacking cytosolic 3′→5′ decay ( Figure 3H ) , their growth is unaffected by tunicamycin ( Figure 1B ) , indicating that 3′→5′ decay of the excised intron does not contribute substantially to intron-mediated HAC1s repression . We also found evidence that incompletely processed HAC1s mRNA is cleaved , which is ligated but contains an internal 2′-PO4 moiety . Cleavage of HAC1s leads to 5′ and 3′ fragments that are degraded; however , the 3′-fragment is only partially degraded by kinase-mediated decay , producing a 5′- and 2′-phosphorylated molecule that cannot be degraded by Xrn1 . Consistent with these findings , a recent study also showed that an RNA with an internal 2′-phosphate group is protected from 3′→5′ decay by E . coli PNPase in vitro ( Munir et al . , 2018 ) ; those and our results together indicate that site-specific installation of a 2′-PO4 is an effective strategy to protect an RNA from complete exonucleolytic degradation in vivo or in vitro . Decay intermediates produced from incompletely processed , 2′-phosphorylated HAC1s have not been previously observed and suggest a plausible regulatory role for Tpt1 in regulating HAC1s fate . We have yet to determine the impact of 2′-phosphorylation on HAC1s translation , but given that tpt1∆ cells grow on tunicamycin ( Figure 1B and Cherry et al . , 2018 ) and activate KAR2 gene expression ( Figure 3B ) , it is likely that some Hac1 protein is produced from 2′-PO4 HAC1s mRNA . It also remains to be determined how and why incompletely processed HAC1s is cleaved . Insofar as the HAC1s cleavage substrate is initially produced by tunicamycin-dependent Ire1 cleavage and ligation , we conjecture that Ire1 incises ligated , 2′-phosphorylated HAC1s upstream of the original ligation junction , yielding smaller 5′-exon and larger 3′-exon fragments . Formally , we cannot currently rule out the possibility that another endonuclease catalyzes secondary HAC1s cleavage; however , Ire1 is the only endoribonuclease known to site-specifically incise HAC1 mRNA . We note that ire1∆ mutants are unable to initiate processing of HAC1u and therefore do not make HAC1s in the first place ( Sidrauski and Walter , 1997 ) , precluding direct analysis of HAC1s cleavage in ire1∆ mutants . We showed that the 2′-PO4 is required for cleavage , as expression of ‘pre-spliced’ HAC1s does not lead to cleavage ( Figure 4F and G ) . It is possible that the presence of a 2′-PO4 in the context of a composite Ire1 splice site ( i . e . , formed from two halves of the original 5′- and 3′-splice sites ( Hooks and Griffiths-Jones , 2011; Sidrauski and Walter , 1997 ) ) on HAC1s is recognized by Ire1 , but because a 2′-OH is the nucleophile for transesterification by metal-independent Ire1 ( Gonzalez et al . , 1999 ) , the 2′-PO4 inhibits the chemical step of incision . This model also provides a plausible mechanism to explain how Ire1 incises a neighboring , non-canonical site . We propose that a failure of Ire1 to release 2′-phosphorylated HAC1s would enable the active site of a nearby Ire1 molecule—in the context of its activated , oligomeric form ( Korennykh et al . , 2009 ) —to catalyze site-specific incision at the second , upstream site . Our results also raise the question of why Ire1 would cut incompletely processed HAC1s . Here , cleavage of incompletely processed ( ligated , but 2′-phosphorylated ) HAC1s could be a means to inactivate HAC1s after prolonged stimulation to attenuate the UPR ( Chawla et al . , 2011; Rubio et al . , 2011 ) .
Yeast strains and sources used in this study are listed in Table 3 . Plasmids were created as indicated in Table 3 . Single colonies were inoculated in drop-out media supplemented with relevant amino acids and incubated at 30°C overnight with rotation . Cultures were diluted to an OD600 of 0 . 2 in yeast-extract , peptone , dextrose ( YPD ) media , and UPR induction was carried out when yeast were growing at mid-log phase with a 2 hr treatment ( unless otherwise indicated ) with tunicamycin ( final concentration of 2 . 5 µg/mL , Sigma-Aldrich ) or DMSO mock treatment . Cells were harvested by centrifugation , and total RNA was isolated by hot acid phenol extraction . For RT-PCR and RT-qPCR experiments , total RNA was treated with TURBO DNase ( 2 U , Ambion ) to degrade contaminating genomic DNA . DNase-treated RNA was reverse transcribed with 200 U of SuperScript III reverse transcriptase ( Invitrogen ) using a gene-specific reverse primer ( Table 2 ) . Products analyzed on a 1 . 5% agarose TBE gel , stained with 1x GelRed ( Sigma ) and imaged with a Bio Rad GelDoc . Densitometry was performed with Bio-Rad Image Analysis software and splicing quantifications were computed and visualized in R using ggplot2 and cowplot R Packages . Quantitative PCR ( qPCR ) for KAR2 was also performed on cDNA as generated above , and assayed for KAR2 and PGK1 using Sso Advanced Universal SYBR Green Supermix ( Bio Rad ) and cycled on a Bio Rad C1000 384-well thermal cycler and plate reader . Output Ct values were analyzed in Microsoft Excel and plotted in R using ggplot2 and cowplot R Packages . Primers specific for HAC1 mRNA and U6 snRNA were PAGE-purified and ethanol precipitated . Oligonucleotide primers were 5′-end-labeled with PNK ( Enzymatics ) and γ-32P-ATP ( Perkin Elmer ) and purified with Sephadex G-25 spin columns ( GE Healthcare resin , Thermo empty columns ) . Radiolabeled primers and total RNA ( 15 µg ) were heated to 65°C for 5 min and cooled to 42°C . SuperScript III reverse transcriptase ( 200 U , Invitrogen ) was added and reverse transcription reactions were run with a final concentration of 500 µM dNTPs . Primers were extended for 30 min at 42°C , 15 min at 45°C , and 15 min at 50°C . SuperScript III RT was inactivated by heating for 20 min at 75°C . RNA was destroyed with in 10 mM NaOH at 90°C for 3 min and neutralized with HCl . Formamide loading dye was added and products were run on a 8% acrylamide TBE 7M Urea gel . Gels were dried ( Bio Rad ) and exposed on a phosphor-imager screen and imaged on a Typhoon 9400 ( GE Healthcare ) . Total RNA ( 3 µg ) was electrophoresed on 6% acrylamide TBE 7M urea gels and transferred to nylon membrane ( Hybond N+ , GE ) by electroblotting . Membranes were UV-crosslinked ( 254 nm , 120 mJ dose ) , blocked in ULTRAhyb-Oligo Buffer ( Ambion ) , and incubated with 5′-32P-labeled oligonucleotide probes ( Table 2 ) in ULTRAhyb-Oligo at 42°C for 18 hr . Membranes were washed with 2X SSC/0 . 5% SDS washing buffer two time for 30 min each , exposed on a phosphor-imager storage screen , and imaged on a Typhoon 9400 ( GE Healthcare ) . Membranes were stripped of original probe with three washes in stripping buffer ( 2% SDS ) at 80°C for 30 min per wash . Membranes were re-blocked and probed a second time for the loading control , SCR1 ( Table 2 ) . | Like any economical factory , cells tune the size of their protein assembly line to suit demand . Proteins consist of strings of amino acids , built from template molecules called mRNAs , that must be folded into specific 3D structures for them to work correctly . If these protein strings are produced faster than they can be folded , the cell triggers the unfolded protein response . This response slows protein production , gets rid of any misshapen proteins , and increases the size of the protein assembly line . It is not clear exactly how the unfolded protein response is tuned , though an mRNA molecule called HAC1 is known to signal the response . First , enzymes remove a short section of HAC1 and join the remaining parts back together in a process called splicing . Spliced HAC1 is then used as a template to make a protein that activates the unfolded protein response . To understand more about this processing of HAC1 , Cherry et al . studied yeast cells that had mutated , non-working versions of some of the enzymes that repair and degrade RNA . This revealed that the splicing of HAC1 competes with another process that breaks down mRNA . Under normal conditions , this means that HAC1 is degraded before it can trigger the unfolded protein response . In addition , for the cell to trigger the unfolded protein response , it needs to break down the part of HAC1 that is removed during splicing . Otherwise , the removed section interferes with the spliced HAC1 mRNA , preventing it from being a signal to activate the unfolded protein response . Cherry et al . also found that a unique , chemically modified fragment of HAC1 mRNA was protected from degradation . They do not know how the unique chemical modification regulates the unfolded protein response , but stabilizing modifications are generally useful in RNA biology . Understanding how the unfolded protein response is tuned could help researchers to find new ways to treat conditions where it does not work correctly , such as neurodegeneration , diabetes and cancer . Additionally , researchers are already trying to develop treatments for a number of diseases that work by inserting new RNA molecules into cells . Understanding how the chemical modification discovered by Cherry et al . protects RNAs from degradation could therefore improve the effectiveness of such treatments . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2019 | Multiple decay events target HAC1 mRNA during splicing to regulate the unfolded protein response |
Color , an important visual cue for survival , is encoded by comparing signals from photoreceptors with different spectral sensitivities . The mouse retina expresses a short wavelength-sensitive and a middle/long wavelength-sensitive opsin ( S- and M-opsin ) , forming opposing , overlapping gradients along the dorsal-ventral axis . Here , we analyzed the distribution of all cone types across the entire retina for two commonly used mouse strains . We found , unexpectedly , that ‘true S-cones’ ( S-opsin only ) are highly concentrated ( up to 30% of cones ) in ventral retina . Moreover , S-cone bipolar cells ( SCBCs ) are also skewed towards ventral retina , with wiring patterns matching the distribution of true S-cones . In addition , true S-cones in the ventral retina form clusters , which may augment synaptic input to SCBCs . Such a unique true S-cone and SCBC connecting pattern forms a basis for mouse color vision , likely reflecting evolutionary adaptation to enhance color coding for the upper visual field suitable for mice’s habitat and behavior .
Topographic representation of the visual world in the brain originates from the light-sensitive photoreceptors in the retina ( Rhim et al . , 2017 ) . Although the neuronal architecture of the retina is similar among different vertebrates , the numbers and distributions of photoreceptors vary considerably ( Hunt and Peichl , 2014 ) . Such patterns have been evolutionarily selected , adapting to the animal’s unique behavior ( diurnal or nocturnal ) and lifestyle ( prey or predator ) for better use of the visual information in the natural environment ( Dominy and Lucas , 2001; Gerl and Morris , 2008; Peichl , 2005 ) . Color , an important visual cue for survival , is encoded by comparing signals carried by photoreceptors with different spectral preferences ( Baden and Osorio , 2019 ) . While amongst mammals , trichromatic color vision is privileged for some primates ( Jacobs et al . , 1996; Nathans et al . , 1986; Yokoyama and Yokoyama , 1989 ) , most terrestrial mammals are dichromatic ( Marshak and Mills , 2014; Puller and Haverkamp , 2011; Jacobs , 1993 ) . The mouse retina expresses two types of cone opsins , S- and M-opsin , with peak sensitivities at 360 nm and 508 nm , respectively ( Jacobs et al . , 1991; Nikonov et al . , 2006 ) . The expression patterns of these two opsins form opposing and overlapping gradients along the dorsal-ventral axis , resulting in a majority of cones expressing both opsins ( herein either ‘mixed cones’ or M+S+ ) ( Applebury et al . , 2000; Ng et al . , 2001; Wang et al . , 2011 ) . Thus , S-opsin enrichment in the ventral retina better detects short-wavelength light from the sky , and M-opsin in the dorsal retina perceives the ground ( e . g . , a grassy field ) ( Baden et al . , 2013; Gouras and Ekesten , 2004; Osorio and Vorobyev , 2005; Szél et al . , 1992 ) , while co-expression of both opsins ( herein either mixed cones or M+S+ ) ( Röhlich et al . , 1994 ) broadens the spectral range of individual cones and improves perception under varying conditions of ambient light ( Chang et al . , 2013 ) . This unusual opsin expression pattern poses a challenge for color-coding , particularly so for mixed cones . However , it has been discovered that a small population of cones only expresses S-opsin ( ‘true S-cones’ , or S+M- ) . These true S-cones are thought to be evenly distributed across the retina ( Franke et al . , 2019; Haverkamp et al . , 2005; Szatko et al . , 2019; Wang et al . , 2011 ) and to be critical for encoding color , especially in the dorsal retina where they are quasi-evenly distributed in a sea of cones expressing only M-opsin ( M+S- ) , a pattern akin to mammalian retinas in general ( Haverkamp et al . , 2005; Wang et al . , 2011 ) . Nonetheless , subsequent physiological studies revealed that color-opponent retinal ganglion cells ( RGCs ) are more abundant in the dorsal-ventral transition zone ( Chang et al . , 2013 ) and the ventral retina ( Joesch and Meister , 2016 ) . Recent large scale two-photon imaging results further demonstrated that color opponent cells were mostly located in the ventral retina ( Szatko et al . , 2019 ) . Intriguingly , a behavior-based mouse study demonstrated that their ability to distinguish color is also restricted to the ventral retina ( Denman et al . , 2018 ) . These results prompt us to study , at the single-cell level and across the whole retina , the spatial distributions of cone types with different opsin expression configurations and , more importantly , with regard to S-cone bipolar cell connections in order to better understand the anatomical base for the unique color-coding scheme of the mouse retina .
In mouse retina , the gradients of S- and M-opsin expression along the dorsal-ventral axis have been well documented ( Figure 1A–B; Applebury et al . , 2000; Calderone and Jacobs , 1995; Chang et al . , 2013; Haverkamp et al . , 2005; Jelcick et al . , 2011; Lyubarsky et al . , 1999; Ortín-Martínez et al . , 2014; Szél et al . , 1992; Wang et al . , 2011 ) , but the distribution of individual cone types with different combinations of opsin expression across the whole retina has not been characterized ( but see Baden et al . , 2013; Eldred et al . , 2020 ) , which we discuss below ) . We developed a highly reliable algorithm to automatically quantify the different opsins ( S and M ) and cone types ( M+S- , true S , and mixed cones , Figure 2 , Figure 2—figure supplement 1 ) based on high-resolution images of entire flat-mount retinas immunolabeled with S- and M-opsin antibodies ( Figure 2—figure supplement 1 ) . As demonstrated in examples of opsin labeling from dorsal , medial , and ventral retinal areas of the pigmented mouse ( Figure 1B , left ) , while M opsin-expressing cones ( M+: M+S+ + M+S- ) were relatively evenly distributed across three regions , S opsin-expressing cones ( S+: M+S+ + S+M- ) showed considerable anisotropy , with a high density in the ventral retina and a precipitous drop in the dorsal retina , confirming previous observations ( Haverkamp et al . , 2005; Jelcick et al . , 2011; Ortín-Martínez et al . , 2014 ) . Surprisingly , instead of finding an even distribution of true S-cones as previously presumed ( Baden et al . , 2013; Haverkamp et al . , 2005; Wang et al . , 2011 ) , we found the ventral region had much more numerous true S-cones ( ~30% of the local cone population; Figure 1C left , Supplementary file 1A ) than did the dorsal region ( ~1% ) . This result is evident from density maps of cone types from three examples of pigmented mice , showing highly concentrated true S-cones in the ventral retina ( Figure 2A , left column , bottom row ) . In addition , M+S--cones were concentrated in the dorsal retina , whereas mixed cones dominated the medial and ventral retina ( Figure 1C left and Figure 2A , left column , 4th and 5th rows ) . Such a highly skewed distribution of true S-cones conflicts with the general notion that true S-cones only account for ~5% of cones and are evenly distributed across the mouse retina ( Baden et al . , 2013; Franke et al . , 2019; Haverkamp et al . , 2005; Szatko et al . , 2019; Wang et al . , 2011 ) ; however , it is not unprecedented considering the diverse S-cone patterns seen in mammals ( Ahnelt et al . , 2000; Ahnelt and Kolb , 2000; Calderone et al . , 2003; Hendrickson et al . , 2000; Hendrickson and Hicks , 2002; Kryger et al . , 1998; Müller and Peichl , 1989; Nadal-Nicolás et al . , 2018; Ortín-Martínez et al . , 2014; Ortín-Martínez et al . , 2010; Peichl , 2005; Schiviz et al . , 2008; Szél et al . , 2000 ) . Therefore , we also examined an albino mouse line to determine whether this observation persists across different mouse strains . Overall , albino retinas had slightly smaller cone populations ( Figure 2B , Supplementary file 1B; Ortín-Martínez et al . , 2014 ) . Interestingly , while M-opsin expressing cones had similar distributions in both strains , S-opsin expression extended well into the dorsal retina of the albino mouse , exhibiting a greatly reduced gradient of S-opsin expression toward the dorsal retina compared to that seen in pigmented mice ( Figure 1B–C , Figure 2A second row; Applebury et al . , 2000; Ortín-Martínez et al . , 2014 ) . Consequently , most cones in the dorsal retina were mixed cones , and M+S- cones were very sparse ( 7% , compared to 97% in pigmented mouse , Figure 1C right , Supplementary file 1A , Figure 2A right ) . However , despite these differences , the percentage and distribution of true S-cones were remarkably conserved between strains . In both strains , true S-cones were extremely sparse in the dorsal retina ( 1% ) but highly concentrated in the ventral retina ( 33% vs 29% , Figure 1C and Supplementary file 1A ) . Notably , the density maps of true S-cones are nearly identical in both strains ( Figure 2A , bottom row ) . Evaluating the distribution of three main cone populations ( mixed , M+S- , and true S-cone ) in four retinal quadrants centered upon the optic nerve head reveals different profiles between pigmented and albino strain for mixed and M+S- cones ( Figure 2C ) . For example , in the dorsotemporal ( DT ) quadrant , we observed an increase of M+S- cones from the center to the periphery ( green line ) in pigmented mice , compared to a majority of mixed cones ( gray line ) in albino mice . However , true S-cone profiles ( magenta lines ) were similar between the two strains in all quadrants , except for a slightly increased density along the edge of the ventronasal ( VN ) quadrant in pigmented mice . A recent study successfully modeled cone opsin expression and type determination according to graded thyroid hormone signaling in a pigmented mouse strain ( C57BL/6 ) ( Eldred et al . , 2020 ) . It would be interesting to see whether a different pattern of thyroid hormone and/or receptor distribution could recapitulate a similar true S-cone distribution with a very different form of S-opsin expression . One major concern regarding cone classification based on opsin immunolabeling is that some S+M- cones may instead be mixed cones with low M-opsin expression ( Applebury et al . , 2000; Baden et al . , 2013; Nikonov et al . , 2006; Röhlich et al . , 1994 ) . Even though a similar cone-type distributions have been observed in mouse retina , it has been assumed that only a fraction of the S+M- cones are ‘true’ S-cones ( Baden et al . , 2013; Eldred et al . , 2020 ) . Out of caution , S+M- cones were only referred to as ‘anatomical’ S-cones due to a lack of confirmation regarding their bipolar connections ( Baden et al . , 2013 ) . Thus , both true S-cones and S-cone bipolar cells have been generally acknowledged to be evenly distributed across the retina ( Haverkamp et al . , 2005; Wang et al . , 2011; Baden et al . , 2013; Szatko et al . , 2019; Franke et al . , 2019; Eldred et al . , 2020 ) . In order to confirm the distribution of true S-cones , it is critical to uncover the distribution and dendritic contacts of S-cone bipolar cells ( type 9 , or SCBCs ) . Previously , SCBCs have only been identified among other bipolar , amacrine and ganglion cells in a Thy1-Clomeleon mouse line , rendering the quantification of their distribution across the entire retina impractical ( Haverkamp et al . , 2005 ) . We generated a Copine9-Venus mouse line , in which SCBCs are specifically marked ( Figure 2—figure supplement 1C ) , owing to the fact that Cpne9 is an SCBC-enriched gene ( Shekhar et al . , 2016 ) . In retinal sections , these Venus+ bipolar cells have axon terminals narrowly ramified in sub-lamina 5 of IPL ( Figure 3A ) , closely resembling type 9 BCs as identified in EM reconstructions ( Behrens et al . , 2016; Stabio et al . , 2018a ) . In flat-mount view , these bipolar cells are often seen to extend long dendrites to reach true S-cones , bypassing other cone types ( Figure 3B–C ) . The majority of dendritic endings formed enlarged terminals beneath true S-cones pedicles ( Figure 3C–c’ ) , but occasional slender ‘blind’ endings were present ( arrow in Figure 3C–c” ) , which have been documented for S-cone bipolar cells in many species ( Haverkamp et al . , 2005; Herr et al . , 2003; Kouyama and Marshak , 1992 ) . Unexpectedly , we found that the distribution of SCBCs was also skewed toward VN retina , albeit with a shallower gradient ( Figure 3D–E ) . To examine the connections between true S-cones and SCBCs , we immunolabeled S- and M-opsins in Copine9-Venus mouse retinas . Because M-opsin antibody signals did not label cone structures other than their outer segments , we first identified true S-cones at the outer segment level and then traced S-opsin labeling to their pedicles in the outer plexiform layer ( OPL ) , where they connect with SCBCs ( Figure 3C , for more details see material and methods ) . Although convergent as well divergent connections were found between true S-cones and SCBCs in both dorsal and ventral retina ( see the source data ) , we noted different connectivity patterns . While in the dorsal retina , a single true S-cone connected to approximately 4 SCBCs ( 3 . 8 ± 0 . 2 , see material and methods ) , in the ventral retina , a single SCBC contacted approximately 5 true S-cones ( 4 . 6 ± 0 . 4; Figure 3C , Supplementary file 2 ) . These results agree well with the true S-cone to SCBC ratios calculated from cell densities in the DT and VN retina . Specifically , in the dorsal retina , the true S-cone to SCBC ratio was approximately 1:3 . 6 , compared to 5 . 3:1 in the ventral retina ( Supplementary file 3 ) . Accordingly , both data sets support the presence of a prevalent divergence of true S-cone to SCBC connections in the dorsal retina , in comparison to a prominent convergence of contacts from true S-cones to SCBCs in VN retina . Critically , the specificity of wiring from true S-cones to SCBCs also confirms the identity of true S-cones as revealed by opsin labeling and further supports the finding that true S-cones are highly concentrated in VN mouse retina . As demonstrated above , in the mouse retina , despite a large population of mixed cones , SCBCs precisely connect with true S-cones , preserving this fundamental mammalian color circuitry motif ( Behrens et al . , 2016; Breuninger et al . , 2011; Haverkamp et al . , 2005; Mills et al . , 2014 ) . However , the increased density of SCBCs in the ventral retina does not match that of true S-cones ( compare Figure 3D and Figure 2a , last row ) . Thus , individual SCBCs in the ventral retina may be required to develop more dendrites to maximize the number of contacts made with different S-cone terminals ( Supplementary file 2 , graphs in Figure 3C ) . Intriguingly , we discovered in both strains that true S-cones in the ventral retina appeared to cluster together rather than forming an even distribution , as revealed by K-nearest neighbor analysis ( Figure 4A–B , Supplementary file 2 ) . Ideally , such true S-cone clustering may increase the availability of targets for individual SCBCs in a reduced space . To quantify the spatial patterning of true S-cone populations ( or their lack thereof ) , we compared the observed true S-cone distributions within 1 mm diameter VN and DT retinal samples to artificially generated alternative populations ( Figure 4C ) . To this end , we considered two extreme patterning rules: First , one in which the space between true S-cone locations was maximized within the set of actual locations for all cones , creating a relatively uniform ( evenly ‘distributed’ ) mosaic of true S-cones . At the other extreme , cone identities were permuted randomly ( ‘shuffled’ ) among observed cone locations ( Figure 4C ) . Repetition of these algorithms generated distributions of patterning metrics for true S-cones ( see below ) that remain constrained by the observed cone locations and proportions of cone types for each 1 mm sample . To quantitatively compare the patterning of real true S-cone populations to their artificial counterparts , we first computed two measures of regularity for true S-cones: nearest neighbor and Voronoi diagram regularity indices ( NNRI and VDRI , respectively; Reese and Keeley , 2015; Figure 4C–D ) ; larger values of these metrics indicate smaller variability in the spacing between cones and thus more regular patterns . Interestingly , far from being regularly distributed , true S-cone placement was quite irregular and nearly indistinguishable from shuffled populations ( including a slight trend toward regularity measures lower than random , which may indicate a tendency toward clustering , Figure 4D; see Reese , 2008 ) . To further probe the possibility of true S-cone clustering , we computed the ratios of true S-cone neighbors for each cone ( denoted here as the S-cone neighbor ratio [SCNR]; see Methods for the calculation of the SCNR search radius for each retinal sample ) . Intriguingly , SCNRs were significantly larger for true S-cones than for other cone types , which were equal to expected ratios due to random chance—especially so in ventral retinas , further indicating a clustering of true S-cones in those areas ( Figure 4E ) . Notably , a more extreme form of clustering of S-cones has been observed in the ‘wild’ mouse ( Warwick et al . , 2018 ) and with much lower densities in some felids ( Ahnelt et al . , 2000 ) . Here , such clustering may reflect the mode of true S-cone development in the ventral retina , for example , by ‘clonal expansion’ to achieve unusually high densities ( Bruhn and Cepko , 1996; Reese et al . , 1999 ) . It is tempting to speculate that it may also facilitate the wiring of true S-cones with sparsely distributed SCBCs , which were not observed to cluster in the ventral retina ( Figure 3E ) . Indeed , we observed examples of groups of true S-cones forming clusters whose pedicles in the OPL were tightly congregated in a patch and contacted by a nearby SCBC ( Figure 4F ) . Despite being nocturnal and having a rod-dominated retina ( Carter-Dawson and LaVail , 1979; Jeon et al . , 1998 ) , mice can detect color ( Denman et al . , 2018; Jacobs et al . , 2004 ) . Although it remains uncertain whether the source of long-wavelength sensitive signals for color opponency arises in rods or M-cones ( Baden and Osorio , 2019; Ekesten et al . , 2000; Ekesten and Gouras , 2005; Joesch and Meister , 2016; Reitner et al . , 1991 ) , it is clear that true S-cones provide short-wavelength signals for color discrimination . Given the previously-held notion that true S-cones are evenly distributed across the retina ( Baden et al . , 2013; Franke et al . , 2019; Haverkamp et al . , 2005; Szatko et al . , 2019; Wang et al . , 2011 ) , whereas M+S- cones are concentrated in the dorsal retina of pigmented mouse , it is intuitive to speculate that color coding is prevalent in the dorsal retina . However , previous physiological and behavioral studies indicate that , although luminance detection can occur across the mouse retina , color discrimination is restricted to the ventral retina ( Breuninger et al . , 2011; Denman et al . , 2018; Szatko et al . , 2019 ) . Thus , our discovery of high enrichment of true S-cones in the ventral retina provides a previously missed anatomical feature for mouse color vision that could help to re-interpret these results . From projections mapping true S-cone densities into visual space ( Figure 4—figure supplement 1; Sterratt et al . , 2013 ) , it is conceivable that high ventral true S-cone density will provide a much higher sensitivity of short-wavelength signals , thus facilitating color detection for the upper visual field . Although the true S-cone signals carried by SCBCs in the dorsal retina might not be significant for color detection , they could certainly participate in other functions , such as non-image forming vision , that are known to involve short-wavelength signals ( Altimus et al . , 2008; Doyle et al . , 2008; Patterson et al . , 2020 ) . Interestingly , the overall true S-cone percentage in the mouse retina remains approximately 10% ( Figure 2B ) , and the average true S-cone to SCBC ratio across the whole retina is about 1 . 7:1 ( Supplementary file 1B-C ) , similar to what has been reported in other mammals ( Ahnelt et al . , 2006; Ahnelt and Kolb , 2000; Bumsted et al . , 1997; Bumsted and Hendrickson , 1999; Curcio et al . , 1991; Hendrickson and Hicks , 2002; Hunt and Peichl , 2014; Kryger et al . , 1998; Lukáts et al . , 2005; Müller and Peichl , 1989; Ortín-Martínez et al . , 2010; Peichl et al . , 2000; Schiviz et al . , 2008; Shinozaki et al . , 2010; Szél et al . , 1988 ) . Such a spatial rearrangement of true S-cones and SCBCs likely reflects evolutionary adaptation to enhance short-wavelength signaling and color coding for the upper visual field as best suited for the habitat and behavior of mice ( Baden et al . , 2020 ) . For example , it may facilitate aerial predator detection during daytime ( Yilmaz and Meister , 2013 ) . Similarly , skewed S-cone arrangement has been reported for other terrestrial prey mammals ( Famiglietti and Sharpe , 1995; Juliusson et al . , 1994; Röhlich et al . , 1994 ) , while zebrafish possess a UV-enriched ventral retina that enhances their predation ( Zimmermann et al . , 2018 ) . In addition , we observed that the clustering of true S-cones in the ventral retina may allow several neighboring cones of the same type to converge onto the same SCBC ( Figure 4F ) , which could potentially enhance signal-to-noise ratios for more accurate detection , as described recently in human fovea ( Schmidt et al . , 2019 ) . It is also remarkable that despite the very different S-opsin expression patterns in both mouse strains , the true S-cone population and distribution are strikingly similar between pigmented and albino mice , suggesting a common functional significance .
Three months old male pigmented ( C57BL/6J , n = 5 ) , albino ( CD1 , n = 5 ) mice were obtained from the National Eye Institute breeding colony . The Venus-Cpne9 mouse line ( n = 5; based on previous single cell sequencing data [Shekhar et al . , 2016] ) carries a reporter ( Venus ) allele under the control of the mouse Cpne9 locus . The reporter allele was created directly in B6 . SJL ( F1 ) zygotes using CRISPR-mediated homologous recombination ( HR ) ( Yang et al . , 2013 ) . Briefly , a HR targeting template was assembled with PCR fragments of 5’ and 3’ homology arms of 910 bp and 969 bp respectively , flanking exon one , and a Venus expression cassette carrying the bovine growth hormone polyadenylation ( bGH-PolyA ) signal sequence as the terminator . Homology arms were designed such that integration of the reporter cassette would be at the position right after the first codon of the Cpne9 gene in exon one . A pair of guide RNAs ( gRNA ) , with outward orientation ( 38 bp apart ) , were synthesized by in vitro transcription as described ( Yang et al . , 2013 ) and tested for their efficiency and potential toxicity in a zygote differentiation assay where mouse fertilized eggs were electroporated with SpCas9 protein and gRNA ribonuclear particles . Eggs were cultured in vitro for 4 days in KSOM ( Origio Inc , CT ) until differentiated to blastocysts . Viability and indel formation were counted respectively . gRNA sequences are ( 1 ) Copine9_gRNA_L ( 73/25 ) , 5’GAGACATGACTGGTCCAA3’; ( 2 ) Copine9_gRNA_R ( 62/4 . 40 ) , 5’GCCTCGGAGCGTAGCGTCC3’ . A mixture of the targeting plasmid ( super coiled , 25 ng/µl ) with two tested gRNAs ( 25 ng/µl each ) and the SpCas9 protein ( Life Science technology , 30 ng/µl ) were microinjected into mouse fertilized eggs and transferred to pseudopregnant female recipients as described elsewhere ( Yang et al . , 2013 ) . With a total of 15 F0 live births from 6 pseudopregnant females , 11 were found to carry the knockin allele by homologous recombination , a HR rate of 73% . F0 founders in B6 . SJL F1 ( 50% C57BL6 genome ) were crossed consecutively for 3 generations with C57BL6/J mice to reach near congenic state to C57BL6/J . Mice were housed a 12:12 hr light/dark cycle . All experiments and animal care are conducted in accordance with protocols approved by the Animal Care and Use Committee of the National Institutes of Health and following the Association for Research in Vision and Ophthalmology guidelines for the use of animals in research . All animals were sacrificed with an overdose of CO2 and perfused transcardially with saline followed by 4% paraformaldehyde . To preserve retinal orientation , eight retinas per mouse strain/line were dissected as flat whole-mounts by making four radial cuts ( the deepest one in the dorsal pole previously marked with a burn signal as described [Nadal-Nicolás et al . , 2018; Stabio et al . , 2018b] ) . The two remaining retinas were cut in dorso-ventral orientation ( 14 μm ) after cryoprotection in increasing gradients of sucrose ( Sigma-Aldrich SL ) and embedding in optimal cutting temperature ( OCT; Sakura Finetek ) . Immunodetection of flat-mounted retinas or retinal sections was carried out as previously described ( Nadal-Nicolás et al . , 2018 ) . Importantly , the retinal pigmented epithelium was removed before the immunodetection . First , whole-retinas were permeated ( 4 × 10’ ) in PBS 0 . 5% Triton X-100 ( Tx ) and incubated by shaking overnight at room temperature with S-opsin ( 1:1200 ) and M-opsin ( 1:1000 ) or cone arrestin ( 1:300 ) primary antibodies diluted in blocking buffer ( 2% normal donkey serum ) . Cpne9-Venus retinas were additionally incubated with an anti-GFP antibody ( 1:100 ) to enhance the original Venus signal . Retinas were washed in PBS 0 . 5% Tx before incubating the appropriate secondary antibodies overnight ( 1:500 ) . Finally , retinas were thoroughly washed prior to mounting with photoreceptor side up on slides and covered with anti-fading solution . Retinal sections were counterstained with DAPI . Retinal whole-mounts were imaged with a 20x objective using a LSM 780 Zeiss confocal microscope equipped with computer-driven motorized stage controlled by Zen Lite software ( Black edition , Zeiss ) . M- and S-opsins were imaged together to allow the identification and quantification of different cone types . Magnifications from flat mounts and retinal cross-sections ( Figure 1 ) were taken from dorsal , medial and ventral areas using a 63x objective for opsin co-expression analysis . Images from retinal cross-sections were acquired ~1 . 5 mm dorsally or ventrally from the optic disc . In four retinas per strain , we acquired images from three 135 × 135 μm samples ( 63x ) per each area of interest ( dorsal , medial and ventral ) . These areas were selected according to the S-opsin gradient in wholemount retinas ( see scheme in Figure 1C ) . Cone outer segments were manually classified as M+S- , true S- ( S+M- ) or mixed ( M+S+ ) cones depending on their opsin expression . Data representation was performed using GraphPad Prism 8 . 3 software . To characterize the distribution of the different cone photoreceptor types in the mouse retina , we developed and validated an automatic routine ( ImageJ , NIH ) to identify , quantify the total number of outer segments and finally extract the location of each individual cone ( Figure 2—figure supplement 1A ) . Briefly , maximum-projection images were background-subtracted and thresholded ( background-noise mean value , 9 . 6 ± 1 . 2% and 15 . 2 ± 3 . 2% for S- and M-opsin respectively , the threshold was applied at 15 . 7% ) to create a binary mask that was then processed using watershed and despeckle filters to isolate individual cones and reduce noise . The ‘3D Objects Counter’ plugin was applied to such images to count cones within fixed parameters ( shape and size ) and extract their xy coordinates for further analysis . This automation was validated by statistical comparison with manual counting performed by an experienced investigator ( Pearson correlation coefficient R2 = 96–99% for M- or S-opsin respectively , Figure 2—figure supplement 1B ) . To count cone subtypes , images were pre-processed with image processing software ( Adobe Photoshop CC ) to isolate the desired subtype and then manually marked using Photoshop , or automatically counted using ImageJ as described above . Total cone populations were determined by combining M- and S-opsin channels , while mixed M+S+ cones were obtained by masking the M-opsin signal with the S-opsin channel . M+S- cones in pigmented mice were obtained by subtracting the S-opsin signal from the M-opsin photomontage . Finally , M+S- cones ( in albino samples ) , true S-cones ( both strains ) ( Figure 2—figure supplement 1C ) and Venus+ SCBCs ( Cpne9-Venus mouse line ) were manually marked on the retinal photomontage ( Adobe Photoshop CC ) . Topographical distributions of cone population densities were calculated from cone locations identified in whole-mount retinas using image processing ( see above ) . From these populations , isodensity maps were created using Sigmaplot 13 . 0 ( Systat Software ) . These maps are filled contour plots generated by assigning to each area of interest ( 83 . 3 × 83 . 3 μm ) a color code according to its cone density , ranging from 0 ( purple ) to 17 , 300 cones/mm2 for all cone types except for true S-cones and M+S--cone in the albino strain ( 5000 cones/mm2 ) , as represented in the last image of each row of Figure 2A , or 1 , 400 SCBCs/mm2 ( Figure 3D ) within a 10-step color-scale . These calculations allow as well , the illustration of the number of cones at a given position from the ON center . To analyze the relative opsin expression along the retinal surface , we have considered three cone populations ( mixed , M+S-- and true S-cones ) dividing the retina in four quadrants: dorsotemporal , dorsonasal , ventrotemporal and ventronasal ( DT , DN , VT and VN respectively , scheme in Figure 2C ) . The relative percentage of cone-types is represented in line graphs from four retinas/strain ( SigmaPlot 13 . 0 ) . To characterize the connectivity of Venus+ S-cone bipolar cells ( Venus+ SCBCs ) with true S-cone terminals , we acquired images from the same area ( 260 × 260 μm , 63x ) at two focal planes: First , we focused upon the INL+OPL , then the corresponding photoreceptor outer segment ( OS ) layer , respectively , for two areas of interest ( DT and VN ) . To verify connectivity between Venus+ SCBC dendrites and true S-cone pedicles in the OPL , in addition to S-opsin immunodetection , we also labeled retinas using cone arrestin antibodies to discriminate mixed cone pedicles from true S-cone pedicles , because true S-cone pedicles contain either low or no cone arrestin ( Figure 3B , Haverkamp et al . , 2005 ) . In other retinas , SCBC contacts were verified by tracking each cell body from cone pedicles to their respective OS to confirm S+M- opsin labeling ( Figure 3C ) . In five retinas ( with S- and M-opsin double immunodetection ) , we analyzed the connectivity between 186 Venus+ SCBCs ( 133 and 53 for DT and VN respectively ) and 263 true S-cone pedicles ( 74 and 189 , DT and VN respectively ) . The number of synaptic contacts was assessed by tracking manually each SCBC-branch from the cell body using the Zen lite black visualization package ( Z-stack with 1 μm interval ) . Multiple branch contacts in one true S-cone pedicle from a single SCBC were considered a single contact and counted only once ( Figure 3B ) , while secondary bifurcations were considered as multiple contacts ( Figure 3c’ ) . SCBC-blind endings were not counted ( Figure 3c” ) . The average number of contacts per retina was used to calculate the DT and VN means ( Supplementary file 2 and graphs in 3C ) . To characterize the true S-cone cluster connectivity in the VN retina , retinal whole-mounts were imaged with a 63x objective , from the photoreceptor outer segments to the OPL , in a Z-stack image with 0 . 5 μm interval . To visualize the true S-cone clustering and Venus+ SCBC connectivity , we identified numerically , and color coded each true S-outer segment form a cluster . The corresponding true S-pedicles were identified by tracking the cell body from their S+M-OSs . Focusing on the outer plexiform layer ( OPL ) , each individual true S-cone pedicle -that form a cluster- was manually outlined and color coded accordingly . Lastly , the SCBC synaptic terminals that belong to a single SCBC , were identified by their specific contacts to the respective true S-cone pedicle ( Figure 4F ) . Retinal images were reconstructed and projected into visual space using R software v . 3 . 5 . 2 for 64-bit Microsoft Windows using Retistruct v . 0 . 6 . 2 as in Sterratt et al . , 2013 . Reconstruction parameters from that study were used: namely , a rim angle of 112° ( phi0 = 22° ) , and eye orientation angles of 22° ( elevation ) and 64° ( azimuthal ) . For Figure 4—figure supplement 1 , true S-cone density contour lines and heatmaps were computed in MATLAB and overlaid onto flat-mount retina opsin labeling images using ImageJ prior to processing by Retistruct . Statistical comparisons for the percentage of cones/retinal location , the total cone quantifications ( Supplementary file 1 ) and the DT or VN true S-cones and Venus+SCBCs ( Supplementary file 2 ) were carried out using GraphPad Prism v8 . 3 for Microsoft Windows . Data are presented as mean ± standard deviation . All data sets passed the D'Agostino-Pearson test for normality , and the comparisons between strains were performed with Student’s t-test . For each 1 mm retinal sample , VDRI , NNRI , and SCNR values were normalized and compared to the distributions of ‘shuffled’ cone populations . Such comparisons were not performed against ‘distributed’ populations , because in those populations , VDRI and NNRI values were consistently much higher—and SCNR much lower—than in real samples ( see Figure 4D–E ) . The ‘shuffled’ populations for each retinal region produced measurements that were well described by normal distributions ( Kolmogorov-Smirnov test , MATLAB ) . Thus , to allow comparisons across samples , we converted each measurement into a Z-score using the mean and standard deviation of those measures from shuffled populations . One-tailed Student’s t-tests were performed to compare the normalized measures to the distribution of ‘randomly shuffled’ cone population measures , and significance was determined at the p<0 . 05 level . | Many primates , including humans , can see color better than most other mammals . This difference is due to the variety of light-detecting proteins – called opsins – that are produced in the eye by cells known as cones . While humans have three , mice only have two different opsins , known as S and M , which detect blue/UV and green light , respectively . Mouse cones produce either S-opsins , M-opsins or both . Fewer than 10 percent of cone cells in mice produce just the S-opsin , and these cells are essential for color vision . Mice are commonly used in scientific research , and so their vision has been well studied . However , previous research has produced conflicting results . Some studies report that cone cells that contain only S-opsin are evenly spread out across the retina . Other evidence suggests that color vision in mice exists only for the upper field of their vision , in other words , that mice can only distinguish colors that appeared above them . Nadal-Nicolás et al . set out to understand how to reconcile these contrasting findings . Molecular tools were used to detect S- and M-opsin in the retina of mice and revealed large differences between the lower part , known as the ventral retina , and the upper part , known as the dorsal retina . The ventral retina detects light coming from above the animal , and about a third of cone cells in this region produced exclusively S-opsin , compared to only 1 percent of cones in the dorsal retina . These S-opsin cone cells in the ventral retina group into clusters , where they connect with a special type of nerve cells that transmit this signal . To better understand these findings , Nadal-Nicolás et al . also studied albino mice . Although albino mice have a different distribution of S-opsin protein in the retina , the cone cells producing only S-opsin are similarly clustered in the ventral retina . This suggests that the concentration of S-opsin cone cells in the ventral retina is an important feature in mouse sight . This new finding corrects the misconception that S-opsin-only cone cells are evenly spread throughout the retina and supports the previous evidence that mouse color vision is greatest in the upper part of their field of vision . Nadal-Nicolás et al . suggest this arrangement could help the mice to detect predators that may attack them from above during the daytime . Together , these new findings could help to improve the design of future studies involving vision in mice and potentially other similar species . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"short",
"report",
"neuroscience"
] | 2020 | True S-cones are concentrated in the ventral mouse retina and wired for color detection in the upper visual field |
SynGAP is a Ras/Rap GTPase-activating protein ( GAP ) that is a major constituent of postsynaptic densities ( PSDs ) from mammalian forebrain . Its α1 isoform binds to all three PDZ ( PSD-95 , Discs-large , ZO-1 ) domains of PSD-95 , the principal PSD scaffold , and can occupy as many as 15% of these PDZ domains . We present evidence that synGAP-α1 regulates the composition of the PSD by restricting binding to the PDZ domains of PSD-95 . We show that phosphorylation by Ca2+/calmodulin-dependent protein kinase II ( CaMKII ) and Polo-like kinase-2 ( PLK2 ) decreases its affinity for the PDZ domains by several fold , which would free PDZ domains for occupancy by other proteins . Finally , we show that three critical postsynaptic signaling proteins that bind to the PDZ domains of PSD-95 are present in higher concentration in PSDs isolated from mice with a heterozygous deletion of synGAP .
We propose a new model for regulation of trapping of AMPARs and other synaptic proteins in 'slots' at the postsynaptic membrane . Numerous experiments have shown that AMPARs become immobilized at the postsynaptic membrane by a three stage process involving insertion of receptors into the perisynaptic membrane , diffusion into the synapse and 'trapping' within the structure of the postsynaptic density ( PSD; Opazo and Choquet , 2011 ) . The sites at which the receptors are trapped have been referred to as 'slots , ' ( Shi et al . , 2001 ) and they are believed to consist principally of PDZ domains on PSD-95 ( Opazo et al . , 2012 ) . Phosphorylation by CaMKII of TARPs ( Transmembrane AMPAR Regulatory Proteins; Bredt and Nicoll , 2003 ) increases their affinity for PDZ domains of PSD-95 suggesting that the phosphorylation may promote trapping of new AMPARs in the PSD ( Tomita et al . , 2005; Opazo et al . , 2010 ) . Indeed the PDZ domains of PSD-95 act as docking sites for several synaptic regulatory proteins ( Figures 1 and 2B ) ; including NMDA-type glutamate receptors ( NMDARs; Kornau et al . , 1995 ) , as well as neuroligins ( Varoqueaux et al . , 2006 ) and LRRTMs ( Leucine Rich Repeat TransMembrane proteins; Linhoff et al . , 2009 ) , which nucleate new synapse formation and contribute to clustering of AMPARs ( Siddiqui et al . , 2010 ) . AMPARs , NMDARs , TARPs , LRRTMs , and neuroligins comprise the most highly enriched transmembrane proteins precipitated together with PSD-95 from the PSD fraction of mouse forebrain ( Dosemeci et al . , 2007 ) . The multiplicity of PDZ binding proteins at the synapse raises the question of whether and how competition for binding to the PDZ domains is regulated at individual synapses . 10 . 7554/eLife . 16813 . 003Figure 1 . Competition among synaptic proteins for binding to PDZ domains of PSD-95 . Each of the three PDZ domains of PSD-95 binds only one protein at a time . The composition of the PSD-95 complex is determined by a dynamic equilibrium that depends on the relative affinities of the proteins that bind to it and their relative concentrations . The figure illustrates competition among three prominent membrane proteins ( TARP , LRRTM , and neuroligin ) and cytosolic synGAP-α1 . Note that LRRTMs and neuroligin also bind across the cleft to presynaptic neurexins . Other PSD proteins that bind to PDZ1 and PDZ2 include the GluN2A and GluN2B subunits of the NMDARs , plasma membrane Ca2+ pumps ( PMCA2B and 4B ) , the ErbB4 receptor ( Lim et al . , 2002 ) , and BRAG1 ( an ArfGEF; Sakagami et al . , 2008; Myers et al . , 2012 ) . PDZ3 binds a smaller group of additional proteins , including β1-adrenergic receptors , and CRIPT ( Lim et al . , 2002 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 00310 . 7554/eLife . 16813 . 004Figure 2 . Domain diagrams of synGAP-α1 and PSD-95 . ( A ) SynGAP-α1 . The N-terminus of r-synGAP-α1 is indicated , as are the locations of the major sites phosphorylated by CaMKII ( black ) , CDK5 ( green ) ( Walkup et al . , 2015 ) , and PLK2 ( blue ) ( Walkup IV et al . ; Lee et al . , 2011 ) , most of which are in the 'disordered domain . ' 'MPD' indicates a region in which several nearby serines ( 808 , 810 , 821 , 825 , and 827 ) are phosphorylated by both CaMKII and PLK2 . Numbering is based on rat isoform synGAP A1-α1 . For comparison , phosphorylation sites identified in Lee et al . , ( 2011 ) were numbered according to the B isoform and can be compared to ours by subtracting 44 residues from our numbering . Phosphorylation sites identified in Araki et al . , ( 2015 ) were numbered according to the A isoform and can be compared to ours by adding 15 residues to our numbering . The 5 residue PDZ ligand is located at the C-terminus . ( B ) PSD-95 . The five major domains of PSD-95 , including the approximate relationships of its three N-terminal PDZ domains are indicated ( Cho et al . , 1992 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 004 SynGAP , a postsynaptic GTPase activating protein , is unusually abundant in the PSD scaffold . One prominent alternatively spliced isoform , synGAP-α1 ( Li et al . , 2001 ) contains a PDZ-domain ligand and binds to all three of the PDZ domains of PSD-95 ( Kim et al . , 1998; Figures 1 , 2 ) . Here we propose that association of synGAP-α1 with PDZ domains of PSD-95 contributes to regulation of docking of AMPARs within the PSD and , therefore , to regulation of its overall protein composition . This function is distinct from its function as a Ras/Rap GAP ( Walkup et al . , 2015 ) . We support this hypothesis by showing that phosphorylation by at least two protein kinases on several sites in the regulatory region of synGAP-α1 reduces the affinity of synGAP-α1’s PDZ-ligand for the three PDZ domains of PSD-95 . One of these protein kinases is CaMKII , which is activated when synaptic plasticity is initiated by activation of NMDA-type glutamate receptors ( NMDARs ) . The other is PLK2 , a constitutively active kinase that is induced by neuronal activity and mediates homeostatic scaling ( Seeburg et al . , 2005 ) . We further show that binding of Ca2+/CaM to synGAP-α1 selectively reduces its affinity for PDZ3 of PSD-95 . Finally , we show that reduction of the total amount of synGAP in heterozygous knockout mice alters the composition of PSDs in the mutant mice in a way that is most directly explained by reduced competition from synGAP-α1 for binding to PDZ domains of PSD-95 . The PSD is an organized complex of signaling proteins attached to the postsynaptic membrane of excitatory glutamatergic synapses in the central nervous system . It comprises a network of scaffold proteins , the most prominent of which is PSD-95 , a member of the MAGUK family ( Membrane-Associated GUanylate Kinase-like proteins ) ( Kennedy , 2000; Sheng and Kim , 2011 ) . An average PSD ( ~360 nm in diameter ) is estimated to contain ~300 molecules of PSD-95 with 900 PDZ domains ( Chen et al . , 2005; Sugiyama et al . , 2005 ) . SynGAP is nearly as abundant in the PSD as PSD-95 itself ( Chen et al . , 1998; Cheng et al . , 2006; Dosemeci et al . , 2007 ) . It is expressed as several isoforms with four different C-termini , α1 , α2 , β , and γ ( Li et al . , 2001; McMahon et al . , 2012 ) . The α1-containing isoforms are abundantly expressed at synapses and contain the PDZ ligand . Assuming that α1 isoforms make up 30–50% of the total synaptic synGAP , and that one synGAP-α1 molecule can bind to any one of the three PDZ domains in each molecule of PSD-95 , synGAP-α1 could occupy 10–15% of the PDZ domains of PSD-95 in an average PSD . Thus , it could help to regulate the size and strength of the synapse by limiting the availability of 'slots' that can bind AMPAR complexes ( Hayashi et al . , 2000; Shi et al . , 2001; Opazo et al . , 2012 ) . This proposed function could explain its unusually high abundance in the PSD which , until now , has been mysterious ( Chen et al . , 1998; Sheng and Hoogenraad , 2007 ) . It has been proposed that , in general , PDZ domains act as flexible protein interaction points that can be modified to support changes in cytoplasmic organization ( Kurakin et al . , 2007 ) . Complexes formed by PDZ domain interactions are examples of linked , multiple equilibria , the stable configurations of which are determined by the concentrations of each component and by their affinities for the relevant PDZ domains . Evidence has indicated that PSD-95 protein complexes exist in dynamic equilibrium permitting continual turnover and potential rearrangement of their composition ( Sturgill et al . , 2009; Schapitz et al . , 2010 ) . Because synGAP-α1 is an abundant protein in the PSD with a relatively high affinity for all three PDZ domains of PSD-95 , reduction of its affinity after phosphorylation will allow other components to compete more effectively for binding; thus , the composition of the PSD-95 complex will shift to a new equilibrium . This proposed mechanism can account for many previous experimental observations . For example , two other groups used imaging to show that phosphorylation by CaMKII in living neurons triggers movement of synGAP away from the PSD ( Yang et al . , 2013; Araki et al . , 2015 ) . In the Araki study , the authors suggested that phosphorylation of synGAP results in 'dispersion' of synGAP away from the PSD and therefore has the effect of upregulating Ras near the PSD . We propose here that an additional important consequence of the decrease in binding of synGAP to the PDZ domains of PSD-95 is readjustment of the composition of the PSD resulting from increased availability of the PDZ domains of PSD-95 . Indeed , the dynamics of movements of synGAP and AMPARs visualized in living neurons following synaptic stimulation are consistent with this hypothesis . Activation of NMDARs and CaMKII causes dispersal of synGAP away from the PSD within a few minutes ( Yang et al . , 2013; Araki et al . , 2015 ) . The same stimuli produce an equally rapid increase in the rate of trapping of AMPARs at synaptic sites ( Makino and Malinow , 2009; Opazo et al . , 2010; Opazo et al . , 2012 ) . Thus , the rates of these two processes , observed in living neurons , are compatible with the notion that reduced binding of synGAP to PSD-95 during induction of LTP opens up binding slots for AMPAR complexes . Dispersal of synGAP away from the postsynaptic membrane would be expected to result in increased activation of Ras and Rap , which would lead to increases in the rates of exocytosis and/or endocytosis of AMPARs , respectively ( Zhu et al . , 2002 ) . However , experiments show that increased exocytosis of AMPARs does not contribute to their increased trapping in the minutes following synaptic stimulation ( Makino and Malinow , 2009 ) . Instead , exocytosis replenishes surface AMPARs in the dendrite and perisynaptic spine membrane , and this process occurs with a slower time course than enhanced trapping . An additional example of a previous result that is explained by this model , is the observation that absence of synGAP in hippocampal neurons cultured from synGAP KO fetuses leads to accelerated ( 'precocious' ) maturation of spines , including early movement of PSD-95 into spine heads , and ultimately larger clusters of PSD-95 in individual synapses compared to wt neurons ( Vazquez et al . , 2004 ) . In this study , expression of wt-synGAP-α1 in the mutant neurons rescued all of these phenotypes; however , expression of synGAP-α1 with a deletion of the five residue PDZ ligand ( ΔSXV ) failed to rescue any of the effects of synGAP deficiency on precocious maturation of spines . In fact , expression of synGAPΔSXV caused an increase in the size of clusters of PSD-95 in spines compared to wt neurons . This failure to rescue the phenotypes was not a result of mislocalization of synGAP-α1; synGAPΔSXV localized like wt-synGAP-α1 to developing spine heads . Instead , the data are consistent with the idea that synGAP-α1 normally competes with several proteins for binding to the PDZ domains of PSD-95 , and thus limits the size of clusters of PSD-95 and its associated proteins , as well as their movement into spine heads ( Vazquez et al . , 2004 ) . Yet another result supporting the model is the finding of McMahon et al . ( 2012 ) that the α1 and α2 isoforms have markedly different effects on synaptic strength when expressed in primary cultures of forebrain neurons . For example , 12 to 36 hr after transfection of the neurons with the A1α1 isoform of synGAP , 73% of the neurons were 'silent' , that is , had no miniature end-plate potentials ( mEPSCs ) , compared with 11% of control neurons . In contrast , transfection with the A1α2 isoform , which differs from α1 only in its C-terminal 48 residues and lacks a PDZ ligand , had no effect on the proportion of silent neurons , which remained at 11% . Like Vazquez et al . , ( 2004 ) , they found that the differential effect did not arise from mislocalization of the A1α2 isoform . Both isoforms of synGAP localized normally to dendrites and spines . The most straightforward explanation of this result is consistent with our hypothesis , which predicts that overexpression of the synGAP-α1 isoform bearing the PDZ ligand restricts binding of AMPAR auxiliary proteins to PDZ domains and thus interfers with their localization to the synaptic site . Neurons cultured from synGAP deficient mice have been reported by several groups to have a higher average number of AMPARs at their synapses than wt neurons ( Kim et al . , 2003; Vazquez et al . , 2004; Rumbaugh et al . , 2006 ) . The data presented here suggests that the increase in AMPARs in Syngap+/- mice may be a direct result of increased binding of TARPs and LRRTMs to PDZ domains that are made available by the reduced amount of synGAP ( Tomita et al . , 2005; de Wit et al . , 2009 ) . Finally , our proposed model and supporting results may help to explain the mechanism underlying a form of developmental intellectual disability ( ID ) resulting from synGAP haploinsufficiency . Mutations in a single copy of synGAP have been causally implicated in sporadic cases of non-syndromic ID , often associated with either autism ( ASD ) or epilepsy ( Berryer et al . , 2013 ) . The frequency of developmental ID worldwide is estimated at 1 to 3% , and 25 to 50% of cases are sporadic , meaning that the parents are not affected . Although data is still sparse , mutations causing SynGAP haploinsufficiency appear to account for 2–9% of sporadic cases ( Hamdan et al . , 2011; Berryer et al . , 2013 ) , suggesting that its prevalence in the population could be as high as 0 . 03–0 . 1% and placing it in the same range of frequency as Fragile-X syndrome . The amount of synGAP in the brains of mice with synGAP haploinsufficiency is reduced by 50% ( Vazquez et al . , 2004 ) . We show here that this reduction leads to a shift in the composition of the PSD scaffold , apparently resulting from the decrease in synGAP’s ability to compete for binding to PDZ domains of PSD-95 . This derangement is likely a significant factor in the human pathology leading to ID , ASD , and epilepsy .
SynGAP-α1 can be expressed in bacteria and purified in a soluble form by deleting the first 102 residues of its N-terminus ( Walkup et al . , 2015 ) . This version of synGAP , termed r-synGAP-α1 , retains all of the identified functional domains , the regulatory domain , and the C-terminal PDZ ligand ( Figure 2A ) . In a previous study we showed that r-synGAP-α1 is phosphorylated by CaMKII at several residues including S1283 , which is 7 residues upstream of the PDZ domain ligand located at residues 1290–1293 ( Walkup et al . , 2015 ) . Because this phosphorylation site is so near the PDZ ligand , we wondered whether its phosphorylation , or phosphorylation of other sites by CaMKII , would interfere with binding of synGAP-α1 to PDZ domains of PSD-95 . To test this , we incubated r-synGAP-α1 with affinity resins substituted with recombinant PDZ domains as described under Materials and methods . The beads contained PDZ1 , PDZ2 , PDZ3 , a fragment containing PDZ1 and PDZ2 ( PDZ12 ) , or a fragment containing all three PDZ domains ( PDZ123 ) ( Figure 3—figure supplement 1; Walkup and Kennedy , 2014 ) . Binding of r-synGAP-α1 to the beads was tested with or without a prior 10 min phosphorylation by CaMKII . As expected , without phosphorylation , r-synGAP-α1 binds specifically to each of the three PDZ domains ( Figure 3A ) . In this assay , its binding is highest to PDZ3 . Binding of r-synGAP-α1 to PDZ123 reveals a substantial avidity effect; that is , the amount bound per individual PDZ domain is twice that bound to PDZ3 alone and four times that bound to either PDZ1 or PDZ2 alone . 10 . 7554/eLife . 16813 . 005Figure 3 . Phosphorylation by CaMKII regulates association of r-synGAP-α1 with PDZ domains of PSD-95 . ( A ) Association of r-synGAP-α1 with PDZ domains of PSD-95 before and after phosphorylation by CaMKII . R-synGAP-α1 was incubated in a phosphorylation mix for 10 min with either 0 CaMKII and 0 Ca2+/CaM ( control ) or 10 nM CaMKII and 0 . 7 mM CaCl2 /3 . 4 μM CaM ( + CaMKII ) before binding to PDZ domain resins for 60 min at 25°C , as described under 'Materials and methods . ' There is no detectable binding of synGAP to unsubstituted resin , and no detectable non-specific binding of proteins to the PDZ resins under these conditions ( Walkup and Kennedy , 2014 ) . ( B ) Both Ca2+/CaM and CaMKII are required in the phosphorylation reaction to reduce binding of r-synGAP-α1 to PDZ123 resin . R-synGAP-α1 was incubated in the phosphorylation reaction without either Ca2+/CaM or CaMKII or with both before binding to PDZ resin . The final bar shows that phosphorylation of the PDZ123 domain resin itself doesn’t alter binding of r-synGAP-α1 . PDZ123 domain affinity resin was phosphorylated for 60 min in the presence of CaMKII and 0 . 7 mM CaCl2/3 . 4 μM CaM before incubation with control r-synGAP ( 500 nM ) for 60 min at 25°C . ( C ) Stoichiometry of phosphorylation of r-synGAP-α1 by CaMKII . R-synGAP-α1 ( 725 nM ) was phosphorylated in the presence of CaMKII ( 10 nM ) , as described under 'Materials and methods . ' At the indicated times , reactions were quenched by addition of 3x Laemmli sample buffer . Radiolabeled r-synGAP-α1 was isolated by SDS-PAGE and quantified as described under 'Materials and methods . ' ( D ) Change in affinity of r-synGAP-α1 for PDZ123 after phosphorylation by CaMKII for times corresponding to those measured in C . R-synGAP-α1 was phosphorylated for 0 . 5 to 10 min as described in C before incubation with PDZ123 domain affinity resin for 60 min as described under 'Materials and methods . ' Control ( -CaMKII , -Ca2+/CaM ) is r-synGAP-α1 incubated in the phosphorylation reaction in the absence of CaMKII and Ca2+/CaM . ( E ) Change in affinity of r-synGAP-α1 for PDZ123 after phosphorylation for 10 min by CDK5 or PLK2 . or by a combination of the two , as described in 'Materials and methods . ' The reduction in binding after phosphorylation for 10 min by CaMKII is shown for comparison . Data shown in A-E are plotted as mean ± S . E . ( n = 4 ) . For A , B , D , and E , the statistical significance of differences in binding to PDZ domain resin relative to unphosphorylated r-synGAP-α1 control ( -Ca2+/CaM ) was determined by ordinary one way ANOVA ( uncorrected Fisher’s LSD ) . **p<0 . 01; ***p<0 . 001; ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 00510 . 7554/eLife . 16813 . 006Figure 3—source data 1 . Source data for Figure 3A . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 00610 . 7554/eLife . 16813 . 007Figure 3—figure supplement 1 . Purification of recombinant PDZ domains of PSD-95 . PDZ domains of PSD-95 were expressed in E . coli individually or combined in a single peptide , as indicated , and purified as described under 'Materials and methods . ' Proteins in the starting soluble fraction ( S ) and eluted from PDZ ligand affinity columns ( E ) were fractionated on 12% ( PDZ1 , PDZ2 , PDZ3 , PDZ12 ) or 4–12% gradient ( PDZ123 ) SDS-polyacrylamide gels and stained as described under 'Materials and methods . 'DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 007 Phosphorylation by CaMKII reduces binding of r-synGAP-α1 to all of the individual PDZ domains and to PDZ12 and PDZ123 ( Figure 3A ) . The reduction in binding requires the presence of both Ca2+/CaM and CaMKII in the phosphorylation reaction mixture ( Figure 3B ) . The fourth bar of Figure 3B shows that the reduction in binding is not caused by phosphorylation of PDZ domains on the column by residual CaMKII . We have shown previously that as many as 10 sites on r-synGAP-α1 are phosphorylated by CaMKII ( Walkup et al . , 2015 ) . Approximately 5 of these , including site S1123 , are fully phosphorylated after an 0 . 5 min reaction ( Walkup et al . , 2015 and Figure 3C ) . To test whether the reduction in binding depends primarily on phosphorylation of the rapidly phosphorylated sites or requires phosphorylation of most of the sites , we tested binding of r-synGAP-α1 to PDZ123 after phosphorylation for various times ( Figure 3D ) . The reduction in binding is maximal after 0 . 5 min , indicating that phosphorylation of the more rapidly phosphorylated sites is sufficient for full reduction of affinity . R-synGAP-α1 can also be phosphorylated by CDK5 ( Walkup et al . , 2015 ) and by PLK2 ( Lee et al . , 2011 ) both of which increase its RasGAP activity . We tested the effect of 10 min of phosphorylation by each of these kinases on affinity of r-synGAP-α1 for PDZ123 . Phosphorylation by CDK5 had no significant effect on affinity; whereas , phosphorylation by PLK2 decreased its affinity by ~40% , less strongly than does phosphorylation by CaMKII ( Figure 3E ) . Both CaMKII and PLK2 can phosphorylate several sites in a region marked MPD ( multiple-phosphorylation domain ) in Figure 2A , including serines at 808 , 810 , 821 , 825 , and 827 ( Walkup et al . , 2015; Walkup et al . , in preperation ) . This result means that phosphorylation of multiple sites at various locations in the regulatory region can decrease binding of r-synGAP-α1 to PDZ domains , but to varying degrees . The three PDZ domains of PSD-95 are located in the N-terminal half of the protein from residues 61 to 402 . The first two PDZ domains are separated by 4 residues and the third is 53 residues downstream . We determined the affinities of r-synGAP-α1 for individual PDZ domains , for PDZ12 ( residues 61–249 ) and for PDZ123 ( residues 61–402 ) by a competition assay in which SPR is used to detect the amount of free r-synGAP-α1 in solutions containing a constant amount of r-synGAP-α1 and varying amounts of recombinant PDZ domains ( Nieba et al . , 1996; Lazar et al . , 2006; Abdiche et al . , 2008 ) . To detect the free r-synGAP-α1 , recombinant PDZ domains are immobilized on the Biacore chip as described under 'Materials and methods . ' We used the competition method rather than conventional Biacore measurements in which varying concentrations of r-synGAP-α1 are applied to a chip containing immobilized PDZ domains because concentrations of r-synGAP-α1 above ~100 nM produced a large bulk resonance signal caused by high viscosity that obscured the change in resonance produced by its binding to PDZ domains . The competition assay eliminates the need to apply high concentrations of r-synGAP-α1 to the chip . We generated a standard curve in which the maximum resonance responses of a series of concentrations of r-synGAP-α1 from 0 nM to 50 nM ( Figure 4A , grey traces ) were determined and plotted against r-synGAP-α1 concentration ( Figure 4B , large grey dots ) . The data were fit with a hyperbolic curve . The maximum resonance response of a series of mixtures containing 25 nM r-synGAP-α1 and increasing concentrations of PDZ1 from 0 nM to 10 µM were measured , and the concentration of r-synGAP-α1 remaining free to bind to PDZ1 on the chip was then determined from the standard curve ( Figure 4B , small black dots ) . A KD of 220 ± 30 nM ( Table 1 ) was calculated as described under Materials and methods ( Figure 4C ) . We used the same method to measure KDs for PDZ2 and PDZ3 ( Figure 5A and B , respectively ) . Because , PDZ12 and PDZ123 contain more than a single PDZ domain binding site , all of which have similar affinities , we determined single 'apparent' equilibrium dissociation constants ( KDapp ) for these two constructs . As discussed in 'Materials and methods' , it is not possible to derive a unique equation incorporating two or three binding sites when the affinities of the multiple sites are similar . The best fit of the data with the equation for a single binding site provides a KDapp that can be used to characterize the binding behavior of the constructs ( Figure 5C and D , respectively ) . The values are summarized in Table 1 . We obtained an additional value of 730 ± 50 nM for the KD of PDZ3 by a conventional Biacore assay , which is in good agreement with the KD measured by the competition assay . These data show that , under these conditions , PDZ1 has a higher affinity for r-synGAP-α1 than does PDZ3 . 10 . 7554/eLife . 16813 . 008Figure 4 . Measurement of affinity of r-synGAP-α1 for PDZ1 of PSD-95 by the 'competition in solution' method . ( A ) Biacore sensorgrams showing the calibration curves ( grey lines ) for binding of 0–50 nM r-synGAP-α1 and the measurement of free r-synGAP-α1 ( samples; black lines ) in mixtures containing 25 nM r-synGAP-α1 and 0–10 μM PDZ1 domain . Free r-synGAP-α1 was detected by binding to PDZ1 domains immobilized on a Biacore chip as described under 'Materials and methods . ' ( B ) A standard calibration curve was constructed by plotting the maximum calibrated resonance responses ( marked by arrow in A ) against the corresponding concentrations of r-synGAP-α1 ( large grey dots ) . The maximum resonance responses for each sample mixture were plotted on the standard curve to determine the free r-synGAP-α1 concentrations in each mixture ( black dots ) . ( C ) Plot of free r-synGAP-α1 concentrations determined in B against the log of PDZ domain concentrations ( black circles ) . The data were fit to the binding equation shown in 'Materials and methods' with the use of Biacore software . A KD value ( Table 1 ) was calculated from the equation as described under 'Materials and methods . 'DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 00810 . 7554/eLife . 16813 . 009Figure 5 . Affinities and apparent affinities of r-synGAP for PDZ2 , PDZ3 , PDZ12 and PDZ123 domains of PSD-95 determined by the 'competition in solution' method . The concentrations of free r-synGAP-α1 in sample mixtures containing each of the indicated PDZ domains were measured as described in Figure 4A and B , and under 'Materials and methods . ' The values ( black dots ) were plotted against the log of the PDZ domain concentration and fit to a binding curve as described in Figure 4C . Representative experiment for A , PDZ2; B , PDZ3; C , PDZ12; and D , PDZ123 . The calculated KD and KDapp values from all experiments are listed in Table 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 00910 . 7554/eLife . 16813 . 010Table 1 . Affinities of R-synGAP-α1 for PDZ Domains of PSD-95 . Dissociation constants ( KD ) and apparent dissociation constants ( KDapp ) for the interactions of r-synGAP-α1 with the PDZ domains of PSD-95 were determined by the Biacore/SPR 'competition in solution' method as described under 'Materials and methods . ' In one experiment , the KD for PDZ3 was determined by conventional SPR as described under 'Materials and methods . ' Goodness of Fit refers to the fit of data shown in Figures 4 , 5 , and 7 to the equation relating synGAPfree to PDZ domain concentration , assuming a single binding site , as described under 'Materials and methods . ' Because PDZ12 , and PDZ123 contain more than one binding site for r-synGAP-α1 , the affinities are given as apparent dissociation constants . Data are expressed as mean ± S . E . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 010PDZ Domain from PSD-95No . of ExperimentsDissociation Constant ( KD ) for Binding to R-synGAP ( nM ) Goodness of Fit ( R2 ) PDZ13220 ± 300 . 908 - 0 . 947PDZ221500 ± 1000 . 967 , 0 . 969PDZ32620 ± 700 . 951 , 0 . 962PDZ31730 ± 50 ( by conventional SPR ) N . A . Apparent Dissociation Constant ( KDapp ) for Binding to R-synGAP ( nM ) PDZ124350 ± 400 . 931 - 0 . 987PDZ12374 . 7 ± 0 . 60 . 957 - 0 . 987PDZ1233 ( r-synGAP Phosphorylated by CaMKII ) 46 ± 100 . 810 - 0 . 880PDZ1232 ( r-synGAP S1283D ) 16 ± 30 . 953 , 0 . 954 To examine which phosphorylation sites can reduce binding of r-synGAP-α1 to PDZ domains , we measured binding of recombinant mutants of r-synGAP-α1 to PDZ123 affinity resin . We first compared binding of all of the unphosphorylated mutants by ANOVA corrected for multiple comparisons by the Tukey method ( Figure 6 , black bars ) . Single or multiple mutations of the indicated serines to alanine did not significantly alter binding before phosphorylation; nor did the double mutation S1093D/S1123D . Only mutations in which site S1283 was mutated to the phosphomimetic aspartate ( S1283D and S1093D/S1123D/S1283D ) decreased binding of unphosphorylated r-synGAP-α1 to PDZ123 with a significant p-value ( p<0 . 0001 for both ) . This result suggests that phosphorylation of S1283 has the most potent effect of all the sites on binding to PDZ domains . Mutations S1123A and S1283A did not significantly alter the change in binding after a brief , 0 . 5 min phosphorylation . However , they both produced a trend toward lower reduction of binding , suggesting that both sites contribute to reduction of binding after brief phosphorylation . The effect of the double mutation S1093A/S1123A after 0 . 5 min of phosphorylation was not significantly different from the effect of S1123A , suggesting that phosphorylation of S1093 doesn’t influence binding to PDZ domains after a brief phosphorylation . In contrast , the triple mutant S1093A/S1123A/S1283A was the only S to A mutation that significantly interfered with the loss of binding ( p=0 . 0074 ) compared to wild type after 0 . 5 min of phosphorylation . This result reinforces the conclusion that sites S1123 and S1283 both contribute to reduction of binding of r-synGAP-α1 to PDZ domains after brief phosphorylation . 10 . 7554/eLife . 16813 . 011Figure 6 . Effect of phosphorylation by CaMKII on association of PDZ123 domains with phospho-deficient and phospho-mimetic mutants of r-synGAP-α1 . Wild-type and mutant r-synGAP-α1 were incubated with phosphorylation mixtures for 10 min without ( Control ) , or for 0 . 5 or 10 min with 10 nM CaMKII , 0 . 7 mM CaCl2/3 . 4 μM CaM ( CaMKII ) , then incubated with PDZ123 affinity resin for 60 min at 25°C , as described under 'Materials and methods . ' ( A ) Binding to PDZ123 of r-synGAP-α1 and single mutations at S1123 and S1283 . Sites S1123 or S1283 were mutated either to alanine ( S1123A , S1283A ) or to the phosphomimetic aspartate ( S1123D , S1283D ) . ( B ) Binding to PDZ123 of r-synGAP-α1 and double or triple mutations at S1093 , S1123 , and S1283 , as indicated . Data are mean ± S . E . ( n = 4 ) . The statistical significances of differences between wild type and the various mutants in binding to PDZ123 before phosphorylation were determined by ordinary one way ANOVA ( Tukey correction for multiple comparisons ) . Only mutations including S1283D were significantly different from wild type ( indicated by ( **** ) ) . Differences between unphosphorylated and phosphorylated individual mutants were compared individually by ordinary one way ANOVA ( uncorrected Fisher’s LSD ) . *pp<0 . 05; **p<0 . 01; ***p<0 . 005; ****p<0 . 0001 . See the text for additional statistical comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 011 Notably , none of the mutations interfere with the loss of binding to PDZ domains after ten min of phosphorylation ( Figure 6A and B ) . Taken together , these results and the results of phosphorylation by PLK2 mean that phosphorylation of S1283 and S1123 significantly reduces binding of r-synGAP-α1 to PDZ domains , however maximum loss of binding can also be accomplished by cumulative phosphorylation over ten min of several sites within the regulatory domain ( Figure 2A and Walkup et al . , 2015 ) . This mechanism is consistent with the finding of Holt et al . that clusters of sites in rapidly evolving disordered regions appear to shift position in evolutionary time suggesting that regulation by phosphorylation often involves disruption or enhancement of protein-protein interactions by addition of multiple negative charges ( Holt et al . , 2009 ) . We measured the apparent dissociation constants ( KDapp’s ) for binding to PDZ123 of r-synGAP-α1 phosphorylated for 10 min by CaMKII , and of the phosphomimetic mutant r-synGAP-α1 S1283D , as described for Figure 5 and under 'Materials and methods' . Phosphorylation for 10 min by CaMKII increases the KDapp of r-synGAP-α1 approximately ten-fold ( Figure 7A , Table 1 ) ; whereas mutation of S1283 to aspartate increases the KDapp approximately four-fold ( Figure 7B , Table 1 ) . Thus , cumulative phosphorylation of several residues can reduce affinity for PDZ domains by an order of magnitude; whereas , addition of a negative charge at S1283 alone can reduce the affinity by a factor of four . 10 . 7554/eLife . 16813 . 012Figure 7 . Apparent affinities of phosphorylated r-synGAP-α1 and r-synGAP-α1-S1283D for PDZ123 determined by the 'competition in solution' method . Representative plots of the concentrations of ( A ) free phospho-r-synGAP-α1 phosphorylated as described for PDZ Binding Assays under 'Materials and methods , ' and ( B ) r-synGAP-α1-S1283D , measured in sample mixtures containing PDZ123 as described in Figure 4A and B , and under 'Materials and methods . ' The values ( black dots ) were plotted against the log of the PDZ123 concentration in the mixture and fit to a binding curve as described in Figure 4C . The calculated values of KDapp are listed in Table 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 012 While studying phosphorylation of r-synGAP-α1 by CDK5 ( Walkup et al . , 2015 ) , we found that the presence of Ca2+/CaM in reactions with either CDK5/p35 or CDK5/p25 doubled the rate and stoichiometry of the phosphorylation ( Figure 8A and B ) . Inclusion of Ca2+ or CaM alone in the phosphorylation reactions did not alter the rates or stoichiometry . 10 . 7554/eLife . 16813 . 013Figure 8 . Effect of Ca2+/CaM on stoichiometry of phosphorylation of r-synGAP-α1 and histone H1 by CDK5 . Stoichiometry of phosphorylation of r-synGAP-α1 ( A and B ) and histone H1 ( C and D ) by CDK5/p35 or CDK5/p25 . R-synGAP-α1 ( 286 nM ) or histone H1 ( 4 . 3 μM ) were incubated with CDK5/p35 or CDK5/p25 as described under 'Materials and methods' in the presence or absence of 0 . 7 mM CaCl2 or 3 . 4 μM CaM , as indicated in each panel . Reactions were quenched at the indicated times by addition of 3x Laemmli sample buffer and radiolabeled r-synGAP-α1 and histone H1 were quantified as described under 'Materials and methods . ' Data are plotted as mean ± S . E . ( n = 4–7 ) . The statistical significance of differences in phosphorylation in the presence of Ca2+ and CaM were determined by ordinary one way ANOVA ( uncorrected Fisher’s LSD ) . **p<0 . 01; ***p<0 . 001; ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 01310 . 7554/eLife . 16813 . 014Figure 8—figure supplement 1 . R-synGAP-α1 binds to CaM affinity resin . Clarified E . coli lysate ( Load ) containing r-synGAP-α1 was incubated with CaM-Sepharose 4B or control Sepharose 4B resin in the presence of 0 or 5 mM CaCl2 and 0 or 10 mM EGTA , as described under 'Materials and methods . ' After washing , bound protein was eluted from the resin with 100 mM EGTA ( Eluate ) , fractionated by SDS-PAGE , and visualized by staining with Gel Code Blue ( Total Protein ) or transferred to a PVDF membrane . R-synGAP-α1 was detected on the immunoblots with anti-synGAP or anti-TetraHis antibodies , as described under 'Materials and methods . ' In the absence of exogenous calcium , r-synGAP-α1 bound weakly to the CaM-Sepharose , but not to control Sepharose beads . When 5 mM Ca2+ was included in the binding and wash buffers its binding to CaM-Sepharose increased , while addition of 10 mM EGTA to the buffers nearly abolished binding . CaM-Sepharose eluate was quantified by scanning the stained blots and gel on a Li-Cor Imager as described under 'Materials and methods . 'DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 01410 . 7554/eLife . 16813 . 015Figure 8—figure supplement 2 . Effect of Ca2+/CaM on GAP activity of r-synGAP-α1 . The GAP activity of r-synGAP-α1 ( 250 nM ) for the indicated GTP-binding protein was assayed as described in Walkup et al . ( 2015 ) except that 1 mM Ca2+ , 3 . 4 µM CaM , or both were added to the GAP assay . A , HRas GAP activity; B , Rap1 GAP activity; and C , Rap2 GAP activity . Data are mean ± S . E . The statistical difference of the GTPase activity of each GTP-binding protein in the absence of synGAP from all of the other conditions was determined by ordinary one way ANOVA ( uncorrected Fisher’s LSD ) . ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 015 We tested whether this effect resulted from binding of Ca2+/CaM to CDK5/p35 ( e . g . He et al . , 2008 ) , or CDK5/p25 by comparing the rates of phosphorylation of histone H1 , a well-known substrate of CDK5 in the presence and absence of Ca2+/CaM ( Figure 8C and D ) . Phosphorylation of histone H1 by either CDK5/p35 or CDK5/p25 was unaffected by Ca2+/CaM . This result suggests that Ca2+/CaM binds directly to r-synGAP-α1 , causing a substrate-directed enhancement of its phosphorylation . To further verify that Ca2+/CaM binds directly to r-synGAP-α1 , we showed that r-synGAP-α1 binds to a CaM-Sepharose affinity resin in a Ca2+-dependent manner , as would be expected if it binds Ca2+/CaM specifically and with significant affinity ( Figure 8—figure supplement 1 ) . We found that the presence of Ca2+/CaM alone in a Ras- or Rap-GAP assay has no effect on the GAP activity of r-synGAP-α1 ( Figure 8—figure supplement 2 ) . We measured the affinity of binding of Ca2+/CaM to r-synGAP-α1 by the conventional SPR method on the Biacore as described under Materials and methods . CaM was immobilized on a chip , and r-synGAP-α1 was applied to it at concentrations from 0 to 75 nM ( Figure 9A ) . Analysis of the equilibrium phase of association at each concentration ( Figure 9B ) yielded a KD of 9 ± 1 nM , indicative of high affinity binding . 10 . 7554/eLife . 16813 . 016Figure 9 . Affinity of r-synGAP-α1 and sr-synGAP for Ca2+/CaM determined by equilibrium analysis . ( A and B ) The affinity of r-synGAP-α1 for Ca2+/CaM was measured by SPR with CaM immobilized on the chip and r-synGAP-α1 injected at 0–75 nM onto the chip surface as described under 'Materials and methods . ' ( A ) Sensorgrams with the blank and reference flow cell readings subtracted show the response upon injection of r-synGAP-α1 onto the chip surface ( 0–75 s ) and its dissociation from the chip surface ( 75–150 s ) . ( B ) RUs at equilibrium ( marked by arrow in A ) were plotted against the corresponding concentrations of r-synGAP-α1 and fitted to a hyperbolic curve . A KD of 9 ± 1 nM was calculated as described under 'Materials and methods . ' ( C and D ) The affinities of r-synGAP-α1 ( C ) and sr-synGAP ( D ) , ( 0–500 nM ) for Ca2+/CaM were measured by incubation with CaM-Sepharose resin as described under 'Materials and methods . ' Integrated intensities of bound r-synGAP-α1 and sr-synGAP were measured from immunoblots as described under 'Materials and methods' and plotted versus the corresponding concentrations incubated with resin . Integrated intensities from Western blots were linear over the range of r-synGAP-α1 and sr-synGAP concentrations used in the assays . Data in C and D are plotted as mean ± S . E . ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 016 To begin to define the location of the high affinity Ca2+/CaM binding site , we compared the affinities for Ca2+/CaM of r-synGAP-α1 and a C-terminal truncated protein , sr-synGAP ( residues 103–725 ) by a bead-binding assay as described under Materials and methods . We measured the amount of each protein bound to a fixed amount of CaM-Sepharose after incubation with increasing concentrations ( Figure 9C and D ) . Both r-synGAP-α1 ( Figure 9C ) and sr-synGAP ( Figure 9D ) showed saturable binding to the CaM-Sepharose resin , and did not bind to control Sepharose beads lacking CaM ( data not shown ) . The data were fit to hyperbolic curves and the KD’s for binding of r-synGAP-α1 and sr-synGAP to Ca2+/CaM were calculated to be 31 ± 3 nM and 210 ± 30 nM , respectively . Thus , the high affinity site appears to be located in the regulatory disordered region of r-synGAP-α1 , which is missing in sr-synGAP . The KD‘s determined by the bead-binding assay ( 31 ± 3 nM ) and Biacore equilibrium binding ( 9 ± 1 nM ) are in the range of those reported for calcineurin ( PP2B ) and CaMKII , 1–10 nM ( Hubbard and Klee , 1987; Cohen and Klee , 1988 ) and 40–80 nM ( Miller and Kennedy , 1985; Meyer et al . , 1992; Hudmon and Schulman , 2002 ) , respectively . We did not detect any binding when sr-synGAP was injected onto the CaM-substituted Biacore chip at concentrations from 10–2500 nM . Thus , the relatively weak binding of sr-synGAP observed in the bead-binding assay is not reproducible when measured on the Biacore chip . These data indicate that Ca2+/CaM binds only weakly , if at all , to the N-terminal half of synGAP . A meta-analysis algorithm for detecting potential CaM-binding domains ( Mruk et al . , 2014 ) predicts two Ca2+/CaM binding sites in the C-terminal half of r-synGAP-α1 , one from residues 1000–1030 and another in the putative coiled coil domain from residues 1229–1253 . The SPR measurements do not allow us to confirm or to rule out the presence of two high affinity sites of similar affinity . We tested whether binding of Ca2+/CaM alters the binding of r-synGAP-α1 to PDZ domains by comparing binding to each affinity resin in the presence or absence of Ca2+/CaM ( Figure 10A ) . The presence of Ca2+/CaM during incubation with resin significantly reduces binding of r-synGAP-α1 to PDZ3 and to PDZ123 , but not to PDZ1 and/or PDZ2 . Thus , binding of Ca2+/CaM has a more specific , but weaker , effect on binding to the PDZ domains than does phosphorylation . The effects of phosphorylation and of the presence of Ca2+/CaM during incubation with resin are not additive ( Figure 10B ) ; that is , the presence of Ca2+/CaM during the incubation with resin does not further decrease binding of phosphorylated r-synGAP-α1 to PDZ123 . 10 . 7554/eLife . 16813 . 017Figure 10 . Effect of Ca2+/CaM binding on association of r-synGAP-α1 with PDZ domains of PSD-95 . ( A ) Association of control and Ca2+/CaM bound r-synGAP-α1 with PDZ domains of PSD-95 . R-synGAP-α1 ( 500 nM ) without ( Control ) or with ( + CaM ) 0 . 7 mM CaCl2/3 . 4 μM CaM was incubated with PDZ domain resins ( PDZ1 , PDZ2 , PDZ3 , PDZ12 , and PDZ123 ) for 60 min at 25°C and bound r-synGAP-α1 was measured as described under 'Materials and methods . ' ( B ) Effects of bound Ca2+/CaM and phosphorylation by CaMKII on association of r-synGAP-α1 with PDZ123 domain are not additive . The association with PDZ123 domain resin of Control , Ca2+/CaM bound ( plus CaM ) , phosphorylated r-synGAP-α1 ( plus CaMKII ) and phosphorylated r-synGAP-α1 bound to Ca2+/CaM ( plus CaM and CaMKII ) was measured as described in A . Data are plotted as mean ± S . E . ( n = 4 ) . The statistical significance of differences in PDZ domain binding relative to Control was determined by ordinary one way ANOVA ( uncorrected Fisher’s LSD ) . **p<0 . 01; ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 01710 . 7554/eLife . 16813 . 018Figure 10—source data 1 . Source data for Figure 10A . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 018 The physiological significance of the finding that phosphorylation by CaMKII decreases the affinity of r-synGAP-α1 for the PDZ domains of PSD-95 is best considered in the context of the high copy number of synGAP in the PSD . In molar terms , synGAP-α1 is 30–50% as abundant in the PSD as PSD-95 itself ( Chen et al . , 1998; Cheng et al . , 2006; Dosemeci et al . , 2007; Sheng and Hoogenraad , 2007 ) . Our data suggest that phosphorylation of synGAP-α1 by CaMKII , triggered by activation of NMDARs , would promote dissociation of synGAP-α1 from the PDZ domains , reducing its ability to compete with other proteins for binding . Syngap+/- mice have been shown to have about half as much synGAP in homogenates from forebrain as wt littermates . Because binding equilibria are driven not only by the intrinsic affinities of the binding partners , but also by their concentrations , one prediction of our proposed hypothesis is that synGAP haploinsufficiency , which reduces the amount of synGAP in the brain by 50% ( Vazquez et al . , 2004 ) , will cause a significant increase in binding to PSD-95 of other prominent PSD-95 binding proteins , such as TARPs , LRRTMs , or neuroligins . Thus , PSDs isolated from Syngap+/- mice would be predicted to have less synGAP and more TARPs , LRRTMs , and/or neuroligins bound to PSD-95 than do PSDs isolated from wt mice . We prepared PSD fractions from the forebrains of six Syngap+/- mice and from six wt litter mates and measured the ratios of synGAP , TARPs , LRRTM2 , neuroligin-1 , and neuroligin-2 to PSD-95 in the two fractions by quantitative immunoblot as described in 'Materials and methods' ( Figure 11 ) . The recoveries of total protein and the amount of PSD-95 per µg protein were identical in the two preparations . As predicted , the level of synGAP is decreased in relation to PSD-95 by ~ 24% in PSDs ( p=0 . 0007 , d = 1 . 75 ) from the Syngap+/- mice ( Figure 11A ) . Furthermore , the ratios of TARPs 2 , 3 , 4 , ɣ8 , and of LRRTM2 to PSD-95 are significantly increased ( Figure 11B , C; TARP/PSD-95 , ~12% , p=0 . 017 , d = 0 . 93; LRRTM2/PSD-95 , ~14% , p=0 . 0035 , d = 0 . 66 ) . This result strongly suggests that , as predicted , the increase in availability of PDZ1/2 domains on PSD-95 in the Syngap+/- mice enhances steady-state binding of TARPs and LRRTMs to those sites . Interestingly , the ratio of neuroligin-1 to PSD-95 is unchanged in the Syngap+/- mice ( d = 0 . 07 , Figure 11D ) , suggesting that increased availability of PDZ3 on PSD-95 is not a strong driver of association of neuroligin-1 with the PSD fraction . However , the ratio of neuroligin-2 to PSD-95 ( Figure 11E ) is increased by ~9% ( p=0 . 019 , d = 0 . 64 ) . Neuroligin-2 normally associates mostly with inhibitory synapses and mediates their maturation ( Varoqueaux et al . , 2004 ) . However , Levinson et al . ( 2005 ) reported that over-expression of PSD-95 in neurons causes a redistribution of neuroligin-2 , increasing the proportion associated with excitatory synapses . Thus , the effect of reduction of synGAP on the distribution of neuroligin-2 shown here is the same as the effect of over-expression of PSD-95 , suggesting that both manipulations increase the number of PDZ3 domains available for binding . Taken together , these results verify the prediction that a decrease in availability of synGAP in the PSD scaffold , increases the association of TARPs , LRRTMs , and neuroligin-2 with the PSD in vivo by releasing a restriction on binding to PDZ domains of PSD-95 . 10 . 7554/eLife . 16813 . 019Figure 11 . Altered composition of the PSD in mice with heterozygous deletion of synGAP . Ratios of amounts of the indicated proteins to PSD-95 in each lane were measured as described in 'Materials and methods' and are reported as mean ± S . E . For all blots except those for neuroligin-1 , PSD-95 was detected with a secondary Ab labeled with AlexaFluor680 and the binding protein was detected with secondary Ab labeled with IRDye 800 . On the neuroligin-1 blot , both PSD-95 and neuroligin-1 were detected with AlexaFluor680; the two bands were well-separated in each lane . Representative sets of visualized bands for WT and HET from the same blot are shown below the graphs . Individual points represent the ratio of the indicated protein to PSD-95 in a single lane ( n refers to these technical replicates ) . The WT and HET PSD preparations were each made from six animals; thus , each of the two preparations represents six biological replicates . ( A ) SynGAP to PSD-95 ratio . Data were collected for 22 lanes from two blots containing 5 µg total PSD fraction per lane . One blot contained six lanes WT and six lanes HET samples , the other contained five lanes of each . The mean ratio of synGAP to PSD-95 was 0 . 234 ± 0 . 012 for WT ( n = 11 ) and 0 . 179 ± 0 . 005 ( n = 11 ) for HET ( −24% ) . Means were compared by unpaired , one-tailed t-test with Welch correction , p = 0 . 0007 . Effect size , d = 1 . 75 . ( B ) TARP ɣ -2 , 3 , 4 , 8 to PSD-95 ratio . Data were collected for 24 lanes from two blots containing 10 µg total PSD fraction per lane . Each blot contained six lanes WT and six lanes HET samples . Densities of all four TARPs were pooled . The mean ratio of TARPs to PSD-95 was 0 . 066 ± 0 . 003 ( n = 12 ) for WT and 0 . 075 ± 0 . 003 ( n = 12 ) for HET ( +12% ) . Means were compared by unpaired , one-tailed t-test with equal variance , p = 0 . 017 . Effect size , d = 0 . 93 . ( C ) LRRTM2 to PSD-95 ratio . Data were collected for 36 lanes from three blots containing six WT and six HET samples , alternating 5 and 10 µg ( 3 each ) . The mean ratio of LRRTM2 to PSD-95 was 0 . 051 ± 0 . 003 ( n = 17 ) for WT and 0 . 059 ± 0 . 003 ( n = 17 ) for HET ( +14% ) . Means were compared by paired , one-tailed t-test , p=0 . 0035 . Effect size , d = 0 . 66 . ( D ) Neuroligin-1 to PSD-95 ratio . Data were collected for 47 lanes from four blots two of which contained 5 µg and two 10 µg total PSD fraction per sample . Each blot contained 6 lanes WT and 6 lanes HET samples . The mean ratio of neuroligin-1 to PSD-95 was 0 . 114 ± 0 . 005 ( n = 24 ) for WT and 0 . 115 ± 0 . 004 ( n = 23 ) for HET ( no difference ) . Means were compared by unpaired one-tailed t-test , p = 0 . 413 . Effect size , d = 0 . 07 . ( E ) Neuroligin-2 to PSD-95 ratio . Data were collected for 44 lanes from four blots containing 10 µg total PSD fraction per lane . Each blot contained six lanes WT and six lanes HET samples . The mean ratio of neuroligin-2 to PSD-95 was 0 . 071 ± 0 . 002 ( n = 20 ) for WT and 0 . 078 ± 0 . 003 ( n = 24 ) for HET ( +9% ) . Means were compared by unpaired , one-tailed t-test with Welch correction , p=0 . 019 . Effect size , d = 0 . 64 . *p<0 . 005; **p<0 . 01; ***p<0 . 001DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 01910 . 7554/eLife . 16813 . 020Figure 11—source data 1 . Source data for ratios determined in Figure 11 A through E . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 020
We have presented a new model for regulation of the composition of the spine postsynaptic membrane and PSD . It builds on previous findings that the PSD scaffold , in particular the PSD-95 complex , is dynamically regulated by activity ( Gray et al . , 2006; Sturgill et al . , 2009 ) , that activation of NMDARs and CaMKII leads to enhanced trapping of AMPARs within the PSD by binding to PDZ domains of PSD-95 ( Opazo et al . , 2010; 2012 ) , and that synGAP moves away from the PSD after phosphorylation by CaMKII ( Yang et al . , 2013; Araki et al . , 2015 ) . We propose that binding of the C-terminus of synGAP-α1 to the PDZ domains of PSD-95 restricts binding of TARPs , LRRTM2 , neuroligin-2 , and perhaps additional proteins . As a result , regulation of synGAP-α1’s concentration and binding affinity for PDZ domains helps to determine the precise composition of the PSD at individual excitatory synapses ( Figure 12 ) . A corollary of this hypothesis is that synGAP plays an important role in limiting the size and strength of excitatory synapses by limiting and helping to regulate the available 'slots' that can bind AMPAR complexes ( Hayashi et al . , 2000; Shi et al . , 2001; Opazo et al . , 2012 ) . We have provided support for this notion by demonstrating that phosphorylation of several sites in synGAP’s regulatory domain by CaMKII or by PLK2 reduces the affinity of its PDZ ligand for all three of the PDZ domains of PSD-95 . Both of these protein kinases are important regulators of synaptic strength ( Seeburg et al . , 2008; Hell , 2014 ) . We have also verified a strong prediction of the hypothesis which is that PSDs from mice with a deletion of one copy of the synGAP gene will contain fewer copies of synGAP and more copies of other proteins that bind to PSD-95 . Indeed , we have shown that PSD fractions from young Syngap+/- mice have ~24% less synGAP per molecule of PSD-95 than those from wt mice; and , in contrast , they have significantly more TARP proteins ( ~12% ) , LRRTM2 ( ~14% ) , and neuroligin-2 ( ~9% ) per molecule of PSD-95 . The ratio of neuroligin-1 to PSD-95 is not altered . 10 . 7554/eLife . 16813 . 021Figure 12 . Cartoon model of rearrangement of PSD caused by synGAP haploinsufficiency or by phosphorylation of synGAP-α1 by CaMKII . Unphosphorylated synGAP-α1 ( Top ) binds to PDZ1 , PDZ2 or PDZ3 of PSD-95 , occupying as many as ~15% of its PDZ domains . The PDZ domains of PSD-95 are shown in blue and their numbers are indicated on the left pair of PSD-95 molecules . AMPARs that have been inserted into the extrasynaptic membrane by exocytosis associate with TARPs and with LRRTMs , both of which can bind to PDZ1 and PDZ2 of PSD-95 . Neuroligin-1 ( NLG-1 ) binds to PDZ3 of PSD-95 . LRRTMs and NLG-1 also bind across the synaptic cleft to presynaptic neurexins . Induction of LTP ( Lower Right ) causes flux of calcium through NMDARs that activates CaMKII leading to phosphorylation of synGAP-α1 on sites in the regulatory domain . The affinity of synGAP-α1 for the PDZ domains decreases , allowing TARPs , LRRTMs , and perhaps NLG-2 to displace synGAP-α1 by binding to the PDZ domains . The shift in affinity of synGAP-α1 creates 'slots' that can be occupied by TARPs and LRRTMs bound to AMPARs , and by NLG-2 , leading to strengthening of the synapse . SynGAP haploinsufficiency ( Lower Left ) results in a reduced amount of synGAP-α1 in the PSD . Similar to the model for LTP , this leaves more PDZ domains available to bind TARPs , LRRTMs , and NLG-2 . DOI: http://dx . doi . org/10 . 7554/eLife . 16813 . 021 In the mouse brain , synGAP begins to be expressed at birth and its expression increases rapidly as synapses are forming during the first weeks after birth ( Vazquez et al . , 2004; Barnett et al . , 2006 ) . Newborns with complete deletion of Syngap appear normal , but die a few days after birth with movement defects and apparent seizure activity , indicating a defect as synapses are forming ( Kim et al . , 2003; Vazquez et al . , 2004 ) . Syngap heterozygous mice also appear normal and survive into adulthood with no apparent alterations in gross brain morphology , but with defects in behavior and synaptic plasticity ( Komiyama et al . , 2002 ) . Syngap+/- neurons have higher average numbers of AMPARs at their synapses than wt ( Kim et al . , 2003; Vazquez et al . , 2004 ) , and , have , on average , larger spine heads ( Vazquez et al . , 2004; Carlisle et al . , 2008 ) . They form excitatory synapses precociously in culture ( Vazquez et al . , 2004 ) and also during postnatal cortical and hippocampal development ( Clement et al . , 2012 ) , leading to an elevated E/I ratio ( excitatory/inhibitory synaptic ratio ) and derangement of critical developmental periods ( Clement et al . , 2013 ) . While all of these influences could contribute to alterations in the composition of PSDs in Syngap+/- animals , several arguments suggest that the increased presence of TARPs , LRRTMs , and neuroligin-2 that we report in PSDs of young animals ( Figure 11 ) is likely a primary , rather than a secondary effect of the mutation . The effect of overexpression of synGAP-α1 in cultured forebrain neurons on development of spines ( Vazquez et al . , 2004 ) and on numbers of synaptic AMPARs ( McMahon et al . , 2012 ) is rapid , measurable within 12–24 hr of transfection and is not caused by mislocalization of the exogenous synGAP- α1 . These effects are predicted by our model and do not depend on the in vivo context of the neurons . They are also not dependent on the GAP activity of synGAP and they require the PDZ ligand . While it is possible that a presently unknown direct effect on protein expression produces these changes , that effect would have to be specific for synGAP-α1 and require its PDZ ligand . No evidence has emerged for a direct effect of the PDZ ligand on protein expression . Thus , the hypothesis we have put forward is the most parsimonious explanation for the specific effects of synGAP-α1 on PSD composition . Similarly , the change in protein composition of PSDs in Syngap+/- animals ( Figure 11 ) cannot be explained by a two-fold reduction in RasGAP and RapGAP activity . SynGAP produces a much larger fold increase in GTPase activity of Rap than of Ras ( Walkup et al . , 2015 and references therein ) . Therefore , loss of half of the GAP activity would be predicted to increase the level of active Rap even more than that of active Ras . Thus , one would expect a relative increase in endocytosis of AMPARs compared to exocytosis ( Zhu et al . , 2002 ) . That would result in fewer AMPAR binding proteins in the PSD; not more , as we observe . For now , application of Occam’s razor strongly favors the hypothesis we have presented . Future experiments will be necessary to determine the quantitative significance of synGAP-α1’s restriction of PDZ domain binding sites at various times during development and during acute stimulation of synapses . Because binding between molecules is driven by their concentrations and also by the inherent affinity between the binding components , we predict that phosphorylation of synGAP-α1 by CaMKII or PLK2 can alleviate the restriction , enabling reconfiguration of the PSD scaffold . Thus , acute phosphorylation of synGAP-α1 by CaMKII following activation of NMDARs during induction of LTP could initiate rearrangement of the composition of PSDs of individual synapses by causing an increase in equilibrium binding in the PSD of AMPARs associated with TARPs and LRRTMs , and of other PSD-95 binding proteins , as they diffuse from perisynaptic locations . It will be important to test this prediction in future experiments . Dephosphorylation of synGAP-α1 after its movement away from PDZ domains might be expected to allow synGAP-α1 to displace the newly added TARPs and LRRTMs and reverse the addition of new AMPARs . Thus , if this model is correct , additional processes occurring later in the consolidation of LTP would be needed to stabilize the newly added AMPARs in the synapse and/or to permanently decrease the number of synGAP-α1 molcules per PSD-95 in potentiated synapses . Such processes could include degradation of the phosphorylated synGAP-α1 and its replacement by newly synthesized alternatively-spliced isoforms that lack the C-terminal PDZ binding domain ( Li et al . , 2001; McMahon et al . , 2012 ) . Several studies have shown that an increase in localization of neuroligin-2 at excitatory synapses is mediated by binding to PSD-95 and increases the ratio of excitatory to inhibitory synaptic contacts ( Prange et al . , 2004; Levinson et al . , 2005; 2010 ) . It seems likely that the restriction by synGAP-α1 of binding of neuroligin-2 to PDZ3 of PSD-95 would be more significant during development , during formation of new synapses , or perhaps in later phases of consolidation of LTP in adults . It is not clear whether a pool of perisynaptic neuroligin-2 exists at adult synapses that could be recruited to new synaptic sites over a few minutes after phosphorylation of synGAP-α1 . Nonetheless , the small but significant increase in neuroligin-2 that we observe in excitatory PSDs from Syngap+/- mice , as well as the increased steady state amounts of TARPs and LRRTMs would increase the overall excitatory/inihibitory ( E/I ) balance of synapses onto neurons in the mutant mice , and perhaps also in humans with SYNGAP1 haploinsufficiency . We note that the reduction in affinity of synGAP-α1 for PDZ domains of PSD-95 after phosphorylation by CaMKII is apparently not sufficient for complete dispersal of synGAP-α1 away from the PSD , although it is likely necessary . Mutated synGAP-α1 missing the PDZ ligand ( synGAPΔSXV ) cannot bind to PDZ domains , yet still localizes to synaptic spines ( Vazquez et al . , 2004 ) . Furthermore , Yang et al . ( 2013 ) showed by immunoelectron microscopy that both α1 and α2 synGAP isoforms localize to the PSD core , and McMahon et al . ( 2012 ) showed by mass spectrometry that α2 isoforms are present in isolated PSDs . These data mean that reduction of affinity of synGAP for PDZ domains may be sufficient to decrease its ability to compete for binding slots on PSD-95; however , complete detachment of synGAP from the PSD in vivo likely requires additional events . The functional significance of our finding that r-synGAP-α1 contains a high affinity binding site for Ca2+/CaM is less clear . We have shown that binding of Ca2+/CaM alters the conformation of the carboxyl terminal regulatory domain of r-synGAP-α1 allowing CDK5 to phosphorylate additional sites; the binding also reduces the affinity of r-synGAP-α1 for PDZ3 by ~25% . However , the consequences of these two effects for synaptic function are not known . Once again , the high copy number of synGAP in the PSD may provide a clue . The biochemical events initiated by Ca2+ flux through NMDARs that lead to changes in synaptic strength ( Sjostrom and Nelson , 2002 ) are initiated by formation of transient and limiting concentrations of Ca2+/CaM in the spine ( Markram et al . , 1998; Pepke et al . , 2010 ) . Approximately ten regulatory enzymes compete for binding of , and activation by , this Ca2+/CaM ( Kennedy , 2013 ) . Because of the abundance of synGAP in the PSD , the high affinity binding site for Ca2+/CaM on synGAP will compete effectively for the newly formed Ca2+/CaM and may act as a Ca2+/CaM buffer . Haploinsufficiency of SYNGAP1 in humans is the cause of ~2–9% of cases of nonsyndromic cognitive disability with co-morbid Autism Spectrum Disorder or Epilepsy ( Berryer et al . , 2013 ) . The reduced amount of synGAP and resulting decrease in its ability to compete for PDZ domains are likely as significant as the reduction in synaptic Ras/Rap GAP activity in the pathology of these disorders . An increase in the E/I balance of synapses onto neurons in the forebrain of affected individuals is predicted by our results with Syngap+/- mice , and could be responsible for the symptoms of cognitive disability , ASD , and/or epilepsy . It would be worth considering whether pharmaceutical agents could be designed that would bind to PDZ domains of PSD-95 ( e . g . Cui et al . , 2007 ) and compensate for the reduced level of synGAP .
Soluble , recombinant synGAP-α1 ( r-synGAP-α1 ) , comprising residues 103–1293 in synGAP A1-α1 ( 118–1308 in synGAP A2-α1 ) , or sr-synGAP , comprising residues 103–725 in synGAP A1-α1 ( 118–740 in synGAP A2-α1 ) , was purified from E . coli as previously described ( Walkup et al . , 2015 ) . The isoform names and residue numbering are taken from ref . ( Walkup et al . , 2015 ) . Henceforth , except where indicated , we use residue numbering corresponding to synGAP A1-α1 . Briefly , a pET-47b ( + ) plasmid ( EMD Millipore , Temecula , CA; catalog no . 71461 ) containing r-synGAP-α1 cDNA ( AF048976 ) fused to an N-terminal 6x Histidine Tag and a PreScission Protease cleavage site was transformed into the Rosetta2 ( DE3 ) E . coli strain ( EMD Millipore , catalog no . 71397 ) for protein expression . Bacterial pellets were harvested by centrifugation and lysed by microfluidization in a ML-110 microfluidizer ( Microfluidics ) . Soluble r-synGAP-α1 was purified on Talon Metal Affinity Resin ( Clontech , Mountain View , CA; catalog no . 635503 ) , and concentrated by ultrafiltration through a 30 kDa cutoff-filter ( Thermo Scientific , Waltham , MA; catalog no . 88531 ) for r-synGAP-α1 or 9 kDa cutoff-filter ( Thermo Scientific , catalog no . 89885A ) for sr-synGAP . Concentrated samples of r-synGAP-α1 were exchanged into storage buffer ( 20 mM Tris , pH 7 . 0; 500 mM NaCl , 10 mM TCEP , 5 mM MgCl2 , 1 mM PMSF , 0 . 2% Tergitol Type NP-40 , and Complete EDTA-free protease inhibitor ) by ultrafiltration , flash-frozen in liquid nitrogen , and stored at -80°C . Sr-synGAP was further purified on a size exclusion column prior to storage ( Walkup et al . , 2015 ) . Soluble recombinant PDZ domains , comprising residues 61–151 ( PDZ1 ) , 155–249 ( PDZ2 ) , 302–402 ( PDZ3 ) , 61–249 ( PDZ12 ) , and 61–403 ( PDZ123 ) from murine PSD-95 ( Q62108 ) were purified from E . coli as previously described ( Walkup and Kennedy , 2014 , 2015 ) with the modifications below . Briefly , pJExpress414 plasmids ( DNA2 . 0 , catalog no . pJ414 ) containing codon optimized PDZ domains were transformed into the BL21 ( DE3 ) E . coli strain ( EMD Millipore , catalog no . 70235–3 ) for protein expression . Single colonies of BL21 ( DE3 ) cells harboring pJExpress414 plasmids were grown overnight at 37°C in lysogeny broth ( LB ) ( Teknova , Hollister , CA; catalog no . L9110 ) supplemented with 100 μg/ml carbenicillin . Overnight cultures were diluted 1:500 into LB medium and grown at 37°C until cultures reached an O . D . 600 of 1 . 0 . IPTG was added to a final concentration of 0 . 2 mM , and cultures were grown for an additional 4 . 5 hr at 37°C . Bacterial pellets were harvested by centrifugation and lysed using non-ionic detergent ( BugBuster , EMD Millipore ) and Ready-Lyse ( Epicentre , Madison , WI ) . Soluble PDZ1 , PDZ2 and PDZ12 domains were purified on GluN2B peptide ( GAGSSIESDV ) PDZ Ligand Affinity Resin ( Walkup and Kennedy , 2014 ) by eluting with 400 µg/ml SIETEV peptide . PDZ3 and PDZ123 were purified on CRIPT peptide ( GAGNYKQTSV ) PDZ Ligand Affinity Resin ( Walkup and Kennedy , 2014 ) by eluting with 400 µg/ml YKQTSV peptide . PDZ domains were concentrated by ultrafiltration through a 3 kDa Amicon Ultracentrifugal Filter Unit ( EMD Millipore , catalog no . UFC 900396 ) . The PDZ peptide ligands were removed from PDZ domains by dialysis into storage buffer ( 50 mM HEPES , pH 7 . 5; 100 mM NaCl , 5 mM TCEP , 1 mM PMSF , and Complete EDTA-free protease inhibitor ) . Purified PDZ domains ( >99% pure; 45–610 µM; Figure 1 ) were flash-frozen in liquid nitrogen , and stored at -80°C . We used SDS-PAGE to determine purity of proteins and to quantify binding to PDZ domain resin . Protein samples were diluted 1:3 into 3x Laemmli buffer ( 100 mM Tris HCl , pH 6 . 8; 2 . 1% SDS , 26% glycerol , 7 . 5% β-mercaptoethanol , and 0 . 01% bromophenol blue ) and heated to 95°C for 3 min before fractionation on 8% SDS-PAGE gels at 165 V in 25 mM Tris base , 192 mM glycine , 0 . 1% SDS . Proteins were stained with Gel Code Blue ( Thermo Scientific , catalog no . 24592 ) , imaged on a Li-Cor Odyssey Classic Infrared Imaging System ( Li-Cor Biosciences , Lincoln , NE ) at 700 nm , and quantified with Licor Image Studio Software ( v4 . 0 . 21 ) against standard curves of BSA ( catalog no . A7517-1VL ) and lysozyme ( catalog no . L4631-1VL ) purchased from Sigma-Aldrich , St . Louis , MO . The protein standards were loaded onto each gel in lanes adjacent to the protein samples . Molecular weights of stained proteins were verified by comparison to Precision Plus Protein All Blue Standards ( BioRad , Irvine , CA; catalog no . 161–0373 ) . For immunoblotting , proteins fractionated by SDS-PAGE were electrically transferred to low fluorescence PVDF membranes ( Thermo Scientific , catalog no . 22860 ) in 25 mM Tris , 200 mM glycine , and 20% methanol . Membranes were washed with 50 mM Tris-HCl , pH 7 . 6; 150 mM NaCl ( TBS ) followed by blocking with Odyssey Blocking Buffer ( Li-Cor Biosciences , catalog no . 927–50000 ) . Membranes were washed in TBS supplemented with 0 . 1% Tween 20 ( TBS-T ) before incubation in Odyssey Blocking Buffer containing 1:1000 diluted rabbit anti-synGAP ( ThermoFisher Pierce , Waltham , MA catalog no . PA1-046 ) or 1:1500 BSA-free anti-TetraHis ( Qiagen , Hilden , Germany , catalog no . 34670 ) . Bound antibodies were detected with 1:10 , 000 goat anti-mouse Alexa-Fluor 680 ( Life Technologies , catalog no . A-21057 ) or 1:10 , 000 goat anti-rabbit Alexa-Fluor 680 ( Life Technologies , Carlsbad , CA; catalog no . A-21109 ) visualized with a Li-Cor Odyssey Classic Infrared Imaging System and quantified with Li-Cor Image Studio Software . PDZ domain affinity resins ( PDZ1 , residues 61–151; PDZ2 , residues 155–249; PDZ3 , residues 302–402; PDZ12 , residues 61–249; PDZ123 , residues 61–402 from murine PSD-95 ) were prepared by the HaloTag-HaloLink method as previously described ( Walkup and Kennedy , 2014 , 2015 ) . Briefly , bacterial cell pellets containing PDZ domain-HaloTag fusion proteins were resuspended in 10 ml/g of Purification Buffer , and lysed by three passes through a ML-110 microfluidizer . The lysate was clarified by centrifugation , added to HaloLink resin ( Promega , Madison , WI; catalog no . G1915 ) , and mixed with continuous agitation for 1 . 5 hr at 4°C on an end-over-end mixer . Unbound protein was separated from the resin by centrifugation and the PDZ-HaloTag-HaloLink resin was resuspended , transferred to a column , and allowed to settle . The resin was extensively washed and then stored at 4°C in a buffer supplemented with 0 . 05% NaN3 . The resin was used or discarded within 1 week of preparation . The densities of PDZ domains on the resin varied from 50 to 100 pmol of PDZ123 domain and from 200 to 500 pmol of PDZ1 , PDZ2 , PDZ3 , or PDZ12 domains per µl resin . Phosphorylated or nonphosphorylated r-synGAP-α1 ( 500 nM , 200 µl ) was mixed with 20 µl of Affinity Resin containing PDZ1 , PDZ2 , PDZ3 , PDZ12 , or PDZ123 domains , pre-equilibrated with Binding/Wash Buffer ( 25 mM Tris , pH 7 . 0; 150 mM NaCl , 1 mM MgCl2 , 0 . 5 mM TCEP , 0 . 2% Tergitol , 0 . 5 mM EDTA ) in a cellulose acetate spin cup ( ThermoFisher Pierce , catalog no . 69702 ) for 60 min on an end-over-end mixer . In some experiments , 2 . 5 µM CaM and 0 . 5 mM CaCl2 were included to test the effect of binding of Ca2+/CaM to r-synGAP-α1 on synGAP’s affinity for PDZ domains . After the incubation , the resin in the spin cup was centrifuged for 2 min at 1500 x g to remove unbound protein , and the resin was washed 4 times with 200 µl of Binding/Wash Buffer . To elute bound protein , 100 µl of 1x Laemmli Buffer was added and the resin was incubated for 5 min at room temperature . The eluted protein was collected by centrifugation at 6000 x g for 2 min , fractionated by SDS-PAGE , stained with Gel Code Blue , and quantified on a Li-Cor Classic as described above . Integrated intensities reflecting the amount of bound r-synGAP-α1 were determined with Li-Cor software and plotted with Prism ( v6 . 0d , GraphPad Software , La Jolla CA ) . There is no detectable non-specific binding of r-synGAP-α1 to unsubstituted resin or of proteins lacking PDZ domain ligands to the affinity resins ( Walkup and Kennedy , 2014 ) . Phosphorylation of r-synGAP-α1 by CaMKII , CDK5 , and PLK2 was carried out immediately prior to PDZ binding assays , as previously described ( Walkup et al . , 2015 ) . Reaction mixtures contained 50 mM Tris-HCl , pH 8 . 0; 10 mM MgCl2 , 0 or 0 . 7 mM CaCl2 , 0 . 4 mM EGTA , 30 μM ATP , 0 or 3 . 38 μM calmodulin , 10 mM DTT , 725 nM r-synGAP-α1 . Reactions contained no kinase , 10 nM CaMKII , 230 nM CDK5/p35 ( EMD Millipore , catalog no . 14-477M ) or 230 nM PLK2 ( Life Technologies , catalog no . PV4204 ) . Mixtures for CDK5 and PLK2 did not contain CaCl2 or CaM . Samples were quenched by addition of 1/3 volume of 50 mM Tris , pH 8 . 0; 0 . 4 M NaCl , 10 mM MgCl2 , 0 . 8% tergitol ( Type NP-40 ) , 6 µM autocamtide-2 related inhibitory peptide ( Genscript , Piscataway , NJ; catalog no . RP10271 ) , 90 µM roscovitine ( Sigma-Aldrich ) and 40 mM EGTA at the indicated times . When we planned to add Ca2+/CaM during the subsequent incubation with resin , the EGTA was omitted . Samples were stored on ice until their use in PDZ domain binding assays . Phosphorylation of 725 nM r-synGAP-α1 by 10 nM CaMKII was carried out as previously described ( Walkup et al . , 2015 ) in reaction mixtures containing 50 mM Tris-HCl , pH 8 . 0 , 10 mM MgCl2 , 0 or 0 . 7 mM CaCl2 , 0 . 4 mM EGTA , 500 μM [γ-32P]-ATP ( 375 cpm/pmol ) ( 6000 Ci/mmol , Perkin Elmer , Waltham , MA; catalog no . BLU002Z/NEG002Z ) , 0 or 3 . 4 μM calmodulin , 10 mM DTT . Phosphorylation of 286 nM r-synGAP-α1 and 4 . 3 μM Histone H1 ( New England Biolabs , Ipswich , MA; catalog no . M2501S ) , by CDK5/p35 ( EMD Millipore , catalog no . 14-477M ) or CDK5/p25 ( EMD Millipore , catalog no . 14–516 ) was carried out in the same reaction mixture containing 110 nM CDK5/p35 or CDK5/p25 but no CaMKII . After fractionation by SDS-PAGE , phosphorylated proteins were quantified with a Typhoon LA 9000 phosphorimager ( GE Healthcare Life Sciences , Pittsburgh , PA ) as previously described ( Walkup et al . , 2015 ) . Relative densities were converted to pmol phosphate by comparison to densities of standard amounts of [γ-32P]-ATP . The stoichiometry of phosphorylation was calculated by dividing mol of incorporated phosphate by mol of r-synGAP-α1 loaded per lane . We used a 'competition in solution' method ( also called 'affinity in solution' ) ( Nieba et al . , 1996; Lazar et al . , 2006; Abdiche et al . , 2008 ) to measure the affinity of r-synGAP-α1 for PDZ domains . In this method , PDZ domains are immobilized on the chip surface and used to capture and measure the concentration of free r-synGAP-α1 in pre-equilibrated mixtures of a constant amount of r-synGAP-α1 with varying amounts of soluble recombinant PDZ domains . Experiments were performed on a Biacore T200 ( GE Healthcare Life Sciences ) . Purified PDZ domains ( PDZ1 , PDZ2 , PDZ3 , PDZ12 , PDZ123 from PSD-95 ) were coupled to Series S CM5 Sensor Chips ( GE Healthcare Life Sciences , catalog no . BR-1005-30 ) by the amine coupling protocol specified in the Biacore T200 Control Software with reagents purchased from GE Healthcare Life Sciences . Sensor surfaces were activated by applying a 1:1 mixture of 50 mM N-hydroxysuccinimide ( NHS ) : 200 mM 1-ethyl-3- ( 3-dimethylaminopropyl ) carbodiimide hydrochloride ( EDC ) provided in the Biacore Amine Coupling Kit ( GE Healthcare Life Sciences , catalog no . BR-1000-50 ) dissolved in HBS-N running buffer ( degassed 0 . 01 M HEPES pH 7 . 4 , 0 . 15 M NaCl ) ( GE Healthcare Life Sciences , catalog no . BR-1006-70 ) . PDZ domains were diluted to 0 . 1–5 µM in Biacore Sodium Acetate Buffer [10 mM sodium acetate , pH 4 . 0 ( GE Healthcare Life Sciences , catalog no . BR-1003-49 ) for PDZ1 and PDZ3; pH 4 . 5 for PDZ2; pH 5 for PDZ12 and PDZ123] . PDZ domains were injected into flow cells 2 and 4 until 200 to 400 RU ( resonance units ) of PDZ domain were immobilized . Flow cells 1 and 3 were left blank to be used as reference surfaces . Ethanolamine ( 1 M , pH 8 . 5 ) was injected for 7 min at 10 μl/min to block remaining active sites on all four flow cells . A calibration curve was prepared by applying samples of 0 to 50 nM r-synGAP-α1 prepared by two-fold serial dilution of 50 nM r-synGAP-α1 into 1x HBS-EP+ buffer ( degassed 0 . 01 M HEPES , pH 7 . 4; 0 . 15 M NaCl , 3 mM EDTA , 0 . 005% v/v Surfactant P20; GE Healthcare Life Sciences , catalog no . BR-1006-69 ) to the chip and recording the maximum RU for each concentration . Samples for calibration were incubated for 2 hr at room temperature before randomized injection onto the chip surface at 25°C at 10 μl/min for 200 s over all four flow cells . Between each sample injection , the chip was regenerated by injecting 1 M MgCl2 at 100 μl/min for 60 s , waiting 180 s for the baseline to stabilize , then injecting a second pulse of MgCl2 solution , waiting 300 s for the baseline to stabilize , and finally executing a 'carry over control injection' in which HBS-EP+ buffer is flowed over the chip surface at 40 µl/min for 30 s . Mixtures of r-synGAP-α1 and PDZ domains were prepared by 1:1 dilution of 50 nM r-synGAP-α1 with two-fold serial dilutions of PDZ domains ( 0–20 µM PDZ1 , PDZ2 , PDZ3 , PDZ12 or 0–160 µM PDZ123 ) in HBS-EP+ buffer to produce mixtures containing 25 nM r-synGAP-α1 and 0–10 µM PDZ1 , PDZ2 , PDZ3 , PDZ12 or 0–80 µM PDZ123 . For each mixture of r-synGAP-α1 and PDZ domain , the different concentrations were injected randomly and a series of sensorgrams were recorded as described for the calibration curve . Sensorgrams were processed with Biacore T200 Evaluation Software , ( ver . 3 . 0 , GE Healthcare Life Sciences ) . The y-axes were zeroed at the baseline for each cycle and x-axes were aligned at the injection start . Bulk refractive index changes and systematic deviations in sensorgrams were removed by subtracting the responses in reference flow cells corresponding to the sample flow cells ( e . g . 2–1 , 4–3 ) . The averaged sensorgrams for 0 nM r-synGAP-α1 were then subtracted from sensorgrams for all other concentrations . The concentrations of free r-synGAP-α1 in each mixture with PDZ domains was determined from the calibration curve , exported into Prism , and plotted against the log of PDZ domain concentration . The curve was fit to the equation:synGAPfree= ( synGAPtot−PDZtot−KD ) 2± ( PDZtot+synGAPtot+KD ) 24− ( PDZtot×synGAPtot ) This equation ( from the Biacore T200 Software Handbook ) assumes the existence of a single binding site between PDZ domain and r-synGAP-α1 . Equilibrium dissociation constants ( KD ) for binding to PDZ1 , PDZ2 , and PDZ3 were determined from the best fit curves as described in the Biacore T200 Software Handbook ( p . 210 ) . The experimental curves deviated slightly from the equation at higher concentrations of soluble PDZ domain because of a low affinity association between PDZ domains bound to the chip and soluble free PDZ domains . The deviation is most obvious for phosphorylated r-synGAP-α1 which has a relatively low affinity for PDZ123 . We determined that these deviations had only a small effect ( <~5% ) on the calculated KD’s by comparing KD’s after fitting the curves including or excluding points for high concentrations of PDZs at which the r-synGAP-α1 concentration appeared to plateau above zero . PDZ12 and PDZ123 contain more than a single PDZ domain binding site . It is not possible to derive a unique equation incorporating two or three binding sites when the affinities of the multiple sites are similar . Therefore , single 'apparent' equilibrium dissociation constants ( KDapp ) were determined by obtaining the best fit of the data with the equation for a single binding site . Rosetta2 ( DE3 ) cells containing sr-synGAP or r-synGAP-α1 were lysed in Lysis buffer as described in Walkup et al . ( 2015 ) except that the buffer contained 200 mM NaCl , 0 or 5 mM CaCl2 , and 0 or 10 mM EGTA . The resuspended cells were lysed by sonication with a Digital Sonifier 450 Cell Disruptor ( Branson , Wilmington , NC ) for two passes at 90 s/pass ( 15% power , 1 . 0 s on , 1 . 5 s off ) , and insoluble material was removed by centrifugation at 16 , 000 × g for 40 min at 4°C . Clarified cell lysate ( 1 . 7 ml ) containing sr-synGAP or r-synGAP-α1 ( ~6 mg/ml total protein ) was incubated with end-over-end mixing for 60 min with 0 . 3 ml CaM-Sepharose 4B ( GE Healthcare Life Sciences , catalog no . 17-0529-01 ) or control Sepharose 4B ( GE Healthcare Life Sciences , catalog no . 17-0120-01 ) . The resin was pipetted into a BioSpin column ( Bio-Rad , catalog no . 732–6008 ) and washed with 12 ml ( 40 column volumes ) of Lysis/Wash Buffer . Bound protein was eluted with 1 . 2 ml ( 4 column volumes ) of Lysis/Wash Buffer containing 100 mM EGTA . Eluted proteins ( 30 µl aliquot ) were resolved by SDS-PAGE and transferred to a PVDF membrane which was probed with anti-synGAP or BSA-free anti-TetraHis as described above . Direct binding of r-synGAP-α1 to Ca2+/CaM immobilized on a chip was assayed on a Biacore T200 with a Series S Sensor Chip CM5 . CaM ( Enzo Life Sciences , Farmingdale , NY; catalog no . BML-SE325-0001 ) was coupled to the chip by the amine coupling protocol specified in the Biacore T200 Control Software , as described above . Purified , lyophilized CaM ( 250 μg ) was resuspended in water and exchanged into Biacore Sodium Acetate , pH 4 . 0 Buffer ( 10 mM sodium acetate , pH 4 . 0 ) with an Amicon Ultra-0 . 5 ml centrifugal filter with a 3 kDa molecular weight cutoff ( EMD Millipore , catalog no . UFC500396 ) . CaM was further diluted to 0 . 5 nM in 10 mM Sodium Acetate , pH 4 . 0 , and injected into flow cells 2 and 4 until 50 RU of CaM were immobilized ( ~7 min at a flow rate of 10 μl/min ) . Flow cells 1 and 3 were left blank to be used as reference surfaces . Ethanolamine ( 1 M , pH 8 . 5 ) was injected for 7 min at 10 μl/min to block remaining active sites on all four flow cells . R-synGAP-α1 ( 0 nM to 75 nM ) in 1x HBS-EP+ running buffer supplemented with 10 mM CaCl2 , was injected in triplicate at 25°C at 100 μl/min for 75 s over all four flow cells . Different concentrations of r-synGAP-α1 were applied in randomized order . After injection ended , dissociation was monitored in each flow cell for 500 s . Regeneration of the sensor chip was performed by injecting 50 mM NaOH ( GE Healthcare Life Sciences , catalog no . BR-1003-58 ) at 100 μl/min for 30 s , waiting 180 s for the baseline to stabilize , then injecting a second pulse of NaOH , waiting 240 s for the baseline to stabilize , and finally executing a 'carry over control injection . ' Sensorgrams were processed using the Biacore T200 Evaluation Software , version 3 . 0 , as described above . Resonance units of bound r-synGAP-α1 at equilibrium were exported into Prism and plotted against the concentrations of r-synGAP-α1 . The data were fit globally to a hyperbolic curve by nonlinear regression to determine equilibrium dissociation constants ( KD ) . Purified sr-synGAP and r-synGAP-α1 were diluted to 1 to 500 nM in Binding/Wash Buffer ( 50 mM Tris , pH 7 . 5; 200 mM NaCl , 5 mM TCEP , 2 mM CaCl2 ) . Aliquots of diluted r-synGAP-α1 ( 300 µl ) were incubated with end-over-end mixing for 60 min with 50 μl of CaM-Sepharose 4B in a screw cap spin column ( Thermo Scientific , catalog no . 69705 ) . Concentrations of sr-synGAP and r-synGAP-α1 were 20–3 , 000 fold below the ligand binding capacity of the CaM-Sepharose resin . Unbound protein was removed by centrifugation at 4000 x g for 30 s . Beads were washed in Binding/Wash Buffer ( 250 μl , 5 volumes ) , and bound protein was eluted with 50 μl of 1x Laemmli buffer with 10% β-mercaptoethanol . Eluted proteins were fractionated by SDS-PAGE and transferred to a PVDF membrane as described above . Blots were probed with 1:1000 anti-synGAP or 1:1500 BSA-free anti-TetraHis anti-bodies and quantified on a Li-Cor Classic as described above . Integrated intensities reflecting the amount of bound r-synGAP-α1 were determined with Li-Cor software and plotted against the corresponding concentrations of r-synGAP-α1 with Prism software . The data were fit to a hyperbolic curve by nonlinear regression to determine the dissociation constant ( KD ) . Syngap+/- mice were bred from a knockout strain created in our lab ( Vazquez et al . , 2004 ) . The mutant strain has been deposited in the Mutant Mouse Resource & Research Center at University of California , Davis , listed as MMRRC:037374-UCD . All procedures were approved by the Caltech Institute Animal Care and Use Committee . PSD fractions were prepared as previously described ( Cho et al . , 1992 ) from six wt and six Syngap+/- mouse litter mates matched by age ( 7–12 weeks ) , and sex ( wt , 1 female , 5 males; Syngap+/- , 2 female , 4 male ) . The mice were killed by cervical dislocation and forebrains were dissected and rinsed in Buffer A ( 0 . 32 M sucrose , 1 mM NaHCO3 , 1 mM MgCl2 , 0 . 5 mM CaCl2 , 0 . 1 mM PMSF , 1 mg/l leupeptin ) . Forebrains from each set of six mice were pooled and homogenized with 12 up and down strokes at 900 rpm in 14 ml Buffer A . Homogenates were diluted to 35 ml in Buffer A and centrifuged at 1400 × g for 10 min . The pellet was resuspended in 35 ml Buffer A , homogenized ( 3 strokes ) , and centrifuged at 710 g for 10 min . Supernatants from the two centrifugations were combined and centrifuged at 13 , 800 g for 10 min . The pellet was resuspended in 8 ml of Buffer B ( 0 . 32 M sucrose , 1 mM NaH2CO3 ) , homogenized with 6 strokes and layered onto a sucrose gradient ( 10 ml each of 0 . 85 M , 1 . 0 M , and 1 . 2 M sucrose in 1 mM NaH2CO3 buffer ) . The gradient was centrifuged for 2 hr at 82 , 500 g in a swinging bucket rotor . The synaptosome-enriched layer at the interface of 1 . 0 and 1 . 2 M sucrose was collected , diluted to 15 ml with Solution B and added to an equal volume of Buffer B containing 1% Triton . The mixture was stirred for 15 min at 4°C and centrifuged for 45 min at 36 , 800 g . The pellet containing the PSD-enriched , Triton-insoluble fraction was resuspended in 800–1000 μl of 40 mM Tris pH 8 with a 21 gauge needle and 1 ml syringe , and further solubilized by hand in a teflon glass homogenizer . Samples were aliquoted and stored at −80°C . Equal amounts of protein from each PSD sample ( 5–15 μg ) were dissolved in SDS-PAGE sample buffer ( 33 mM Tris HCl , pH 6 . 8; 0 . 7% SDS , 10% glycerol , 2 . 5% β-mercaptoethanol , and 0 . 003% bromophenol blue ) , heated at 90°C for 5 min , fractionated on polyacrylamide gels ( 8% or 10% ) , and electrically transferred to PVDF membranes in 25 mM Tris , 150 mM glycine , 2% methanol at 250V for 2 . 5 hr at 4°C , as described above . Membranes were blocked with Odyssey blocking buffer ( Licor Biosciences ) and then incubated in a primary antibody solution of 5% BSA in TBS-T overnight at 4C , as described above . Primary antibodies included mouse-anti-PSD-95 ( ThermoFisher , catalog no . MA1-046 [clone 7E3-1B8] , RRID AB_2092361 , dilution 1:10 , 000 ) , rabbit-anti-SynGAP ( Pierce , catalog no . PA1-046 , RRID AB_2287112 , dilution 1:3500 ) , rabbit-anti-TARP ( ɣ-2 , 3 , 4 , and 8; EDM Millipore , catalog no . Ab9876 , RRID AB_877307 , dilution 1:300 ) , rabbit-anti-LRRTM2 ( Pierce , catalog no . PA521097 , RRID AB_11153649 , dilution 1:3000 ) , mouse-anti-neuroligin-1 ( Sigma , catalog no . sab5201464 , RRID AB_2570548 , dilution 1:250 ) , and rabbit-anti-neuroligin-2 ( Synaptic Systems , Gottingen , Germany , catalog no . 129202 , RRID AB_993011 , dilution 1:1000 ) . The membranes were then washed 3-times in TBS-T . The membrane was incubated with secondary antibodies ( Alexa Fluor 680-goat-anti-mouse IgG ( Life Technologies , catalog no . A21057; 1:10 , 000 ) or IRDye800-goat-anti-rabbit IgG ( Rockland , Limerick , PA; catalog no . 611-132-122; 1:10 , 000 ) for 45 min at room temperature in 5% nonfat milk in TBS-T , then washed 3 times in TBS-T , then twice in TBS prior to scanning . For most experiments , each blot contained 6 duplicate samples of PSD fractions from wt and the same number from Syngap+/- mice . Each blot was incubated with a mixture of two primary antibodies; mouse-anti-PSD-95 and the antibody against neuroligin-1 , neuroligin-2 , TARPs , LRRTM2 , or synGAP . Then the blots were incubated with a mixture of the appropriate secondary antibodies . For measurement of neuroligin-1 , both PSD-95 and neuroligin-1 were detected by the same goat anti-mouse secondary; the bands were physically separated on the gel and were quantified independently . Bound antibodies were visualized in the appropriate fluorescent channels with an Odyssey Classic Infared Imaging System ( Li-Cor Biosciences ) . Before running samples for quantification , we determined the amount of PSD sample that would result in signals for each protein that were strong enough for measurement and not saturated . To quantify the densities of the bands , each visual image was first set to high brightness in order to capture the boundaries of the signals for each band . The images were then used as a template in Li-Cor software to draw rectangular regions of interest around protein bands , and around identically sized background regions in the same lane . Background densities were subtracted from each protein signal . The digital data read by the Li-Cor software is unchanged by the visualization settings and is linear over several orders of magnitude . For each lane , the ratio of the integrated density of each of the five proteins to the integrated density of PSD-95 was calculated . For three gels , one outlier measurement ( defined as greater than 2 times the standard deviation of the mean [S . E . ] ) was excluded from the calculation . The mean and S . E of the ratios were determined for wt and Syngap+/- PSD fractions . The means were compared by t-tests performed with Prism software as indicated in the legend of Figure 11 . The effect size was measured by Cohen’s d ( ratio of the difference in the mean to the pooled standard deviations of the measurements ) . Cohen’s d between 0 . 5 and 0 . 8 is generally considered a 'medium' effect size , greater than 0 . 8 is generally considered a large effect size . For four of the proteins , the number of measurements was sufficient to determine a significant difference between wt and Syngap+/- . In the case of neuroligin-1 , the means were identical after 24 individual measurements of each sample . | The formation of memories is believed to depend on the strengthening of connections , called synapses , between neurons in the brain . When neurons are activated together , their synaptic connections become permanently strengthened to record the memory . This strengthening is called activity-dependent long-term potentiation . As long-term potentiation develops , more protein receptors are added to the receiving side of the synapse . This allows the receiving neuron to produce a larger electrical response to the signaling chemicals it receives from the neuron on the sending side of the synapse . The addition of receptors is regulated by a set of enzymes held near the membrane of the synapse by a protein scaffold known as the postsynaptic density . A major scaffold protein called PSD-95 contains binding sites , known as PDZ domains , that hold protein receptors and regulatory enzymes in place . One regulatory enzyme called synGAP is present in large numbers in the postsynaptic density and binds to the same PDZ domains as the receptors . Humans that have just one copy ( instead of the usual two ) of the gene that encodes synGAP have cognitive disabilities that are often accompanied by autism and epilepsy . By studying purified proteins , Walkup et al . found that adding phosphate groups to synGAP reduces the enzyme’s ability to bind to the PDZ domains . This reduced binding ability could make more PDZ domains available to bind to protein receptors and hold them at the synapse . To measure the effect of reduced synGAP levels on the proteins found at postsynaptic densities , Walkup et al . used mice that had just one copy of the synGAP gene in their neurons . These mice have less synGAP in their postsynaptic densities and more of three proteins that bind to PDZ domains . These proteins hold receptors in the synapse and help synapses to form . Thus , synGAP may restrict the binding of other proteins to the PDZ domains in order to regulate the strength of the synapse . Further experiments are now needed to investigate the importance of restriction by synGAP of binding to PDZ domains under a variety of circumstances in which the activity of neurons alters the strength of synapses . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"neuroscience"
] | 2016 | A model for regulation by SynGAP-α1 of binding of synaptic proteins to PDZ-domain 'Slots' in the postsynaptic density |
In eukaryotes , intra-chromosomal recombination generates DNA circles , but little is known about how cells react to them . In yeast , partitioning of such circles to the mother cell at mitosis ensures their loss from the population but promotes replicative ageing . Nevertheless , the mechanisms of partitioning are debated . In this study , we show that the SAGA complex mediates the interaction of non-chromosomal DNA circles with nuclear pore complexes ( NPCs ) and thereby promotes their confinement in the mother cell . Reciprocally , this causes retention and accumulation of NPCs , which affects the organization of ageing nuclei . Thus , SAGA prevents the spreading of DNA circles by linking them to NPCs , but unavoidably causes accumulation of circles and NPCs in the mother cell , and thereby promotes ageing . Together , our data provide a unifying model for the asymmetric segregation of DNA circles and how age affects nuclear organization .
Homologous recombination plays an important role in DNA repair in all organisms and acts through the formation of a Holliday junction between the damaged molecule and its repair template . When recombination involves two homologous chromosomes , the resolution of the Holliday junction leads to a cross-over , the exchange of arms between two chromosomes , in 50% of the cases ( Heyer , 2004 ) . However , when recombination occurs between two loci on the same chromosome , i . e . if the break affects a repeated region of the genome , recombination is intra-chromosomal . In 50% of these events , the resolution of the Holliday junction leads to the excision of a DNA circle containing one or several of the involved repeats ( Gaubatz , 1990 ) . Accordingly , due to the highly repeated nature of the ribosomal DNA ( rDNA ) locus , rDNA circles have been found in all eukaryotes tested so far ( Cohen et al . , 2010 ) . Similarly , double-minute , satellite- and t-circles , as well as a multitude of micro DNAs have been identified in animal cells ( Cohen and Segal , 2009; Shibata et al . , 2012 ) . However , how cells react to these molecules is not understood . Particularly , it is unclear whether these cells recognize , sort , and eliminate these molecules or let them passively disappear by dilution . In Saccharomyces cerevisiae , extrachromosomal rDNA circles ( ERCs ) form spontaneously but fail to accumulate in the population , owing to their highly asymmetric segregation upon cell division ( Sinclair and Guarente , 1997 ) . Indeed , the asymmetry of the budding process generates a mother and a daughter cell with distinct sizes and fate . While most daughter cells are born young , i . e . with a full replicative potential , mother cells age with each budding cycle and therefore generate only a limited number of daughter cells ( ( Mortimer and Johnston , 1959 ) and reviewed in ( Denoth Lippuner et al . , 2014 ) ) . Since rDNA circles contain a replication origin they duplicate in S-phase , and due to the asymmetric segregation they accumulate with time in the mother cell . Several arguments indicate that they contribute to replicative ageing of these mother cells . Artificially introducing an ERC into the bud reduces its life span ( Sinclair and Guarente , 1997 ) . Likewise , mutations that enhance ERC formation shorten the longevity of mother cells ( Kaeberlein et al . , 1999 ) . Conversely , mutant cells with reduced rates of ERC formation , such as fob1∆ , or those defective in ERC retention , such as bud6∆ mutant cells , live longer ( Defossez et al . , 1999; Shcheprova et al . , 2008 ) . Interestingly , any artificial circle that replicates in vivo but lacks a partitioning sequence ( e . g . a centromere ) segregates similarly and promotes replicative ageing when introduced into yeast ( Falcon and Aris , 2003 ) . Thus , non-chromosomal DNA circles segregate asymmetrically , affect cellular physiology , and promote ageing in a manner that is independent of their DNA sequence . How these DNA circles contribute to ageing is not known . One approach to characterize the effects of DNA circles on cellular physiology is to study the mechanism of their segregation . Two models have been proposed to explain the retention of circular DNA molecules in the mother cells ( Ouellet and Barral , 2012 ) . The morpho-kinetic model proposes that circles freely diffuse in the nucleus and that their retention results from the morphology of the dividing yeast nucleus and the short duration of anaphase , which together limit the probability that DNA circles diffuse into the bud ( Gehlen et al . , 2011 ) . Using measured parameters for nuclear geometry and division speed , this model predicts a retention frequency of 0 . 75–0 . 90 per individual plasmid . However , mathematical modeling indicates that observed ageing curves require retention frequencies above 0 . 99 per individual ERC ( Gillespie et al . , 2004 ) , which is higher than what the morpho-kinetic model can achieve . A second model , the barrier model , is based on the observation that a lateral diffusion barrier in the outer membrane of the nuclear envelope impedes the diffusion of membrane proteins through the bud neck and hence their exchange between mother and bud parts of the nucleus ( Shcheprova et al . , 2008; Boettcher et al . , 2012; Clay et al . , 2014 ) . This model proposes that DNA circles attach to a receptor in the nuclear envelope to ensure their subsequent confinement into the mother cell by the lateral diffusion barrier ( Shcheprova et al . , 2008; Clay et al . , 2014 ) . The main difference between these models is whether confinement of the circle within the mother cell is purely passive or relies on mechanisms that are able to distinguish non-chromosomal DNA circles from bona-fide chromosomes to promote their specific anchorage and asymmetric segregation . However , no such mechanism is known yet . Whether DNA circles passively diffuse or are recognized by the cell would be predicted to have distinct consequences on the localization of the circles and their effects on nuclear organization . A passive model predicts that DNA circles do not interact specifically with any nuclear structure . Therefore , their accumulation should have little impact on nuclear organization . On the other hand , if cells recognize DNA circles , accumulating circles would increasingly interact with the corresponding structure and should progressively affect its size and organization . Thus , in order to better understand whether and how DNA circles are recognized by the cell , and to shed light on how they interfere with cellular physiology , we investigated how accumulating DNA circles localize and whether they affect nuclear organization .
To investigate the localization and effects of non-centromeric DNA circles on nuclear organization , we used the plasmid pPCM14 ( Figure 1A ) , containing a replication origin ( ARS1 ) and 224 repeats of the TetO sequence ( Megee and Koshland , 1999 ) . In cells expressing a TetR-mCherry fusion protein , which binds the TetO sequence , the plasmid is observed as a focus of red fluorescence . Additionally , the plasmid contains an excisable centromere , leading to the formation of a labeled non-centromeric DNA circle upon expression of the R-recombinase . The plasmid also contains two auxotrophic selection markers: URA3 located between the two recombination-sites to select against accidental centromere excision and LEU2 on the residual backbone , allowing selection for the plasmid after centromere excision . In budding yeast , all centromeres co-localize with spindle pole bodies ( SPBs ) throughout the cell cycle ( Goshima and Yanagida , 2000 ) . Accordingly , we observe the centromeric plasmid in close proximity to the SPBs ( Figure 1A ) . 3 hr after addition of estradiol to induce centromere excision , plasmids are localized away from the SPBs in 68% of the cells . Most of those cells displayed one or two small plasmid foci ( 61% or 33% , respectively; Figure 1A ) and only 6% of the cells showed more than two foci . These foci were on average 2 . 5 times more intense than the nucleoplasmic background ( Figure 1B ) and localized to the vicinity of the nuclear rim ( see below , and Figure 5 ) . 10 . 7554/eLife . 03790 . 003Figure 1 . Accumulation of DNA circles induces the formation of an NPC cap . ( A ) Schematic overview of the plasmid pPCM14 and fluorescent images of pPCM14 before and 3 or 18 hr after excision of the centromere by addition of estradiol ( ED ) in cells expressing TetR-mCherry , Nup82-3sfGFP , and Spc42-GFP . ( B ) Quantifications of the TetR-mCherry intensity in the circle area ( Ic ) and the residual nuclear area ( Ir ) 3 or 18 hr after centromere excision ( all box plots represent minimum to maximum , line represents median , N ≥ 30 cells ) . ( C ) Fluorescent images of the nuclear proteins Fob1 , Spc42 , and Nup82 ( green ) in cells with or without accumulated plasmids ( red , 18 hr after addition of ED ) . Quantifications of total fluorescence of different nuclear pore markers in cells with or without accumulated plasmids , normalized by the median of cells without plasmids ( N > 30 cells ) . ( D ) Quantifications of fluorescence intensity of Nup82 in the vicinity of the DNA circle ( Ic ) and the rest of the nucleus ( Ir ) , normalized by the median Ir ( N = 50 cells ) . ( E ) Time lapse images of a nucleus with accumulated plasmids ( red ) and the NPC cap ( green ) . Images were taken every 5 min . ( A-E ) ***p < 0 . 001; images are max Z-projections; arrow depicts the NPC cap , arrow head the accumulated circles; scale bars always represent 2 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 003 In contrast , when the cells were grown in selective media for 16–18 hr , allowing the circles to accumulate through multiple rounds of replication and division , the majority of the cells ( 93% ) contained no plasmid , consistent with the plasmids segregating asymmetrically and accumulating in only few retaining cells . A small fraction of the cells ( 1% ) contained few plasmids visualized as small foci as above , whereas the rest of the cells ( 6% ) carried one large and intense signal , 8-fold above background ( Figure 1B ) . Consistent with the idea that this fluorescent mass corresponded to the accumulated plasmids , these cells were the few initial mother cells that had already proceeded through several division cycles , as determined by the visualization of bud scars with calcofluor white ( 5 . 2 ± 1 . 9 bud scars on cells with fluorescent mass compared to 0 . 7 ± 0 . 9 bud scars in cells without plasmids ) . After five rounds of duplication , cells contain approximately 50–100 plasmids ( Hyman et al . , 1982 ) . Remarkably , these accumulated plasmids did not spread throughout the nucleoplasm but rather were collected to a single place at the nuclear periphery . In order to characterize this structure in more detail , the accumulated plasmids were visualized in wild-type cells co-expressing TetR-mCherry with different nuclear markers fused to GFP . The nucleoporin Nup82 tagged with three copies of a super-folder GFP ( Nup82-3sfGFP ) was used to visualize the morphology of the nuclear envelope; Spc42-GFP indicated the position of the SPB and Fob1-GFP labeled the nucleolus . In most cases , the accumulated DNA circles localized adjacent to , but were excluded from , the nucleolus ( Fob1-GFP; Figure 1C ) . Accordingly , they localized opposite to the SPB . Labeling of the nuclear pore complexes ( NPCs ) indicated that the nuclei loaded with DNA circles were larger than those with no or few circles ( average volume increased 1 . 7-fold; N = 50 cells ) . Remarkably , accumulation of DNA circles drastically affected the NPC distribution . The circle-loaded cells contained on an average 2 . 7-fold more nuclear pores compared to cells without plasmids ( Figure 1C ) . This increase was independent of the NPC reporter used ( Nup82-3sfGFP , 2 . 7-folds , N = 50 cells; Nup82-GFP , 2 . 6-fold , N = 25 cells; Nup49-GFP , 3 . 0-fold , N = 25 cells , p < 0 . 001; Figure 1C ) and was exclusively observed in the cells that contained DNA circles . Five generations old wild-type cells lacking plasmids did not show such changes ( see Figure 9 ) . Most strikingly , NPC distribution over the nuclear envelope was no-longer uniform; a substantial fraction of the NPCs accumulated in a single cap that covered one side of the nucleus ( Figure 1A , C ) . In all cells , this cap covered the DNA circles . Furthermore , all circle-loaded cells exhibited such a cap . As a consequence , the fluorescence intensity of Nup82-3sfGFP was on an average two-fold higher in the vicinity of the circles ( Ic = 2 . 02 ± 0 . 80 A . U . ) compared with the rest of the nucleus ( Ir = 1 . 0 ± 0 . 35 A . U . , N = 50 cells , p < 0 . 001; Figure 1D ) . In time-lapse movies ( 5 min intervals for 60 min ) , the cap and the mCherry patch moved together ( Figure 1E , Video 1 ) , suggesting that DNA circles and the NPC cap were linked to each other . Thus , the accumulated circles failed to randomly diffuse in the nucleoplasm . Instead , they appeared to interact with some structure in the nuclear envelope , and thereby affected the morphology of the nucleus . 10 . 7554/eLife . 03790 . 005Video 1 . The NPC cap and accumulated plasmids are simultaneously retained in the mother cell during mitosis . A cell expressing TetR-mCherry ( red ) and Nup82-3sfGFP ( green ) , containing accumulated plasmids ( 18 hr ED ) , followed through nuclear division using time lapse microscopy ( one Z-stack every 5 min for 1 hr , maximal projection , scale bar represents 2 μm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 005 Interestingly , these changes in NPC distribution had a clear impact on cell division , as observed in time-lapse movies ( Video 1 and Figure 2A ) . While all dividing cells formed two daughter nuclei of different sizes , the mother nucleus being generally larger , this asymmetry was enhanced in cells with accumulated circles ( Figure 2B ) . Furthermore , the density of NPCs was increased in these mother nuclei ( Figure 2C ) . Together , this results in an increase in the asymmetry of NPC segregation in circle-loaded cells ( Figure 2D ) . Mother cells containing accumulated plasmids retained a larger fraction of their NPCs ( median = 78% , N = 50 ) , compared to cells containing no circles ( 63% , N = 50 cells , p < 0 . 001 ) . Interestingly , in these cells both the TetR-mCherry signal and the GFP cap localized to the vicinity of the nuclear bridge linking mother and bud lobes of the dividing nucleus on its mother side ( Figures 1A and 2A and Video 1 ) . Only in very rare cases a fraction of the plasmid mass passed to the bud ( 2% , N > 50 cells ) . Qualitatively , the same effects were observed when using Nup82-3sfGFP , Nup82-GFP or Nup49-GFP as reporters . Irrespective of plasmid accumulation , tagging Nup49 slightly enhanced NPC retention ( Figure 2D ) , as previously reported ( Chadrin et al . , 2010; Makio et al . , 2013 ) . Consistent with resulting from DNA circle accumulation , cells in which the centromere was not excised from the plasmid and untransformed cells of a similar age did not show such NPC distribution and segregation ( Figure 1 and see below ) . Together , these data indicate that the accumulation of non-chromosomal DNA circles affects the size of the nucleus , the organization of the nuclear envelope , and causes the mitotic retention of NPCs in yeast mother cells . These data argue against DNA circles freely diffusing in the nucleoplasm . 10 . 7554/eLife . 03790 . 006Figure 2 . Accumulated plasmids retain the NPC cap in the mother cell during mitosis leading to an increased asymmetric segregation of NPCs . ( A ) Time lapse images of a dividing nucleus in a cell expressing Nup82-3sfGFP and containing accumulated plasmids . ( B–D ) Percentage of the nuclear area ( B ) , the mean fluorescence density ( C ) , and the total green fluorescence ( D ) segregated to the mother cell in telophase cells with and without plasmids using different nuclear pore markers ( N > 25 cells , ***p < 0 . 001 , **p < 0 . 01 ) . ( E ) Representative pictures of bud6∆ cells containing plasmids and quantifications of Nup82-3sfGFP intensity in the vicinity of the DNA circle ( Ic ) and the rest of the nucleus ( Ir ) normalized by the median Ir ( N = 50 cells ) . ( A–E ) Arrow depicts the NPC cap , arrow head the accumulated circles . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 006 The diffusion barrier present in the envelope of anaphase nuclei depends on the septins and the protein Bud6 ( Shcheprova et al . , 2008 ) . In cells lacking Bud6 , the NPC cap still formed around the plasmid leading to an increase in NPC signal in the circle area ( Ic = 2 . 29 ± 0 . 67 A . U . ) compared to the residual area ( Ir = 1 . 0 ± 0 . 33 A . U . , N = 50 cells , p < 0 . 001; Figure 2E ) . However , in 11% of the cells , part of the plasmid mass was transmitted to the daughter cell ( compared to 2% in wild-type cells , N = 50 cells ) . The propagated plasmid foci were generally small , but when a bigger part was transmitted , an NPC cap was visible around it ( see representative image in Figure 2E ) . Thus , retention of the circles and the NPC cap partially depends on the lateral diffusion barrier separating the mother and bud parts of the nuclear envelope . Together , these data suggest that DNA circles interact with some structure at the nuclear periphery , affecting NPCs . Next , we wondered what caused the interaction of the circles with the nuclear periphery and if this contributed to circle retention . Therefore , we asked whether any of the protein complexes already reported to mediate chromatin interaction with the nuclear periphery facilitate the retention of DNA circles . We screened a collection of knockout strains for defects in circle partitioning , using the labeled plasmid described above . Three hours after centromere excision , individual plasmid dots were visualized and the frequency at which they segregated to the bud was determined ( propagation frequency , pf , see ‘Materials and methods’ ) . In wild-type cells , non-centromeric circles rarely passed to the daughter cells ( pf = 3 . 9 ± 1 . 1; Figure 3A ) , that is 2 . 5- to 4-fold less frequently than predicted by the passive model ( pf = 10–15; ( Gehlen et al . , 2011 ) ) . Circle propagation into the bud was significantly increased in yku70∆ and bud6∆ cells , as previously reported ( ( Shcheprova et al . , 2008; Gehlen et al . , 2011 ) ; pf = 10 . 3 ± 1 . 3 and 11 . 8 ± 0 . 8 , respectively , N = 3 clones , >50 cells each , p < 0 . 001; Figure 3A ) , validating our approach . 10 . 7554/eLife . 03790 . 007Figure 3 . The SAGA complex mediates retention of DNA circles including ERCs . ( A ) Representative images and quantifications of plasmid propagation frequencies ( pf ) in wt and different mutant cells ( mean ± SD N ≥ 3 clones ) . ( B ) Mother-bud distribution of the rDNA containing plasmid pDS163 in wt , gcn5∆ and sgf73∆ mutant cells ( mean ± SD N = 10 independent experiments ) . ( C ) Detection of ERC levels in wt , sgf73∆ and gcn5∆ mutant cells by Southern blotting ( « depict ERCs ) in 11 and 14 generations old cells . ( D ) Quantifications of half-sectors representing recombination events within the rDNA repeats in wt and mutant cells ( mean ± SD N = 3 clones ) . ( A–D ) ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 007 Using this assay , no effect on circle retention was observed in cells lacking any of the sirtuin proteins , Sir1-4 , which are involved in chromatin silencing and telomere anchorage to the inner nuclear membrane ( Gotta et al . , 1996 ) , in cells lacking the nucleoplasmic domain of the Sun-domain protein Mps3 , involved in NPC and SPB insertion ( Jaspersen and Ghosh , 2012 ) and telomere tethering to the nuclear envelope ( Bupp et al . , 2007 ) , in cells lacking the lamin-binding-related proteins Src1 or Heh2 ( Grund et al . , 2008 ) , and in cells lacking Slx5 , a protein involved in DNA double strand break repair ( Mullen et al . , 2001 ) and forming perinuclear foci ( Cook et al . , 2009 ) ( Figure 3A ) . In contrast , cells lacking Gcn5 , a component of the Spt-Ada-Gcn5 acetyltransferase ( SAGA ) complex , showed a clear increase in plasmid propagation ( pf = 11 . 1 ± 2 . 8 , N = 3 clones , >100 cells each , p < 0 . 001 ) , similar to the defect observed in bud6∆ mutant cells and fitting well with propagation frequencies predicted by the passive model . The SAGA complex regulates chromatin organization of actively transcribed genes and their recruitment to the nuclear periphery through anchorage to NPCs ( Vinciguerra and Stutz , 2004; Cabal et al . , 2006; Luthra et al . , 2007 ) . All SAGA mutants tested , including spt3Δ and sgf73Δ , displayed similar defects in DNA circle retention ( pf = 9 . 7 ± 1 . 7 and 11 . 8 ± 1 . 6 , respectively , N = 3 clones , >50 cells each , p < 0 . 001; Figure 3A ) . Furthermore , replacement of the GCN5 gene with a point mutant allele abrogating its acetyl-ransferase activity , gcn5-E173A ( Wang et al . , 1998 ) led to a similar increase in plasmid propagation to the daughter cell as observed in gcn5Δ mutant cells ( pf = 11 . 0 ± 1 . 5 , N = 3 clones , >100 cells each , p < 0 . 001 ) . Thus , SAGA and its acetyl–transferase activity , which are known to link chromatin to NPCs , promote the retention of DNA circles in the mother cell . In order to determine whether SAGA played a general role in circle retention , we next wondered whether it was also involved in the retention of endogenous ERCs in the mother cell . We first monitored the inheritance of an ERC marked with the ADE2 gene ( pDS163; ( Sinclair and Guarente , 1997 ) ) in wild-type and SAGA mutant cells , by dissecting daughters off their mothers and following the inheritance of the ADE2 marker between them . Following the segregation of the ADE2 gene did not determine the segregation of single ERCs but solely whether all ERCs remained in the mother cell or not . In 38% ( ±6% ) of the wild-type cells , all ERCs that had accumulated in the mother cell remained in the mother cell , as demonstrated by the fact that their daughters lacked the ADE2 marker and formed a red colony ( see material and methods , ( Sinclair and Guarente , 1997 ) ) . This percentage was reduced 2 . 5-fold in the gcn5∆ and the sgf73∆ single mutant cells ( 16 ± 2% and 15 ± 2% , N ≥ 3 clones , p < 0 . 001; Figure 3B ) , indicating that high-fidelity retention of ERCs in the mother cell requires SAGA function . To further test this conclusion , we next investigated whether the accumulation of ERCs in aged yeast mother cells depended on SAGA function . We purified wild-type , sgf73∆ and gcn5∆ mutant mother cells after 11 and 14 generations using the mother enrichment program ( Lindstrom and Gottschling , 2009 ) , normalized the extracted DNA using qPCR on the ACT1 gene , and monitored ERC levels using Southern blotting . While no ERCs were detected in young wild-type cells , substantial ERC accumulation was observed in aged wild-type mothers . In contrast , SAGA mutant cells of the same age contained fewer ERCs ( Figure 3C ) . To determine whether this lower rate of ERC accumulation was due to a defect in ERC formation or ERC retention , we next estimated the impact of SAGA inactivation on recombination in the rDNA . A strain containing one rDNA repeat marked with the ADE2 gene was used to score the rate of excision in the rDNA locus ( Jacobson and Pillus , 2009; McCormick et al . , 2014 ) . Excision of the ADE2 marker and loss of the ADE2 marked ERC leads to the formation of red-white sectored colonies . By scoring the frequency of half sectored colonies , we observed that the rate of ERC formation was decreased 2 folds in the sgf73∆ , similar but less pronounced to what McCormick et al . ( 2014 ) reported ( Figure 3D and ( McCormick et al . , 2014 ) ) . However , ERC formation increased 1 . 5-fold in gcn5∆ mutant cells ( Figure 3D ) . Cells lacking SIR2 and FOB1 served as positive and negative controls in this assay ( Gottlieb and Esposito , 1989; Defossez et al . , 1999; McCormick et al . , 2014 ) . Together , these results indicate that , at least in the gcn5∆ mutant mother cells , the slow rate of ERC accumulation results from the failure in their retention rather than reduced ERC formation . Together , these data establish that SAGA promotes the retention and accumulation of both endogenous ERCs and non-chromosomal DNA circles of exogenous origins in the yeast mother cells . Whereas the function of SAGA in chromatin anchorage to NPCs was intriguing in this context , we also considered that SAGA might contribute to the retention of non-chromosomal DNA circles indirectly through its role in transcription . Therefore , we next investigated whether SAGA function was indirect , through its effect on processes already known to promote DNA circle retention in the mother cell , such as nuclear morphology , anaphase duration , and diffusion barriers . Measuring anaphase duration using time lapse movies of cells expressing Nsg1-GFP revealed that deleting YKU70 prolonged anaphase , as expected ( Gehlen et al . , 2011 ) , whereas deletion of BUD6 , GCN5 , or SGF73 did not ( N ≥ 180 cells , p < 0 . 001; Figure 4A ) . Furthermore , nuclear bridges were not wider in gcn5∆ cells ( N ≥ 70 cells; Figure 4B ) . Thus , we next asked whether SAGA inactivation affected the compartmentalization of the nuclear envelope during anaphase . The strength of the diffusion barrier at the bud neck of early anaphase nuclei was measured using fluorescence loss in photobleaching ( FLIP ) , as previously described ( Luedeke et al . , 2005; Shcheprova et al . , 2008 ) . The barrier strength was measured in wild-type , gcn5∆ mutant and bud6∆ mutant cells , using Nup49-GFP as a reporter . Although the bud6∆ mutation reduced compartmentalization of the envelope , as reported previously ( Shcheprova et al . , 2008 ) , no effect was observed in gcn5∆ mutant cells compared to wild-type cells ( N ≥ 20 cells; Figure 4C ) . Thus , SAGA does not affect nuclear geometry , anaphase duration , and the diffusion barrier . Therefore SAGA must play a different role in circle retention . 10 . 7554/eLife . 03790 . 008Figure 4 . Previously described mechanisms for plasmid retention are not affected in SAGA deficient cells . ( A ) Duration of early anaphase in wt cells and cells lacking Bud6 , Yku70 , Gcn5 , or Sgf73 ( mean ± SD N ≥ 180 cells ) . ( B ) Measurements of the width of early anaphase nuclei at the bud neck , categorized by different length of the nucleus in wt and gcn5∆ mutant cells ( mean ± SEM , N ≥ 70 cells ) . ( C ) Photobleaching analysis of wt , bud6∆ , and gcn5∆ cells expressing Nup49-GFP . Graph shows the Barrier Index ( ratio of the time to decay to 70% of initial fluorescence in the mother compartment to the bud compartment of early anaphase nuclei , mean ± SEM , N ≥ 20 cells ) . ( A–C ) ***p < 0 . 001 . Images are max Z-projections ( A ) or represent one focal plane ( B and C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 008 Therefore , we next asked whether SAGA's function in tethering chromatin to nuclear pores was directly relevant for its role in circle retention . SAGA's function in chromatin anchorage to NPCs involves TREX-2 , a multiprotein complex that localizes to nuclear pores , facilitates mRNA export , and physically interacts with SAGA ( Fischer et al . , 2002; Rodriguez-Navarro et al . , 2004 ) . Thus , we examined whether TREX-2 also contributed to circle retention . Inactivation of the core TREX-2 gene , SAC3 , caused circles to propagate to the bud 3 times more frequently than in wild-type cells , i . e . at a frequency similar to SAGA defective cells ( pf = 11 . 5 ± 3 . 0 , N = 3 clones , p < 0 . 001; Figure 5A ) . Deletion of SUS1 , which inactivates both SAGA and TREX-2 complexes ( Pascual-Garcia and Rodriguez-Navarro , 2009 ) , and deleting both the GCN5 and SAC3 genes led to similar effects ( pf = 11 . 3 ± 1 . 4 , N = 3 clones , p < 0 . 001 and pf = 11 . 1 ± 1 . 3 , N = 3 clones p < 0 . 001 , respectively ) , indicating that the effects of inactivating SAGA and TREX-2 were not additive . We conclude that the SAGA complex and the NPC component TREX-2 act together in a pathway mediating the targeted retention of non-chromosomal circles in the mother cell . 10 . 7554/eLife . 03790 . 004Figure 5 . SAGA attaches DNA circles to NPCs via TREX-2 . ( A ) pf in wt and different mutant cells ( gcn5∆ from Figure 3A for comparison , mean ± SD , N ≥ 3 clones ) . Model of how non-centromeric plasmids including ERCs might be attached to NPCs via the SAGA and TREX-2 complexes . ( B ) Percentage of plasmids at a resolvable distance to the nuclear periphery in wt , gcn5∆ and sus1∆ mutant cells ( mean ± SD , N = 3 clones ) . ( C ) Normalized fluorescence intensity of Nup82-3sfGFP in green ( NPCs ) and TetR-mCherry in red ( plasmid ) aligned by the peak of TetR intensity . Quantifications of average Nup82-3sfGFP intensity 130 nm around the plasmid ( mean ± SEM , N = 50 cells ) . ( A-C ) ***p < 0 . 001 , **p < 0 . 01 . Images represent one focal plane . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 004 Thus , we next wondered whether TREX-2 and SAGA promoted the linkage of DNA circles to NPCs ( see model Figure 5A ) . In order to investigate this possibility , we first tested whether individual non-centromeric plasmids co-localize with nuclear pores and whether this required SAGA function . Consistent with this idea , the vast majority of non-centromeric circles localized to the nuclear rim in wild-type cells: only 15% ( ±2 . 8% ) of plasmid foci were at a resolvable distance from the nuclear envelope ( Figure 5B ) . In contrast , this fraction was substantially increased in cells lacking SAGA ( gcn5∆ mutant cells ) or both SAGA and TREX-2 function ( sus1∆ mutant cells; 31 . 9% ± 1 . 3% and 24 . 1% ± 2 . 6% , respectively , N = 3 clones , p < 0 . 001 ) . To determine whether the plasmids localizing to the periphery co-localized with the nuclear pores , Nup82-3sfGFP intensity traces ( N > 50 cells ) were measured along the nuclear envelope in equatorial focal sections of the nuclei containing a single labeled acentric DNA circle at the rim . High LED-based illumination and fast image acquisition minimized blur due to NPC movement . Aligning all traces with the position of the DNA circle showed that the average Nup82-3sfGFP intensity increased in the vicinity of the circle compared to elsewhere ( +33 . 2% ± 0 . 09% , Figure 5C ) . No such increase in nuclear pore signal was observed near DNA circles in gcn5∆ and sus1∆ mutant cells ( +4 . 6% ± 0 . 05% and −2 . 0% ± 0 . 07% , p < 0 . 01 ) . Thus , we conclude that SAGA indeed promotes the interaction of non-chromosomal DNA circles to nuclear pores . SAGA is involved in many different processes , including transcription , thus we wondered whether detachment from the pores was the main reason for the increased circle propagation observed in SAGA deficient cells . To test this , we artificially tethered circles to the NPC components Nup49 or Nup170 , as previously reported ( Khmelinskii et al . , 2011 ) , and asked whether this was sufficient to bypass the requirement for SAGA function in circle retention . Neither fusion protein had any effect on circle retention when replacing the corresponding endogenous nucleoporin in otherwise wild-type cells ( pf = 5 . 5 ± 0 . 7 in NUP170-TetR and 4 . 6 ± 1 . 6 in NUP49-TetR cells; Figure 6A ) , as reported . In contrast , when expressed in gcn5∆ , sgf73∆ , or sus1∆ mutant cells , these fusion proteins restored the retention of the reporter circle in the mother cell fully or largely so . In support of the TetR-fusion proteins mediating circle attachment to NPCs , the labeled circles localized close to the envelope in these cells ( Figure 6B ) . Importantly , tethering of the circles to the nuclear pores did not bypass the requirement for the diffusion barrier in plasmid retention , since bud6∆ mutant cells expressing Nup170 or Nup49 fused to TetR displayed plasmid propagation frequencies similar to bud6∆ cells ( pf = 11 . 9 ± 1 . 8 and 10 . 8 ± 1 . 2 , respectively; Figure 6A ) . Thus , non-chromosomal DNA circles artificially tethered to NPCs no longer need SAGA for being efficiently retained in the mother cell by the diffusion barrier . 10 . 7554/eLife . 03790 . 009Figure 6 . SAGA-dependent attachment of circles to stable NPC components ensures their asymmetric segregation . ( A ) pf in wt and different mutant cells expressing Nup170 or Nup49 fused to TetR or the fusion proteins Gcn5-Sac3 or Spt7-Nup49 ( mean ± SD , N ≥ 3 clones ) . Scheme of the fusion proteins . ( B ) Percentage of plasmids at a resolvable distance to the nuclear periphery in wt , gcn5∆ and sus1∆ mutant cells expressing Nup170 fused to TetR ( mean ± SD , N = 3 clones ) ( C ) Fluorescence recovery after bleaching the bud part of early anaphase nuclei measured in cells expressing several NPC components tagged with GFP . Quantification of the time to recover 15% of fluorescence intensity ( t15 , mean ± SEM , all compared with Nup2 for statistics ) . ( D ) Correlation between the fluorescence recovery in the bleached bud compartment ( t15 ) of the depicted proteins fused to GFP and the pf in cells expressing the same proteins fused to TetR . Dashed line represents pf in wt cells . ( A–C ) ***p < 0 . 001 , **p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 009 The study by Khmelinskii et al . ( 2011 ) indicated that anchoring of a non-centromeric plasmid to 11 different NPC components had no effect on plasmid retention , but that targeting of the two basket proteins Mlp1 and Nup2 to the reporter plasmid was sufficient to impair its retention in the mother cell ( Khmelinskii et al . , 2011 ) . We reasoned that the first set of observations is expected if circles are already tethered to the same NPCs by SAGA . To better understand the effect of the two other fusion proteins , we first investigated whether they localized stably at NPCs , using fluorescence recovery after photobleaching ( FRAP; Figure 6C ) . Indeed , anchorage would require a stable interaction , whereas at least Nup2 has been reported to shuttle between NPCs and the nucleoplasm ( Dilworth et al . , 2001 ) . Accordingly , we found that both Nup2 and Mlp1 , as well as Nup60 , diffuse more rapidly than the core nucleoporins Nup1 , Nup49 , and Nup170 , at least during anaphase ( Figure 6C ) , indicating that their residence at NPCs is transitory . Remarkably , the stable association of nucleoporins with NPCs was highly correlated with their ability to mediate circle retention ( Figure 6D ) . Together , our data indicate that artificially attaching non-centromeric plasmids to NPC components bypasses the need for SAGA function , provided that the used nucleoporin is stable at NPCs . Thus , SAGA functions in retention through promoting stable anchorage of the circles to NPCs . To investigate how direct the effect of SAGA on circle anchorage to NPCs might be , we next asked whether a physical link between SAGA and NPCs was required for SAGA's function in mediating the anchorage of circles to NPCs and whether SAGA interacted with the DNA circles themselves . To address the first question , we generated fusion proteins to covalently link the SAGA complex to the TREX-2 complex ( Gcn5-Sac3 ) or directly to NPCs ( Spt7-Nup49 ) and asked whether this was sufficient to restore the retention of DNA circle in cells lacking Sgf73 or Sus1 . Expression of either of these two fusion proteins had no effect on plasmid retention in otherwise wild-type cells ( pf = 5 . 2 ± 0 . 4 and 4 . 1 ± 1 . 0 , respectively; Figure 6A ) . However , it largely bypassed the need for Sgf73 ( pf = 6 . 1 ± 1 . 3 with Gcn5-Sac3 and 6 . 8 ± 2 . 0 with Spt7-Nup49 , compared to 11 . 8 ± 1 . 6 in sgf73∆ cells ) and Sus1 ( pf = 6 . 5 ± 1 . 3 with Gcn5-Sac3 and 5 . 6 ± 0 . 2 with Spt7-Nup49 , compared to 11 . 3 ± 1 . 4 in sus1∆ cells , N = 3 clones , p < 0 . 001 ) . Thus , the main function of Sgf73 and Sus1 in plasmid retention is to physically link SAGA to nuclear pores . These data establish that SAGA needs to be at NPCs in order to function in circle retention . To address whether SAGA interacts with non-centromeric DNA circles , we then analyzed the localization of Gcn5-GFP and Sgf73-GFP fusion proteins in cells accumulating non-centromeric DNA circles labeled with TetR-mCherry . Both SAGA components were strongly enriched in the plasmid area , indicating that SAGA interacts more readily with non-chromosomal DNA than with the rest of the genome ( Figure 7A ) . Accordingly , chromatin immunoprecipitation ( ChIP ) experiments ( N = 5 independent experiments ) reproducibly showed a clear enrichment of the SAGA proteins Gcn5 and Spt20 on a non-centromeric reporter plasmid ( pYB1670/CEN- ) , as judged by probing for the autonomous replication sequence ( ARS ) and two non-coding sequences of bacterial origin . Strikingly , this enrichment was nearly lost ( p < 0 . 05 ) when the same plasmid contained a centromere ( CEN+; Figure 7B ) . Thus , SAGA is recruited to higher levels on non-centromeric than on centromeric DNA circles . Together , these data indicate that SAGA more stably associates with non-chromosomal than with chromosomal DNA . 10 . 7554/eLife . 03790 . 010Figure 7 . SAGA preferentially interacts with non-centromeric DNA circles ( A ) . Fluorescent images ( deconvolved max Z-projections ) of Gcn5-GFP and Sgf73-GFP ( green ) in cells with and without accumulated plasmids ( red ) . ( B ) ChIP-qPCR analysis to test the binding of the SAGA components Gcn5 and Spt20 to three sequences on pYB1670 ( ARS , non-coding sequences ( NCS ) 1 and 2; see scheme of plasmid ) and a control sequence on chromosome V ( mean ± SD , N = 5 independent experiments , ***p < 0 . 001 , **p < 0 . 01 , *p < 0 . 05 ) . ( C ) pf in cells expressing Gcn5 or Sir2 fused to TetR ( mean ± SD , N = 3 clones , ***p < 0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 010 To further test the role of SAGA recruitment on the circles in circle retention , we tested the effect of increasing SAGA levels on the plasmid . We increased Gcn5 levels on our TetO circle through expression of a Gcn5-TetR fusion protein and asked whether this affected its retention . Strikingly , this construct reduced the propagation of the circle to the bud 3-fold compared to wild-type ( pf = 1 . 1 ± 0 . 3; N = 3 clones , p < 0 . 001; Figure 7C ) . In contrast , cells expressing the deacetylase Sir2 fused to TetR segregated the circle to the bud as frequently as SAGA defective cells . We concluded that artificial loading of Sir2 to the plasmid either interfered with normal SAGA recruitment to the plasmid or that Sir2 constantly counteracts SAGA's function on the plasmid . Thus , we concluded that binding of SAGA to the circle is required for efficient retention of the circle in the mother cell . So far , our data have established that SAGA promotes the attachment of non-chromosomal DNA circles to NPCs and that this attachment is essential for the efficient retention of the circles in the mother cell . This observation is at odds with data by several labs indicating that NPCs are not nearly as well retained in the mother cell than DNA circles ( Khmelinskii et al . , 2010; Colombi et al . , 2013; Menendez-Benito et al . , 2013 ) . Possibly accounting for these conflicting data , we observed above ( Figure 2A ) that the mother cells that were loaded with circles retained more pores . Thus , we reasoned that DNA circles might promote the retention of the NPCs to which they were attached . In favor of this idea , and in contrast to what we observed in wild-type cells ( Figure 2A ) , the accumulation of DNA circles had no effect on pore segregation in gcn5∆ and sgf73∆ single mutant cells ( Figure 8A ) . Thus , SAGA-dependent attachment of the circles to pores promotes the retention not only of the circles but also of those NPCs . 10 . 7554/eLife . 03790 . 011Figure 8 . DNA circles and NPCs reciprocally reduce their dynamics in a SAGA-dependent manner . ( A ) Fluorescent images of Nup82-3sfGFP ( green ) in cells with accumulated plasmids ( red ) . Quantification of the percentage of NPCs segregated to the mother cell in wt , gcn5∆ and sgf73∆ cells with or without plasmids ( box plots: min to max , median ( line ) , N = 50 cells ) . Images are max Z-projections . ( B ) Time lapse images of a photobleaching experiment in wt cells expressing Nup170-GFP ( green ) and containing plasmids ( red ) . Rectangle depicts the bleaching area . Plotted Nup170-GFP intensity in the circle area ( Ic ) and an equidistant area opposite of the circles ( Io ) over time , set to 100% prior to bleaching . Images represent one focal plane . Quantification of the time 30% of fluorescence decays ( t70; mean ± SD , N = 20 cells ) . ( C ) Average total fluorescence of the whole nucleus plotted over time in wt and gcn5∆ mutant cells with ( dark green ) and without plasmids ( light green ) , set to 100% prior to bleaching ( mean ± SD , N ≥ 20 cells , ***p < 0 . 001 ) . ( B and C ) Images represent one focal plane . ( D ) Quantification of plasmid clustering in wt and SAGA deficient cells . Cells were divided in 3 categories: plasmids localized always , partially , or never to one focus throughout a 1 hr time lapse movie ( mean , N ≥ 160 cells ) . ( E ) Speed of individual plasmids in wt and gcn5∆ cells expressing TetR-GFP measured from time lapse movies with 3 s intervals ( mean ± SD , N ≥ 50 plasmids ) . ( A–E ) ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 011 To shed light on why circle-bound NPCs are better retained in the mother cell , we next asked whether these NPCs showed different dynamic behavior as unbound NPCs . FLIP experiments revealed that SAGA-mediated attachment of circles to NPCs strongly affected their diffusion speed . When fluorescence intensity was measured both in the vicinity of the circles ( Ic ) and at opposite side of the nucleus ( Io ) in wild-type cells loaded with circles , with both measurement regions equidistant from the bleached spot , the signal decayed slower and to a lesser extent in regions associated with the circles ( t70 = 82 ± 39 s ) than in those on the opposite side ( t70 = 25 ± 8 s , N = 20 cells , p < 0 . 001; Figure 8B ) . We could not perform the same experiment in SAGA deficient cells because circles did not cluster to one focus and the cells formed no NPC cap . NPC distribution remained uniform over the nuclear surface , and non-centromeric circles were scattered throughout the nucleoplasm ( Figure 8D ) . We concluded that SAGA mediates both circle clustering and the formation of the NPC cap . Accordingly , in FLIP experiments the fluorescence decayed homogenously over the entire nuclear envelope , whether they contained circles or not ( t50 = 46 ± 9 . 1 s for gcn5∆ and 41 ± 6 . 7 s for sgf73∆ mutant cells; Figure 8C ) , as in wild-type cells devoid of circles ( t50 = 35 ± 5 . 9 s ) . Thus , SAGA-dependent attachment of non-centromeric DNA circles substantially , and specifically , reduces the mobility of the corresponding NPCs . These data establish two points . First , in circle-free exponentially growing cells only very few NPCs , if any , are stably attached to and slowed down by chromatin . Second , NPCs that are attached to DNA circles show very slow dynamics , possibly explaining their enhanced retention in the mother cell during mitosis . Consistent with the attachment of circles to NPCs affecting each other's dynamics , SAGA also slowed down the non-centromeric DNA circles . In time-lapse movies ( one stack every 3 s , 20 optical slices per stack , for 3 min ) , the average speed of the DNA circles was significantly higher in cells lacking Gcn5 ( 0 . 09 ± 0 . 027 μm/s , N ≥ 50 plasmids , p < 0 . 001 ) than in wild-type cells ( 0 . 05 ± 0 . 019 μm/s; Figure 8E ) . Thus , our data establish that non-chromosomal DNA circles and NPCs reciprocally constrain the diffusion speed of each other , in a SAGA-dependent manner . Together , our data suggest that SAGA targets non-chromosomal DNA circles and mediates their retention in the mother cell by linking them to NPCs . However , since these results were all obtained with reporter plasmids , we next asked whether SAGA function was also relevant for endogenous circles . We reasoned that , if ERCs rely on an interaction with NPCs for their retention in the mother cell , then the NPCs attached to these ERCs should also be retained and accumulate with age . To test this possibility , we monitored nuclear pores in ageing cells using Nup49- , Nup170- , or Nup82-GFP fusion proteins as a reporter . To allow imaging them throughout their lifespan , the cells were grown in a microfluidic device ( Lee et al . , 2012 ) that selectively retains the mother cells , due to their larger size , while their daughter cells are removed by the medium flow . Transmission pictures were taken for 64 hr at 20 min intervals , providing information about the age of all cells present ( Video 2 ) . In addition , 2h fluorescence movies at 15 min intervals were recorded after 48 hr and 64 hr . At both time points , the original cells and some coincidentally trapped cells born in the chip were present , allowing simultaneous monitoring of pores in cells of different ages ( ranging from 0 to 38 generations; Figure 9A ) . Strikingly , a large majority of the aged cells showed a non-uniform distribution of NPCs on the nuclear surface . A substantial fraction of very old cells formed an even more intense cap ( Figure 9B ) . These cells divided maximally three more times before dying . Thus , cap formation was observed upon ageing in wild-type cells . 10 . 7554/eLife . 03790 . 012Video 2 . A microfluidic device to follow individual cells throughout their entire life . A trapped cell is visualized for 67 hr taking a transmission picture every 20 min . The trapped cell reaches age 29 before it dies . Coincidently , other cells are trapped under the same micropad , whereas most daughter cells are washed away by a constant flow of media . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 01210 . 7554/eLife . 03790 . 013Figure 9 . NPCs segregate increasingly asymmetric with age leading to their accumulation in old cells in a barrier dependent manner . ( A ) Transmission image of cells trapped in the microfluidic device after 64 hr . Cells of different age are trapped ( ➤ , ° , and * depicts 0- , 27- , and 31-generation old cells , respectively ) . ( B ) Cells expressing Nup170-GFP showing a pronounced NPC cap . ( C ) Total fluorescence of the depicted NPC marker in the mother ( dark green ) and daughter cell ( light green ) grouped by age categories of the mother cells ( N ≥ 50 cells , ***p < 0 . 001 , **p < 0 . 01 , *p < 0 . 05 ) . ( D ) Percentage of NPC fluorescence segregated to the mother cell plotted against their age in wt cells expressing different NPC markers and bud6∆ cells . Lines show fitted curves; dots represent the average of 10 data points grouped by age ( mean ± SD , N ≥ 50 cells ) . ( E–G ) Quantifications of total fluorescence ( E ) , nuclear radii ( F ) , and mean fluorescence intensity ( G ) in G1 cells of increasing age ( mean ± SD , N ≥ 50 cells ) quantified as in ( D ) . ( A–E ) Numbers in white depict the age of the corresponding cell ( g = generations ) . Images are sum Z-projections . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 013 To characterize how NPC inheritance evolved as the cells aged , the total fluorescence of the Nup49- , Nup170- , and Nup82-GFP reporters was measured in image stacks through the entire volume of mother and daughter nuclei of telophase cells and plotted as a function of the age of the mother cell . For all three NPC markers , fluorescence intensity increased significantly in the mother nuclei with age ( Figure 9C ) . This effect was strongest for Nup49-GFP , as above ( Figure 2 ) . The increasing amount of NPCs in the mother cell did not happen at the cost of the daughter cells . On the contrary , occasionally an increased amount of nuclear pores was also observed in daughter cells of old mother cells ( >15 generations; Figure 9C ) . Thus , like in young cells loaded with artificial circles , the number of NPCs increases in ageing mother cells whereas the number of pores transmitted to the daughter cells remains fairly constant . Accordingly , the fraction of nuclear pores that segregated towards the mother cell during mitosis increased as the cells aged , indicating that these cells partitioned NPCs in an increasingly asymmetric manner ( Figure 9D ) . Young mother cells expressing Nup170-GFP or Nup82-GFP retained 56 ± 1 . 9% and 57 ± 2 . 5% of the signal in the mother cell . The retention of pores in the mother cell increased after 25 generations to 73 ± 5 . 0% ( Nup82-GFP ) and 78 ± 3 . 8% after 35 generations ( Nup170-GFP; Figure 9D ) . Cells expressing Nup49-GFP showed the same trend , although asymmetry was higher in both young ( 63% ± 3 . 3% ) and older cells ( 81% ± 16 . 1% after 20 generations; Figure 9D ) . Analysis of nuclei in G1 cells of increasing age confirmed that old mothers contained more NPCs than young cells ( on average 4 . 7 ± 1 . 2-fold more than young mothers in cells expressing Nup49-GFP after 25 generations , 4 . 0 ± 1 . 1-fold in Nup170-GFP expressing cells after 30 generations , and 4 . 1 ± 1 . 4-fold in Nup82-GFP expressing cells after 35 generations; Figure 9E ) . These nuclei were both larger ( radius = 1 . 5 fold larger; Figure 9F ) and more densely populated with pores ( 1 . 9-fold increase in density; Figure 9G ) compared to the nuclei of young cells . Remarkably , the increase with age in NPC asymmetry , NPC retention , and NPC number was much slower in cells lacking a diffusion barrier ( bud6∆ mutant cells; Figure 9C–E ) . Thus , consistent with the possibility that ERC accumulation affects NPC segregation , mother cells retain an increased number of NPCs as they age , in a diffusion barrier-dependent manner . In order to test whether NPC accumulation in old cells indeed depended on ERCs , we next characterized the NPC content of ageing sir2∆ mutant cells , which form ERCs earlier , and fob1∆ mutant cells , which form ERCs much later ( Defossez et al . , 1999; Kaeberlein et al . , 1999 ) . Remarkably , the proportion of NPCs retained in the mother cell at mitosis and the total amount of pores in G1 cells increased much slower with age in fob1∆ mutant cells compared to wild-type , irrespective of the reporter used ( Nup49-GFP or Nup170-GFP ) , and faster in the sir2∆ mutant mother cells ( Figure 10A–C ) . Thus , the presence of ERCs largely drives NPC retention in ageing cells . 10 . 7554/eLife . 03790 . 014Figure 10 . The accumulation of NPCs in old mother cells depends on ERCs and SAGA . ( A ) Total Nup170-GFP intensity in G1 cells plotted against their age in wt and depicted mutant cells . ( B ) Percentage of fluorescence segregated to the mother cell in wt and depicted mutant cells expressing Nup170-GFP or Nup49-GFP . ( A and B ) Images are sum Z-projections . ( C ) Total fluorescence of Nup170-GFP in mother ( dark green ) and daughter cells ( light green ) grouped by age categories of the mother cells ( N ≥ 50 cells , ***p < 0 . 001 , **p < 0 . 01 , *p < 0 . 05 ) . ( D ) Percentage of total Nup170-GFP fluorescence segregated to the mother cell in cells lacking Sgf73 and expressing either Spt7-Nup49 or Gcn5-Sac3 . ( A , B , D ) Lines show fitted curves; dashed lines for comparison; dots represent the average of 10 data points grouped by age ( mean ± SD , N ≥ 50 cells ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 014 Consistent with the SAGA-dependent effect of non-centromeric DNA circles on pore segregation in young cells , the sgf73∆ mutant cells accumulated NPCs slowly as they aged ( Figure 10A–C ) . Furthermore , artificially linking SAGA to NPCs using the Gcn5-Sac3 and Spt7-Nup49 fusion proteins restored NPC accumulation in ageing sgf73∆ mutant cells ( Figure 10D ) . Based on these and our findings above , we conclude that SAGA mediates the interaction of ERCs and NPCs , leading to the joined retention and accumulation of both ERCs and NPCs in ageing yeast mother cells . The retention of DNA circles contributed to the replicative ageing in yeast and SAGA deficient cells accumulated less ERCs and NPCs . Therefore , we next wondered whether these cells lived longer . Using microdissection , the lifespan of cells lacking Gcn5 and Spt20 , a structural component of the SAGA complex , as well as Sgf73 , linking SAGA to TREX-2 , was determined . Wild-type cells showed a median lifespan of 27 generations , whereas the sgf73∆ mutant cells lived 60% longer , ( 44 generations , N = 150 cells , p < 0 . 001; Figure 11A ( Schleit et al . , 2013 ) ) . Strikingly , gcn5∆ and spt20∆ mutant cells did not show an increased longevity ( 27 and 17 generations , N = 50–150 cells; Figure 11A ) . Since SAGA is involved in many cellular processes , we wondered whether the observed lifespan was the result of a balance between the advantages and disadvantages of losing Gcn5 and Spt20 . We rationalized that if this was the case , the longevity of these mutant cells would be limited by other factors than ERC accumulation , and thus decreasing ERC formation should not increase the longevity of these cells . Accordingly , removal of Fob1 failed to extend the lifespan of the gcn5∆ mutant cells ( median lifespan of gcn5∆ fob1∆ double mutant cells = 27 generations , N = 150 cells; Figure 11B ) and spt20∆ mutant cells ( spt20∆ fob1∆ = 16 generations , N = 75 cells; Figure 11C ) . Decreased ERC accumulation in gcn5Δ mutant cells , as suggested by our Southern Blot experiments , and in gcn5Δ fob1Δ double mutant cells did not prolong lifespan of these cells , suggesting that these cells die of reasons independent of ERC accumulation . Additionally , the fact that we observed increased ERC formation in cells lacking Gcn5 ( Figure 3D ) but not a shorter lifespan suggests that indeed the increased ERC formation is compensated by poor circle retention in the mother cells . Thus , cells lacking Gcn5 and Spt20 age through mechanisms independent of ERC formation and accumulation . 10 . 7554/eLife . 03790 . 015Figure 11 . The SAGA complex promotes ageing . ( A ) Replicative lifespan ( RLS ) of wt , gcn5∆ , spt20∆ , and sgf73∆ cells . ( B ) RLS of fob1∆ and gcn5∆ fob1∆ cells . ( C ) RLS of fob1∆ and spt20∆ fob1∆ cells . ( D ) RLS of wt cells expressing Gcn5-Sac3 and Spt7-Nup49 . ( E ) RLS of cells lacking Sgf73 and expressing Gcn5-Sac3 or Spt7-Nup49 . ( A–D ) Median lifespan is depicted in brackets; dashed lines represent RLS of depicted wt and mutant cells of ( A ) for comparison; N = 50–150 cells; ***p < 0 . 001 , **p < 0 . 01 , *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 015 Finally , cells lacking Sgf73 are extremely long-lived . The fact that sgf73Δ mutant cells show reduced ERC formation probably contributes to this phenotype ( Figure 3D and ( McCormick et al . , 2014 ) ) . However , the rate of ERC formation is similar in fob1∆ and sgf73∆ mutant cells , yet the fob1∆ mutation has a much more modest effect on lifespan ( 32 generations ) than sgf73∆ ( 44 generations ) . Therefore , Sgf73 also contributes to ageing by additional mechanisms beyond reducing ERC formation . To test whether Sgf73 promotes ageing via the retention of ERCs and NPCs in the ageing cell , we asked whether targeting SAGA back to NPCs restored ageing in the sgf73∆ mutant cells . Expression of the fusion proteins Gcn5-Sac3 and Spt7-Nup49 did not alter the longevity of the wild-type cells ( 27 generations , N = 50 cells; Figure 11D ) . However , it substantially shortened the lifespan of the sgf73∆ mutant cells ( 16 and 34 generations , respectively , N ≥ 50 cells , p < 0 . 01; Figure 11E ) . Therefore , we conclude that decreased levels of SAGA at pores largely contribute to the extended lifespan observed in sgf73∆ cells . Based on the data presented here , we further conclude that SAGA promotes ageing , at least in part , through promoting the anchorage of non-chromosomal DNA circles to NPCs and promoting their joint retention and accumulation with age . Furthermore , non-chromosomal DNA circles alone do not accumulate and do not promote ageing if they are not at NPCs .
Our data provide a unifying model for how non-chromosomal DNA circles are retained ( Model Figure 12A ) . First , we propose that both morpho-kinetic parameters ( Gehlen et al . , 2011 ) as well as tethering to NPCs are required for efficient circle retention in vivo . The morpho-kinetic constraints provide the baseline for circle retention and ensure that unattached plasmids are still fairly well retained , as observed in SAGA defective cells ( this study ) or for unattached episomes ( Gehlen et al . , 2011 ) . Above this baseline , the barrier-tether mechanism further tightens circle retention to achieve high fidelity . Indeed , supporting this model mutations that prolong anaphase ( yku70∆ ) and mutations that weaken the nuclear diffusion barrier without prolonging anaphase ( bud6∆ ) both affect circle retention . Second , we propose that tight confinement of the circles by the diffusion barrier requires the attachment of the circles to NPCs . Indeed , what the barrier efficiently retains in the mother cell is neither NPCs alone nor the unbound circles but rather the circle-bound NPCs . Accordingly , anchorage of the circles to NPCs strongly enhances the retention of both NPCs and circles . 10 . 7554/eLife . 03790 . 016Figure 12 . Model . ( A ) Model of how SAGA prevents DNA circles from segregating into the daughter cell: SAGA attaches DNA circles to NPCs via the TREX-2 complex , which reduces the diffusion of the circle-bound NPCs , preventing them to pass the diffusion barrier at the bud neck . ( B ) The retention of circle-bound NPCs leads to a SAGA-dependent accumulation of ERCs and NPCs in old mother cells . DOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 016 This finding helps solving a debate in the field . Although NPC attachment had already been proposed to mediate the retention of DNA circles ( Shcheprova et al . , 2008 ) , the observation that NPCs were themselves inefficiently retained in exponentially growing cells strongly argued against this model ( Khmelinskii et al . , 2010; Colombi et al . , 2013; Menendez-Benito et al . , 2013 ) . Populations of exponentially growing cells are primarily composed of young cells , with only about 5% of the cells having divided more than three times . Since only very few , if any , young cells contain an ERC or other DNA circle , our model accounts well for why most exponentially growing cells only weakly retain NPCs in the mother cell . Two types of mechanisms may underlie the selective retention of circle-bound NPCs . First , it might result from the reduced mobility of the circle-bound NPCs , which necessarily reduces the frequency at which they approach the barrier , and therefore their chance of passing it . This reduced mobility may be due to the drag that attached plasmids might oppose to NPCs . In addition , it might be exacerbated by the propensity of circle-bound NPCs to cluster , by a yet unknown mechanism . Alternatively , post-translational modifications of the NPC , perhaps by SAGA , may affect the interaction of the NPC with myosin and actin ( Steinberg et al . , 2012; Colombi et al . , 2013 ) or with the lipids in the envelope , impeding their dynamics . Second , the barrier might be selective . Previous data have established that the barrier has a much larger impact on large entities than on small ones . Remodeling and clustering of the NPCs might interfere particularly strongly with their ability to pass the barrier . Strikingly , the immobilization and segregation effects observed upon circle attachment to NPCs are reminiscent of the phenotypes caused by removing the nucleoporin Nsp1 ( Colombi et al . , 2013; Makio et al . , 2013 ) or expressing defective alleles of the nucleoporins Nic96 and Nup133 ( Shcheprova et al . , 2008; Chadrin et al . , 2010 ) , Nic96 being a key binding partner for Nsp1 at the pore . Therefore , SAGA may regulate NPC behavior by affecting specific parts of the pore itself . Together , our observations explain why the diffusion barrier in the envelope of anaphase nuclei , which has limited effects on bulk NPC distribution , has a crucial impact on ageing . Indeed , ageing cells accumulate circles with increasing speed . When the circles cannot attach or when the diffusion barrier is defective , the morpho-kinetic model still promotes a retention level of about 88–85% . However , our data indicate that the loss of 12–15% of the circles per division cycle has a substantial impact on ERC accumulation and the longevity of the cells , the bud6∆ and the sgf73∆ mutant cells being remarkably long-lived . This large impact on longevity for apparently small defects in retention is fully in line with the idea that high-fidelity retention is a crucial parameter , as predicted by in silico simulations ( Gillespie et al . , 2004; Shcheprova et al . , 2008 ) . Our data indicate that , among its many functions , the role of SAGA in linking chromatin to NPCs is especially crucial for circle retention and ageing . Thus , while the gene-gating hypothesis designed to account for SAGA-dependent transcription at NPCs is debated ( Cabal et al . , 2006; Green et al . , 2012 ) , our findings may provide an alternative model for SAGA's function at the pore . How SAGA promotes circle attachment to NPCs is not fully understood at this point . SAGA might modify chromatin or some chromatin-binding component of the NPC and thereby regulate the interaction of specific DNA pieces with the pore . Alternatively , it may function itself as a physical tether . Although our data do not allow any definitive conclusion , it is remarkable that SAGA needs to be both on the NPC and on the chromatin to mediate its function in circle attachment to pores . Thus , we favor the idea that SAGA is an integral part of the tether . The fact that its acetyl transferase activity is also needed for its function might reflect a dual role of SAGA , both as a tether and as a regulator of tethering . Identifying the acetylation targets of SAGA and the role of these modifications in circle retention are now required in order to address how SAGA facilitates chromatin anchorage to NPCs . One intriguing aspect of DNA circle retention is that it affects a very large spectrum of molecules . Together with the current knowledge , the identification of SAGA as a player in this process suggests an explanation for this observation . The SAGA complex , which associates broadly with chromatin and interacts with every expressed sequence ( Ohtsuki et al . , 2010; Bonnet et al . , 2014 ) , promotes DNA anchorage to NPCs ( Luthra et al . , 2007 ) . Current data suggest that this association occurs on both chromosomal and non-chromosomal DNA . Nevertheless , we observed an enrichment of SAGA on non-centromeric DNA circles both by using microscopy and ChIP experiments . Therefore , we propose that either SAGA is more efficiently recruited to or more stably associated with acentric DNA molecules or , alternatively , that during mitosis SAGA interaction or function is specifically repressed on chromosomal DNA . Strikingly , such a mechanism has the potential of discriminating self from foreign DNA . Since circles containing transcribed sequences are expected to recruit more SAGA , they may be more efficiently marked and retained than non-transcribed circles . In addition , our ChIP data are consistent with previous data indicating that ARSs are also SAGA targets ( Espinosa et al . , 2010 ) . In fitting , DNA circles with strong replication origins are better retained in mother cells than those with weak origins ( Futcher and Cox , 1984 ) . Thus , replicating circles may also be better retained than non-replicating circles . Therefore , a SAGA-based mechanism to discriminate between chromosomal and non-chromosomal DNA would primarily target non-chromosomal DNA molecules of internal or foreign origin based on their expression and replication potential , i . e . on their potential toxicity . Hence , we suggest that this pathway fulfills the functions of a genomic immunity system capable of preventing the spread of extra-chromosomal DNA molecules of both foreign and endogenous origin in a population . Since DNA circles form in all organisms tested so far ( Gaubatz , 1990; Cohen et al . , 2003 , 2010 ) , there must exist some mechanism to restrict their propagation in higher eukaryotes as well . Together , our data suggest a mechanism for the effect of non-chromosomal DNA circles on cellular longevity . This mechanism relies not on the presence of the circles themselves but on their SAGA-dependent association with NPCs ( Model Figure 12B ) . Therefore , replicative ageing in yeast might be at least in part the secondary consequence of a mechanism involved in restricting the propagation of non-chromosomal DNA , but inevitably altering nuclear organization and NPC number . Human diseases leading to premature ageing phenotypes , such as Bloom and Werner syndromes , originate from dysfunctions in two RecQ helicases involved in DNA recombination , BLM , and WRN ( Burtner and Kennedy , 2010 ) . RecQ helicases inhibit the formation of Holliday junctions ( Sharma et al . , 2006 ) and therefore of circles . Thus , an increased rate of circle formation might underlie at least some aspects of the Werner and Bloom syndromes . Interestingly , the Hutchinson–Gilford progeria syndrome ( HGPS ) , which is caused by the defects of the nuclear lamina and causes alterations of the nuclear structure ( Scaffidi and Misteli , 2006 ) , leads to very similar ageing phenotypes as Werner and Bloom syndromes . These different observations indicate a link between genomic instability , DNA circle formation , and changes in nuclear organization during ageing and suggest that NPCs might be at the center of these relationships . Together , our data indicate that yeast is a powerful system to dissect how NPC number and modifications affect nuclear homeostasis and cellular viability .
Yeast strains used in this study are listed in Table 1 . All strains are isogenic to S288C , except the strain containing the ADE2 gene in the rDNA locus to check the rate of ERC formation , which is derived from W303 and was a gift from Lorraine Pillus ( Jacobson and Pillus , 2009 ) . GFP and knock out strains derived from collections ( http://web . uni-frankfurt . de/fb15/mikro/euroscarf/index . html and http://clones . lifetechnologies . com/cloneinfo . php ? clone=yeastgfp ) or were manually created as described ( Janke et al . , 2004 ) . Cells were grown using the standard conditions: YPD at 30°C unless otherwise indicated . pYB1601 containing 3sfGFP was created by Mathias Bayer . pDS316 containing one rDNA repeat was a gift from D . A . Sinclair ( Sinclair and Guarente , 1997 ) and pYB1382 containing gcn5-E173A was a gift from S . L . Berger ( Wang et al . , 1998 ) . gcn5-E173A was integrated into the genome to replace GCN5 . To allow the visualization of SPBs and the expression of the recombinase upon addition of estradiol , spc42::SPC42-CFP:kanMX4 and the estradiol binding domain fused to Gal4 trp1::GAL4-EBD:TRP1 were introduced into the plasmid visualization strain used in Shcheprova et al . ( 2008 ) . To create TetR fusion proteins , a plasmid pYB1666 containing linker-TETR-mCHERRY-hphNT1 was created by homologous recombination into pRS314 . PCR products amplifying linker-TETR-mCHERRY-hphNT1 and containing homologous overhangs were used to tag endogenous genes with TetR and correct integrations were checked by microscopy . To generate the Gcn5-Sac3 and Spt7-Nup49 fusion proteins , pYB1665 containing GCN5-linker-SAC3-SAC3 ( 3′UTR ) -hphNT1-GCN5 ( 3′UTR ) and pYB1739 containing SPT7-linker-NUP49-NUP49 ( 3′UTR ) hphNT1-SPT7 ( 3′UTR ) were created by homologous recombination into pRS314 . PCR fragments containing the fusions were treated with DpnI and gel-purified before integration into yeast . Integrations were verified by PCR . For all fusion proteins the linker sequence is as described in Gruber et al . ( 2006 ) . 10 . 7554/eLife . 03790 . 017Table 1 . Strain listDOI: http://dx . doi . org/10 . 7554/eLife . 03790 . 017yYB numberGenotype6637anup82::NUP82-3sfGFP:kanMX4; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; PURA3-TETR-mCherry:kanMX4; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-10110 , 071 , 10 , 072 , 10 , 073afob1::FOB1-yeGFP:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; PURA3-TETR-mCherry:kanMX4; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-36271anup82::NUP82-yeGFP:natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; PURA3-TETR-mCherry:kanMX4; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-1019038anup49::NUP49-yeGFP:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; PURA3-TETR-mCherry:kanMX4; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-1019415 , 9416 , 10 , 454anup49::NUP49-yeGFP:hphNT1; bud6::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; PURA3-TETR-mCherry:kanMX4; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101 clones 1-34221apPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-1014222 , 5547 , 6521abud6::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35516 , 5517 , 8099ayku70::hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-34124 , 4161 , 7237asir1::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-34165 , 4260 , 4261asir2::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-34121 , 7239 , 7240asir3::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-34235 , 4236 , 7243asir4::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-38100 , 8101 , 8102amps3::mps3-aa75-100Δ:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-36524 , 6652 , 8059asrc1::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-37957 , 7958 , 7960aheh2::hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-37954 , 7955 , 7956aslx5::hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-34120 , 4263 , 4264agcn5::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-34118 , 4119 , 4262aspt3::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-34287 , 4577 , 5518asgf73::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35319 , 5320 , 5321agcn5::gcn5E173A:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-34166 , 4237 , 4238asac3::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-34122 , 4573 , 8103asus1::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-3384aura3-52; his3∆200; leu2; lys2-801; ade2-101; trp1∆633870agcn5::natNT2; ura3-52; his3∆200; leu2; lys2-801; ade2-101; trp1∆634287asgf73::natNT2; ura3-52; his3∆200; leu2; lys2-801; ade2-101; trp1∆633415ahoD::PSCW11-Cre-EBD78:natMX; ubc9::loxP-UBC9-loxP:LEU2; cdc20::loxP-CDC20-intron-loxP:hphMX; ade2::hisG; his3; lys2; ura3; trp1∆636580agcn5::kanMX4; hoD::PSCW11-Cre-EBD78:natMX; ubc9::loxP-UBC9-loxP:LEU2; cdc20::loxP-CDC20-intron-loxP:hphMX; ade2::hisG; his3; lys2; ura3; trp1∆636195asgf73::kanMX4; hoD::PSCW11-Cre-EBD78:natMX; ubc9::loxP-UBC9-loxP:LEU2; cdc20::loxP-CDC20-intron-loxP:hphMX; ade2::hisG; his3; lys2; ura3; trp1∆635457apPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-redStar:natNT2; nup170::NUP170-mCherry:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101;5502 , 5503 , 5504agcn5::hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-redStar:natNT2; nup170::NUP170-mCherry:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-38115 , 8116 , 8117asus1::hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-redStar:natNT2; nup170::NUP170-mCherry:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-36648apPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; nup82::NUP82-3sfGFP:kanMX4; PURA3-TETR-mCherry:kanMX4; spc42::SPC42-yeGFP:hphNT1; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-1016752 , 6765 , 6895agcn5::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; nup82::NUP82-3sfGFP:kanMX4; PURA3-TETR-mCherry:kanMX4; spc42::SPC42-yeGFP:hphNT1; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-38093 , 8094 , 8095asus1::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; nup82::NUP82-3sfGFP:kanMX4; PURA3-TETR-mCherry:kanMX4; spc42::SPC42-yeGFP:hphNT1; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35881 , 5882 , 5883anup170::NUP170-linker-TETR-mCherry:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35886 , 7253 , 7254anup170::NUP170-linker-TETR-mCherry:hphNT1; gcn5::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35962 , 5973 , 7863anup170::NUP170-linker-TETR-mCherry:hphNT1; sus1::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35933 , 5934 , 5935anup49::NUP49-linker-TETR-mCherry:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-37959 , 8032 , 8033anup49::NUP49-linker-TETR-mCherry:hphNT1; gcn5::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-37865 , 7866 , 7867anup49::NUP49-linker-TETR-mCherry:hphNT1; sgf73::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35884 , 5885 , 5961anup170::NUP170-linker-TETR-mCherry:hphNT1; sgf73::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35966 , 5967 , 7864anup49::NUP49-linker-TETR-mCherry:hphNT1; sus1::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35968 , 5969 , 5970anup49::NUP49-linker-TETR-mCherry:hphNT1; bud6::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35963 , 5964 , 5965anup170::NUP170-linker-TETR-mCherry:hphNT1; bud6::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35852 , 5853 , 5854agcn5::GCN5-linker-SAC3:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-37141 , 7142 , 7143aspt7::SPT7-linker-NUP49:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35888 , 5889 , 7856agcn5::GCN5-linker-SAC3:hphNT1; sus1::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-37859 , 7869 , 7870aspt7::SPT7-linker-NUP49:hphNT1; sus1::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-36048 , 7858 , 7868agcn5::GCN5-linker-SAC3:hphNT1; sgf73::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-37557 , 7860 , 7861aspt7::SPT7-linker-NUP49:hphNT1; sgf73::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-39520 , 9521 , 9522agcn5::GCN5-yeGFP:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; PURA3-TETR-mCherry:kanMX4; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-39523 , 9524 , 9525asgf73::SGF73-yeGFP:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; PURA3-TETR-mCherry:kanMX4; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-37311αpYB1670 ( REC-URA3-CEN-REC-ARS1-LEU2-AmpR ) ; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; leu2∆0; ura3∆0; met15∆07555agcn5::GCN5-yeGFP:HIS3; pYB1670 ( REC-URA3-CEN-REC-ARS1-LEU2-AmpR ) ; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; leu2∆0; ura3∆0; met15∆07524αspt20::SPT20-yeGFP:HIS3; pYB1670 ( REC-URA3-CEN-REC-ARS1-LEU2-AmpR ) ; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; leu2∆0; ura3∆0; met15∆08831 , 8832 , 8833anup82::NUP82-3sfGFP:kanMX4; gcn5::hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; PURA3-TETR-mCherry:kanMX4; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-39002 , 9003 , 10 , 455anup82::NUP82-3sfGFP:kanMX4; sgf73::hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; PURA3-TETR-mCherry:kanMX4; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1&29039anup170::NUP170-yeGFP:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; PURA3-TETR-mCherry:kanMX4; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-39392anup170::NUP170-yeGFP:hphNT1; gcn5::natNT2; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; PURA3-TETR-mCherry:kanMX4; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-1017828anup49::NUP49-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆08112anup170::NUP170-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆08182anup82::NUP82-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆09337 , 9338 , 9339anup170::NUP170-yeGFP:HIS3; bud6::kanMX4; leu2∆0; ura3∆0; met15∆0; clones 1-39612 , 9613 , 9614anup170::NUP170-yeGFP:HIS3; fob1::kanMX4; leu2∆0; ura3∆0; met15∆0; clones 1-39616 , 9617 , 9618anup170::NUP170-yeGFP:HIS3; sir2::kanMX4; leu2∆0; ura3∆0; met15∆0; clones 1-39372 , 9373 , 9374anup170::NUP170-yeGFP:HIS3; sgf73::hphNT1; leu2∆0; ura3∆0; met15∆0; clones 1-39171 , 9173 , 9174anup49::NUP49-yeGFP:HIS3; fob1::kanMX4; leu2∆0; ura3∆0; met15∆0; clones 1-39181 , 9182 , 9183anup49::NUP49-yeGFP:HIS3; sir2::kanMX4; leu2∆0; ura3∆0; met15∆0; clones 1-34499 , 4500 , 4501anup49::NUP49-yeGFP:HIS3; sgf73::kanMX4; leu2∆0; ura3∆0; met15∆0; clones 1-39360 , 9361 , 9362anup170::NUP170-yeGFP:HIS3; sgf73::kanMX4; spt7::SPT7-linker-NUP49:hphNT1; leu2∆0; ura3∆0; met15∆0; clones 1-39340 , 9341 , 9342anup170::NUP170-yeGFP:HIS3; sgf73::kanMX4; gcn5::GCN5-linker-SAC3:hphNT1; leu2∆0; ura3∆0; met15∆0; clones 1-35520ahis3∆1; leu2∆0; ura3∆0; met15∆0 ( EUROSCARF wt ) 5142agcn5::kanMX4; his3∆1; leu2∆0; ura3∆0; met15∆08108asgf73::kanMX4; his3∆1; leu2∆0; ura3∆0; met15∆010 , 453afob1::kanMX4; his3∆1; leu2∆0; ura3∆0; met15∆09915 , 9916 , 9917afob1::kanMX4; gcn5::hphNT1; his3∆1; leu2∆0; ura3∆0; met15∆0; clones 1-36626agcn5::GCN5-linker-SAC3:hphNT1; his3∆1; leu2∆0; ura3∆0; met15∆07252aspt7::SPT7-linker-NUP49:hphNT1; his3∆1; leu2∆0; ura3∆0; met15∆06632 , 6633 , 6634agcn5::GCN5-linker-SAC3:hphNT1; sgf73::kanMX4; his3∆1; leu2∆0; ura3∆0; met15∆0; clones 1-37540 , 7541 , 7542aspt7::SPT7-linker-NUP49:hphNT1; sgf73::kanMX4; his3∆1; leu2∆0; ura3∆0; met15∆0; clones 1-35532ansg1::NSG1-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆04233 , 7303 , 7304abud6::natNT2; nsg1::NSG1-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆0; clones 1-36897 , 8105 , 8106ayku70::hphNT1; nsg1::NSG1-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆0; clones 1-36766 , 6767 , 8104agcn5::natNT2; nsg1::NSG1-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆0; clones 1-34601 , 4615 , 4616asgf73::kanMX4; nsg1::NSG1-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆0; clones 1-32177αnup49::NUP49-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆03403 , 3583 , 4223αbud6::kanMX4; nup49::NUP49-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆0; clones 1-34094 , 4095 , 4096αgcn5::natNT2; nup49::NUP49-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆0; clones 1-38109anup2::NUP2-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆08110anup60::NUP60-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆08111amlp1::MLP1-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆08112anup170::NUP170-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆07828anup49::NUP49-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆07336anup1::NUP1-yeGFP:HIS3; leu2∆0; ura3∆0; met15∆07503 , 8113 , 8114anup1::NUP1-linker-TETR-mCherry:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35933 , 5934 , 5935anup49::NUP49-linker-TETR-mCherry:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35881 , 5882 , 5883anup170::NUP170-linker-TETR-mCherry:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35930 , 5931 , 5932amlp1::MLP1-linker-TETR-mCherry:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-35843 , 5844 , 5845anup60::NUP60-linker-TETR-mCherry:hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-310 , 789aspt20::kanMX4; his3∆1; leu2∆0; ura3∆0; met15∆010 , 804 , 10 , 805 , 10 , 806aspt20::kanMX4; fob1::hphNT1 his3∆1; leu2∆0; ura3∆0; met15∆0;clones 1-311 , 120arDNA::ADE2-CAN1; ade2-1; his3-11; leu2-3-112; ura3-1; trp1-111 , 137 , 11 , 252 , 11 , 253agcn5::HIS3; rDNA::ADE2-CAN1; ade2-1; his3-11; leu2-3-112; ura3-1; trp1-1 clones 1-311 , 144 , 11 , 145 , 11 , 146afob1::hphNT1; rDNA::ADE2-CAN1; ade2-1; his3-11; leu2-3-112; ura3-1; trp1-1 clones 1-311 , 150 , 11 , 151 , 11 , 152asgf73::hphNT1; rDNA::ADE2-CAN1; ade2-1; his3-11; leu2-3-112; ura3-1; trp1-1 clones 1-311 , 153 , 11 , 154 , 11 , 155asir2::hphNT1; rDNA::ADE2-CAN1; ade2-1; his3-11; leu2-3-112; ura3-1; trp1-1 clones 1-310 , 998 , 10 , 999 , 11 , 000asac3::natNT2; gcn5::hphNT1; pPCM14 ( 224 tetO-REC-URA3-CEN-REC-LEU2 ) ; spc42::SPC42-CFP:kanMX4; leu2::TETR-GFP:LEU2; his3::PGAL-REC:HIS3; trp1::GAL4-EBD:TRP1; ade2-101; clones 1-3 For microscopy , cells were resuspended in non-fluorescent medium ( SD–TRP ) and either mounted on a glass cover slip ( for still images or movies of max . 5 min duration ) or immobilized on a 2% agar pad containing non-fluorescent medium . To visualize , the accumulated plasmids cells were grown on SD–URA to maintain the centromere on the plasmid , then shifted to SD -LEU containing β-Estradiol ( 1 µM final concentration; Sigma-Aldrich , St . Louis , MO ) to induce excision of the centromere ( rec-CEN4-URA3-rec ) and to select for the presence of the plasmid . After 16–18 hr , a Deltavision microscope ( Applied Precision ) equipped with a CCD HQ2 camera ( Roper ) , 250W Xenon laps , Softworx software ( Applied Precision ) , and a temperature chamber set to 30°C was used to visualize the plasmids and the depicted proteins tagged with GFP . Time lapse movies of 5 min , intervals for 1h taking 20 Z-stacks of 0 . 4-μm spacing , were recorded with a 100×/1 . 4 NA objective and 2x2 binning . To determine the total NPC fluorescence , integrated density was measured of the nucleus and background in sum projections using Image J ( National Institutes of Health ) . After subtraction of the background , the fluorescence was normalized by the median fluorescence of cells without plasmids . To determine fluorescence intensity , the mean gray value of the NPC area adjacent to the accumulated circles ( Ic ) and the residual NPC area ( Ir ) was measured and normalized by the median intensity of Ir . For the percentage fluorescence segregated to the mother cell , the integrated density of the nucleus in the mother and the bud was measured after nuclear division , and after subtraction of the background the fluorescence in the mother was divided by the total fluorescence . The nuclear area depicts the 2D area of the nucleus in a sum projection in the mother and daughter of telophase cells . The mean intensity is the total intensity divided by the nuclear area . Box plots represent the distribution of the measurements in single cells whereas whiskers reach from min to max , the box covers 50% of the data and the line depicts the median . To investigate plasmid propagation , the cells were kept on SD–URA medium and diluted in YPD medium containing estradiol only 3 hr before imaging . The cells were washed once using 1× PBS buffer and then resuspended in non-fluorescent medium . Images were obtained using an Olympus BX50 microscope , equipped with a piezo motor , a monochromator light source , a CCD camera ( Andor ) , and the TillVision software ( TillPhotonics ) . Images were acquired with a 100x / 1 . 4 NA objective , 2 × 2 binning , and stacks of 20 focal slices . Maximal intensity projections were used to determine plasmid propagation ( described below ‘plasmid retention assay’ ) . To determine individual plasmid localization , accumulated plasmid co-localization with Gcn5 and Sgf73 , nuclear geometry , and anaphase duration; a Deltavision microscope was used ( described above ) . Prior to analysis , the images were deconvolved . For plasmid localization , images were taken with an Olympus 100×/1 . 4 NA objective and 1 × 1 binning . 15 focal sections with 0 . 3-µm spacing were acquired . The focal plane in which the plasmid was the brightest was used to determine whether the plasmid was at the nuclear rim or at a clearly resolvable distance from the nuclear periphery . Nuclear geometry was analyzed using an Olympus 100×/1 . 4 NA objective , 1 × 1 binning , and a 1 . 6× auxiliary magnification . Stacks of 15 sections and 0 . 2-µm spacing were obtained . Nuclear length and width was measured as illustrated in Figure 4B using ImageJ , and average nuclear width was calculated for different categories of nuclear length . Anaphase duration was determined using an Olympus 60×/1 . 42 NA objective and 1 × 1 binning . 30 min movies were taken with 1 min between each stack of 8 sections and 0 . 6-µm spacing . Early anaphase was measured form the first time point the nucleus crossed the bud neck until the nucleus contracts as shown in the cartoon and representative pictures in Figure 4A . To investigate the enrichment of NPCs close to non-centromeric plasmids , a Core Deltavision containing a CCD HQ camera ( Roper ) and solid state LEDs was used . An Olympus 100×/1 . 4 NA objective , 1 × 1 binning was used to create stacks of 15 focal sections and 0 . 2-µm spacing . Using ImageJ fluorescence intensity was measured along the nuclear rim . Nup82-3sfGFP intensity was normalized by its average intensity and TetR-mCherry intensity was normalized by its highest intensity in each cell . Afterwards , all obtained traces were aligned by the peak of TetR-mCherry ( plasmid ) intensity . To determine the speed of the non-centromeric plasmids , cells were imaged 3h after centromere excision using the same Deltavision with LEDs . First a Z-stack of 16 sections and 0 . 3-μm spacing was recorded in CFP and YFP to visualize both the plasmid and the SPBs in order to select for the non-centromeric plasmids . Then the plasmid was followed in a time lapse movie , taking the same stacks but only in GFP every 3 s for 3 min . The plasmids were tracked using Imaris ( Bitplane ) to determine their three dimensional coordinates . The coordinates were then used to determine the distance covered by the plasmid for each time point , which was deviated by three to determine the distance per second . All speed values were averaged for every plasmid , then averaged over all non-centromeric plasmids recorded and plotted in the graph . Both FLIP and FRAP experiments were essentially done as described in Boettcher et al . 2012 . Briefly , to measure the dynamics of NPCs in cells with and without accumulated plasmids , cells were grown in SD -URA , restreaked on SD–LEU containing β-estradiol ( 1 µM final concentration; Sigma ) 16–18 hr prior to imaging , and then immobilized on a 2% agar pad containing non-fluorescent medium . A Zeiss LSM 780 laser confocal microscope ( Zeiss Microimaging ) with a 63×/1 . 4 NA objective and a multiarray 32PMT GaAsP detector controlled by the ZEN 2011 software ( Carl Zeiss ) was used for imaging . Bleaching was performed using a 488 nm argon laser at 2 . 6% laser intensity and 45% output . Repeated photobleaching was performed in the depicted area with 9 s intervals for 30 time points . Afterwards the fluorescence intensity of Nup170-GFP was measured in the vicinity of the circles ( Ic ) and an area of the same size equidistant from the bleaching region ( Io ) using ImageJ . Both intensities were set to 100% prior to the first bleaching . Fluorescence decay was plotted for each cell using Prism ( GraphPad Software ) and t70 was defined as the time after which 70% of the initial intensity was still recorded . Finally , the average t70 of several cells was calculated for both Ic and Io and shown in the graph . The same settings were used to investigate the fluorescence decay in the whole nucleus , whereas the whole nuclear area was measured and set to 100% prior to bleaching . The mean and standard deviation of individual cells was plotted in the graph . To determine the strength of the nuclear diffusion barrier , a Zeiss LSM 510 microscope was used . Early anaphase cells were selected and photobleaching was applied on the depicted areas . To analyze the FLIP experiments , the total integrated fluorescent density in both the mother and daughter part of the nucleus was measured . After subtracting the average fluorescent loss determined by 5 neighboring cells , the fluorescent signal of the mother and daughter part was set to 100% at the beginning of the experiment . All experiments were pooled and analyzed using Prism to fit a one-phase decay curve . The barrier index was defined as the ratio of the times needed to lose 40% of the initial fluorescent signal in the bud over the mother compartment . For the FRAP experiments , the whole bud compartment of early anaphase nuclei was bleached and fluorescent recovery was measured in a constant area located in the bud . After subtracting the average fluorescent loss by imaging based on three neighboring cells , the fluorescence after photobleaching was set to 0% . The fluorescence recovery was fitted to a one-phase association curve for each cell using Prism . T15 was defined as the time to recover 15% fluorescence . The average of t15 of all measured cells was calculated and is shown in the graph . Late anaphase or telophase cells containing one , two , or four plasmids that did not co-localize with the spindle pole body were analyzed . Propagation frequency ( pf ) was calculated as; 100- ( xa+2y√b+4z4√c ) / ( x+2y+4z ) , where a , b , and c being the percentages of cells retaining all plasmids in the mother cell and x , y , and z being the number of cells counted for one , two , or four plasmids , respectively . Retention of pDS316 ( Sinclair and Guarente , 1997 ) was analyzed using freshly transformed wt and mutant cells . Cells were grown on SD–ADE plates and then transferred to plates containing low amounts of adenine 3h before micromanipulations . Using microdissection techniques , mother cells were separated from their daughter cells . After 3–5 days , both the mother and the daughter cells formed a colony , whereas the appearance of white sectors depicted whether the original cell contained pDS316 . Old cells were purified following the protocol of the mother enrichment program ( Lindstrom and Gottschling , 2009 ) . Briefly , 2 . 5 × 108 cells were labeled with Sulfo-NHS-LC-Biotin ( Pierce ) and recovered for 2hr at 30°C prior to the addition of β-Estradiol ( 1 µM final concentration ) . After either 26 hr and 27 . 5 hr or 34 hr and 35 . 5 hr ( for wt and mutant cells ) cells were washed , diluted in PBS , and incubated with 25 µl streptavidin-coated magnetic beads ( MicroMACS , Miltenyi Biotec ) for 30 min at 4°C before purification using LS MACS columns ( Miltenyi Biotec ) . Cells were eluted using 1× PBS containing 2 mM EDTA , pelleted , and split into two fractions: 1 ) cells were incubated in 20 µg/ml calcofluor white and bud scars were counted using fluorescence microscopy; 2 ) cells were lysed and DNA was purified using standard methods . DNA content was measured performing qPCR amplifying ACT1 as follows: genomic DNA from wt non-aged cells was used to generate a dilution curve . Simultaneously , three qPCRs for two different dilutions were run for each genomic DNA sample derived from the aged wt , gcn5∆ and sgf73∆ mutant cells . The average of the six qPCR values obtained for each strain was used to calculate the volume loaded on a 0 . 6% agarose gel containing ethidium bromide . The gel was run in TBE containing ethidium bromide for 24 hr at 50 V at 4°C . The gel was blotted to transfer membrane ( magnacharge nylon , GE Water&Process Tech . ) using standard protocols . Membranes were hybridized with a 32P-labeled probe specific to the rDNA as described ( Lindstrom et al . , 2011 ) and visualized using a Typhoon phosphoimager ( GE healthcare ) . Cells containing one copy of the ADE2 gene inserted in the rDNA array were grown over night in SD -ADE medium to prevent premature loss of the marker . After dilution the cells were grown for 4–6 hr , diluted again and plated on YPD plates lacking extra Adenine . Colonies were grown for 3 days at 30°C and then shifted to 4°C for two days before analysis . Half-sectors were counted and the total amount of colonies was assayed using an automated colony counter developed by Sandra Guetg ( Barral lab , unpublished ) . ChIP analysis was essentially done as described ( Chymkowitch et al . , 2012 ) : Cells were fixed with 1% ( wt/vol ) formaldehyde for 30 min and quenched for 5 min using glycine ( 125 mM final concentration ) . Cells were washed with Tris-buffered saline , resuspended in lysis buffer , and lysed in a cooled freezer mill ( Spex ) for 6 cycles ( 2 min with 2 min breaks each at 12 cps ) . Afterwards , the samples were sonicated twice for 15 min by altering cycles of 30 s pulses followed by 30 s cool down periods using a Bioruptor ( Diagenode ) . After centrifugation , supernatants were immunoprecipitated overnight at 4°C on anti-GFP beads ( Chromotek ) . Washing steps and elution were performed as described ( Chymkowitch et al . , 2012 ) . After RNase and proteinase K treatment , DNA fragments were purified using QIAquick PCR purification kit ( Qiagen ) and were analyzed by qPCR using a rotorgene qPCR system ( Qiagen ) . Primer sequences were as follows: ChrV: GAACAAGGTTACAAATCC and CCGATTTGTGAGATTCTTCCT ARS: AGTAAAGATTTCGTGTTCATGCAG and ATAGTGAAGGAGCATGTTCGG non-coding sequence ( NCS1 ) : GATACCTGAGTATTCCCACAG and CACACCGCATATGGTGCACTC non-coding sequence ( NCS2 ) : CCGCTTACCGGATACCTGTC and ATCCTGTTACCAGTGGCTGC Fold enrichment was calculated as IP/Input of protein X normalized by the same ratio measured in wt cells , where no protein was tagged with GFP to correct for unspecific binding . Pore segregation during ageing was investigated using the microfluidics dissection platform ( Lee et al . , 2012; Huberts et al . , 2013 ) with constant flow of synthetic complete medium at 1 μl/min and an Eclipse Ti fluorescent microscope ( Nikon Instruments ) controlled by Micro-Manager 1 . 4 software ( μManager ) . Images were acquired with a Plan Apo lambda 60×/1 . 4 NA objective and a Orca R2 CCD camera ( Hamamatsu ) . Cell growth and budding events were monitored by capturing images using transmitted light every 20 min . Fluorescent images were taken after 24 hr and 36 hr for short-lived strains ( e . g . sir2Δ cells ) and after 48 hr and 64 hr for normal or long-lived strains using a mercury lamp as a light source ( Intensilight , Nikon Instruments ) . Stacks of 14 slices covering a range of 7 μm were captured every 15 min for 2 hr , to follow the nuclear division . Nuclei of cells of different ages were analyzed using sum projections of the fluorescence images from both time points ( 48 hr and 64 hr ) . Total fluorescence ( integrated density ) was measured in the mother and daughter cell after nuclear division using ImageJ . After background subtraction , the values were grouped by the indicated age categories of the mother cells , normalized by the median of the daughter cells in the youngest age category , and plotted using Prism . Alternatively , nuclear fluorescence in the mother cells was divided by the total fluorescence ( mother plus daughter nuclei ) for each cell and plotted against the age of the mother cell to visualize the increase in asymmetry with age . This data were used to fit a one-phase association curve , which is shown in the graph . Additionally , 10 values were grouped age dependently and plotted showing the average age and percentage of fluorescence segregated to the mother cell as well as the standard deviations . For G1 cells , the total amount of NPCs ( integrated density ) was measured , plotted against their age , and an exponential curve was fitted using Prism . The values were then normalized by intersection of the curve with the Y-axis and set to 1 . The dots on the graph were created from 10 values as described for the dividing cells . Similarly , the mean fluorescence intensity ( integrated density divided by area ) was plotted as well as the radius of the G1 nuclei , measured drawing a line through the nucleus at its broadest area using ImageJ , dividing it by two and plotting it against the age of the cell . Pedigree analysis was done as described ( Shcheprova et al . , 2008 ) : freshly streaked cells were restreaked on YPD plates and 2 hr later virgin daughter cells were placed to analyze their lifespan . Over-night the cells were kept at 4°C ( max . 10-12 hr ) . Micromanipulation was performed using a Zeiss Axioscope 40 microdissection microscope ( Zeiss ) . Lifespan curves were analyzed using Prism . One-way ANOVA followed by Dunnett's multiple comparison or an unpaired two-tailed t-test was used to test for significance except the lifespan curves , which were compared using Log-rank ( Mantel–Cox ) tests and the ChIP experiment , for which a paired t-test ( within the same experiment ) was used . | Budding yeast is a microorganism that has been widely studied to understand how it and many other organisms , including animals , age over time . This yeast is so named because it proliferates by ‘budding’ daughter cells out of the surface of a mother cell . For each daughter cell that buds , the mother cell loses some fitness and eventually dies after a certain number of budding events . This process is called ‘replicative ageing’ , and it also resembles the way that stem cells age . In contrast , the newly formed daughters essentially have their age ‘reset to zero’ and grow until they turn into mother cells themselves . Several molecules or factors have been linked to replicative ageing . These are retained in the mother cell during budding , rather than being passed on to the daughters . Non-chromosomal DNA circles , for example , are rings of DNA that detach from chromosomes during DNA repair and that accumulate inside the ageing mother cell over time . How the mother cells retain these circles of DNA is an on-going topic of debate . Similar to plants and animals , chromosomes in yeast cells are confined in a membrane-bound structure known as the cell nucleus . The nuclear membrane is perforated by channels called nuclear pore complexes that ensure the transport of molecules into , and out of , the nucleus . Now , Denoth-Lippuner et al . establish that for the non-chromosomal DNA circles to be efficiently confined in the mother cell , the DNA circles must be anchored to the nuclear pore complexes . Denoth-Lippuner et al . next asked how the DNA circles were anchored to these complexes; and found that another complex of proteins known as SAGA is involved . When components of the SAGA complex were deleted in budding yeast cells , non-chromosomal DNA circles spread into the daughters as well . On the other hand , artificially anchoring these DNA circles to the nuclear pore complex alleviated the need for the SAGA complex , in order to retain these molecules in the mother cell . Denoth-Lippuner et al . also show that SAGA-dependent attachment of the DNA circles to the nuclear pore complexes causes these complexes to remain in the mother cell . As a consequence , these nuclear pore complexes accumulate in the mother cells as they age . The number of nuclear pore complexes in the daughter cells , however , remained fairly constant . Together these data raise the question of whether the effects of DNA circles on the number and activity of the nuclear pores might account for their contribution to ageing , perhaps by affecting the workings of the nucleus . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2014 | Role of SAGA in the asymmetric segregation of DNA circles during yeast ageing |
Conceptual metaphors are linguistic constructions . Such a metaphor is humans’ mental representation of social rank as a pyramidal-like structure . High-ranked individuals are represented in higher positions than low-ranked individuals . We show that conceptual metaphorical mapping between social rank and the representational domain exists in our closest evolutionary relatives , the chimpanzees . Chimpanzee participants were requested to discriminate face identities in a vertical arrangement . We found a modulation of response latencies by the rank of the presented individual and the position on the display: a high-ranked individual presented in the higher and a low-ranked individual in the lower position led to quicker identity discrimination than a high-ranked individual in the lower and a low-ranked individual in the higher position . Such a spatial representation of dominance hierarchy in chimpanzees suggests that a natural tendency to systematically map an abstract dimension exists in the common ancestor of humans and chimpanzees .
“high” vs “low status” , “top of the heap” , “bottom of the barrel”: These or similar expressions are widely observed across cultures and languages ( Pinker , 1997 ) . The cross-modal correspondence between the visuospatial domain ( high or low ) and an abstract domain ( rank ) has been described as a conceptual metaphor ( Lakoff and Johnson , 1980a , b ) and thought to be uniquely human ( Feldman and Narayanan , 2004 ) . A conceptual metaphor takes one concept and connects that to another concept in order to better understand that concept . The way we think and act is largely influenced by conceptual metaphors , even without being fully aware of them ( Lakoff and Johnson , 1980a ) . The question remains if conceptual metaphorical mapping is indeed uniquely human or if it appears in other primates and thus describes a conceptual metaphorical mapping that predates language . To answer this question , we examined if our evolutionary closest relatives , the chimpanzees , have conceptual metaphors as we humans do . Decades of research have shown that neural representations of objects and entities exist in monkeys ( Sigala et al . , 2002 ) , apes ( Fukushima et al . , 2010 ) and humans ( Kourtzi and Kanwisher , 2001 ) . The abilities of object representation and recognition further apply to social domains such as recognition of faces ( Dahl et al . , 2007; Dahl et al . , 2013 ) , conspecifics ( Dahl et al . , 2007 ) , and ingroup-outgroup members ( Pokorny and de Waal , 2009 ) . To avoid or reduce costly social conflicts among individuals , a crucial skill for living in social groups is to express one’s status and to recognize the status of the others via visual or vocal cues ( Paxton et al . , 2010 ) . Recognition of status or rank allows inferences about expected roles of oneself and of others during group situations ( Ridgeway and Diekema , 1989 ) . Access to food resources , information and social respect are facilitated with high rank in the hierarchy , while some degree of protection and care are granted to lower-ranked individuals ( Fiske , 1992 ) . In many non-human species , spatial information , such as perceived physical body size , and facial and body postures , serve as an indicator of social status ( Maestripieri , 1996; Tiedens and Fragale , 2003 ) . Besides these non-verbal cues , we humans developed a conceptual metaphor , which connects social rank to the spatial domain ( e . g . , “high” vs “low status” [Pinker , 1997] ) . Accordingly , humans represent social status in a pyramidal-like structure ( Bosserman , 1983 ) : high-ranked individuals are represented in spatially higher positions than low-ranked individuals in diverse contexts , for example , stratification ( Saunders , 1989 ) , organizations ( Weber , 1991 ) , religion , family , human needs ( Maslow , 1943 ) , and others . In this study , we examined if such conceptual metaphorical mapping between social dominance and the spatial domain is uniquely human or if it appears in our evolutionary closest relative , the chimpanzees . We addressed this question by comparing response latencies on discriminating photographs of familiar chimpanzee faces of high- and low-ranked individuals in a vertically aligned delayed matching-to-sample task . We hypothesize that coherent arrangements , such as a high-ranked individual presented in the higher and a low-ranked individual presented in the lower position , lead to quicker identity discrimination than incoherent arrangements , such as a high-ranked individual in the lower and a low-ranked individual in the higher position .
We sequentially presented a cue ( individual 1 , 750 ms ) followed by an inter-stimulus interval ( 500 ms ) and two vertically arranged stimuli ( match: individual 1; distractor: individual 2 ) ( Figure 1A , B ) . The chimpanzees were required to indicate which of the two simultaneously presented faces ( match , distractor ) corresponds to the initially presented face ( cue ) in identity by touching one of the faces . Note that the participants did not classify the stimuli based on social rank . In addition to coherent and incoherent combinations of stimulus and position , we included a neutral condition combining two pictures of closely ranked individuals in a trial ( close condition ) . In the first step of analyses , we pooled response latencies for coherence ( coherent vs incoherent vs close ) and position ( high vs low ) . Using a mixed model ANOVA with coherence and position as fixed factors , we found a main effect for coherence ( F ( 2 , 30 ) = 5 , m . s . e . = 1 . 10e+004 , p<0 . 001 ) , but not for position ( p>0 . 34 ) or the interaction between the two factors ( p>0 . 83 ) ( Figure 1C ) . Post-hoc t-tests ( Bonferroni-corrected for multiple comparisons ) revealed a significant response facilitation for coherent as opposed to incoherent trials ( t ( 10 ) = -3 . 20 , m . s . e . = 8 . 60e+003 , p<0 . 01 [one-tailed] ) and close trials ( t ( 10 ) = −1 . 92 , m . s . e . = 1 . 14e+004 , p<0 . 05 [one-tailed] ) . However , there was no significant deterioration for incoherent compared to close trials ( p>0 . 11 ) ( Figure 1C ) . In the second step of analyses , we pooled response latencies for position ( high vs low ) and compared the differences of coherent and incoherent trials in one-sample t-tests . We found significant deviations from zero for the high position ( t ( 5 ) = 16 . 64 , m . s . e . = 1 . 11e+004 , p<0 . 001 [one-tailed] ) and for the low position ( t ( 5 ) = 20 . 13 , m . s . e . = 6 . 88e+003 , p<0 . 001 [one-tailed] ) ( Figure 1F ) . Further , there was no significant difference between the positions high and low ( p>0 . 55 ) . 10 . 7554/eLife . 00932 . 003Figure 1 . Task sequence , example stimuli and response latency analyses . ( A ) Typical trial sequence . ( B ) Stimulus exemplars . ( C ) Average response latencies for coherent , incoherent and close trials ( mean ± SEM ) for all participants , ( D ) for high-ranked participants and ( E ) for low-ranked participants . ( F ) Average response latency differences ( coherent–incoherent ) for high and low positions . ( C , F ) The number of independent data points ( N ) is six for each condition . ( G ) Normalized frequency distribution of response latencies of high- and low-ranked participants for coherent , incoherent and close trials . ( H ) Sensitivity index for both stimulus sets and stimuli . Positive values indicate facilitation for coherent relative to incoherent trials . DOI: http://dx . doi . org/10 . 7554/eLife . 00932 . 003 We further split the participants into two groups according to their own ranks . This is equivalent to a separation by the stimulus sets ( set 1 for the high-ranked and set 2 for the low-ranked participants , see ‘Matarials and methods’ ) . High-ranked participants showed the same response patterns as low-ranked participants ( Figure 1D , E ) . Due to the low sample size , we were unable to run statistical tests across participants; however , we collapsed all data samples of high-ranked participants for coherent , incoherent and close conditions and compared them to the corresponding conditions in low-ranked participants ( Figure 1G ) . Two-sample Kolmogorov-Smirnov tests revealed that none of the frequency distributions of response latencies were statistically different across the two participant groups ( all p>0 . 54 ) . In addition , we calculated a sensitivity index reflecting the amount to which two sample distributions are separable from each other ( also referred to as d-prime ) . We again split the data according to the rank of the individuals ( high- vs low-ranked ) ( equivalent to splitting according to the stimulus sets ) . We further binned data samples ( response latencies ) across participants for coherent and incoherent trials . We then calculated the sensitivity index for each stimulus in combination with each other stimulus with which it occurred in the task ( Figure 1H , x-axis = distractor; y-axis = cue and match ) by considering the means and standard deviations of the corresponding data bins of coherent and incoherent trials ( Equation 1 , ‘Materials and methods’ ) . Positive values illustrate response facilitation for coherent trials relative to incoherent trials . As indicated in Figure 1D , E , there is more variance in the responses of the low-ranked ( Figure 1E , four participants ) than the high-ranked participants ( Figure 1D , two participants ) , leading to a lower sensitivity in low-ranked participants ( t ( 14 ) = 2 . 53 , m . s . e . = 8 . 36e+003 , p<0 . 05 , Figure 1H ) . Importantly , all stimuli elicited sensitivity scores in accordance with the hypothesis , that is , coherent stimulus arrangements led to response facilitation and incoherent stimulus arrangements led to response deterioration , indicated by positive values .
We showed that in chimpanzees discrimination performances between familiar conspecific faces are systematically modulated by the location and the social status of the presented individuals , leading to discrimination facilitation or deterioration . Coherent arrangements as opposed to incoherent arrangements led to a facilitation of recognition . Further , both , a high-ranked individual at the higher position and a low-ranked individual at the lower position caused recognition facilitation equivalently . Importantly , the participants were not trained on discriminating the ranks of the presented individuals . Instead , they were substantially affected by the rank while discriminating the identity of those individuals . The modulations are in accordance with a spatial arrangement representing high-ranked individuals at the top and low-ranked individuals at the bottom , hence reflecting the inverse function of response latencies and spatial distance of two individuals on the display and in the mental hierarchy space of the participants . It has to be noted that a confound between perceptual and conceptual determinants might exist . We accounted for potential confounding perceptual cues , such as physical height , gestures and postures ( Ridgeway and Diekema , 1989; Paxton et al . , 2010 ) . However , teasing apart perceptual and conceptual determinants entirely is almost impossible . There might be a co-dependence of the two domains , with the conceptual domain having been established on the basis of the perceptual domain . We applied the following control for a perceptual confound: If perceptual cues cause the effect , for example high-ranked individuals would naturally appear in higher positions than low-ranked individuals , the response characteristics of high-ranked and low-ranked participants would be different due to the rank difference between participant and the presented individuals: A high-ranked participant is more likely to show response facilitation for low-ranked individuals , while a low-ranked participant is more likely to show response facilitation for high-ranked individuals . This , however , is not the case . Response characteristics are similar in high-ranked and low-ranked participants . Thus , even though perceptual determinants cannot be fully excluded with this control , it is still suggestive that the effect is not merely based on perceptual cues but on , to some extent , conceptual components . In addition , we controlled for a differential effect due to one of the two stimulus sets used in the experiment . We showed that individual stimuli elicited a comparable effect within and across stimulus sets by estimating sensitivity indices . In other words , the two stimulus sets contributed equally to the effect . A spatial component of representation has been shown in other domains: for example in humans’ responses to low-digit numbers are faster with a left-side button-press whereas higher digits are categorized faster when right-side button-presses are required ( Dehaene et al . , 1993 ) . Moreover , merely looking at number causes a shift of attention to the left or the right side ( Fischer et al , 2003 ) . In other words , the mental number line reflects a cross-modal mapping of visual cues ( numbers ) and cognitive labels ( values ) . Interestingly , social status and number comparisons recruit to some extent overlapping neural substrates in the intraparietal sulci of the human brain ( Chiao et al . , 2009 ) . There is evidence for a non-verbal , supramodal neural representation of numerosity in the macaque ventral intraparietal sulcus and lateral prefrontal cortex ( Nieder , 2012 ) ; however , neural evidence for cross-modal correspondences is missing . Rare evidence for non-human cross-modal correspondences comes from visuo-auditory mappings between high luminance and high pitch in chimpanzees ( Ludwig et al . , 2011 ) . This relationship between luminance and pitch illustrates a form of sound-symbolism , which refers to the concept that in human language words and referents are not arbitrary ( Nuckolls , 1999 ) . Hence , while the existence of such a systematic mapping between luminance and pitch in chimpanzee suggests the emergence of an early vocabulary of human language ( Ramachandran and Hubbard , 2001 ) , we here expand this finding to a cross-modal correspondence between vision and an abstract domain: the social status . A natural tendency to systematically map an abstract dimension , such as social status , in our closest and nonlinguistic relatives , the chimpanzees , suggests that this tendency had already evolved in the common ancestors of humans and chimpanzees . This tendency might have influenced the emergence of metaphorical linguistics , thinking via image schema , an embodied pre-linguistic structure of experience that motivates conceptual metaphor mappings ( Lakoff and Johnson , 1980a ) . According to Lakoff ( Lakoff and Johnson , 1980b ) , orientational metaphors , such as ‘more is up’ , ‘good is up’ and ‘dominant is up’ , are based on observational correlation between increasing a substance and seeing the level of the substance rise , like adding an element to a pile . Given the strong physical basis , these metaphors are good candidate for universal concepts . Until now , conceptual metaphors have been exclusive human experiences . Our findings point in a different direction .
Six chimpanzees ( Pan troglodytes; 1 male juvenile , 2 female juveniles [both around 11 years] and 3 female adults [both around 31 years] ) participated in this study . The chimpanzees live in groups of 14 individuals with access to environmentally enriched outdoor ( 770 m2 ) as well as indoor compounds . The chimpanzees participated in a variety of computer-controlled tasks in the past ( Matsuzawa , 2003; Matsuzawa et al . , 2006 ) . They are experienced in horizontally aligned delayed matching-to-sample ( DMS ) tasks; however , they are inexperienced in a vertical version of a DMS task . Effects of training in the vertically aligned DMS task can be ruled out . The vertical spacing of the match and distractor stimuli was about 70 mm . Stimuli were presented at a 17-inch LCD touch panel display ( 1280 × 1024 pixels ) controlled by custom-written software under Visual Basic 2010 ( Microsoft Corporation , Redmond , Washington , USA ) . The stimulus size was approximately 4 . 5 by 6° of gaze angel . One degree of gaze angle corresponded to approximately 0 . 86 cm on the screen at 50 cm viewing distance . Below the display a food tray was installed in which pieces of food reward were delivered by a custom-designed feeder after completion of a correct trial . Chimpanzee participants sat in an experimental booth ( 2 . 5 m wide , 2 . 5 m deep , 2 . 1 m high ) , with the experimenter and the participants separated by transparent acrylic panels . We used photographs of faces of chimpanzee individuals with obvious dominant or submissive social ranks . The face pictures were taken from individuals familiar to the participants . All faces were normalized for luminance and contrast . The agreement of twenty independent raters ( researchers and caretakers familiar with the chimpanzees at the Primate Research Institute ) on the social ranks of the chimpanzees was found to be Kappa = 1 . 00 ( p<0 . 001 ) , 95% CI ( 1 . 00 , 1 . 00 ) . For the six participants ( Pan troglodytes; 1 male adolescent , 2 female adolescents and 3 female adults ) , the face stimuli varied according to the group particular participants belonged to . We only presented face stimuli from the same group as the participant . In total , we used two sets of four face stimuli each: stimulus set 1 for two participants and stimulus set 2 for four participants . Under the assumptions that high-rank individuals presented in spatially higher position and low-ranked individuals presented in spatially lower position lead to faster identification , while high-rank individuals in lower position and low-ranked individual in higher position lead to slower identification , we designed the experiment according to the conditions coherent and incoherent , referring to the coherence between social rank and spatial position of presentation on the screen . In addition , we compared stimuli of close distance in social rank , here referred to as the close condition , serving as the baseline condition . For this condition , there is no prediction for a crossmodal association of social rank and spatial position . The order of experimental conditions and the spatial position of match and distractor stimuli were counterbalanced . Each participant did six blocks with 48 trials each . Only correct trials went into the analyses , which is 69% ( ±4 . 5% S . E . M . ) of all trials . The dependent variable was response latencies . We conducted an analysis of variances among the participants using a mixed model ANOVA , with coherence ( coherent , incoherent and close conditions ) and position as a fixed factors ( high , low ) as well as two-sample t-test ( Bonferroni-corrected for multiple comparisons ) to compare individual experimental conditions . For the post-hoc analysis of coherence , we collapsed the data samples from high and low positions for all three conditions ( coherent , incoherent and close ) of each participant ( N = 6 ) . For the post-hoc analysis of position , we subtracted the incoherent condition from the coherent condition for high and low positions of each participant ( N = 6 ) . Further , the distribution of response latencies for each condition was binned into 12 equally sized bins and normalized by dividing the absolute frequency ( i . e . , the number of events in each bin ) by the total number of occurrences , resulting in the relative frequency with values ranging from 0 to 1 . These normalized distributions were compared using two-sample Kolmogorov-Smirnov tests . To compare the outcome by the two stimulus sets , we did the following analytical procedure: We split the data samples according to the stimulus set used for the participants and according the conditions coherent and incoherent . These four data sets were then binned according to all combinations of stimuli as they occurred in the experiment . In other words , we binned the response latencies for each stimulus showing a high-ranked individual in combination with both stimuli showing low-ranked individuals and , visa-versa , for each stimulus showing a low-ranked individual in combination with both stimuli showing high-ranked individuals ( Figure 1H ) . For both stimulus sets we took the distributions of response latencies for each stimulus combination of coherent and incoherent conditions and determined a sensitivity index , describing the separation of these distributions under consideration of the standard deviation using the following equation: ( 1 ) idx = μc1− μc212 ( σc12 + σc22 ) with c1 and c2 being two experimental conditions , µ the mean and σ the standard deviation . Positive values indicate facilitation for coherent above incoherent trials . | It is thought that the ability to connect an abstract concept to something physical helps us to understand abstract ideas more easily . Examples include the use of conceptual metaphors that draw parallels between something abstract , such as social status , and physical position , even though there is no connection between them: familiar examples include phrases such as ‘top dog’ or ‘upper class’ . It has long been assumed that the use of such conceptual metaphors is uniquely human . Many social animals have hierarchies of dominance within groups , with particular individuals being ranked above or below other individuals . Chimpanzees—our closest relatives in the animal kingdom—are a good example of this , and although their cognitive processes are known to be similar to those of humans in many ways , we do not know if they make use of conceptual metaphors . Moreover , we don’t even know if conceptual metaphors can exist in the absence of language . When researchers want to investigate how concepts are cognitively linked in the brain , they often use ‘coherent’ or ‘incoherent’ stimuli . A good example of an incoherent stimulus would be the word ‘red’ printed in blue ink . Because our neural representations of the colour blue and the word blue are linked , it is harder for a person to read the word red when it is printed in blue than when it is printed in red ( which would be a coherent stimulus ) . To test whether chimpanzees use a conceptual metaphor in which social status corresponds to height , Dahl and Adachi showed six chimpanzees photographs of four other chimpanzees who were known to them , and tested whether the relative positions of the photographs affected the ability of the chimpanzees to identify which of the two photographs they had been shown earlier . For example , a photograph of a high-ranked , dominant chimpanzee could be shown above a photograph of a lower-ranked chimpanzee ( a coherent stimulus ) or below a photograph of a lower-ranked chimpanzee ( an incoherent stimulus ) . The chimpanzees doing the tests had to identify which of the photographs they had been shown earlier by touching the correct photograph on a screen . Dahl and Adachi found that it took longer for chimpanzees to complete the task when the photograph was in the ‘wrong’ position . This suggests that the neural representations of social status and physical position might be linked in chimpanzees . If the social status and the physical position of the photograph match , the chimpanzee doing the test can quickly identify the photograph that it has been shown earlier . However , if they do not match , the conflict between the neural representations of social status and physical position slows down the response . These findings suggest that conceptual metaphors are not uniquely human and , moreover , that they could have emerged before the development of language . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2013 | Conceptual metaphorical mapping in chimpanzees (Pan troglodytes) |
Though horizontal gene transfer ( HGT ) is widespread , genes and taxa experience biased rates of transferability . Curiously , independent transmission of homologous DNA to archaea , bacteria , eukaryotes , and viruses is extremely rare and often defies ecological and functional explanations . Here , we demonstrate that a bacterial lysozyme family integrated independently in all domains of life across diverse environments , generating the only glycosyl hydrolase 25 muramidases in plants and archaea . During coculture of a hydrothermal vent archaeon with a bacterial competitor , muramidase transcription is upregulated . Moreover , recombinant lysozyme exhibits broad-spectrum antibacterial action in a dose-dependent manner . Similar to bacterial transfer of antibiotic resistance genes , transfer of a potent antibacterial gene across the universal tree seemingly bestows a niche-transcending adaptation that trumps the barriers against parallel HGT to all domains . The discoveries also comprise the first characterization of an antibacterial gene in archaea and support the pursuit of antibiotics in this underexplored group .
HGT is rampant among prokaryotes and phages and is an important mechanism for acquisition of new genes and functions ( Popa and Dagan , 2011 ) , including the shuttling of antibiotics and antibiotic resistance between bacteria ( Clardy et al . , 2009 ) . Instances of interdomain horizontal transfer of diverse genes between two domains of life or between viruses and their hosts are also increasingly documented ( Nelson et al . , 1999; Husnik et al . , 2013; Dunning Hotopp et al . , 2007; Wu et al . , 2013; Gladyshev et al . , 2008; Bratke and McLysaght , 2008; Danchin et al . , 2010 ) . While a minority of these transfers have been functionally investigated , the biological activity , selective advantages , and ecological contexts of many interdomain HGT events remain poorly characterized ( Dunning Hotopp 2011 , Keeling and Palmer , 2008 ) . In comparison to these intradomain or interdomain highways of HGT ( Beiko et al . , 2005 ) , independent transmission of the same gene family to archaea , bacteria , eukaryotes , and viruses is extremely uncommon and subject to apparently rare events throughout the history of life ( Moran et al . , 2012; Lundin et al . , 2010; Koonin et al . , 2003; McClure , 2001; McDonald et al . , 2012 ) . When taken together , genome-enabled studies suggest that horizontal gene transfers ( HGTs ) are biased and experience a frequency gradient that decreases from within domain > between two domains > between all domains of life ( Bruto et al . , 2013; Zhaxybayeva and Doolittle , 2011; Puigbo et al . , 2009; Andam and Gogarten , 2013 , 2011 ) . However , the fact that a large number of organisms from all domains of life have been artificially transformed with genes evolved in other taxa indicates that there is no fundamental obstruction to interdomain HGT when barriers to transfer are deliberately removed . One significant question then is why do highways of intra- or interdomain transfers occur more frequently than transfers to all domains in the universal tree of life ? There are at least two explanations . First , recurrent transfer of the same gene family may be limited by incompatible mechanics of gene transfer ( e . g . , transduction , transfection , plasmid exchange , isolation of eukaryotic genome in nucleus , separation of somatic and germline tissues in multicellular eukaryotes ) between domains compared to within domains . However , the individual success of gene transfers between any two domains of life , for example archaea and bacteria ( Nelson et al . , 1999; van Wolferen et al . , 2013 ) , bacteria and eukaryote ( Andersson , 2005; Bordenstein , 2007; Gladyshev et al . , 2008; Danchin et al . , 2010 ) , and archaea and eukaryote ( Andersson et al . , 2003; Schonknecht et al . , 2013 ) , suggests that these barriers may be minimal . Second , the selective barriers against HGT of the same gene to multiple taxa and preservation of the gene through evolutionary time are multifaceted given the potential costs associated with HGT ( Baltrus , 2013 ) , and that each recipient may not benefit from the trait conferred . In the case of selfish genetic elements , HGT provides a strong benefit to the gene itself , yet these genes can be detrimental to the host organism and select for the evolution of countermeasures to gene propagation ( Werren , 2011 ) . Thus , it may be necessary for a transferred gene to confer a benefit to its new host in order to be stably maintained in the host genome over the long term . Given these evolutionary dynamics , there may be very few niche-transcending genes ( Wiedenbeck and Cohan , 2011 ) , defined as genes that are useful in different physiological capabilities , cellular structures , and ecological niches , which repeatedly increase fitness of each recipient across the whole diversity of life and can be stably and repeatedly transferred between very divergent organisms . Among the few putative cases , there is a pore-forming toxin domain that appears to have been anciently transferred between diverse lineages ( Moran et al . , 2012 ) . However , the distribution of the transfer across the tree of life is unclear because archaeal sequences were not included in phylogenetic analyses due to low support values . Other candidate genes encode proteins involved in nucleotide metabolism , intramembrane proteolysis , or membrane transport , but the transfer events defy clear interpretations due to their deep antiquity in evolutionary time and the confounding issues of ancient paralogy ( Lundin et al . , 2010; Koonin et al . , 2003; McClure , 2001; McDonald et al . , 2012 ) . Moreover , these transfers are often not functionally validated in the recipient taxa . Here we demonstrate for the first time , to our knowledge , that a functional antibacterial gene family scattered across the tree of life in diverse ecological contexts . This bacterial gene encodes a glycosyl hydrolase 25 ( GH25 ) muramidase , a peptidoglycan-degrading lysozyme that hydrolyzes the 1 , 4-β-linkages between N-acetylmuramic acid and N-acetyl-D-glucosamine in the bacterial cell wall . Typically found in bacteria ( Cantarel et al . , 2009 ) , the lytic enzyme classically functions in cell division and cell wall remodeling ( Vollmer et al . , 2008 ) , while in bacteriophages they lyse bacterial peptidoglycan at the end of the phage life cycle ( Fastrez , 1996 ) . Although members of the GH25 muramidase family have been noted in other taxa ( Korczynska et al . , 2010; Nikoh et al . , 2010 ) , extensive analysis of their evolutionary history and functions have not been undertaken . We hypothesized that the transfer of antibacterial genes from bacteria to archaea and to eukaryotes bestows a niche-transcending adaptation that overcomes the barriers against repeated and evolutionarily stable HGT of the same type of gene across the tree of life .
During a homology search , we uncovered 75 nonredundant sequences ( E-values ≤ 10−12 ) of a bacterial GH25 muramidase in disparate taxa across the tree of life , indicating possible HGT of a bacterial gene to both eukaryotic and archaeal species as well as to phages . Putative HGT events were identified in the genomes of the plant Selaginella moellendorffii ( Banks et al . , 2011 ) , the deep-sea hydrothermal vent archaeon Aciduliprofundum boonei ( Reysenbach et al . , 2006 ) , the pea aphid Acyrthosiphon pisum ( International Aphid Genomics Consortium 2010 , Nikoh et al . , 2010 ) , and several species of fungi such as Aspergillus oryzae ( Machida et al . , 2005 ) . We verified the presence of the lysozyme gene in natural populations of selected HGT recipients by PCR and sequencing of the GH25 muramidase domain ( Figure 1—figure supplement 1 ) , including Aciduliprofundum field samples harvested from hydrothermal vents worldwide . We detected lysozyme genes in 9 out of 12 field isolates of Aciduliprofundum from deep-sea vents in the Atlantic and Pacific oceans , 5 out of 6 species in the plant genus Selaginella , and 8 out of 9 aphid species in the subfamily Aphidinae ( Supplementary file 1 ) . However , it is possible that PCR-negative strains actually do possess lysozyme genes and were simply more divergent than could be detected with our primers . At the protein level , sequenced field isolate lysozymes were relatively similar to each other in each clade , with 74% pairwise identity ( 385aa alignment ) amongst Aciduliprofundum sequences , 85 . 1% ( 203aa alignment ) amongst three intact Selaginella sequences , and 87 . 3% identity ( 93aa alignment ) amongst Aphidinae sequences , with more variable divergence between clades ( full identity matrix available in Supplementary file 2 ) . We also found lysozymes in two additional WO phages as part of an ongoing next generation sequencing project of Wolbachia viruses ( unpublished data ) . To rule out spurious bacterial contamination in these genomes and to confirm genomic integration of the lysozyme gene , we employed direct sequencing of PCR products amplified using primers inside the lysozyme gene paired with primers outside the gene for Aciduliprofundum field samples and S . moellendorffii . Incorporation of the lysozyme gene was verified in all cases tested ( Figure 1—figure supplement 2 ) . Additionally , Aciduliprofundum samples were grown in strict monocultures as determined by 16S amplicon monitoring . Integration of the A . pisum lysozyme has been previously established ( Nikoh et al . , 2010 ) . Flanking genes in the recipient genomes were non-bacterial on either side of the transferred lysozyme in each case ( Figure 1 , Supplementary file 3 ) , with two exceptions . A bordering gene in A . boonei , ADP-ribose-1”-monophosphatase ( App-1 ) , possesses both bacterial and archaeal homologs and a phylogenetic analysis suggests HGT unrelated to the lysozyme transfer ( Figure 1—figure supplement 3A ) . This transfer was likely between archaea and Thermotogae bacteria . The second exception is a GH2 hydrolase gene adjacent to the lysozyme in A . oryzae . This hydrolase has bacterial homologs in the phylum Actinobacteria , and recapitulates the same phylogenetic pattern seen in the GH25 muramidase ( see below , Figure 1—figure supplement 3B ) . Thus , it is likely that the lysozyme and GH2 hydrolase were transferred to fungi in a single event . GH2 hydrolases have a number of carbohydrate-degrading enzymatic abilities ( Cantarel et al . , 2009 ) , including β-galactosidase , β-mannosidase , and β-glucuronidase activities , suggesting that this gene may benefit the fungi by adding additional digestive/nutritive capabilities in Dikarya . 10 . 7554/eLife . 04266 . 003Figure 1 . Architecture of HGT candidates and surrounding genes . Each arrow represents an open reading frame transcribed from either the plus strand ( arrow pointing right ) or the minus strand ( arrow pointing left ) . The color of the arrow indicates the taxa the gene is found in based on its closest homologs . Black = Eubacteria , purple = virus , red = Archaea , green = Plantae , Orange = Fungi , Blue = Insecta , white = no known homologs , dashed line = present in multiple domains . The length of the arrows and intergenic regions are drawn to scale except where indicated with broken lines . The four paralogs of the lysozyme in S . moellendorffii occur on two genomic scaffolds with light green bands connecting homologous genes . Vertical arrows indicate the location of introns in the A . oryzae and A . pisum lysozymes . Abbreviations: Lys: lysozyme , gpW = phage baseplate assembly protein W , SH3: Src homology domain 3 , App-1 = ADP-ribose-1”-monophosphatase , PRT = phosphoribosyltransferase , LD = leucoanthocyanidin dioxygenase; IMP = integral membrane protein . A protein diagram for each lysozyme is drawn to scale with the light gray regions highlighting a conserved protein domain . *A . pisum diagram is based on Acyr_1 . 0 assembly and transcription data ( Nikoh et al . , 2010 ) ; the annotation in Acyr_2 . 0 is different . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 00310 . 7554/eLife . 04266 . 004Figure 1—figure supplement 1 . Presence of HGT lysozyme genes in field samples . ( A ) PCR amplifications of portions of the GH25 muramidase domain in the indicated taxa . All amplifications were Sanger sequenced to confirm integration . Primers used are listed in Table S3 . Abbreviations: Sb: S . braunii , Sm: S . moellendorffii , Su: S . uncinata , Ssa: S . sanguinolenta , Sst: S . stauntoniana , Sl: S . lepidophylla , E: East Pacific Rise , L: Lao Spreading Center , M: Mid-Atlantic Ridge , Pu: Pleotrichophorus utensis , Aa: Artemisaphis artemisicola , Ue: Uroleucon erigeronensis , Av: Aphis varians , Ap: Acyrthosiphon pisum , Al: Aphis lupini , Cs: Cedoaphis sp . , As: Aphthargelia symphoricarpi , Bs: Braggia sp . , - denotes water only control . ( B ) World map with approximate locations of A . boonei field samples . Those that tested positive for the GH25 muramidase domain are indicated by green stars and those that tested negative are indicated by red stars . Map is a public domain image from Wikimedia Commons . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 00410 . 7554/eLife . 04266 . 005Figure 1—figure supplement 2 . PCR amplifications testing genomic integration with primers within and outside of lysozyme genes . Primers used are listed in Supplementary file 1 and binding sites are indicated in gene diagrams with small black arrows . All integrations were confirmed with Sanger sequencing . Abbreviations: Sm: S . moellendorffii , L: Lao Spreading Center , - denotes water only control , CHP = conserved hypothetical protein , App-1 = ADP-ribose-1”monophosphatase , PRT = phosphoribosyltransferase . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 00510 . 7554/eLife . 04266 . 006Figure 1—figure supplement 3 . Protein phylogeny of neighboring genes to transferred lysozymes . ( A ) App-1 phylogeny based on alignment of 141aa without indels consisting of top E-value hits to blastp using A . boonei App-1 as the query . Taxon of origin for each amino acid sequence is indicated by color . Posterior probability is indicated at all nodes with values above 50 . Branch lengths represent number of substitutions per site as indicated by scale bar . Tree is arbitrarily rooted . ( B ) GH2 hydrolase phylogeny based on an alignment of 188aa without indels consisting of top E-value hits to blastp using A . oryzae hydrolase as the query . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 006 To establish parallel HGT , i . e . , the independent transfer of the same gene family to multiple lineages , we conducted a phylogenetic analysis on 86 GH25 muramidase sequences using Bayesian and maximum likelihood inference methods ( Figure 2A ) . We combined non-redundant Aciduliprofundum , Selaginella , and WO sequences obtained from PCR and Sanger sequencing with blastp results to reconstruct the phylogeny . Putative instances of HGT are diagrammed in Figure 3 . Additionally , transferred lysozymes in nonbacterial taxa were used as queries to identify homologs and make a second set of phylogenetic trees to confirm the HGT ( Figure 2B–D , Figure 2—figure supplements 1 and 2 , Supplementary file 4 ) . Three key results emerge from these phylogenetic analyses: ( i ) at least three independent instances of interdomain HGT of the bacterial GH25 muramidase occurred in nonbacterial taxa ( Aciduliprofundum , Selaginella , and Insecta ) as well as a number of transfers to bacteriophages , ( ii ) vertical transmission of the transferred gene ensues in some descendant taxa ( i . e . , Aciduliprofundum and Selaginella ) , and ( iii ) frequent HGT of the muramidase between bacterial clades accompanies the interdomain transfer , indicating unusually frequent and broad-ranging HGT of this niche-transcending gene family . 10 . 7554/eLife . 04266 . 007Figure 2 . Phylogeny of GH25 muramidase . ( A ) Phylogeny based on alignment of 113aa without indels consisting of top E-value hits to blastp using WORiA phage lysozyme as a query . Taxon of origin for each amino acid sequence is indicated by color . Posterior probability ( Bayesian phylogeny ) and bootstrap values ( maximum likelihood phylogeny ) are indicated at all nodes with values above 50 . Branch lengths represent number of substitutions per site as indicated by scale bar . Tree is arbitrarily rooted . Iterative phylogenies based on top E-value blastp hits to A . boonei lysozyme ( B ) , A . pisum lysozyme ( C ) , and S . moellendorffii lysozyme ( D ) are also shown . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 00710 . 7554/eLife . 04266 . 008Figure 2—figure supplement 1 . Iterative HGT analysis alignments . Consensus alignment of GH25 muramidases without indels used in iterative phylogenies in Figure 2 is shown for ( A ) A . boonei , ( B ) A . pisum , and ( C ) S . moellendorffii . Conservation is indicated by amino acid symbol size and bar graphs below the consensus sequence . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 00810 . 7554/eLife . 04266 . 009Figure 2—figure supplement 2 . Protein phylogeny of A . oryzae GH25 muramidase and relatives . Phylogeny based on alignment of 186aa without indels consisting of top E-value hits to blastp using A . oryzae lysozyme as a query . Taxon of origin for each amino acid sequence is indicated by color . Posterior probability ( Bayesian phylogeny ) and bootstrap values ( maximum likelihood phylogeny ) are indicated at all nodes with values above 50 . Branch lengths represent number of substitutions per site as indicated by scale bar . Tree is arbitrarily rooted . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 00910 . 7554/eLife . 04266 . 010Figure 2—figure supplement 3 . DNA phylogeny of A . oryzae GH25 muramidase and relatives . Phylogeny based on alignment of 282 bp without indels consisting of top E-value hits to blastn using A . oryzae lysozyme exon 2 as a query . Taxon of origin for each nucleic acid sequence is indicated by color . Posterior probability ( Bayesian phylogeny ) and bootstrap values ( maximum likelihood phylogeny ) are indicated at all nodes with values above 50 . Branch lengths represent number of substitutions per site as indicated by scale bar . Tree is arbitrarily rooted . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 01010 . 7554/eLife . 04266 . 011Figure 3 . Schematic of HGT events . Bayesian phylogeny based on the 16S rRNA gene from selected taxa is shown . Colored lines indicate putative horizontal gene transfer events , although other possible HGT patterns cannot be definitively excluded . Posterior probabilities are noted at each node . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 011 To statistically validate parallel HGT across the tree of life , we performed a Shimodaira–Hasegawa test ( SH-test ) ( Shimodaira and Hasegawa , 1999 ) by comparing our consensus tree ( Figure 2A ) against a hypothetical tree with a binary constraint in which bacterial sequences are monophyletic and separate from monophyletic nonbacterial sequences . In this hypothetical tree consistent with the tree of life , lineage relationships within the bacterial and nonbacterial groups were permissively set as unconstrained . Results indicate that the hypothetical tree is significantly worse than the HGT tree , as expected ( p<0 . 01 , D ( LH ) = −133 . 9 , SD = 31 . 5 ) . We repeated this analysis with the hypothetical tree compared to 100 randomly sampled HGT trees from maximum likelihood bootstrapping and found the hypothetical tree was also worse than each of these trees ( p<0 . 01 ) . Finally , we performed SH tests between the HGT tree and either 1 ) a three-domain constraint tree or 2 ) a monophyletic eukaryote branch constraint tree , and again found that the constraint trees were significantly worse than the HGT tree ( p<0 . 01 ) . Thus , the null hypothesis of vertical descent is rejected , even under the most permissive conditions . We observed that each interdomain HGT event ( Figure 2 ) occurred between taxa that coexist in the same ecological niche , a likely prerequisite for HGT . For instance , the A . boonei lysozyme is in a clade dominated by Firmicutes whose members can be common in deep ocean sediments ( Orcutt et al . , 2011a ) . Indeed , Bacillus species have even been found in hydrothermal vents of the same fields in which Aciduliprofundum strains were isolated ( Reysenbach et al . , 2000 ) . The A . pisum lysozyme clade includes Wolbachia prophages and Proteobacteria , which are common endosymbionts of aphids and other insects ( Gomez-Valero et al . , 2004; Augustinos et al . , 2011; Wang et al . , 2014 ) . The S . moellendorffii plant lysozyme is closely related to Actinobacteria , which are dominant microbes in soil ( Bulgarelli et al . , 2013 ) . These associations , while not proof of HGT , establish interactions that may have facilitated the transfers , although any number of intermediate gene carriers is possible . While the phylogenetic pattern of the GH25 muramidase found in fungi is consistent with HGT ( Figure 2A , Figure 2—figure supplement 2 ) , the transfer occurred anciently in fungal evolution prior to the divergence of Dikarya , as the domain is present in both Basidiomycota and Ascomycota , but not other fungal phyla . As is the case with most putative ancient transfers , the deep branches of the tree are poorly supported and a definitive donor taxon cannot be established . Additionally , the possibility of multiple ancient transfers between bacteria and fungi or among fungi and plants cannot be excluded . However , a nucleotide-level phylogeny also supports HGT from an ancestral Actinobacterium ( Figure 2—figure supplement 3 ) . Interestingly , the lysozyme gene in the aphid A . pisum consists of a fusion of a bacterial GH25 muramidase domain and a eukaryotic carboxypeptidase domain . The gene includes five introns ( Nikoh et al . , 2010 ) , none of which interrupt the GH25 domain , consistent with a relatively recent HGT event and the absence of the gene from most sequenced insects ( Figure 1 ) . The lysozyme in the fungus A . oryzae , meanwhile , contains only a single intron , but it does interrupt the GH25 domain , consistent with the domain’s long association with fungi from the subkingdom Dikarya ( Figure 1 ) . We found no evidence of a GH25 muramidase in 323 sequenced archaeal genomes spanning all the major phyla and sister taxa to A . boonei ( Reysenbach et al . , 2006; Flores et al . , 2012 ) . This lack of homology does not appear to be due to insufficient representation of archaeal diversity , as the 323 members span all of the major phyla: Crenarchaeota ( 56 ) , Euryarchaeota ( 205 ) , Nanoarchaeota ( 10 ) , and Thaumarchaeota ( 39 ) . Indeed , if vertical descent were assumed for a recent Bayesian phylogeny of archaea with sequenced genomes ( Brochier-Armanet et al . , 2011 ) , this would require at least 10 independent losses of the lysozyme gene , an assumption that is certainly less parsimonious than a single HGT event . Moreover , the relative divergence of the small subunit rRNA gene in A . boonei compared to the putative bacterial HGT donors is greater than the relative divergence of the lysozyme gene ( Figure 4 ) , a finding that is inconsistent with both genes being transmitted by vertical descent . Also , there are no other homologs beyond those presented in this study in 132 plant genomes , and only one insect species with additional homologs out of 109 insect genomes . Thus , if the lysozyme were present in the last common ancestor of all domains , it would require the unlikely loss of the gene in dozens of lineages while maintaining it in an exceedingly small number of species . In summary , the presence of a GH25 muramidase in nonbacterial species represents a series of recurrent , independent horizontal gene transfer events derived from diverse , ecologically associated bacteria . 10 . 7554/eLife . 04266 . 012Figure 4 . Comparison of GH25 muramidase and rRNA divergence . ( A ) Unrooted Bayesian phylogeny of the GH25 muramidase from A . boonei and selected relatives , based on an alignment of 185aa without indels . Taxon of origin for each nucleic acid sequence is indicated by color . Posterior probability is indicated at all nodes with values above 50 . Branch lengths represent number of substitutions per site as indicated by scale bar . ( B ) Unrooted Bayesian phylogeny of the 16S rRNA gene for the same taxa as in ( A ) , based on an alignment of 1 , 156 bp without indels . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 012 We next undertook a series of experiments to test the hypothesis that the transferred muramidase functions as an antibacterial molecule . Since HGT frequently results in pseudogenized and nonfunctional genes ( Kondrashov et al . , 2006; Nikoh et al . , 2010 , 2008; Dunning Hotopp et al . , 2007 ) , we first investigated the amino acid sequences for preserved antibacterial action of the transferred lysozymes in nonbacterial genomes . We aligned all 86 GH25 muramidase sequences to identify conserved sites ( Figure 5A ) . We then mapped the conserved amino acids to a three-dimensional structure prediction of the A . boonei GH25 muramidase domain ( Figure 5B ) . Highly conserved residues ( >85% identity between all taxa ) invariably mapped to the previously identified active site pocket ( Martinez-Fleites et al . , 2009 ) . Conservation was also evident for structure predictions of other GH25 muramidases in the phylogeny such as S . moellendorffii ( Figure 5C ) . 10 . 7554/eLife . 04266 . 013Figure 5 . Conservation of A . boonei GH25 muramidase domain . ( A ) Consensus alignment of 86 GH25 muramidases with insertions and deletions removed . Conservation is indicated by amino acid symbol size and bar graphs below the consensus sequence . Active site residues and highly conserved amino acids modeled below are indicated with red and orange asterisks , respectively . ( B ) Space-filling model of the active site face of the predicted structure of A . boonei GH25 muramidase domain and ( C ) S . moellendorffii GH25 muramidase domain . Active site residues are indicated in red and the eight additional residues most highly conserved across all 86 proteins are orange . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 013 Second , we cloned , expressed , and purified the GH25 muramidase domain from the archaeon A . boonei as well as from closely related homologs in Paenibacillus polymyxa and PhiBP . We obtained each muramidase in a pure elution ( Figure 6—figure supplement 1 ) and tested for antibacterial action against a range of bacterial species . As predicted , the A . boonei GH25 muramidase efficiently killed several species of bacteria in the phylum Firmicutes - the putative donor group of the gene ( Figure 6A ) . The bacterial inhibition by A . boonei GH25 muramidase was more potent than the positive control , chicken egg white lysozyme , and was dose-dependent ( Figure 6B ) . Bacterial and phage muramidases did not elicit antibacterial killing , similar to cyan fluorescent protein and buffer-only negative controls . Bacteria typically use a large protein complex to limit their lysozymes’ activity to the septum during cell division ( Uehara and Bernhardt , 2011 ) , and PhiBP phage has a documented spectrum of activity limited only to a P . polymyxa strain unavailable for our analyses ( Halgasova et al . , 2010 ) . As expected , the A . boonei GH25 muramidase did not exhibit antibacterial activity against Gram-negative species or Gram-positive species outside of the families Bacillaceae and Paenibacillaceae , which was equivalent to the killing range of chicken egg white lysozyme with the exception of the Actinobacterium Micrococcus luteus ( Figure 6—figure supplement 2 ) . 10 . 7554/eLife . 04266 . 014Figure 6 . Antibacterial action of A . boonei GH25 muramidase domain against Firmicutes . ( A ) Bacteria of the specified strain/species incubated overnight on tryptic soy agar after a 20-min liquid preincubation with the proteins indicated . Genera: B = Bacillus , P = Paenibacillus . Proteins: CEWL = chicken egg white lysozyme , P . poly = P . polymyxa lysozyme , PhiBP = bacteriophage PhiBP lysozyme , A . boo = GH25 domain of A . boonei lysozyme , CFP = cyan fluorescent protein . Images are representative of at least three independent experiments . ( B ) Dose-dependence of A . boonei GH25 muramidase antibacterial action . B . subtilis colony survival is shown after incubation with A . boonei GH25 muramidase at the indicated concentrations for 20 min at 37°C . N = 10 for each concentration . p < 0 . 001 for linear model fit . Error bars are ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 01410 . 7554/eLife . 04266 . 015Figure 6—figure supplement 1 . Lysozyme purifications . PAGE gel stained with GelCode blue before and after purification of 6x-histidine tagged enzymes using nickel affinity chromatography . L = crude E . coli lysate expressing the indicated lysozyme , E = elution after lysozyme purification . P . poly = P . polymyxa lysozyme , PhiBP = bacteriophage PhiBP lysozyme , A . boo = A . boonei GH25 domain . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 01510 . 7554/eLife . 04266 . 016Figure 6—figure supplement 2 . Antibacterial test of A . boonei GH25 muramidase on non-Firmicutes bacteria . Bacteria of the specified strain/species incubated overnight on tryptic soy agar after a 20-min liquid preincubation with the proteins indicated . Genera: L = Listeria , S = Staphylococcus , E . saccharolyticus = Enterococcus , M = Micrococcus , E . cloacae = Enterobacter , E . coli = Escherichia , S = Serratia , D = Deinococcus . Proteins: CEWL = chicken egg white lysozyme , P . poly = P . polymyxa lysozyme , PhiBP = bacteriophage PhiBP lysozyme , A . boo = GH25 domain of A . boonei lysozyme , CFP = cyan fluorescent protein . Images are representative of at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 016 Third , the A . boonei muramidase domain is part of a larger gene ( 1725 bp ) composed of other domains that may broaden or constrain the range of antibacterial activity . To test the full-length gene's function in the absence of genetic tools in this system , we cloned the entire gene into an expression plasmid in E . coli and discovered that bacterial colonies grew poorly , with tiny , slow-growing colonies on solid media , and substantial cell death coinciding with a small amount of leaky expression in liquid culture ( Figure 7A ) . However , two colonies grew to normal size and upon sequencing , we determined that their expression plasmids were disrupted by insertions of 774 bp ( mutant 1 ) and 768 bp ( mutant 2 ) of a native IS1 family transposase from E . coli at 21 bp or 266 bp from the start of the lysozyme gene , respectively . These insertions also resulted in a number of premature stop codons in the lysozyme reading frame , disrupting production of the full-length gene ( Figure 7B ) . Thus , E . coli death requires intact and full-length lysozyme , and toxicity is not due to the expression construct itself . In sum , expression of the complete lysozyme resulted in E . coli death , while cloned genes with insertion sequences and premature stop codons abolished the lytic capacity of these proteins from within E . coli cells , providing evidence for an expanded host range to the antibacterial action . 10 . 7554/eLife . 04266 . 017Figure 7 . E . coli death following full-length A . boonei lysozyme expression . ( A ) Live/dead stain of BL21 ( DE3 ) E . coli transformed with expression constructs for the full-length lysozyme from A . boonei or a control lysozyme WORiA , a bacteriophage infecting Wolbachia pipientis strain wRi , after overnight growth without induction . PAGE gels of crude E . coli lysates from E . coli expressing the indicated lysozyme after 6 hr of induction are also shown with the expected sizes of lysozymes indicated with arrows . ( B ) Structure of original full-length A . boonei lysozyme expression plasmid and two spontaneous knockout mutants caused by insertion of 774 bp ( mutant 1 ) and 768 bp ( mutant 2 ) of IS1 transposase sequences . These insertions also resulted in a number of stop codons in the reading frame of the lysozyme . Knockout mutants grew to normal colony size , while all wild type colonies had intact expression plasmids , grew poorly , and died over time in liquid culture . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 017 Fourth , if horizontally transferred lysozymes serve as antibacterials to fend off bacterial niche competitors , two predictions follow: the lysozyme will be upregulated in response to bacterial competition and upregulation may correlate with a relative growth advantage in coculture . We thus cultured A . boonei cells in anaerobic marine media ( Reysenbach et al . , 2006 ) with and without cohabiting Mesoaciditoga lauensis ( phylum Thermotogae ) that was isolated from the same hydrothermal vent field as Aciduliprofundum in the Eastern Lau Spreading Center ( Reysenbach et al . , 2013 ) . As expected , we observed a significant increase in A . boonei lysozyme expression at four ( up to 127% increase ) and 12 hr ( up to 43% increase ) of coculture with M . lauensis in comparison to negative control cultures of the singular A . boonei ( Figure 8A ) . Ideally , A . boonei wild type and lysozyme knockouts would be employed to test relative fitness and bacterial inhibition . However , genetic manipulation of A . boonei is not currently feasible . 10 . 7554/eLife . 04266 . 018Figure 8 . Lysozyme expression and relative fitness during A . boonei and M . lauensis coculture . ( A ) Expression of A . boonei GH25 muramidase relative to the control gene elongation factor 1α , after the indicated time of coculture with M . lauensis ( M . l ) at the specified ratio relative to A . boonei . *p < 0 . 05 , **p < 0 . 01 , by Mann–Whitney U pairwise comparisons . N = 6 for all samples . Primers are listed in Supplementary file 1 . ( B ) Relative fitness of A . boonei vs M . lauensis in monoculture ( N = 5 ) and coculture ( N = 4 ) . ( C ) Growth of A . boonei ( red ) and M . lauensis ( blue ) monocultures over time . Significant differences in cell abundance occur at 24 , 52 , and 64 hr ( p < 0 . 05 ) , and 56 and 60 hr ( p < 0 . 01 ) based on pairwise Wilcoxon tests . ( D ) Growth of A . boonei and M . lauensis in coculture over time . Significant differences in cell abundance occur at 48 , 52 , and 64 hr ( p < 0 . 05 ) based on pairwise Wilcoxon tests . Error bars are ±SEM for all panels . DOI: http://dx . doi . org/10 . 7554/eLife . 04266 . 018 Growth experiments of A . boonei and M . lauensis were continued for 72 hr , during which there was a relative Malthusian fitness ( Lenski et al . , 1991 ) increase for A . boonei in coculture vs . monoculture ( Figure 8B ) across the exponential growth phase . This difference is marginally non-significant , perhaps due to low sample sizes ( p = 0 . 11 , N = 5 , MWU two-tailed test ) . When the species are cultured separately for 72 hr , M . lauensis cell abundance is greater than that of A . boonei during 14 out of the 19 sampling points ( Figure 8C , blue circles ) , indicating that bacteria outperform archaea in monoculture conditions . However , when the two species are cocultured , the cell abundances reverse and A . boonei outperforms M . lauensis for 14 out of the 19 time points ( Figure 8D , red circles ) . This competitive frequency difference is significant ( Chi-square test , p = 0 . 0035 ) , complementing the Malthusian fitness increase . Additionally , for each monoculture time point , there are 4 . 43% fewer A . boonei cells on average than M . lauensis , while in coculture there are 6 . 22% more A . boonei cells per time point ( Mann Whitney U . p = 0 . 023 ) . Thus , A . boonei outcompetes its bacterial competitor in coculture despite a higher monoculture growth rate for the bacteria , although whether lysozyme upregulation is directly responsible for this effect cannot be definitively proven . Other possible explanations for this finding include A . boonei scavenging of bacterial waste products , possessing a superior ability to obtain a rate-limiting nutrient , or deploying alternative antibacterial defense mechanisms . The universal tree represents the evolutionary relationships between cellular domains and establishes the modern foundation for benchmarking the magnitude of HGT across life . Indeed , HGTs have been described between each domain including archaea and bacteria ( Nelson et al . , 1999; van Wolferen et al . , 2013 ) , bacteria and eukaryote ( Andersson , 2005; Bordenstein , 2007; Gladyshev et al . , 2008; Danchin et al . , 2010 ) , and archaea and eukaryote ( Andersson et al . , 2003; Schonknecht et al . , 2013 ) . Despite these cases and others ( Brown , 2003; Zhaxybayeva and Doolittle , 2011 ) , HGTs are not without limits and often succumb to the selective costs of genomic rearrangements , cytotoxic effects , disruptive insertions , and functional inefficiencies upon integration ( Baltrus , 2013 ) . It follows then that HGTs do not occur at equal rates across the universal tree , but rather experience preferential routes in which the costs of HGT are easier to overcome . The resulting pattern of HGT can be understood as a gradient of decreasing frequency from within domain > between two domains > between all domains of life ( Bruto et al . , 2013; Zhaxybayeva and Doolittle , 2011; Puigbo et al . , 2009; Andam and Gogarten , 2011 , 2013 ) . In support of this pattern , the overwhelming evidence of gene transfers between bacteria is counterbalanced by the extreme lack of parallel gene transfers across all extant groups of life . As these parallel transfers are usually ancient and occur in non-model organisms ( Lundin et al . , 2010; Koonin et al . , 2003; McClure , 2001; McDonald et al . , 2012; Moran et al . , 2012 ) , they can defy clear interpretations due to their deep antiquity and lack of functional validation . One feature that parallel HGTs have in common is that the gene's phenotype must transcend different physiological capabilities , cellular structures , and ecological niches to repeatedly increase the fitness of each recipient across the whole diversity of life . While not traditionally used in the context of parallel HGT across all cellular domains , the term niche-transcending gene appropriately captures these conditions ( Wiedenbeck and Cohan , 2011 ) . The lysozyme gene family we describe in archaea , bacteria , eukaryotes , and viruses provides one such example because the adaptive benefit of an antibacterial muramidase has repeatedly surmounted the obstacles against recurrent HGT . Indeed , horizontally transferred homologs of the GH25 muramidase exhibit differential tissue expression in A . pisum ( Nikoh et al . , 2010 ) and bacteriolytic activity in the fungus Aspergillus nidulans ( AN6470 . 2 ) ( Bauer et al . , 2006 ) . Thus , the horizontally transferred homologs in eukaryotes confer the same transcriptional and enzymatic activity as in the archaea . The muramidase in a thermophilic archaeon is of special note as archaea do not possess murein cell walls ( Albers and Meyer , 2011 ) , and genes encoding an antibacterial peptide have never before been identified ( Cantarel et al . , 2009 ) . Members of the genus Aciduliprofundum are widespread thermoacidophiles in deep-sea hydrothermal vent chimney biofilms ( Flores et al . , 2012 ) in which bacteria are frequent inhabitants ( Orcutt et al . , 2011a; Miroshnichenko and Bonch-Osmolovskaya 2006 ) , including the M . lauensis species tested above . Archaea have been largely ignored in the context of antibiotic discovery , likely because of the conjecture that archaea do not compete with bacteria in nature . However , given that they coexist with diverse bacterial species in the environment ( Oren , 2002; Kato and Watanabe , 2010; Orcutt et al . , 2011b ) and can compete for similar resources , there may be significant , unexploited potential for antibiotics in this domain . Only a handful of antimicrobial peptides produced by archaea have been characterized , and those are active only against other archaea ( O'Connor and Shand 2002 ) despite the fact that archaea are known to inhibit bacteria in diverse environments ( Atanasova et al . , 2013; Shand and Levya , 2008 ) . It is also possible that since Aciduliprofundum strains metabolize peptides , the lysozyme enables a nutritive strategy in which lysed bacteria provide nutrients for the archaeon to scavenge . Based on this work , we suspect that systematic surveys of archaeal gene products will likely uncover a broad range of antibacterial activities , and may eventually offer novel peptide or small molecule therapeutics . Such antibacterial products may have naturally evolved thermostability that would increase their attractiveness as therapeutics . GH25 muramidases have been demonstrated as effective antibacterials against biofilms of Streptococcus pneumoniae ( Domenech et al . , 2011 ) and related enzymes have proven efficacious in mouse models of bacterial mucosal colonization ( Fenton et al . , 2010 ) , sepsis ( Loeffler et al . , 2003 ) , and endocarditis ( Entenza et al . , 2005 ) . In summary , we infer that the evolutionary path to this parallel HGT was paved by the universal drive for nonbacterial taxa to compete in a bacterial world . We predict that similar to the cascade of antibiotic gene transfer discoveries that followed their initial reporting , parallel transfers of genes to all cellular domains and viruses might regularly have antimicrobial functions .
PCR was performed using GoTaq DNA Polymerase ( Promega , Madison , WI ) with primers listed in Supplementary file 1 . PCR products were electrophoresed using 1% agarose gels in sodium boric acid buffer . Following electrophoresis , gels were dyed with GelRed ( Phenix Research , Candler , NC ) and imaged on an Alpha Innotech GelRed Imager ( Alpha Innotech , San Leandro , CA ) . Amplified bands were excised from the gels and purified with an SV Wizard Gel Cleanup kit ( Promega ) . Following purification , DNA concentration was measured using the Qubit DNA high sensitivity kit ( Life Technologies , Grand Island , NY ) and sequencing reactions were performed by Genewiz ( South Plainfield , NJ ) . The lysozyme protein from Wolbachia prophage WORiA ( ZP_00372884 ) was used as a query in a blastp search of the NCBI nonredundant protein database using Geneious Pro v5 . 5 . 6 . All hits with E-values below 10−12 were collected and duplicate entries were removed . Sequences from field and laboratory samples were added to this collection and aligned with MUSCLE ( Edgar , 2004 ) , insertions and deletions were removed , and the eight most highly conserved residues from the MUSCLE alignment were mapped to a structure prediction of A . boonei lysozyme using PyMOL . Structure prediction was performed using the homology-based modeling tool Phyre2 ( Kelley and Sternberg , 2009 ) . For phylogenetic analyses , ProtTest ( Abascal et al . , 2005 ) was used to determine the best model of protein evolution based on the corrected Akaike information criterion ( AICc ) . MrBayes ( Ronquist et al . , 2012 ) and PhyML ( Guindon et al . , 2010 ) were used to build a phylogenetic tree with Bayesian and maximum likelihood methods , respectively . For the global lysozyme phylogeny , the best model chosen by ProtTest ( LG + I + G ) was used to generate the maximum likelihood tree , while the third best model ( WAG + I + G; ΔAICc: 74 . 82 ) was used to generate the Bayesian tree due to a lack of LG model availability in MrBayes . S . sanguinolenta and S . stauntoniana lysozymes were excluded from this analysis because frameshift mutations suggest the genes may be evolving in the absence of selection , while Aphidinae lysozymes were not included because of shorter sequences of the GH25 muramidase domain obtained through the use of degenerate primers that would have limited resolution of the tree . In an iterative approach , each candidate example of HGT was used as a blastp query against the nr database and the top 15 ( A . boonei , A . pisum , S . moellendorffii ) or top 75 ( A . oryzae ) E-value hits were subjected to the same phylogenetic analysis as above . Evolutionary models used were A . boonei: WAG + I + G on a 148aa indel-free alignment , A . pisum: CpREV + I + G on a 190aa indel-free alignment , S . moellendorffii: WAG + I + G ( ΔAICc: 44 . 28 ) on a 200aa indel-free alignment , A . oryzae: WAG + G ( Bayesian , ΔAICc: 4 . 49 ) or LG + G ( maximum likelihood ) on a 186aa indel-free alignment . The fungal lysozyme was also phylogenetically analyzed on the DNA level using the top 25 E-value blastn hits to exon 2 of the A . oryzae lysozyme gene . jModelTest 2 ( Darriba et al . , 2012 ) was used to determine the best model of nucleic acid evolution ( GTR + I + G , ΔAICc: 7 . 33 ) of a 282 bp indel-free alignment . The archaea HGT clade was also analyzed phylogenetically with a Bayesian tree of selected taxa using lysozyme protein sequences ( WAG + I + G , 185aa ) and compared to 16S rRNA ( GTR + G , 1 , 156 bp ) for the same strains obtained from SILVA ( Quast et al . , 2013 ) . A schematic representation of putative HGT events was plotted on a Bayesian phylogeny based on the 16 s rRNA gene ( GTR + G , 1 , 226 bp ) . Representative species used for this phylogeny were Magnetospirillum magneticum ( Proteobacteria ) , Paenibacillus polymyxa ( Firmicutes ) , Arthrospira maxima ( Cyanobacteria ) , Crenarchaeota archaeon SCGC AAA471-B05 , Chthoniobacter flavus ( Verrucomicrobia ) , Thermotoga maritima ( Thermotogae ) , Pedobacter saltans ( Bacteroidetes ) , Streptomyces violaceusniger ( Actinobacteria ) . Statistical support for the HGT hypothesis was assessed with the Shimodaira–Hasegawa test ( SH-test ) ( Shimodaira and Hasegawa , 1999 ) as implemented in RAxML v . 8 . 0 . 20 ( Stamatakis , 2014 ) . An unresolved binary constraint tree was generated in MacClade v4 . 08 , in which bacterial sequences are monophyletic , as are nonbacterial sequences , with all other topology unconstrained . This constraint tree was used to generate a maximum likelihood best tree with RAxML , using the same evolutionary models as above . The SH-test was then run comparing the maximum likelihood constrained tree to the unconstrained consensus Bayesian tree or to 100 bootstrap trees from the maximum likelihood analysis from PhyML . This procedure was repeated for a three-domain constraint tree consistent with the tree of life and with a constraint tree in which eukaryotic sequences were monophyletic and other sequences were unconstrained . A . boonei GH25 muramidase domain ( ZP_04874596 ) , P . polymyxa lysozyme ( YP_003869492 ) , and PhiBP lysozyme ( CBA18122 ) were cloned and expressed with a 6x C-terminal histidine tag using an Expresso T7 Cloning and Expression System ( Lucigen , Middleton , WI ) according to the manufacturer's instructions . We also cloned the S . moellendorffii and A . oryzae GH25 muramidases , however recombinant 6× histidine-tagged proteins were insoluble when expressed in either E . coli or sf9 insect cells and attempts to solubilize them were unsuccessful . Sequence-confirmed expression plasmids and a control plasmid expressing cyan fluorescent protein ( CFP ) were transformed into HI-Control BL21 ( DE3 ) E . coli cells . Cultures at an OD600 of ∼0 . 5 were induced with 1 mM IPTG for 6 hr , centrifuged , and frozen at −80°C until purification . Frozen pellets were resuspended in lysis buffer containing 10 mM Tris–HCl , pH 7 . 5 , 300 mM NaCl , 0 . 5% Triton X-100 , 0 . 3% sodium dodecyl sulfate , and 1 mM phenylmethylsulfonylfluoride and sonicated 5 times for 30 s with at least 1 min on ice between sonications . Samples were centrifuged and recombinant proteins were purified from supernatant using HisPur Ni-NTA chromatography cartridges ( Thermo Scientific , Waltham , MA ) according to manufacturer's instructions . Glycerol at a final concentration of 40% was added to enzymes in elution buffer for storage at −20°C for a maximum of 3 weeks before use in antibacterial assays . Purifications were analyzed with denaturing polyacrylamide gel electrophoresis and stained with GelCode Blue ( Thermo Scientific ) . Full-length A . boonei lysozyme and WORiA lysozyme were cloned into a pET-20b vector ( EMD Millipore , Darmstadt , Germany ) with a C-terminal 6× histidine tag and sequence-confirmed plasmids were transformed into BL21 ( DE3 ) E . coli ( EMD Millipore ) . Three colonies from each transformation were inoculated into LB media and grown to an OD600 of ∼0 . 5 , induced for 4 hr with 1 mM IPTG and harvested for analysis on PAGE gels . Overnight cultures without induction were examined for bacterial death with a BacLight Live/Dead Stain ( Life Technologies ) . Purified A . boonei GH25 muramidase , P . polymyxa lysozyme , PhiBP lysozyme , CFP , and commercially purchased CEWL ( Sigma-Aldrich , St . Louis , MO ) were diluted to 100 μg/ml in buffer EG ( 60% nickel column elution buffer , 40% glycerol ) and filter sterilized . Bacteria to be tested were grown overnight in tryptic soy broth , split 1:10 , and incubated to exponential growth before being diluted into each enzyme solution . Samples were incubated with shaking for 20 min at 37°C and then 5 μl was spotted onto tryptic soy agar and incubated overnight at 37°C . To evaluate whether antibacterial activity is dose-dependent , B . subtilis was incubated with A . boonei GH25 muramidase at 100 μg/ml , 75 μg/ml , 50 μg/ml , 25 μg/ml and 0 μg/ml and 100 μl was spread on tryptic soy agar plates . Replicates of 10 were performed for each concentration , plates were incubated overnight at 37°C , and colonies were counted the following morning . Bacterial strains used in these experiments are listed in Supplementary file 1 . A . boonei and M . lauensis cultures were performed as previously described ( Reysenbach et al . 2006 ) with the following modifications: yeast extract was added at 2 . 0 g/l , pH was adjusted to 4 . 8 , and cultures were incubated at 65°C . For gene expression studies , 8 . 2 × 105 cells were inoculated into 5 ml cultures in 6 replicates each of monocultures and cocultures at 0 . 1:1 , 1:1 , and 1:0 . 1 ratios and 500 μl samples were collected after 4 and 12 hr of co-incubation and frozen for expression analysis . RNA was isolated from frozen samples using an RNeasy Mini Kit ( Qiagen ) and QIAshredder ( Qiagen ) , DNA contamination was removed with a Turbo DNAfree Kit ( Life Technologies ) , and reverse transcription was performed using a Superscript III first Strand Synthesis System ( Life Technologies ) along with no-reverse transcriptase controls . Quantitative PCR was performed with GoTaq qPCR Master Mix ( Promega ) using a CFX96 Real-Time System ( Bio-Rad , Hercules , CA ) . Primers are listed in Supplementary file 1 . For competition studies , 5 replicates of 5 ml cultures were inoculated as monocultures or 1:1 cocultures and 175 μl was collected every 4 hr for counting of relative species abundance with a hemocytometer . Relative fitness was calculated based on Malthusian parameters over the period of exponential growth as previously described ( Lenski et al . , 1991 ) . | Living things inherit most of their genetic material from their parents , so genes tend to be passed on from one generation to the next—from ancestors to descendants . Sometimes , however , DNA is transferred from one organism to another by other means . These events , collectively called horizontal gene transfer , are fairly common in nature; genes have been passed between different species as well as between different groups of organisms . For example , genes that confer resistance to antibacterial drugs have transferred from one species of bacteria to another , and other genes have also ‘jumped’ from bacteria to plants or animals . Now Metcalf et al . have studied a gene that first arose in bacteria and that encodes an enzyme called a lysozyme . This enzyme breaks down the outer casing of a bacterial cell: a step that is required for a bacterium to reproduce and divide in two . When Metcalf et al . searched for relatives of the lysozyme gene , they found copies in many other species of bacteria and revealed that this gene has been repeatedly transferred between different bacteria . Members of the lysozyme gene family have also ‘jumped out’ of bacteria and into other organisms at least four times . Metcalf et al . found related lysozyme genes in a plant , an insect , many species of fungi , and a single-celled microbe ( called an archaeon ) that lives at hot , deep-sea vents . A gene family being spread this widely across the tree of life has not been seen before . Nevertheless , as DNA is a common biological language to all living things , it is likely that all the different species that have received a lysozyme gene might use it for similar purposes . Metcalf et al . reveal that the lysozyme could be being used as an antibacterial molecule . The archaeon lysozyme can kill a broad range of bacteria; and when the gene was transferred into Escherichia coli bacteria , only the bacteria that mutated the lysozyme gene to render it useless were able to survive . Metcalf et al . also revealed that the archaeon microbe produces more of the enzyme if bacteria are present , which allows it to outcompete these bacteria . These findings suggest that there may be a number of horizontally transferred genes that have antibacterial activity against a wide range of bacteria . Searching for these genes—particularly in the largely underexplored group of archaea—might reveal new sources for antibiotic drugs to treat bacterial infections . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology"
] | 2014 | Antibacterial gene transfer across the tree of life |
ISWI family chromatin remodelers typically organize nucleosome arrays , while SWI/SNF family remodelers ( RSC ) typically disorganize and eject nucleosomes , implying an antagonism that is largely unexplored in vivo . Here , we describe two independent genetic screens for rsc suppressors that yielded mutations in the promoter-focused ISW1a complex or mutations in the ‘basic patch’ of histone H4 ( an epitope that regulates ISWI activity ) , strongly supporting RSC-ISW1a antagonism in vivo . RSC and ISW1a largely co-localize , and genomic nucleosome studies using rsc isw1 mutant combinations revealed opposing functions: promoters classified with a nucleosome-deficient region ( NDR ) gain nucleosome occupancy in rsc mutants , but this gain is attenuated in rsc isw1 double mutants . Furthermore , promoters lacking NDRs have the highest occupancy of both remodelers , consistent with regulation by nucleosome occupancy , and decreased transcription in rsc mutants . Taken together , we provide the first genetic and genomic evidence for RSC-ISW1a antagonism and reveal different mechanisms at two different promoter architectures .
Genomic DNA is packaged into chromatin , a dynamic material that exhibits numerous changes in post-translational modifications , composition , and protein interactions . One aspect of chromatin modulation involves the assembly or disassembly of chromatin through active remodeling , which can confer either occlusion or access to the DNA—a process that is associated with virtually all DNA-mediated transactions , including transcription , replication , and repair . Each remodeling action , either assembly or disassembly , is mediated ( in part ) by specialized ATP-dependent chromatin-remodeling complexes ( Vignali et al . , 2000; Clapier and Cairns , 2009; Narlikar et al . , 2013; Bartholomew , 2014 ) . Certain chromatin remodelers align with these two general categories: those that restrict DNA access by chromatin assembly and organization and those that promote DNA access by chromatin disassembly and disorganization . This broad separation in function can be partially illustrated by studies of individual chromatin remodelers and their effects on gene expression ( Angus-Hill et al . , 2001; Fazzio et al . , 2001; Vary et al . , 2003 ) ; in general , remodelers associated with chromatin disassembly promote DNA access and gene expression , while remodelers associated with chromatin organization more often repress gene expression , though there are exceptions to this simplified view ( e . g . , increased accessibility can promote repressor access to chromatin ) . The SWI/SNF family of chromatin remodelers provides a well-studied example of remodelers associated with nucleosome disorganization and/or disassembly . In yeast , the RSC chromatin-remodeling complex is an essential and abundant paralog of the canonical SWI/SNF remodeler ( Cairns et al . , 1996 ) . The central subunit of RSC , Sth1 , is a DNA-dependent ATPase that translocates DNA , pumping DNA around the surface of a nucleosome , and effectively mobilizing the nucleosome with respect to the underlying sequence ( Saha et al . , 2002 , 2005 ) . This property enables RSC to shift nucleosome positions , as well as completely eject nucleosomes ( Lorch et al . , 1999; Boeger et al . , 2004; Clapier and Cairns , 2009; Dechassa et al . , 2010 ) . In vivo , RSC facilitates transcription by all three RNA polymerases , primarily by enabling promoter access ( Parnell et al . , 2008 ) . RSC maintains proper promoter chromatin structure , as RSC mutants exhibit alterations in nucleosome occupancy and spacing at promoters ( Badis et al . , 2008; Hartley and Madhani , 2009; Ganguli et al . , 2014 ) . RSC activity appears regulated , in part , by the presence of histone modifications ( Kasten et al . , 2004; Ferreira et al . , 2007 ) . RSC contains seven bromodomains on four subunits , implying a key role of acetylation in regulation . Thus , gene activation often involves the recruitment and activation of remodelers such as RSC to act on specific modified nucleosomes and promote promoter accessibility . The converse of gene activation , silencing , is expected to be the reverse process , where nucleosomes are re-positioned and organized to occlude transcription factor access . This reconfiguration of chromatin to a less active or repressive state is a function of other chromatin remodelers , including members of the ISWI family . In yeast , these include two highly conserved ATPase paralogs , ISW1 and ISW2 , related to the Drosophila ‘Imitation SWitch’ ( ISWI ) protein , which is the catalytic component of multiple chromatin-remodeling complexes with roles in nucleosome assembly and gene repression ( Tsukiyama et al . , 1999; Vary et al . , 2003 ) . Similar to the family of SWI/SNF remodelers , the ISWI family of remodelers uses DNA translocation to mobilize nucleosomes , though ISWI remodelers are typically restricted to movement/sliding only and not ejection ( Whitehouse et al . , 1999; Clapier and Cairns , 2009 ) . Importantly , ISWI generates regularly spaced nucleosome arrays by ‘measuring’ the length of DNA linker between nucleosomes , and this property is thought to enable gene repression by ordering nucleosomes into closely spaced regular arrays that can restrict access to DNA ( Grune et al . , 2003; Whitehouse and Tsukiyama , 2006; Gangaraju and Bartholomew , 2007; Tirosh et al . , 2010; Bartholomew , 2014 ) . Studies of remodeler antagonism have been limited . ISW2 function was shown in one study to restrict the binding of the SWI/SNF chromatin remodeler at a target gene in yeast ( Tomar et al . , 2009 ) . Another study showed antagonistic roles by two alternative assemblies of mammalian SWI/SNF complex ( BRG and BRM ) , where BRM appeared to repress BRG activation functions ( Flowers et al . , 2009 ) . A third noted attenuation of BRG activation by the CHD family remodeler Mi-2 ( Ramirez-Carrozzi et al . , 2006 ) at a set of target genes . Although notable , none of the prior studies provide a conceptual view of how two remodelers might antagonize one another at a large number of loci and how antagonism relates to nucleosome occupancy and positioning at co-occupied loci . Here , we examine remodeler antagonism explicitly , providing the first evidence for an antagonistic relationship between ISWI and RSC . We demonstrate the suppression of growth rate phenotypes and the impact of these remodelers on both transcription and chromatin architecture at a genome scale . These studies uniquely reveal important activities of these two chromatin remodelers at particular promoter architectures—‘open’ and ‘closed’—and the requirement for remodeler antagonism for proper regulation .
Rsc7 is a non-essential subunit of the RSC complex that is required for growth at elevated temperatures and for full growth under particular conditions ( Wilson et al . , 2006 ) . We previously used a synthetic genetic array ( SGA ) screen ( Tong et al . , 2001 ) to identify genes that induced lethality in combination with rsc7Δ ( Wilson et al . , 2006 ) . We again used the SGA array to screen for genes whose null mutation would allow for growth of rsc7Δ at an otherwise non-permissive temperature . To accomplish this , a strain containing rsc7Δ was crossed to a haploid deletion library comprised 4700 strains , each bearing a deletion in a single nonessential gene ( SGA array ) . Double mutants were isolated and screened for growth at a restrictive temperature . Four viable combinations were obtained , including combinations of rsc7Δ with rpl20bΔ , lsm7Δ , bud31Δ , or isw1Δ ( Figure 1A , Figure 2C ( 33°C ) , and data not shown ) . The first three genes are associated with various ribonucleoprotein complexes , including ribosomes and snRNPs , which may represent interesting pathways that involve RSC function . However , the identification of isw1Δ was particularly intriguing , as it suggested a possible antagonistic relationship between ISWI complex ( es ) and RSC . Suppression was linked to isw1Δ , as the rsc7Δ temperature sensitivity returned with the introduction of a wild-type copy of ISW1 on a plasmid ( Figure 1A ) . The specificity of this observation is notable , as virtually all combinations of rsc7Δ with mutations in other chromatin-related genes typically resulted in lethality ( Wilson et al . , 2006 ) ; thus , isw1Δ was the sole suppressing mutation with a chromatin/transcription function isolated in our genome-wide format . 10 . 7554/eLife . 06073 . 003Figure 1 . Suppressors of rsc2 and rsc7 alleles obtained by genetic screen . ( A ) Suppression of the rsc7Δ temperature sensitivity by the isw1Δ mutation . Wild-type ( YBC62 ) and rsc7Δ ( YBC2039 ) were transformed with plasmids containing RSC7 , ISW1 , or empty vector and spotted as fivefold serial dilutions to SC-URA media and grown at 35° . ( B ) Histone H3 and H4 suppressors of rsc2-V457M . YBC2140 ( rsc1Δ rsc2-V457M hht1hhf1 hht2-hhf2 [H3-H4 . URA] ) was transformed with TRP1-marked plasmids bearing histone mutations , streaked to SC-TRP + 5FOA to force loss of the WT histone plasmid , and then spotted as 10-fold serial dilutions to SC-TRP at 30°C or 33°C . ( C ) 10-fold dilutions of YBC2140 transformed with H3 or H4 mutations and spotted to SC-TRP + 5FOA at 30°C or 33°C . Figure 1—figure supplement 1 shows that Rsc2 alleles are not suppressed by dot1Δ or sir3Δ . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 00310 . 7554/eLife . 06073 . 004Figure 1—figure supplement 1 . Rsc2 mutations are not suppressed by dot1Δ or sir3Δ . ( A ) dot1Δ does not suppress rsc2-V457M . YBC803 ( rsc1Δrsc2Δ [RSC1 . URA3] ) and YBC1683 ( rsc1Δ rsc2Δ dot1Δ [RSC1 . URA3] ) were transformed with TRP1-marked RSC2 ( p604 ) or rsc2-V457M ( p776 ) and spotted to SC-TRP 30°C and SC-TRP + 5FOA 30°C . ( B ) A null mutation in SIR3 does not suppress rsc2-V457M . YBC803 ( rsc1Δrsc2Δ [RSC1 . URA3] ) and YBC3185 ( rsc1Δ rsc2Δ sir3Δ [RSC1 . URA3] ) were transformed with TRP1-marked RSC2 ( p604 ) or rsc2-V457M ( p776 ) and spotted to SC-TRP 30°C and SC-TRP + 5FOA 32°C . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 00410 . 7554/eLife . 06073 . 005Figure 2 . A null mutation of isw1 suppresses RSC mutations . ( A ) rsc2 Ts− alleles are suppressed by isw1Δ . An ISW1+ strain ( YBC1231; rsc1Δ rsc2Δ [RSC1 . URA3] ) and an isw1Δ strain ( YBC1479; rsc1Δ rsc2Δ isw1Δ [RSC1 . URA] ) were transformed with TRP1-marked RSC2 ( p604 ) , rsc2-V457M ( p776 ) , or rsc2-D461G ( p777 ) , streaked to SC-TRP + 5FOA to force loss of the RSC1 plasmid , and then spotted as 10-fold dilutions to YPD at 30°C , 33°C , and YPD containing 1 . 5% formamide ( Form ) or 12 μg/ml benomyl . ( B ) isw1Δ suppresses 6-azauracil ( 6AU ) and MPA phenotypes of rsc2 mutations . YBC1231 ( ISW1+ ) and YBC1479 ( isw1Δ ) were transformed with TRP1-marked RSC2 ( p776 ) , rsc2-V457M ( p776 ) , or rsc2- ( YBC777 ) , streaked to SC-TRP + 5FOA to force loss of the RSC1 plasmid , and then transformed with URA3-marked vector . Strains were then streaked to SC-URA medium containing 20 μg/ml MPA or 150 μg/ml 6AU . ( C ) isw1Δ suppresses additional RSC mutations but does not suppress snf2 . WT ( YBC62 ) , isw1Δ ( YBC1416 ) , rsc2-V457M ( YBC1111 ) , rsc2-V457M isw1Δ ( YBC2810 ) , rsc33 ( YBC906 ) , rsc3-3 isw1Δ ( YBC1485 p817 ) , rsc4-2 ( YBC1278 ) , rsc4-2 isw1Δ ( YBC2867 ) , rsc7Δ ( YBC1333 ) , rsc7Δ isw1Δ ( YBC2233 ) , snf2Δ ( YBC26 ) , and snf2Δ isw1Δ ( YBC2812 ) were spotted as 10-fold serial dilutions to YPD 30°C , 33°C , 35°C , 38°C , YPD containing 150 mM Hydroxyurea ( HU ) , and YPGal + Antimycin A ( AA ) . Figure 2—figure supplement 1 shows suppression of rsc2 alleles by catalytic isw1 and isw2Δ mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 00510 . 7554/eLife . 06073 . 006Figure 2—figure supplement 1 . rsc2 mutations are suppressed by an ISW1 ATPase mutation and an ISW2 null mutation . ( A ) Growth ability of rsc2-V457M isw1Δ at the non-permissive temperature in the presence of ISW1+ , or ISW1-K227A . ( B ) rsc2 Ts− alleles in combination with isw2Δ . An ISW2+ strain ( YBC1231; rsc1Δ rsc2Δ [RSC1 . URA3] ) and an isw2Δ strain ( YBC1480; rsc1Δ rsc2Δ isw2Δ [RSC1 . URA3] ) were transformed with TRP1-marked RSC2 ( p604 ) , rsc2-V457M ( p776 ) , or rsc2-D461G ( p777 ) , streaked to SC-TRP + 5FOA to force loss of the RSC1 plasmid , and then spotted as 10-fold dilutions to YPD at 30°C , 32°C , and YPD containing 12 μg/ml Benomyl . ( C ) isw2Δ suppresses 6-azauracil ( 6AU ) and MPA phenotypes of rsc2 mutations . YBC1231 ( ISW1+ ) and YBC1480 ( isw2Δ ) were transformed with TRP1-marked RSC2 ( p776 ) , rsc2-V457M ( p776 ) , or rsc2-D461G ( YBC777 ) , streaked to SC-TRP + 5FOA to force loss of the RSC1 plasmid , and then transformed with URA3-marked vector . Strains were then streaked to SC-URA medium containing 20 μg/ml MPA or 150 μg/ml 6-azauracil 6AU . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 006 The suppression relationship between RSC and ISW1 was further strengthened through a second independent genetic screen involving RSC2 . The Rsc1 and Rsc2 proteins are two homologous mutually exclusive subunits of RSC that define two distinct RSC sub-complexes ( Cairns et al . , 1999 ) . Loss of either separately confers distinct phenotypes , while loss of both is lethal , suggesting both unique nonessential and redundant essential functions within the complex ( Cairns et al . , 1999 ) . Rsc1 and Rsc2 share the same domain structure , which consists of one nonessential and one essential bromodomain ( Cairns et al . , 1999 ) , a BAH domain that binds histone H3 ( Tsankov et al . , 2011 ) , and an AT hook ( Cairns et al . , 1999 ) . We previously isolated mutant alleles in the BAH domain of Rsc2 , including rsc2-V457M and rsc2-D461G , that confer temperature sensitivity in rsc1Δ strains ( Schlichter and Cairns , 2005 ) . To identify whether histone mutations might suppress RSC mutations , we used a histone mutagenesis screen . We integrated the rsc2-V457M allele into an rsc1Δ strain bearing histone H3-H4 deletions ( hht1-hhf1Δ and hht2-hhf2Δ ) that was covered by a URA3-marked plasmid bearing wild-type histones , HHT2-HHF2 . We then introduced TRP1-marked plasmids containing hydroxylamine mutagenized HHT2-HHF2 genes and screened for suppression of the temperature sensitivity phenotype upon loss of the wild-type histone plasmid ( using 5-FOA negative selection ) . From 20 , 000 transformants screened , we isolated seventeen suppressors that were verified by isolating and retransforming the plasmid containing the histone mutation . Of these , most contained single mutations: eight had either H3 A7V or H3 A7T mutations , seven had an H3 T6I mutation , and one bore an H3 G33V mutation . However , one mutant bore an H4 RH17 , 18CY double mutation ( Figure 1B ) . All of these histone mutations were also tested for suppression of other temperature-sensitive RSC alleles , including rsc2-D461G , rsc2-Y337H , and rsc4-2 , and each was suppressed ( data not shown ) , suggesting that these mutations generally suppress RSC defects and not a specific defect of rsc2-V457M . We focused on the H4 RH17 , 18CY mutant for subsequent studies as it caused the most robust suppression ( Figure 1B ) . The H4 RH17 , 18CY mutations are adjacent to H4 K16 , a residue whose acetylation serves as a mark for active chromatin ( Millar et al . , 2004 ) . As RSC contains several bromodomains and may be regulated by histone acetylation ( reviewed in Josling et al . , 2012 ) , we considered whether loss of H4 K16 acetylation may underlie the suppression . However , no H4 K16 acetylation was detected by Western blot in the H4 RH17 , 18CY mutant , but this could either be the result of loss of acetylation or failure of the antibody to recognize the mutated epitope . We therefore combined rsc2-V457M with H4 K16Q , H4 K16R , and H4 K16G mutants to determine if loss of K16 acetylation was responsible for the suppression . However , combining these mutants resulted in a slight synthetic sickness instead of suppression ( Figure 1C ) , ruling out this simple model . Notably , the H4 RH17 , 18CY mutations define the center of a region of the H4 tail referred to as the ‘basic patch’ , an epitope of known importance for the binding and activity of several chromatin-modifying factors including Isw1 , Sir3 , and Dot1 ( Clapier et al . , 2002; Fazzio et al . , 2005; Fingerman et al . , 2007; Altaf et al . , 2007; Wang et al . , 2013 ) . To test if the suppression by the basic patch mutation was due to an inability of Dot1 to bind or methylate H3K79 , we combined rsc2-V457M with either an H3 K79A mutation or dot1 null mutant . However , we observed no effect with the H3 K79A mutation ( Figure 1C ) , and the combination with dot1Δ resulted in synthetic sickness ( Figure 1—figure supplement 1 ) . Additionally , combination with an sir3Δ failed to suppress ( Figure 1—figure supplement 1 ) . Taken together , our results point strongly to ISW1 as the most likely candidate for RSC mutant suppression , tested further below . The results of these two genetic screens strongly suggested a functional antagonism between RSC and Isw1 . To directly test isw1 mutant suppression of rsc2 alleles , we combined rsc2-V457M or rsc2-D461G with isw1Δ and observed partial suppression of temperature sensitivity and a set of phenotypes associated with the drugs benomyl and formamide ( Figure 2A ) as well as 6-azauracil ( 6AU ) and mycophenolic acid ( MPA ) ( Figure 2B ) . Growth suppression of the double mutant was lost when ISW1 was restored through plasmid transformation . Suppression requires a loss of ISW1 catalytic function , as rsc2 suppression is observed in a strain bearing a mutation in the catalytic site ( K227A ) of ISW1 ( Figure 2—figure supplement 1A ) . Furthermore , when we combined rsc2-V457M with both isw1Δ and H4 RH17 , 18CY , no enhanced suppression was observed ( data not shown ) , suggesting that they act through the same pathway . We also directly tested whether the ISWI paralog , ISW2 , might also suppress rsc2 alleles . Combining the isw2Δ mutation with rsc2-V457M or rsc2-D461G did not confer suppression of the temperature growth defect , although some partial suppression of other phenotypes was observed ( Figure 2—figure supplement 1B ) . We also did not see additional suppression when isw1Δ and isw2Δ were combined ( data not shown ) . We therefore conclude that the rsc2 mutation suppression is due primarily to the loss of Isw1 activity , with minimal contributions from of loss Isw2 activity . We next asked whether isw1Δ suppression was specific to rsc2 and rsc7 mutations or could extend more generally to RSC mutations . To test this , we combined isw1Δ with two additional RSC mutations in separate subunits , rsc3-3 and rsc4-2 . The isw1Δ allele suppressed both RSC mutations tested ( Figure 2C ) , with suppression of rsc4-2 particularly robust and greater than rsc2 mutants ( Figure 2C , YPD 38°C panel ) . We also tested whether isw1Δ could suppress phenotypes associated with loss of SWI/SNF function . Combining isw1Δ with snf2Δ did not allow growth on galactose or raffinose carbon sources or growth on media containing hydroxyurea ( Figure 2C ) , demonstrating specificity for RSC . Together , these results are consistent with a specific antagonistic relationship between RSC and ISW1 . Isw1 is the ATPase for two distinct remodeling complexes , ISW1a and ISW1b ( Vary et al . , 2003 ) . The ISW1a form contains Ioc3 , associates with particular gene promoters , and is implicated in repression by positioning nucleosomes into regularly spaced arrays ( Gangaraju and Bartholomew , 2007; Yamada et al . , 2011 ) . ISW1b contains Ioc2 and Ioc4 , associates with coding regions , plays a greater role in transcription elongation and termination , and does not regularly space nucleosomes ( Morillon et al . , 2003; Vary et al . , 2003; Gangaraju and Bartholomew , 2007 ) . To determine which form of the ISW1 complex is responsible for the suppression of rsc2 , we combined rsc2-V457M with ioc2Δ , ioc3Δ , or ioc4Δ . Surprisingly , synthetic lethality , and not suppression , was observed when rsc2-V457M was combined with either ioc2Δ or ioc4Δ ( Figure 3A ) . In contrast , combining rsc2-V457M with ioc3Δ resulted in partial suppression of rsc2 phenotypes ( Figure 3B ) . Notably , ioc3Δ potently suppressed rsc4-2 ( Figure 3C ) and partially suppressed conditional rsc1 mutations ( Figure 3—figure supplement 1 ) . Together , these results strongly implicate the loss of Isw1a complex function in rsc suppression . 10 . 7554/eLife . 06073 . 007Figure 3 . Suppression of RSC mutants is specific to Isw1a . ( A ) rsc2-V457M is lethal in combination with ioc2Δ and ioc4Δ , but not ioc3Δ . YBC803 ( rsc1Δ rsc2Δ [RSC1 . URA3] ) , YBC2730 ( rsc1Δ rsc2Δ ioc3Δ [RSC1 . URA3] ) , YBC2729 ( rsc1Δ rsc2Δ ioc2Δ [RSC1 . URA3] ) , and YBC2731 ( rsc1Δ rsc2Δ ioc4Δ [RSC1 . URA3] ) were transformed with TRP1-marked RSC2 ( p604 ) or rsc2-V457M ( p776 ) , and spotted as 10-fold dilutions to SC-TRP 30°C or SC-TRP + 5FOA 30°C . ( B ) rsc2 Ts mutations can be partially suppressed by ioc3Δ . YBC803 ( rsc1Δ rsc2Δ [RSC1 . URA3] ) and YBC 2730 ( rsc1Δ rsc2Δ ioc3Δ [RSC1 . URA3] ) were transformed with TRP1-marked RSC2 ( p604 ) , rsc2-V457M ( p776 ) , or rsc2-D461G ( p777 ) , streaked to SC-TRP + 5FOA to force loss of the RSC1 plasmid , and then streaked to SC-TRP at 30°C or 32°C . ( C ) rsc4-2 is suppressed by ioc3Δ . Strain YBC627 ( rsc4 [RSC4 . URA3] ) and YBC3020 ( rsc4Δ ioc3Δ [RSC4 . URA3] ) were transformed with TRP1-marked RSC4 ( p1060 ) or rsc4-2 ( p1083 ) , streaked to SC-TRP + 5FOA to lose RSC4 . URA3 , and then spotted as 10-fold serial dilutions to SC-TRP 30°C or SC-TRP 38°C . Figure 3—figure supplement 1 shows suppression of rsc1 mutants by ioc3Δ . Figure 3—figure supplement 2 shows the suppression of rsc2 synthetic lethality with set1Δ and gcn5Δ by mutations in ISW1a . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 00710 . 7554/eLife . 06073 . 008Figure 3—figure supplement 1 . Mutations in RSC1 are suppressed by ioc3Δ . An IOC3+ strain ( YBC803; rsc1Δ rsc2Δ [RSC1 . URA3] ) and an ioc3Δ strain ( YBC2730; rsc1Δ rsc2Δ ioc3Δ [RSC2 . URA3] ) , were transformed with TRP1 marked RSC1 ( p609 ) , rsc1-F300S ( p1525 ) , rsc1-Y297H ( p1526 ) , rsc1-V417M ( p1527 ) , rsc2-D421G ( p1528 ) , or vector ( pRS314 ) , and spotted as tenfold dilutions to SC-TRP , or SC-TRP + 5FOA to force loss of the RSC1 or RSC2 plasmid , and growth was assessed at 30°C , 33°C , and 35°C . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 00810 . 7554/eLife . 06073 . 009Figure 3—figure supplement 2 . Synthetic lethality of rsc2 mutations with set1Δ and gcn5Δ can be suppressed by isw1 and ioc3 . ( A ) Mutations in rsc2 that are synthetic lethal with loss of Set1 ( the sole H3 K4 methyltransferase in yeast ) are suppressed by null mutations in ISW1 and IOC3 . Strains with rsc1Δ rsc2Δ set1Δ [RSC1 . URA3] ( YBC1245 ) were combined with isw1Δ ( YBC2744 ) or ioc3Δ ( YBC2803 ) , transformed with TRP1-marked plasmids RSC2 ( p604 ) , rsc2-V457M ( p776 ) , or rsc2-D461G ( p777 ) , and spotted as 10-fold serial dilutions to SC-TRP and SC-TRP + 5FOA ( to enforce loss of the RSC1 plasmid ) at 30°C . Additionally , we combined each of the rsc2 isw1Δ mutant combinations with hyperactive Set1D alleles ( Schlichter and Cairns , 2005 ) and did not see further suppression ( data not shown ) . ( B ) Synthetic lethality of rsc2Δ with loss of the histone acetyltransferase Gcn5 is suppressed by isw1Δ . YBC3496 ( rsc2Δ ) , YBC 3494 ( rsc2Δ gcn5Δ ) , and YBC3495 ( rsc2Δ gcn5Δ isw1Δ ) each covered with [p199; RSC2 . URA3] were spotted as 10-fold serial dilutions to SC 30°C and SC + 5FOA 30°C . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 009 Our prior work revealed moderate rsc2 suppression with increased H3K4me3 ( by hyperactive SET1 alleles ) , and conversely , synthetic lethality with rsc2 set1Δ combinations , suggesting that H3K4me3 either promotes or partially bypasses RSC activity ( Schlichter and Cairns , 2005 ) . However , as ISWI activity is affected by H3K4me and Set1 function ( Santos-Rosa et al . , 2003 ) , an alternative hypothesis is that H3K4me affects RSC indirectly through the alteration of Isw1 activity . To test this , we combined rsc2-V457M or rsc2-D461G with set1Δ , in the absence or presence of ISW1 or IOC3 . Interestingly , either isw1Δ or ioc3Δ can suppress the rsc2 set1Δ lethality ( Figure 3—figure supplement 2 ) . We also find that there is no additional suppression of rsc2 temperature sensitivity by combining SET1 hyperactive mutations and isw1Δ and that isw1Δ can still suppress rsc2 phenotypes in an H3 K4A mutant ( data not shown ) . These results suggest that suppression by loss of Isw1a is epistatic to the effects of Set1 loss and can overcome the reliance of RSC on H3K4 methylation . As RSC activity is known to also be promoted by histone acetylation ( e . g . , H3K14ac; Kasten et al . , 2004; Carey et al . , 2006; Ferreira et al . , 2007 ) , we therefore tested whether loss of Isw1 would reduce the reliance of RSC on H3K14ac . GCN5 is a histone acetyltransferase responsible for much of the H3K14ac in vivo ( Howe et al . , 2001; Johnsson et al . , 2009 ) , and loss of GCN5 is lethal in combination with several RSC mutations , including rsc2Δ ( Cairns et al . , 1999; Kasten et al . , 2004 ) . We found isw1Δ suppressed the lethality of rsc2Δ gcn5Δ mutations ( Figure 3—figure supplement 2 ) . These results suggest that removing the chromatin remodeler that antagonizes RSC , notably ISW1a , reduces the need for RSC activation through acetylation . The genetic relationships identified above prompted us to investigate the spatial relationship between RSC and ISW1a . We therefore determined the occupancy of both of these chromatin remodelers by chromatin immunoprecipitation ( ChIP ) , using the RSC subunit Rsc8 and the ISW1 subunit Ioc3 , both tagged with C-terminal Myc epitope tags . We chose the Rsc8 subunit of RSC because it exists as a dimer in the RSC complex , minimizing the low ChIP efficiency observed with chromatin remodelers ( Whitehouse et al . , 2007; Parnell et al . , 2008; Yen et al . , 2012 ) . We analyzed the immunoprecipitated DNA first by hybridization to high-resolution genome-wide microarrays ( 244K probes , ∼50 bp resolution ) and subsequently high-throughput sequencing . RSC occupancy was scored across gene promoters ( −800 to +800 bp ) , and promoters were then sorted into six clusters using a k-means algorithm to visualize those with and without enrichment ( Figure 4A ) . Using the mean occupancy at the transcription start site ( TSS , ±250 bp ) , 43% of promoters ( 2274 of 5337 ) had RSC enrichment corresponding to a false discovery rate ( FDR ) of less than 1% . We also found RSC was highly enriched at all non-coding RNA genes , including tRNA genes , as reported previously ( Ng et al . , 2002 ) . Notably , we find RSC highly enriched at virtually all centromeres ( Figure 4B ) , a localization not previously reported . 10 . 7554/eLife . 06073 . 010Figure 4 . RSC and ISW1a co-occupy many locations , and their loss impacts gene expression in a complex manner . ( A ) Heat map of Rsc8 and Ioc3 protein occupancy as determined by ChIP at all TSS . Each row represents a gene , with occupancy scored in 50 bp windows , ±800 bp relative to the TSS ( bent arrow ) . Windows overlapping neighboring genes are excluded . Occupancy above global mean is indicated in red , below in blue . Genes are clustered by a k-means algorithm into 6 groups . ( B ) The distributions of mean Rsc8 and Ioc3 occupancy values shown as box and whisker plots for different annotation features . ( C ) The correlation between Rsc8 and Ioc3 at promoters shown as a XY plot , either genome-wide or restricted to the 500 coding genes selected for the custom HybMap microarray . ( D ) The distribution of the mean mutant/wild-type gene expression ratios as determined by the HybMap microarray for three classes of gene types are presented as box and whisker plots . Figure 4—figure supplement 1 compares the ChIP results obtained from microarray vs deep sequencing . Figure 4—figure supplement 2 displays the genes that appear suppressed by isw1Δ or ioc3Δ as determined by HybMap . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 01010 . 7554/eLife . 06073 . 011Figure 4—figure supplement 1 . RSC and ISW1a occupancy correlate between microarray and sequence studies . Heat map of RSC ( A ) and ISW1a ( B ) occupancies . RSC occupancy was determined by Rsc8 ChIP applied to microarray ( MA ) , Rsc8 ChIP paired-end sequencing ( Seq ) , and Sth1 paired-end sequencing . ISW1a occupancy was determined by Ioc3 ChIP applied to microarray and Ioc3 ChIP paired-end sequencing . Occupancy , expressed as log2 fold enrichment over input , was measured in 20 bp windows flanking the TSS ±800 bp; windows overlapping neighboring genes were excluded . Genes were organized into six clusters using a k-means algorithm based on the Rsc8 microarray occupancy and is identical to Figure 4 . Only windows with positive ( enriched ) values ( red ) are plotted to simplify visualization . ( C ) To show correlation between the microarray and sequencing data sets , the maximum occupancy value for each gene determined in a 500 bp window encompassing the TSS ( ±250 bp ) was plotted as a pairwise scatter plot , with the microarray data set on the X axis for each plot . A linear regression line is plotted as a thick black line . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 01110 . 7554/eLife . 06073 . 012Figure 4—figure supplement 2 . Some genes show transcriptional suppression in rsc2 isw1Δ double mutants . The change in expression relative to wild type as determined by HybMap are presented . ( A ) The 14 ( ioc3Δ ) genes that appear downregulated in the single rsc2 mutant ( black bars ) and suppressed in the double mutants ( red bars ) . ( B ) The 12 ( isw1Δ ) genes that appear downregulated in the single rsc2 mutant ( black bars ) and suppressed in the double mutants ( red bars ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 012 In comparison to RSC , the Ioc3 enrichment was less robust , perhaps reflecting a difference in chromatin association or difficulty in capturing complexes . We identified 137 or 230 Pol II promoters at an FDR of 1% or 5% , respectively . Strikingly , 224 of these latter promoters also pass the 1% threshold for RSC enrichment . Visual comparison of the enrichment pattern ( log2 fold ChIP/Input ) across all Pol II promoters reveals a high degree of overlap ( Figure 4A ) , while a pairwise plot between the RSC and ISWI mean fold enrichment values at the TSS shows a positive correlation ( r = 0 . 6; Figure 4C ) . This enrichment also extends beyond Pol II promoters , as we observed high ISW1a occupancy at both ncRNA and tRNA genes ( Figure 4B ) . We did not observe significant enrichment of ISW1a at centromeres , although we note that ISW2 is enriched at centromeres ( Zentner and Henikoff , 2013 ) , which may provide any requisite ISWI function at these loci . These results support the notion that RSC and ISW1a share a spatial ( though perhaps not temporal ) occupancy at particular genes . To extend these results , we repeated Rsc8 , Ioc3 , and Sth1 ( the ATPase subunit of RSC ) ChIP using micrococcal nuclease-digested chromatin analyzed by paired-end sequencing . We compared the log2 fold enrichment values obtained from both microarray and sequencing technologies ( Figure 4—figure supplement 1 ) . Despite the differences in resolution and sensitivity between these methods , we observed strong correlations between our microarray and sequencing results . Since a complete loss of RSC function results in a cessation of all transcription from all three polymerases ( Parnell et al . , 2008 ) , a weaker viable mutation ( such as those in rsc2 ) may result simply in an attenuation of transcription of many or all genes , leading to a general phenotype such as temperature sensitivity . This transcription attenuation , as well as any suppression by ISWI , should be evident by expression analysis . To determine whether this suppression is global in nature or restricted to a subset of genes , we performed a HybMap analysis on a sampling of genes in the genome . The HybMap technique measures both sense and anti-sense RNA levels across a genome ( Dutrow et al . , 2008 ) , providing results that are comparable to RNA-Seq ( Ni et al . , 2010 ) . The advantage of this technique is the direct use of total RNA ( enabling the detection of transcripts lacking polyA ) without RNA labeling and/or amplification protocols to obtain absolute expression levels . Although the format restricted our array to 649 genes , it included a large fraction of genes occupied by RSC ( 84% ) , both RSC and ISW1 ( 9% ) , or unoccupied ( 16% ) , using an FDR threshold of 1% . We performed this analysis on rsc2-V457M , isw1Δ , ioc3Δ , rsc2-V457M isw1Δ , and rsc2-V457M ioc3Δ strains and compared them to wild type . Consistent with the general requirement of RSC function for transcription ( Parnell et al . , 2008 ) , the mean expression of both coding and non-coding genes ( but not tRNAs ) was reduced almost twofold following the loss of RSC ( Figure 4D ) . Interestingly , individual ioc3 or isw1 mutants also lowered mean expression but with less magnitude . However , neither the rsc2 ioc3 nor the rsc2 isw1 double mutants generally suppressed the rsc2 effect by restoring global gene expression . Furthermore , we saw little change among tRNA genes from any genotype and no measureable change in anti-sense transcription levels ( data not shown ) . We also did not observe general aberrant transcription from promoters as reported previously in an rsc3 mutant ( van Bakel et al . , 2013 ) . These results suggest that the suppression of RSC phenotypes is not due to a global effect on gene expression but rather due to an effect at a subset of genes . To see if such genes could be identified from our sampling , we selected genes whose expression was at least partially restored by combining rsc2 with isw1 or ioc3 mutations . This analysis revealed 20 genes ( Figure 4—figure supplement 2 ) , which included genes for ribosome function , snoRNA genes , and several essential genes . It is likely that the combined modest change in expression at these and other genes are responsible for the suppression relationship observed . Since RSC and ISW1 are both chromatin remodelers , the most important test for antagonism involves examining whether mutations in ISW1 could suppress the effects of nucleosomal changes due to the loss of RSC function . Loss of RSC results in a gain of nucleosome occupancy at the nucleosome-depleted region ( NDR ) commonly found near the TSS of genes ( Badis et al . , 2008; Parnell et al . , 2008; Hartley and Madhani , 2009; Ganguli et al . , 2014 ) . We therefore constructed strains that included the sth1td degron allele in combination with an isw1Δ allele . Implementation of the sth1td allele allows for precise inducible destruction of the catalytic subunit of RSC , thus abrogating all RSC catalytic function—which we subsequently term ‘rscΔ’ in figures and text . We chose to use both the RSC and isw1 null alleles to maximize the nucleosomal effects due to the loss of catalytic activity in a manner that mutations in regulatory subunits may not . Mono-nucleosomal DNA was isolated from these yeast strains after inducing the degron allele for 2 hr and analyzed by both high-resolution microarray ( rscΔ and isw1Δ ) and paired-end sequencing ( rscΔ only ) . As a reference , we also analyzed mono-nucleosomal DNA from control strains that cannot degrade Sth1 protein . To analyze the chromatin structure around the TSS , we generated nucleosome profiles around the TSS for every promoter by scoring the nucleosomal occupancy for rscΔ and RSC strains . Promoters were organized into clusters based on their rscΔ/RSC ratio profile using a k-means algorithm ( Figure 5A ) . For each cluster , the mean nucleosome profile of both strains was then generated ( Figure 5B ) ( We note that the clusters in Figure 5 bear no relationship to the clustering analysis in Figure 4 , which instead shows similarity in loci occupied by RSC or ISWI ) . 10 . 7554/eLife . 06073 . 013Figure 5 . Loss of ISW1 partially suppresses nucleosomal changes exhibited by loss of RSC function . ( A ) The promoter profile of nucleosome occupancy ratios between sth1td degron ( rscΔ ) and control ( RSC ) strains is presented as a heat map , where red represents a gain in nucleosome occupancy and blue represents a loss . Genes ( rows ) are organized into six groups by k-means clustering . Columns represent 50 bp windows , ±800 bp relative to the TSS . Windows overlapping neighboring genes are excluded . ( B , C , D ) The mean profiles of nucleosome occupancies for all genes within each cluster are shown . Profiles from mutant backgrounds are shown in red , and wild-type profiles are shown in blue . The y-axis represents log2 occupancy relative to genome average . Figure 5—figure supplement 1 compares the nucleosome profiles obtained from microarray and deep sequencing , as well as the predicted nucleosome occupancy . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 01310 . 7554/eLife . 06073 . 014Figure 5—figure supplement 1 . Nucleosome profiles from microarray and sequencing show strong correlation with each other and predicted occupancy . ( A ) The mean nucleosome profiles for each of the six gene clusters derived in Figure 5 are shown using data derived from paired-end sequencing . ( B ) The predicted mean nucleosome profile ( Segal et al . , 2006 ) is shown in orange along with the observed wild-type nucleosome occupancy derived from microarray . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 014 The aggregate nucleosome profiles of wild type ( blue line , Figure 5B ) confirmed published observations ( Yuan et al . , 2005; Lee et al . , 2007; Whitehouse et al . , 2007 ) , showing a clear NDR flanked by positioned nucleosomes ( termed −1 and +1 ) and phased positioned nucleosomes within the proximal coding region . The loss of RSC ( Figure 5A and Figure 5B ) resulted in two major categories: ( 1 ) clusters 1–4 all share strong changes in nucleosome positioning following the loss of RSC and ( 2 ) clusters 5 and 6 show a weak response to the loss of RSC . Closer examination revealed further differences in each category . For example , the cluster 1 rscΔ profile shows a dramatic gain in nucleosome occupancy over the NDR at the expense of the +1 nucleosome relative to the control RSC profile , consistent with prior work ( see ‘Discussion’ ) . There is also a clear ‘leftward’ shift in nucleosome positions over the body of the gene , towards the NDR . This nucleosome shift is particularly prominent in clusters 3 and 4 , while the NDR is filled into a lesser extent and the +1 nucleosome peak is not as depleted ( compared to cluster 1 ) . To confirm that these effects were not due to limitations in the sensitivity and resolution of microarray analyses , the nucleosome profiles from the microarray data were directly compared to those from the sequencing analysis ( Figure 5—figure supplement 1 ) . Nucleosome profiles of the same clusters show remarkable similarity between those derived from array and sequence , validating our approaches and conclusions . Taken together , the filling of the NDR and a strong ‘leftward’ shift of the +1 nucleosome toward the NDR are consistent features that follow loss of RSC function . We next examined the impact due to the loss of ISW1 ( Figure 5D ) . Loss of ISW1 results in modest nucleosomal changes , most notably within the promoter-proximal 5′ coding region , either as changes in density or phasing , and minimal impact at the NDR . While loss of ISW1 alone has been shown to result in nucleosomal shifts towards the TSS ( Tirosh et al . , 2010; Yen et al . , 2012; van Bakel et al . , 2013 ) , these shifts , discernable in cluster 3 , are much smaller and more restricted than those generated by the loss of RSC ( compare Figure 5B , D ) . Importantly , in the double mutant ( Figure 5C ) , the nucleosomal profiles are more similar to wild type than rscΔ alone . Notably , the NDRs are not as filled and the shifts towards the TSS are not as severe . Taken together , these results provide considerable support for an antagonistic relationship between RSC and ISW1 , especially regarding the positioning and phasing of nucleosomes over the promoter-proximal coding region of the gene . Above , we showed that nucleosome architecture at clusters 1–4 shows a strong response to RSC loss , whereas clusters 5 and 6 show apparently limited changes . Clusters 1–4 display a prototypical promoter nucleosomal architecture ( −1 , NDR , and +1 nucleosome ) . In contrast , clusters 5 and 6 lack this stereotypical organization; here , RSC and/or ISW1a may indeed impact nucleosome occupancy and/or positioning , but the effect may be obscured due to architectural heterogeneity . Notably , these two types of architectures have previously been designated as open ( or structured ) vs closed ( or unstructured ) and have been largely correlated with either constitutive or highly regulated gene types , respectively ( Tirosh and Barkai , 2008; Cairns , 2009 ) . We verified these classifications by plotting the mean nucleosome prediction ( Segal et al . , 2006 ) for each of these clusters ( Figure 5—figure supplement 1B ) . While the predictive power for individual nucleosome positions was weak , the algorithm predicted the depth and breadth of NDRs fairly accurately . The strongly responsive clusters 1–4 had a well-defined NDR prediction , matching the observed profile , while the weakly responsive clusters 5 and 6 showed a broad shallow NDR . Since nucleosome phasing is , in part , determined by how well the −1 and +1 nucleosomes are positioned flanking the NDR , this result matches well with the general lack of consistent phasing across clusters 5 and 6 gene bodies . Interestingly , cluster 3 does not show as strong a predictive NDR as clusters 1 , 2 , and 4 , which may partly explain why this cluster shows nucleosomal shifts in both isw1Δ and rscΔ and weak suppression in the double mutant . Taken together , structured/open promoters show the strongest response to RSC loss , whereas unstructured/closed promoters lack a strong response—though we note that the lack of a uniform structure may obscure the response ( see ‘Discussion’ ) . Given the strong impact on chromatin structure at open/structured promoters vs the closed/unstructured promoters , we next examined how the loss of RSC might impact the transcription of these classes . Using our HybMap RNA expression data as a proxy for transcriptional impact , we scored genes from each category for expression . We note , however , that the differences between the HybMap and nucleosome experiments in several parameters , for example , RSC mutation vs depletion , time points , and representation among clusters ( see Figure 6—figure supplement 1A ) , place limitations on these comparisons . Nevertheless , while all promoter classes showed reduced gene expression in rsc2 mutants ( consistent with Figure 4D ) , cluster 2 and especially clusters 5 and 6 ( the two ‘closed’ promoter clusters ) were most severely negatively impacted ( Figure 6A ) . The inclusion of cluster 2 with clusters 5 and 6 is intriguing; however , it is also the only structured cluster to exhibit significant nucleosome occupancy gain over the body of the gene in rscΔ , which is likely related to the reduction in transcription . These impacts are not simply correlated with the level of gene expression , as the distribution of wild-type expression values between clusters is highly consistent ( Figure 6—figure supplement 1 ) . Similar to what we observed previously , the isw1Δ mutants showed little impact on the bulk expression of these genes ( Figure 6A ) . We also examined the chromatin structure of the 17 Pol II-transcribed genes where ISW1a loss provided significant suppression of rsc2Δ ( Figure 4—figure supplement 2 ) . However , the moderate resolution provided by the microarray format limited the fine mapping of nucleosomes , preventing our ability to identify nucleosomes that might be directly responsible for suppression ( data not shown ) . These 17 genes partitioned slightly more to closed ( nine genes ) than the more common open promoter structures ( eight genes ) . Together , these results suggest that while the clearest effects on nucleosome positioning ( restoration in isw1Δ ) are seen with ‘open’ promoters , the largest effects on transcription are more closely associated with ‘closed’ promoters ( Figure 6A ) . 10 . 7554/eLife . 06073 . 015Figure 6 . Gene clusters identified by their response to RSC loss reveal different promoter classes . ( A ) The relative expression in rsc2 ( left ) or isw1Δ ( right ) mutants relative to wild type as measured by the HybMap assay are plotted as box and whisker distribution plots for each of the six gene clusters identified by their response to RSC loss . ( B ) The mean occupancy profile over each of the six gene clusters is presented for six different factors , including RSC , ISW1 , SNF2 , ISW2 ( Zentner and Henikoff , 2013 ) , CHD1 ( Zentner and Henikoff , 2013 ) , and H2AZ ( Albert et al . , 2007 ) . ( C ) The mean profile for histone turnover over the six gene clusters is shown . Higher values represent higher turnover . ( D ) Heat map representing the p-value significance for the intersection between genes in different categories . Open promoters include genes in clusters 1–4 . Closed promoters include genes in clusters 5 and 6 . Figure 6—figure supplement 1 shows the distribution of normal gene expression for each of the clusters . Figure 6—figure supplement 2 shows the occupancy profile for RSC and ISW1a as determined by deep sequencing . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 01510 . 7554/eLife . 06073 . 016Figure 6—figure supplement 1 . Wild-type RNA expression levels are not significantly different between the six clusters . ( A ) A histogram displays the representation of genes from each cluster on the custom mini-HybMap microarray . Cluster 6 is over-represented because of higher levels of RSC occupancy at these genes . ( B ) A box and whisker plot representing the distribution of median log2 coverage from stranded RNA sequencing for genes in each cluster . Data are from ( Parkhomchuk et al . , 2009 ) . ( C ) A box and whisker plot representing the distribution of median log2 coverage from unstranded RNA sequencing for genes in each cluster . Data are from ( Nagalakshmi et al . , 2008 ) . ( D ) A box and whisker plot representing the distribution of median log2 coverage from stranded RNA microarray for genes in each cluster . Data are from ( Xu et al . , 2009 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 01610 . 7554/eLife . 06073 . 017Figure 6—figure supplement 2 . Occupancy of RSC and ISW1a as measured by sequencing . Enrichment profiles for each of the six gene clusters derived in Figure 5 are shown for RSC ( parts ( A ) and ( B ) ) and ISW1a ( part ( C ) ) . The profile for each cluster is drawn in a different color . Values are log2 fold enrichments over input and collected in 20 bp windows flanking the TSS ±800 bp . Windows overlapping neighboring genes were excluded . The Rsc8 enrichment shows strong enrichment for the +1 nucleosome but not upstream locations , possibly due to altered protein configurations or low efficiency . The Sth1 enrichment shows strong enrichment at both −1 and +1 nucleosomes , as well as a broad upstream enrichment for cluster 6 , similar to the Rsc8 microarray . The Ioc3 enrichment also shows a broad enrichment over the upstream region for cluster 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 017 We next addressed the relationship between RSC occupancy and promoter architecture ( open vs closed promoters ) . Here , we plotted the mean occupancy profile for both Rsc8 and Ioc3 ( Figure 6B and Figure 6—figure supplement 2 ) across the promoter for each of the six clusters identified in Figure 5 . One might expect that genes with a strong response in regard to nucleosome positioning would have high RSC occupancy . Somewhat surprisingly , we found the opposite result . Clusters 5 and especially 6 had the highest mean occupancy of both RSC and ISW1a . We also examined other chromatin remodeler occupancies , including SWI/SNF ( this study ) , Isw2 ( Zentner and Henikoff , 2013 ) , and Chd1 ( Zentner and Henikoff , 2013 ) . Notably , Isw2 and SWI/SNF occupancy displayed higher occupancy in cluster 6 , while Chd1 was more equally distributed among the clusters . Interestingly , histone H2AZ demonstrated an inverse relationship , as clusters 5 and 6 bore the least H2AZ . Considering that these two gene clusters have the highest occupancy of chromatin remodelers , we next asked whether these genes also exhibited high histone turnover . We plotted the mean profile of measured histone turnover rate ( Rufiange et al . , 2007 ) over the six gene clusters and found that the degree of histone turnover correlated well with remodeler occupancy , with cluster 6 having the highest turnover , particularly around and upstream of the TSS ( Figure 6C ) . Thus , unstructured/closed promoters have the highest remodeler occupancies and the highest turnover . Together , these observations coalesce around the idea that these gene clusters identified in Figure 5 , based solely on the impact of RSC remodeler loss , also broadly segregate genes into two distinct types of promoter architectures: open ( structured ) promoters and closed ( unstructured ) . The gene clusters with the greatest measurable impact on chromatin organization due to RSC loss , groups 1–4 , represent the open promoters , whereas groups 5 and 6 represent closed promoters , which collectively lack a distinctive organization and therefore a measurable impact . These promoter architectures matched well with the predictions of remodeler occupancy and histone turnover ( Tirosh and Barkai , 2008; Cairns , 2009 ) . These architectures are also predicted to correlate with specific DNA sequence characteristics and nucleosome composition . For example , open promoters typically contain nucleosome exclusion sequences clustered with binding sites for factors that may help exclude or reposition nucleosomes ( Segal et al . , 2006; Badis et al . , 2008; Hartley and Madhani , 2009 ) . To verify these , we scored promoters for the presence of TATA , Reb1 , and Rsc3 binding sites , as well as the number of AAAAA sequences , which antagonize nucleosome formation ( Kaplan et al . , 2009; Segal and Widom , 2009 ) . We then calculated the statistical enrichment of these sequence attributes for both structured ( clusters 1–4 ) and unstructured ( clusters 5–6 ) genes over background by random permutation analysis . We found that clusters 1–4 showed statistically significant enrichments for Reb1 and Rsc3 binding sites and AAAAA sequences , while closed promoters showed an enrichment of TATA binding sites , matching the predictions ( Figure 6D ) . Importantly , it is unstructured/closed and TATA-rich promoters that have been shown previously to mostly rely on chromatin modifiers and remodelers for their activation ( Raser and O'Shea , 2004; Musladin et al . , 2014 ) . As developed in the ‘Discussion’ , we believe our results , combined with others , argue for two different modes of impact of RSC and other remodelers at the two promoter types: open and closed ( Figure 7 ) . 10 . 7554/eLife . 06073 . 018Figure 7 . Model of action by RSC and ISWI remodelers at open and closed promoters . An open or structured promoter is depicted on the left with regularly spaced nucleosomes ( yellow ovals ) and a predominate NDR that frequently contains sequence elements ( colored lines ) , including Rsc3 and Reb1 binding sites as well AT-rich sequence tracts unfavorable to nucleosome formation . Remodelers such as RSC ( blue oval ) help to maintain nucleosome deficiency , while ISWIa ( orange oval ) antagonizes by ‘filling-in’ the NDR . ( Note: Rsc3 is not required for RSC activity nor is Rsc3 required for all RSC recruitment . ) In the absence of RSC , this filling-in occurs and is conducted by ISW1a , as filling-in is not observed in rsc isw1 double mutants . A closed or unstructured promoter is depicted on the right , evidenced by the lack of a clearly defined NDR and obscured promoter sequence elements , such as the TATA . Nucleosome density ( or likelihood of occupancy ) is depicted by the opacity of the nucleosomes . These promoters have increased nucleosome movement and histone turnover ( yellow arrows ) , likely aided by chromatin remodelers such as RSC and ISWI , which eject or reposition nucleosomes , respectively . In the absence of RSC , nucleosome ejection is reduced , leading to higher nucleosome density ( opaque nucleosomes ) and a reduction in transcription . Additional loss of ISW1a may reduce the assembly/organization of nucleosomes in the promoter , partially restoring transcription . DOI: http://dx . doi . org/10 . 7554/eLife . 06073 . 018
Chromatin remodelers represent a set of complexes with different functional roles; some remodelers are primarily involved in transcriptional activation , while others are more dedicated to chromatin assembly and/or transcriptional repression . Here , we describe an antagonistic relationship between two such chromatin remodelers , RSC and ISW1 , through a combination of genetics , gene expression , and genome-wide nucleosome positioning studies . At genes , RSC is primarily utilized for gene activation , providing this function , at least in part , by establishing or maintaining the NDR structure at promoters . We find that this function is partly counteracted by Isw1 activity , which re-positions nucleosomes to ‘fill in’ the NDR and positions nucleosomes over cis regulatory sequences . While there are other remodelers that also act at promoters , we consider the interactions described herein as the strongest evidence to date exemplifying chromatin remodeler antagonism . Evidence for RSC-Isw1 antagonism was revealed through two entirely independent unbiased genetic screens for suppression of RSC mutants . The first screen utilized an SGA method to identify suppressors of rsc7Δ and revealed isw1Δ as the strongest of four identified gene suppressors and the only gene with a chromatin-related function . Indeed , combinations of rsc mutants with mutations in chromatin factors are almost invariably lethal ( rsc4-HDAC combinations are a rare exception [Kasten et al . , 2004] ) . The second screen—involving rsc2 suppression by histone mutations—yielded a small set of mild suppressors in histone H3 and one suppressor of moderate strength , H4 RH17 , 18CY . This region of the H4 tail is known as the ‘basic patch’—an epitope of known importance for the binding and activity of several chromatin-modifying factors including ISWI , Sir3 , and Dot1 ( Clapier et al . , 2001; Clapier et al . , 2002; Altaf et al . , 2007; Fingerman et al . , 2007 ) . Further genetic work focused the impact of this mutation on ISWI function , then on Isw1 function , and finally on Isw1a function ( as opposed to the compositionally distinct Isw1b complex ) . Notably , combinations of isw1Δ with Swi/Snf mutations did not confer suppression , indicating specificity for suppressing RSC function . Taken together , two independent genetic screens , combined with multiple additional genetic approaches , identify a specific suppression relationship between the RSC complex and the Isw1a complex . This RSC-ISW1a suppression is also consistent with a recent report that loss of Isw1a complex can suppress the phenotypes of gcn5Δ mutations combined with loss of another H3 acetyltransferase , Sas3 ( Lafon et al . , 2012 ) . It is possible that the isw1Δ and ioc3Δ suppression of gcn5Δ sas3Δ may be partially due to reducing the phenotypic effects of reduced acetylation by reducing RSC activity , since RSC function is partially dependent on acetylation ( VanDemark et al . , 2007 ) . We then explored whether this suppression relationship resulted from opposing roles of the two remodelers for regulating chromatin structure . A role for RSC at maintaining proper chromatin structure was previously demonstrated through the use of the strong Sth1 degron allele ( Parnell et al . , 2008; Hartley and Madhani , 2009 ) and other RSC alleles ( Badis et al . , 2008; Ganguli et al . , 2014 ) . Loss of RSC function results in a gain of nucleosome density across Pol III genes and at the NDR of many Pol II genes . A clear observation here is the ‘leftward’ shift of the +1 and subsequent nucleosomes towards the NDR . This is consistent with ( and extends ) published models that RSC maintains the NDR by moving and/or ejecting nucleosomes from the TSS . Our work suggests that this movement and ‘fill in’ is , at least in part , performed by the ISWI family of remodelers , as we have demonstrated a reduction of the ‘fill in’ in the rsc isw1 double mutant . Cells lacking ISW1 alone exhibit modest changes in the coding region ( Ni et al . , 2010; Tirosh et al . , 2010; Yen et al . , 2012 ) , which may , in part , be due to the loss of the ISW1b complex , which is thought to act primarily in the coding region , as opposed to the ISW1a complex that acts at promoters ( Morillon et al . , 2003 ) . Our work here provides the first molecular examination of rsc isw1 double mutants ( prompted by our genetic suppression relationships ) demonstrating antagonism between these remodelers regarding the depth of the NDR , the occupancy and positioning of the +1 nucleosome , and the phasing of proximal nucleosomes in the coding region ( Figure 7 ) . The clustering of gene promoters into different classes based on their chromatin response to the loss of RSC function also revealed an interesting insight regarding the organization of promoter chromatin ( Figure 7 ) . More responsive genes have an open/structured promoter , with a classic −1 , NDR , and +1 nucleosome at uniform positions with respect to the TSS . These patterns are evolutionarily conserved and partially imposed by sequence ( Ioshikhes et al . , 2006; Tsankov et al . , 2011 ) , where the open promoters demonstrate a higher enrichment of nucleosome exclusion sequences , such as tracts of AAAAA , and illustrated by the strong NDR in the prediction model . While sequence alone cannot entirely dictate chromatin structure ( Zhang et al . , 2009 ) , chromatin remodelers like RSC are able to reinforce the NDR by moving nucleosomes out of the NDR . The increased likelihood of Rsc3 or Reb1 binding sites occurring within the NDR may help recruit RSC or other factors to promoters that require nucleosome sliding or ejection activity to maintain this open architecture ( Badis et al . , 2008; Hartley and Madhani , 2009 ) . However , we note that the transcriptional output from these open promoters appears less affected following RSC loss than at closed/structured promoters ( see below ) . In contrast , the genes that lack a uniform chromatin response to RSC loss tend to have a closed or covered promoter , where nucleosomes are not uniformly positioned with respect to the TSS ( Figure 7 ) . This is not to say that these promoters have no chromatin structure at all; rather , each promoter has a unique chromatin structure that is not uniformly identical in phasing . In composite measurements , such as those presented in Figure 5B , these promoters appear to have little chromatin structure , when , in reality , they simply lack consensus structure . These promoters have an increased likelihood to have a TATA box and other transcription factor binding sites , whose access may be regulated by the partial occlusion by nucleosomes ( Ioshikhes et al . , 2006; Tirosh and Barkai , 2008 ) . Here , Isw1a may function to help assemble/mature and properly space nucleosomes at these promoters to repress transcription , which then increases their reliance upon remodelers such as RSC and/or SWI/SNF to expose these binding sites for proper activation . Hence , these promoters would have an increased presence of both activating and repressing chromatin remodelers , as well as histone turnover , both of which we observe ( Figure 7 ) . This continual state of flux , as well as lack of uniformity , may help explain why we observe little collective change in the chromatin structure in the absence of RSC function , while also observing a greater reliance on RSC function to maintain an active transcriptional status . Taken together , our study provides the first evidence for an antagonistic relationship between RSC and ISWI , showing the genetic suppression of growth phenotypes and the lessening of chromatin impact due to the loss of RSC function . These effects are revealed on a genome-wide scale and further reveals that particular promoter chromatin architectures can influence the degree of impact . These results reveal the different strategies chromatin used by genes for maintaining and regulating genic transcription through the use of promoter architecture , DNA accessibility , and the antagonism between complexes that act on promoter chromatin .
Rich media ( YPD ) , synthetic complete ( SC ) , minimal synthetic defined ( SD ) , and sporulation media were prepared by standard methods . Standard procedures were used for transformations , sporulation , and tetrad analysis . All strains are derivatives of S288C , and full strain genotypes are listed in Supplementary file 1 . Plasmids used are listed in Supplementary file 2 . Null mutations in ISW1 or IOC3 were obtained from Invitrogen ( Carlsbad , CA ) and crossed in , or made by PCR disruption , and confirmed by PCR and complementation . To isolate mutations in Histone H3 or Histone H4 that could suppress an rsc2 TS− mutant , p1411 [HHT2-HHF2 . TRP1] was mutagenized with hydroxylamine and transformed into YBC2140 ( rsc1Δ rsc2-V457M hht1Δ-hhf1Δ hht2Δhhf2Δ [HHT2-HHF2 . URA3] ) . Approximately 20 , 000 transformants were plated to SC-TRP + 5FOA medium , incubated at 33°C , and screened for colony growth . Resident plasmids conferring suppression were isolated , retransformed , and sequenced . The SGA screen was performed by mating rsc7Δ [RSC7 . URA3] ( YBC2039 ) with the yeast haploid deletion set ( BY4741 ) from Invitrogen and isolating double mutants as described in Wilson et al . ( 2006 ) . Double mutants were scored for the ability to grow at 35°C following RSC7 plasmid loss on 5FOA . RSC8 , SNF2 , and IOC3 genes were tagged endogenously with 13xMyc tags as described ( Longtine et al . , 1998 ) . Yeast strains were grown in either rich media ( YPD ) or minimal media ( SD ) and ChIP performed from both samples as described previously ( Parnell et al . , 2008 ) . ChIP eluates and input DNA were labeled with either Cy5 or Cy3 , and two biological replicates of each were hybridized to Agilent 244K microarrays . The ChIP efficiency was better in cells grown in SD media , perhaps due to increased cross-linking efficiency ( rich media may inherently have a quenching effect relative to minimal media ) . Comparison between YPD- and SD-derived occupancies revealed little differences besides the relative scale of enrichment; therefore , all analysis was performed using the SD data . For ChIP sequencing , the Rsc8-Myc , Ioc3-Myc , and Snf2-Myc strains were used in addition to a strain expressing Sth1 tagged with 2xFlag under a Met25 promoter . ChIP conditions were similar to those used previously , except chromatin was liberated by micrococcal nuclease . Immunoprecipitated products and corresponding input were assembled into a library using Illumina protocols . Library products were size-selected for mono-nucleosomes prior to paired-end sequencing ( 36 bp for Rsc8 , 50 bp for remainder ) using Illumina sequencers . The HybMap microarray was custom designed to represent genes with a range of RSC and ISW1 occupancies and included 448 coding genes , 93 tRNA genes , and 52 non-coding genes . Gene regions were extended by either 300 bp ( coding and non-coding ) or 150 bp ( tRNA ) . Probes were selected from a pool of tiled 60 mers and adjusted for length to match melting temperatures as necessary . Both strands for each probe were included in the design . Probes have a mean spacing of ∼50 bp . As a control , 502 probes with sequences from zebrafish were included as non-hybridizing control probes; sequences were confirmed not to have significant homology to yeast sequences . Microarray designs were submitted to Agilent Technologies for production as 4 × 44K arrays . Total RNA from three biological replicates was prepared from each yeast strain , hybridized to the array , and detected as described ( Dutrow et al . , 2008 ) . Yeast strains were grown under degron-inducing conditions , and mono-nucleosomal DNA was isolated as described previously ( Parnell et al . , 2008 ) . DNA fragments were size-selected by agarose gel electrophoresis , purified , and labeled with either Cy3 or Cy5 . Labeled DNA from three biological replicates was co-hybridized to Agilent 244K microarrays for each strain . For sequencing , mono-nucleosomal DNA was prepared into a library using Illumina kits and subjected to paired-end 50 bp sequencing . Raw microarray data were quantile normalized , averaged , median scaled , and assigned to genomic coordinates . For the HybMap protocol , probe values were median scaled to the median intensity from the zebrafish control probes . Probe sequences were mapped to the Saccharomyces cerevisiae genome version 64 ( Saccharomyces Genome Database ) . Gene transcript models were based on whole-genome transcriptome data ( Xu et al . , 2009 ) . Transcription start and stop sites were generated from processed transcriptome data and compared and merged with published transcript models . Transcripts with discrepancies were manually curated using published occupancy maps for nucleosome and promoter initiation factors as guides . This resulted in a list of 5338 high-quality transcript models . For ChIP sequencing data , including published data sets obtained through NCBI , raw Fastq alignments were aligned using Novoalign and processed using the MACS2 software ( https://github . com/taoliu/MACS ) to generate fold enrichment data . Most analysis was performed using BioToolBox ( https://github . com/tjparnell/biotoolbox ) . Cluster analysis was visualized with Java Treeview ( Saldanha , 2004 ) . Statistics and graphs were generated with GraphPad Prism ( GraphPad Software , Inc . ) . Intersection analysis was performed with the USeq package ( Nix et al . , 2008 ) . ChIP enrichment FDR values were calculated using MACS2 . Supplemental figures and files are available . Raw microarray and sequencing data are available at GEO under accession number GSE65594 . | The genome of an organism can contain hundreds to thousands of genes . However , these genes are not all active at the same time . Instead , mechanisms exist that control when genes are switched off or on . One such mechanism alters how tightly DNA is packaged into a structure called chromatin . To form chromatin , DNA is wrapped around histone proteins at different points along its length; these structures are known as nucleosomes . Once formed , chromatin can either adopt a tightly packed form that represses gene activity or a less compact form associated with gene activation . The proteins that control how chromatin is packed are called ‘chromatin remodelers’ . These proteins work in complexes that either disassemble chromatin—for example , by repositioning nucleosomes or removing histones—or organize chromatin by replacing nucleosomes and making it more compact . Studies in many organisms have identified two key families of chromatin remodelers . In yeast , the ISWI family of complexes assembles chromatin and a protein complex called RSC disassembles chromatin . Parnell , Schlichter et al . used a range of genetic techniques to investigate whether these two chromatin-remodeling complexes work against each other in a species of yeast called Saccharomyces cerevisiae . The results suggest that this is indeed the case . Both the ISWI complex and the RSC complex bind to regions of DNA called promoters , which are found at the start of a gene and help to regulate its activity . Parnell , Schlichter et al . found that the RSC complex helps to activate genes by establishing or maintaining regions of nucleosome-poor chromatin at a promoter . The chromatin is relaxed in these regions , which allows the proteins that activate genes to access the DNA . This effect is partially counteracted by the ISWI complex , which repositions nucleosomes across the promoters to fill the gaps created by the RSC complex . In comparison , Parnell , Schlichter et al . found that promoters that do not have regions of nucleosome-poor chromatin contain a larger number of both of the remodeling complexes and have a high turnover of histone proteins . This suggests that at these sites , the RSC proteins are needed to overcome the assembly of nucleosomes by the ISWI complex in order to activate the gene . Thus , these two chromatin remodelers , ISWI and RSC , compete at promoters to determine whether they contain or lack nucleosomes , which helps determine whether the gene is silent or active , respectively . Future work will look further at how the ‘winner’ is determined: how transcription factors or signaling systems within the cell bias the recruitment or activity of RSC or ISWI at particular promoters , to determine gene activity . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2015 | The chromatin remodelers RSC and ISW1 display functional and chromatin-based promoter antagonism |
A plethora of non-protein coding RNAs are produced throughout eukaryotic genomes , many of which are transcribed antisense to protein-coding genes and could potentially instigate RNA interference ( RNAi ) responses . Here we have used a synthetic RNAi system to show that gene copy number is a key factor controlling RNAi for transcripts from endogenous loci , since transcripts from multi-copy loci form double stranded RNA more efficiently than transcripts from equivalently expressed single-copy loci . Selectivity towards transcripts from high-copy DNA is therefore an emergent property of a minimal RNAi system . The ability of RNAi to selectively degrade transcripts from high-copy loci would allow suppression of newly emerging transposable elements , but such a surveillance system requires transcription . We show that low-level genome-wide pervasive transcription is sufficient to instigate RNAi , and propose that pervasive transcription is part of a defense mechanism capable of directing a sequence-independent RNAi response against transposable elements amplifying within the genome .
Over the past decade , our understanding of the complexity of the eukaryotic transcriptome has been revolutionized . Genome-wide sequencing studies in many organisms have revealed that protein-coding mRNAs are augmented by a multitude of non-protein coding RNAs ( ncRNAs ) , many produced from regions of the genome traditionally considered to be transcriptionally silent ( The ENCODE Project Consortium , 2012; Bertone et al . , 2004; Cheng et al . , 2005; David et al . , 2006; Birney et al . , 2007 ) . Functional data for the vast majority of ncRNAs are currently lacking , with only a few examples characterized in any detail; however , the diversity of mechanisms by which these act suggests that ncRNAs have a rich and varied biology that is largely still to be sampled . Long ncRNAs which overlap protein-coding genes have the potential to modulate the expression of their cognate coding RNA . Early characterized examples in yeast were thought to work by directly disrupting transcription factor or polymerase binding to the promoter of the coding RNA ( Martens et al . , 2004; Hongay et al . , 2006 ) ; however , more recent data implicate specific chromatin structure changes in repression ( Gelfand et al . , 2011; Hainer et al . , 2011 ) , and many other cases of ncRNAs that alter chromatin modifications have been described ( Camblong et al . , 2007; Berretta et al . , 2008; Houseley et al . , 2008; Pinskaya et al . , 2009; van Werven et al . , 2012 ) . Chromatin modifications are not necessarily repressive , and ncRNAs that enhance expression of their overlapping coding gene have also been described ( Uhler et al . , 2007; Hirota et al . , 2008 ) . In these examples , chromatin modifications are deposited during transcription , and therefore the act of transcription rather than the ncRNA itself is important . This is not always the case , and in higher eukaryotes multiple cis-acting ncRNAs have also been characterized , particularly as functional agents in imprinting . For example , Air and Kcnq1ot1 act in cis to deposit repressive chromatin marks and DNA methylation , but these ncRNAs interact with chromatin modifiers and allele specificity is achieved by restriction of the ncRNA to the vicinity of the transcription site , although the importance of the transcription itself remains controversial ( Nagano et al . , 2008; Pandey et al . , 2008; Redrup et al . , 2009; Latos et al . , 2012 ) . Genomes are also replete with low abundance and unstable RNA . The vast majority of ncRNAs in budding yeast are unstable ( Neil et al . , 2009; van Dijk et al . , 2011 ) , limiting the potential action of the RNAs themselves , although the transcription of such RNAs can still alter gene expression ( reviewed in Houseley , 2012 ) . Such unstable RNAs are also widespread in higher eukaryotes , probably with similar functional roles ( Chekanova et al . , 2007; Preker et al . , 2008 ) . More mysterious is the pervasive transcription that permeates eukaryotic genomes; the ENCODE project found that almost all the human genome is transcribed at some point , but the products of this transcription are vanishingly rare ( Cheng et al . , 2005; Birney et al . , 2007; Kapranov et al . , 2007; Goodman et al . , 2012 ) . It appears that regions of the genome which are not actively transcribed for other reasons undergo pervasive transcription; however , it is not known whether this pervasive transcription simply represents transcriptional noise or whether the transcription or RNAs themselves have important but as yet undiscovered functions . Systems in which a ncRNA is transcribed antisense to a sense protein-coding RNA are common and have strong regulatory potential ( Figure 1A ) ( Derrien et al . , 2012; Carninci et al . , 2005; Xu et al . , 2009 ) . It has been suggested that , since antisense ncRNAs are perfectly complementary to their cognate mRNA , the two species could form double stranded RNA ( dsRNA ) that would be a substrate for the RNA interference system ( RNAi ) . During a basic RNAi response , dsRNA is cleaved by the endonuclease Dicer into short interfering RNA ( siRNA ) , of which one strand is then loaded onto an Argonaute protein . The Argonaute–siRNA complex can anneal to complementary sequences in target RNAs , which are then cleaved by the endonuclease activity of Argonaute . RNAi was originally discovered in Caenorhabditis elegans and rapidly linked to the phenomenon of post-transcriptional gene silencing in plants; however , almost all eukaryotes contain Dicer and Argonaute orthologues and therefore have some form of RNAi system ( Hamilton and Baulcombe , 1999; Fire , 1998; Hannon , 2002 ) . RNAi probably evolved to protect cells against dsRNA viruses , a role which is maintained in plants , insects , and lower eukaryotes ( Ding , 2010 ) and has recently been described in mammalian cells ( Li et al . , 2013; Maillard et al . , 2013 ) . RNAi also forms a potent defense against transposons , and high-copy transposon-derived sequences are excellent targets for RNAi , giving rise to copious siRNAs in most eukaryotes including mammals ( Yang and Kazazian , 2006; Slotkin and Martienssen , 2007; Babiarz et al . , 2008 ) . In addition to degrading transposon-derived and viral RNA , siRNAs can mediate transcriptional repression of target RNAs through chromatin modifications and DNA methylation , although this activity is seemingly much stronger in lower eukaryotes and plants than in mammals ( Martienssen et al . , 2005; Lejeune et al . , 2010; Zhang and Zhu , 2011 ) . However , the source of the dsRNA that is processed into siRNA is not always obvious , nor is the mechanism by which cells differentiate host and transposon-derived sequences . siRNA-mediated repression is complemented by PIWI-interacting RNAs ( piRNAs ) , which bind to the Argonaut-related PIWI-domain proteins and enforce transposon repression in the germline of eukaryotes from worms to mammals ( Siomi et al . , 2011 ) . piRNAs are derived from specific genomic clusters , but it is unclear how the transcripts from these clusters are selected for processing into primary piRNAs and many of the processing enzymes remain to be identified ( reviewed in Ishizu et al . , 2012 ) . 10 . 7554/eLife . 01581 . 003Figure 1 . Frequency of annotated antisense non-protein coding RNAs ( ncRNAs ) and effects on mRNA abundance . ( A ) Schematic of an example sense mRNA-antisense ( ncRNA ) system . ( B ) Number of annotated open reading frames ( ORFs ) with antisense transcripts . Positions of CUTs , SUTs , and XUTs were collated with expressed ORFs ( Xu et al . , 2009; van Dijk et al . , 2011 ) , SUTs were later re-classified as XUTs were removed . Overlaps between ORFs expressed in glucose media ( total 5171 , Xu et al . , 2009 ) and other RNAs were calculated and summed for increasing minimum overlaps of 50–500 bp . ORF–ORF overlaps and ORF–ncRNA overlaps were analyzed separately as ORF–ORF overlaps are consistently smaller . Detailed figures are given in Table 1 . ( C ) Abundance of short interfering RNAs ( siRNAs ) in RNA interference ( RNAi ) + strain produced from expressed ORFs with and without an annotated overlapping antisense ncRNA , based on read counts from published high-throughput sequencing data ( Drinnenberg et al . , 2011 ) . Minimum antisense overlap with ORF was set at 250 bp; only ORFs with >100 reads in the wild-type poly ( A ) + library were assessed to remove noise . Stated p value calculated by Student’s t test . ( D ) Abundance of mRNA in RNAi+ cells relative to wild-type; data source and categories as in C , differences were not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 003 Various studies have looked for endogenous sense-antisense RNA pairs that instigate RNAi responses . Efficient generation of siRNAs from endogenous sense-antisense systems ( endo-siRNA ) has been observed in plants under stress ( Borsani et al . , 2005; Katiyar-Agarwal et al . , 2006 ) , and mammalian oocytes generate endo-siRNAs that can mediate mRNA knockdown ( Tam et al . , 2008; Watanabe et al . , 2008 ) . However , although endo-siRNAs have been detected outside the germline in mammals , they are surprisingly under-represented where sense and antisense RNAs are co-expressed ( Faghihi and Wahlestedt , 2006; Okamura et al . , 2008; Carlile et al . , 2009 ) , and overall there is a positive correlation between antisense and sense RNA expression in mammalian genomes , which is inconsistent with RNAi ( Derrien et al . , 2012; Katayama et al . , 2005 ) . This raises questions about whether endogenous sense-antisense systems do in fact form dsRNA in vivo and , if so , whether all dsRNA is equivalently accessible to Dicer . The tight integration of the RNAi system into the physiology of most eukaryotic cells makes it very difficult to disentangle direct and indirect effects of mutating RNAi components ( reviewed in Ketting , 2011 ) . To elucidate factors important for the induction of RNAi by endogenous sense-antisense systems , we therefore used a recently described synthetic system in which RNAi is reconstituted in Saccharomyces cerevisiae by the introduction of Argonaute and Dicer from the related yeast S . castellii ( Drinnenberg et al . , 2009 ) . S . cerevisiae is highly unusual in lacking an endogenous RNAi system , allowing maintenance of the symbiotic dsRNA Killer virus ( Drinnenberg et al . , 2011 ) . The reconstituted system is functional , since RNAi+ S . cerevisiae efficiently degrades exogenous hairpin RNAs and endogenous Ty retrotransposon transcripts; however , no clear mRNA expression changes are detectable in these cells ( Drinnenberg et al . , 2009 , 2011 ) .
Antisense ncRNAs exist for many S . cerevisiae genes; combining published datasets we found that 15–30% of yeast open reading frames ( ORFs ) have an annotated antisense ncRNA depending on the minimum size of overlap considered ( Xu et al . , 2009; van Dijk et al . , 2011 ) , not including overlapping convergent ORFs ( Figure 1B and Table 1 ) . In RNAi+ cells , these ORFs produce more siRNA than ORFs lacking an antisense ( Figure 1C ) , showing that they are transcribed into dsRNA that is targeted by the RNAi machinery . However , these ORFs do not show reduced mRNA levels in RNAi+ cells consistent with published data , suggesting that insufficient siRNAs are produced to elicit a detectable mRNA knockdown ( Derrien et al . , 2012; Katayama et al . , 2005; Drinnenberg et al . , 2011 ) ( Figure 1D ) . 10 . 7554/eLife . 01581 . 004Table 1 . Stability of antisense ncRNAsDOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 004Overlap size ( bp ) Overlap typeTotalsORF–ORFORF–XUTORF–CUTORF–SUTORF–ncRNAORF–unstable ncRNA50700106657552215961448 ( 91% ) 100449100854347515071367 ( 91% ) 15021696750842514231306 ( 92% ) 20013693147840313581249 ( 92% ) 2509689343438012871181 ( 92% ) 3006981239136311891085 ( 91% ) 350627593353511086990 ( 91% ) 40054694292334992899 ( 91% ) 45051637244322904811 ( 90% ) 50048591204302828741 ( 89% ) Number of ORFs overlapping with ORFs and various classes of ncRNAs , with various minimum size cut-offs for the overlapping region . Totals are given for ORFs overlapping with ncRNAs and with unstable ncRNAs , including a percentage of overlapping ncRNAs that are unstable . ncRNA: non-protein coding RNA; ORF: optical reading frame; XUT: Xrn1-sensitive unstable transcript ( degraded in cytoplasm ) ; CUT: cryptic unstable transcript ( degraded by nuclear exosome ) ; SUT: stable unannotated transcript ( not known to be degraded ) . We first asked whether any endo-siRNA pairs are degraded by RNAi in this reconstituted system . RNAi+ cells produce abundant siRNAs from sub-telomeric Y′ elements and from the ribosomal DNA ( rDNA ) intergenic spacers ( Drinnenberg et al . , 2011 and Figure 2—figure supplement 1 ) and , despite transcriptional repression by the histone deacetylase Sir2 , both regions transcribe sense and antisense ncRNAs that could hybridize to form dsRNA ( Aparicio et al . , 1991; Yamada et al . , 1998; Kobayashi and Ganley , 2005; Houseley et al . , 2007 ) ( Figure 2A , D ) . Northern blots revealed full-length ncRNAs from both strands of the Y′ elements in wild-type cells ( Figure 2B lanes 1 , 5 marked with arrows ) ; these were largely absent in the RNAi+ strain , being replaced by heterogeneous degradation products and readily detectable siRNAs ( Figure 2B , C ) . Despite weak transcriptional repression in this genetic background , ncRNAs and siRNAs were more abundant in sir2Δ mutants reinforcing the precursor–product relationship ( Figure 2B , C ) . Equivalent results were seen for the rDNA intergenic spacer region ( Figure 2D–F ) . These data show that endo-siRNA pairs can form RNAi substrates and undergo efficient degradation by a minimal RNAi system . 10 . 7554/eLife . 01581 . 005Figure 2 . High-copy endogenous sense-antisense pairs instigate efficient RNA interference ( RNAi ) . ( A ) Schematic diagram of sub-telomeric Y′ elements . ( B ) Northern analysis of Y′ element non-protein coding RNAs ( ncRNAs ) comparing wild-type and RNAi+ strains in SIR2 and sir2Δ backgrounds . 18S ribosomal RNA is shown as a loading control . Arrows indicate full-length RNA species . ( C ) Northern analysis of Y′ element-derived short interfering RNAs ( siRNAs ) from cells in B , tRNAs are shown as a loading control . ( D–F ) Equivalent analysis of rDNA intergenic spacer ncRNAs . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 00510 . 7554/eLife . 01581 . 006Figure 2—figure supplement 1 . Correlation of short interfering RNAs ( siRNAs ) with RNA abundance and silencing . Published small RNA sequencing data ( Drinnenberg et al . , 2011 ) were re-mapped to the complete Saccharomyces cerevisiae genome including non-unique sequences and reads were then summed in 50 bp bins on each strand . ( A ) Density of small RNAs in sub-telomeric Y′ elements in wild-type and RNA interference ( RNAi ) + strains compared with published data for silencing efficiency of inserted transgenes ( Pryde and Louis , 1999 ) . Separate traces are shown for siRNAs mapping to sense and antisense sequences above and below the x axis , respectively . siRNA density is inversely correlated with transcriptional silencing , as would be expected if siRNAs are produced from sub-telomeric non-protein coding RNAs ( ncRNAs ) . ( B ) siRNA density in a single rDNA repeat . Many small RNA fragments are produced from the 35S ribosomal RNA gene , as previously observed in Schizosaccharomyces pombe ( Buhler et al . , 2008 ) ; however , a discrete cluster of siRNAs is produced from the overlapping region between the IGS1-F and IGS2-R ncRNAs in the intergenic spacer region ( highlighted ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 006 One distinguishing feature of these regions is high copy number; to determine whether copy number amplification can drive RNAi , we examined MAL32 which has a clearly defined antisense RNA that is co-expressed with the sense mRNA ( Figure 3—figure supplement 1 ) . MAL32 is effectively present at two copies in the haploid genome as the orthologous gene MAL12 shares 99 . 5% nucleotide identity , reducing potential transcriptional repression . As for approximately 90% of yeast antisense RNAs ( Table 1 ) , MAL32 antisense RNA is highly unstable and is only detectable in strains lacking the nuclear exosome co-factor Trf4 ( reviewed in Houseley and Tollervey , 2009 ) ( Figure 3—figure supplement 1 ) , but endogenous MAL32 mRNA was not down-regulated in the RNAi+ strain even in trf4Δ cells ( Figure 3B and Figure 3—figure supplement 2A ) . When expressed from a high-copy plasmid , however , MAL32 mRNA was significantly down-regulated by RNAi ( Figure 3C ) with concurrent production of siRNA ( Figure 3D ) . The MAL32 antisense RNA was only detected in these experiments as a smear of degradation products and was not noticeably altered by RNAi , probably because nuclear degradation acts faster than RNAi on this substrate . To confirm that the knockdown of MAL32 mRNA was not an indirect effect of the strain background or an undirected Argonaute cleavage , we reconstituted the RNAi system in the BY4741 background using separate plasmids expressing Dicer and Argonaute . dsRNA from the MAL32 locus was detectable in these cells and was removed by Dicer; however , a significant knockdown of the mRNA was only observed in cells expressing both Dicer and Argonaute , confirming that the knockdown represents a genuine RNAi response ( Figure 3—figure supplement 3 ) . 10 . 7554/eLife . 01581 . 007Figure 3 . Copy number amplification of coding genes can instigate RNA interference ( RNAi ) . ( A ) Schematic of MAL32 locus . ( B ) MAL32 mRNA and antisense non-protein coding RNA ( ncRNA ) comparing wild-type and RNAi+ strains in wild-type and trf4Δ backgrounds; cells grown on YP raffinose ( extended image and quantification shown in Figure 3—figure supplement 2A ) . ( C ) mRNA and antisense ncRNA from MAL32 cloned onto a high-copy plasmid in wild-type and RNAi+ strains . Lanes 3 , 4 show empty vector control . Antisense panel shows degradation products , no full-length antisense is detectable due to Trf4 activity . ( D ) Short interfering RNA ( siRNA ) analysis of cells from C . ( E ) Schematic of GAL4 locus . ( F ) GAL4 mRNA and antisense ncRNA in wild-type and RNAi+ strains; cells grown in YP galactose ( extended image and quantification shown in Figure 3—figure supplement 2B ) . ( G ) mRNA and antisense ncRNA from GAL4 locus cloned onto a high-copy plasmid in wild-type and RNAi + strains . Lanes 5 , 6 show empty vector , signal is from genomic GAL4; note that cells used here are diploids to mitigate defects in galactose response ( see ‘Materials and methods’ ) . Lanes 3 , 4 show a previously described GAL4 antisense mutant ( Geisler et al . , 2012 ) ; this removes detectable antisense RNA for genomic GAL4 , but the mutant sequence still expresses an antisense ncRNA when cloned on the high-copy plasmid ( see Figure 3—figure supplement 4 ) . ( H ) siRNA analysis of cells in G . For quantification , n = 4 biological replicates , error bars represent ± 1se , *p<0 . 05 , ***p<0 . 01 by Student’s t test , y axes in arbitrary units . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 00710 . 7554/eLife . 01581 . 008Figure 3—figure supplement 1 . Characterization of the MAL32 sense/antisense system . Left: diagram of the MAL cluster near telomere at the right hand end of Chr . II . Sites enriched for H3K4 Me3 are indicated for log phase cells grown in glucose ( Kirmizis et al . , 2007 ) , including an enrichment site at the 3′ end of the MAL32 open reading frame ( ORF ) corresponding to the antisense non-protein coding RNA ( ncRNA ) promoter . The Saccharomyces cerevisiae genome contains two MAL clusters; the other is located at the right hand end of Chr . VII and contains the MAL32 paralogue , MAL12 , which differs by only a few single-nucleotide polymorphisms in the ORF ( 99 . 5% nucleotide identity ) . Right: northern blot showing MAL32 antisense and mRNA expression in glucose and raffinose . The antisense ncRNA , a major 2 kb transcript with additional 2 . 7 kb and 4 . 7 kb species , is ubiquitously expressed but completely degraded by Trf4 in wild-type cells . 5′ RACE analysis in trf4Δ cells ( not shown ) reveals that the antisense is produced from both MAL32 and MAL12 genes . The mRNA is repressed in glucose , but in raffinose is produced from both loci as revealed by the reduced but detectable mRNA expression in the mal12Δ strain . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 00810 . 7554/eLife . 01581 . 009Figure 3—figure supplement 2 . Images and quantification of RNA interference ( RNAi ) degradation patterns . ( A ) Degradation pattern of MAL32 mRNA and antisense non-protein coding RNA ( ncRNA ) in cells grown on YPRaf . Note the degradation of MAL32 mRNA in the trf4Δ sample ( lane 3 ) , which is abolished in the RNAi+ strain . This result was seen in multiple experiments and appears to be an indirect effect of RNAi . The large increase in MAL32 mRNA in the trf4Δ RNAi+ strain compared with trf4Δ results from this loss of RNA degradation . All antisense RNA species are included in the quantification; note that RNAi reduces the level of the full-length antisense RNA but slightly increases some truncated species . ( B ) Degradation pattern of GAL4 mRNA and antisense ncRNA in cells grown on YPGal . Note that degradation products are clearly visible in RNAi+ xrn1Δ sample even though quantification does not show a significant change in full-length RNAs ( compare heterogeneous species of 0 . 1–3 kb in lanes 4 , 8 with other lanes ) . For quantification , n = 4 biological replicates , error bars are ±1se , ***p<0 . 01 by Student’s t test , y axes in arbitrary units . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 00910 . 7554/eLife . 01581 . 010Figure 3—figure supplement 3 . Verification of RNA interference ( RNAi ) knockdown . Dicer and Argonaute were cloned out of the RNAi+ strain onto single-copy plasmids and transformed into the BY4741 background along with the MAL32 high-copy plasmid . Cells were grown on raffinose as in Figure 3C . ( A ) 3 µg RNase A-treated RNA was loaded in each lane to detect double stranded RNA ( dsRNA ) , which is largely resistant to degradation by RNase A . ( B ) MAL32 mRNA levels , with L-A viral RNA as a positive control . For quantification , n = 3 biological replicates , error bars represent ±1se , ***p<0 . 01 by Student’s t test . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 01010 . 7554/eLife . 01581 . 011Figure 3—figure supplement 4 . Expression of GAL4 antisense RNA in mutant construct . Left: structure of the 3′ end of wild-type and mutant GAL4 constructs showing primer locations for 5′ RACE . Right: 5′ RACE for capped RNA present in wild-type cells with the GAL4 , GAL4-3HA , or empty plasmids ( RNA samples as in Figure 3G lanes 1 , 3 , 5 ) . The GAL4 primer detects both wild-type and mutant species , and notably detects a product in the GAL4-3HA construct which is not reported to express an antisense RNA . This product is specific for 3HA as a product is also detected only in this construct with the 3HA primer at the predicted size . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 01110 . 7554/eLife . 01581 . 012Figure 3—figure supplement 5 . Plasmid copy number in high-copy plasmid strains . ( A ) Southern blot of wild-type and RNA interference ( RNAi ) + ura3Δ cells containing high-copy plasmids for MAL32 and GAL4 , with a W303 URA3 wild-type as a quantification control . DNA was extracted from cells grown to stationary phase in selective media , separated on a 1% gel , and probed for URA3 with GAL7 as a loading control . Note that the GAL7 signal is stronger in pGAL4 containing strains as these are diploids as described in ‘Materials and methods’ . Plasmid copy numbers were calculated relative to the W303 URA3 signal . ( B ) Plasmid analysis of MAL32 high-copy plasmid in log-phase cells showing that plasmid copy numbers are similar between wild-type and RNAi+ cells in growth conditions used for RNA isolation . A proportion of the total plasmid is always isolated during RNA preparation and is detected as a slow migrating RNase A-insensitive band . For RNase treatment , the sample shown in lanes 3/9 was treated with RNase A or no enzyme followed by purification with phenol:chloroform and ethanol precipitation . A short exposure of the full-length RNA is shown for the sense probe as this signal is over-exposed when the plasmid signal is visible . ( C ) Short interfering RNA ( siRNA ) reads per 100 bp for 2µ plasmid and chromosome I , showing that the endogenous 2µ plasmid produces high levels of siRNA from all genes compared with chromosomal genes; the three clear peaks correspond to FLP1 , REF1 , and RAF1 . The x axes are not to the same scale , data from ( Drinnenberg et al . , 2011 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 01210 . 7554/eLife . 01581 . 013Figure 3—figure supplement 6 . Confirmation of RNA interference ( RNAi ) action by immunoprecipitation and qRT-PCR . Cells were grown under the same conditions as in Figure 3C , G , then lysed and immunoprecipitated with either no antibody or monoclonal antibody J2 against double stranded RNA ( dsRNA ) . Quantitative RT-PCR was used to quantify MAL32 or GAL4 , L-A ( a dsRNA produced by Killer virus ) and ACT1 in each fraction . Signals were normalized to ACT1 , which should produce minimal dsRNA under normal conditions . For quantification , n = 3 biological replicates , error bars represent ±1se , *p<0 . 05 , ***p<0 . 01 by Student’s t test , data were log transformed for t test to avoid issues caused by the very large difference in dsRNA quantification between wt and RNAi+ cells , y axes in arbitrary units . The significant differences in the no antibody controls match the difference in the total lysates and reflect non-specific RNA binding . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 013 We then tested GAL4 , which is a single-copy gene with a co-expressed antisense that is degraded in the cytoplasm by Xrn1 ( Geisler et al . , 2012 ) ( Figure 3E ) . Cells lacking Xrn1 show increased levels of antisense and , unexpectedly , sense RNA , but , as for MAL32 , we did not detect a significant decrease in full-length RNA in RNAi+ xrn1Δ cells ( Figure 3F and Figure 3—figure supplement 2B ) . However , amplification of the locus by cloning on a high-copy plasmid leads to significant degradation of the sense RNA along with the antisense RNA by RNAi ( Figure 3G , H ) . Surprisingly , this occurred even in a known mutant that lacks antisense expression ( Geisler et al . , 2012 ) , but 5′ RACE ( Rapid Amplification of 5′ Complementary DNA Ends ) experiments revealed that antisense RNA is still produced by this mutant when expressed from a high-copy plasmid , even if it is too heterogeneous for detection by northern blot ( Figure 3—figure supplement 4 ) . Taken together , these experiments on MAL32 and GAL4 demonstrate that increasing gene copy number can make the products of a normal gene susceptible to RNAi . One potential confounding factor in these experiments is the copy number of the high-copy plasmids; if the copy number drops in RNAi+ cells , this would provide a trivial explanation for the observed knockdowns . However , Southern blotting revealed that RNAi+ cells contained approximately twofold more plasmid than the controls , which would tend to decrease rather than increase the apparent effect of RNAi ( Figure 3—figure supplement 5A , B ) . It is likely that RNAi degrades the mRNA for the plasmid-encoded selectable marker and 2µ maintenance genes ( 2µ genes produce copious siRNA , Figure 3—figure supplement 5C ) , and the plasmid copy number rises to compensate for this . We also wanted to use a different method to confirm the northern blot results . We therefore lysed wild-type and RNAi+ cells containing the MAL32 and GAL4 plasmids , precipitated dsRNA using a specific monoclonal antibody ( Schonborn et al . , 1991; Gullerova and Proudfoot , 2012 ) , and assayed total and dsRNA levels by quantitative RT-PCR . MAL32 mRNA was knocked down approximately 75% , as observed by northern blot analysis , while dsRNA was reduced 11-fold , consistent with specific removal of dsRNA by Dicer . GAL4 mRNA knockdown was measured at 80% in this assay , again with an 11-fold reduction in dsRNA in the RNAi+ strain ( Figure 3—figure supplement 6 ) . Increasing gene copy number also increases RNA production . To separate the contributions of RNA abundance and copy number , we analyzed existing genome-wide data ( Hobson et al . , 2012; Drinnenberg et al . , 2011 ) . If siRNA formation depends only on precursor RNA abundance , a positive correlation should be observed between total RNA abundance and siRNA abundance , and there should be no difference between distributions of single-copy loci and multi-copy loci . We observed little evidence for such a positive correlation ( Figure 4—figure supplement 1 ) ; however , plots of siRNA versus total RNA abundance for single-copy and multi-copy loci showed strikingly different distributions , with multi-copy loci clearly biased towards higher siRNA production ( Figure 4A ) . To quantify this difference , loci were segregated into eight groups of increasing total RNA abundance and siRNA abundance was assessed for single-copy and multi-copy loci in each group ( Figure 4B ) . siRNA production was significantly higher from multi-copy loci than from single-copy loci in all except the lowest category of RNA abundance . This result was robust to changes in the threshold between low and high copy , and was still observed in a comparison of low to medium copy number , showing that high-copy Ty1 retrotransposons were not distorting the analysis ( Figure 4—figure supplement 2 ) . A normalization step is required in these analyses to deal with mapping of sequence reads to multi-copy loci ( discussed in detail in ‘Materials and methods’ ) ; however , the same differences were observed with no normalization or a different normalization scheme ( Figure 4—figure supplement 3 ) . These surprising results show that multi-copy loci produce more siRNA than single-copy loci with equivalent RNA abundance . If this observation is real , the siRNA:total RNA ratio should be predictive of copy number , an important test since this comparison requires no copy number normalization . As predicted , the 1% of genome with the highest siRNA:total RNA ratio is massively enriched for multi-copy loci ( Figure 4C ) , and when this ratio was plotted across a chromosome , an obvious correlation was observed between regions of high-copy number and regions with high siRNA:total RNA abundance ( Figure 4D ) . These analyses clearly demonstrate that selectivity towards the products of multi-copy loci is an emergent property of a minimal RNAi system . 10 . 7554/eLife . 01581 . 014Figure 4 . Multi-copy loci are preferentially targeted by RNA interference ( RNAi ) . ( A ) Short interfering RNA ( siRNA ) ( Drinnenberg et al . , 2011 ) and total RNA ( Hobson et al . , 2012 ) abundance for loci with copy number <2 ( left , single-copy ) or ≥2 ( right , multi-copy ) . ( B ) Quantification of data from A binned into categories of increasing total RNA level , with p values for pairwise comparisons of siRNA abundance in single-copy and multi-copy datasets using the Wilcoxon Rank Sum test . ( C ) Copy number distribution of the 1% of loci with the highest siRNA:total RNA ratio compared with other loci; difference is significant by Wilcoxon Rank Sum test , p<2 . 2 × 10−16 , loci scoring below noise threshold ( 0–2 category in B ) were removed . n values for tests in B and C are given in Table 2 . ( D ) Comparison of copy number with siRNA:total RNA ratio across chromosome I . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 01410 . 7554/eLife . 01581 . 015Figure 4—figure supplement 1 . Short interfering RNA ( siRNA ) versus total RNA abundance . Plot of genome-wide siRNA abundance ( Drinnenberg et al . , 2011 ) versus total RNA abundance ( Hobson et al . , 2012 ) . Abundance scores represent log-transformed sums of total read counts for each 100 bp segment of the yeast genome . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 01510 . 7554/eLife . 01581 . 016Figure 4—figure supplement 2 . Increased short interfering RNA ( siRNA ) production from multi-copy sequences is robust to different cut-offs . The dataset shown in Figure 4B was re-analyzed with varying parameters; significant differences between multi-copy and single-copy loci are observed in all cases . p values shown were calculated by Wilcoxon Rank Sum test , n values are given in Table 2 . ( A ) Very high-copy sequences ( >5 copies ) were removed from the analysis to avoid the influence of a small number of very high-copy sequences . ( B ) The datasets were split into <5 and ≥5 copies to ensure that the observed difference was not an artefact of the arbitrary two-copy cut-off used in Figure 4B . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 01610 . 7554/eLife . 01581 . 017Figure 4—figure supplement 3 . Increased short interfering RNA ( siRNA ) production from multi-copy sequences is not due to copy number normalization . ( A ) No normalization for copy number was applied to total RNA abundance levels . ( B ) Both total RNA and siRNA abundance were multiplied by copy number for each analyzed locus . See ‘Materials and methods’ section for information on these normalization methods . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 017 We then directly tested the effect of copy number at the MAL32 locus . We constructed strains in which MAL32 sense and antisense RNAs were expressed at similar levels from multi-copy or single-copy loci by over-expressing single-copy sense and antisense ( Figure 5A ) . In this system , both sense and antisense RNAs were produced at higher levels from the single-copy system ( Figure 5B compare lanes 1 and 3 ) but more siRNAs were produced from the multi-copy system ( Figure 5C compare lanes 2 and 4 ) . The over-expression of both RNAs from the single-copy MAL32 locus led to the production of easily detectable siRNA , as would be expected; however , this result directly demonstrates that gene copy number influences the formation of siRNA above and beyond the effect of total RNA abundance . The increased siRNA production in these cells is most likely due to enhanced dsRNA formation in the multi-copy system . To confirm this , we quantified MAL32 RNA in wild-type cells that is resistant to the single-strand specific nuclease RNase A , and observed significantly more RNase-resistant material in cells expressing MAL32 from the multi-copy system than the single-copy system ( Figure 5D ) . This experiment shows that a multi-copy locus produces more dsRNA than an equivalently expressed single-copy locus in wild-type cells without the RNAi system , explaining the increased siRNA formation in RNAi+ cells . 10 . 7554/eLife . 01581 . 018Figure 5 . Single gene analysis of copy number effect on RNA interference ( RNAi ) . ( A ) Schematic of single-copy and multi-copy MAL32 system . ( B ) Northern analysis of MAL32 RNA from single-copy and multi-copy systems . All visible species are included in antisense quantification . ( C ) MAL32 short interfering RNA ( siRNA ) abundance from cells in B . ( D ) RNase A sensitivity of MAL32 in wild-type cells expressing multi-copy and single-copy MAL32 . Cells were lysed on ice , treated with RNase A as indicated and analyzed by northern blot . 25S and L-A ( a double stranded RNA ) are shown as controls for loading and RNase specificity . n = 3 biological replicates , error bars ±1se , *p<0 . 05 , ***p<0 . 01 by Student’s t test , y axes in arbitrary units . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 018 The influence of copy number suggested that dsRNA formation and potentially siRNA production may occur in the nucleus . We initially tested this by immunofluorescence against Dicer and dsRNA in mixed populations of wild-type and RNAi+ cells ( Figure 6—figure supplement 1 ) . Dicer was present in cytoplasmic foci , but also showed a diffuse cytoplasmic staining ( compare the indicated wild-type and RNAi+ cells ) and , under wide-field imaging , appeared to be present in the nucleus . However , super-resolution images of the same cells showed nuclear exclusion of Dicer; therefore , if Dicer is present in the nucleus , it is at low levels . dsRNA staining in these cells revealed many cytoplasmic foci , presumably Killer virus dsRNAs that are known to be incompletely cleared by RNAi ( Drinnenberg et al . , 2011 ) , but did not show unambiguous nuclear staining . As an alternative , we asked whether the spatial configuration of gene copies within the nucleus could affect siRNA formation; such an effect would provide strong evidence for the formation of dsRNA in the nucleus . In systems which undergo efficient RNAi such as the rDNA and 2µ plasmids , all gene copies are clustered together in a small sub-nuclear volume . To test the importance of this clustering , we used the sense-antisense system at the TRP1 locus which produces detectable siRNA even when present in the genome in only two tandem copies ( data not shown ) . We generated strains with three tandem copies of TRP1 on a single plasmid ( Clustered , Cls ) or three unlinked copies ( Dispersed , Dsp ) ( Figure 6A ) , and expressed Dicer without Argonaute to allow siRNA formation but minimize the effect of RNAi on total RNA levels . Quantification of total RNA showed that both systems produced similar amounts of sense and antisense RNA molecules ( Figure 6B ) , although this experiment was complicated by read-through transcripts of antisense TRP1 from the clustered system ( Figure 6B lanes 1 , 2 ) , a behavior that was not prevented by insertion of transcriptional stop cassettes between the repeats . Nonetheless , the clustered system produced fourfold more TRP1 siRNA than the dispersed system ( Figure 6C ) , showing that close nuclear juxtaposition of transcriptional loci enhances dsRNA formation . 10 . 7554/eLife . 01581 . 019Figure 6 . Clustered loci show higher efficiency of short interfering RNA ( siRNA ) formation . ( A ) Comparison of the systems used . Three copies of TRP1 were placed either in tandem on a single-copy plasmid ( Clustered , Cls ) or a single copy was left in the genome at the TRP1 locus and two further copies were placed on different single-copy plasmids ( Dispersed , Dsp ) . Dicer was expressed from a single-copy plasmid . ( B ) Sense and antisense RNA expression in clustered and dispersed systems with and without Dicer . Quantification shows that Dicer alone has little effect on total RNA levels . Read-through species visible in lanes 1 , 2 are included in the quantification; values have been normalized for the different number of probe binding sites in the read-through RNAs . In the absence of this normalization ( i . e . , counting the number of binding sites rather than the number of molecules ) , the clustered antisense is approximately twofold more abundant than the dispersed antisense , which is insufficient to explain the difference in siRNA formation . ( C ) siRNA produced from TRP1 in clustered and dispersed systems . n = 3 biological replicates , error bars ±1se , ***p<0 . 01 by Student’s t test , y axes in arbitrary units . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 01910 . 7554/eLife . 01581 . 020Figure 6—figure supplement 1 . Analysis of Dicer and double stranded RNA ( dsRNA ) localization . Wild-type and RNA interference ( RNAi ) + cells carrying a green fluorescent protein ( GFP ) -tagged Dicer were grown separately in yeast extract peptone dextrose , then mixed in equal proportions before fixation . Cells were stained with anti-GFP for Dicer ( green ) and an antibody to dsRNA ( red ) . Images were acquired on a Nikon N-SIM microscope; the wide-field images and the processed super resolution images are shown for comparison . Three wild-type cells ( white arrows ) and three RNAi+ cells ( purple arrows ) are indicated . Examination of multiple cells in multiple stacks did not reveal any convincing evidence of Dicer or dsRNA foci in the nucleus , and the Dicer staining overlapping the nucleus in the wide-field images was not visible after processing , nor was it detectable by confocal imaging . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 020 While testing the effects of copy number amplification on siRNA production , we noticed that even low abundance sense-antisense ncRNA pairs ( selected from a published dataset , Xu et al . , 2009 ) underwent efficient RNAi when amplified to high copy number . For both the SUT176 and SUT430 systems ( Figure 7A ) , the sense and antisense RNAs are barely detectable by northern blot and are clearly not targeted for degradation by RNAi ( Figure 7B ) . However , after cloning on high-copy plasmids , the full-length RNAs became highly susceptible to RNAi and produced copious siRNAs ( Figure 7C , D ) . This raised the interesting prospect that low abundance pervasive transcription would be sufficient to trigger efficient RNAi responses from sequences that undergo copy number amplification . 10 . 7554/eLife . 01581 . 021Figure 7 . RNA interference ( RNAi ) against transcripts from amplified low-expression systems . ( A ) Schematic diagrams of SUT176 and SUT430 loci . ( B ) Northern analysis of SUT176 and SUT430 transcripts from single-copy genomic loci in wild-type and RNAi+ cells . Ty1 RNA is a positive control for RNAi , ACT1 is a loading control . ( C ) Analysis of SUT176 and SUT430 non-protein coding RNAs ( ncRNAs ) expressed from high-copy plasmids in wild-type and RNAi+ cells . Amplified regions are indicated in A . ( D ) Short interfering RNA analysis of cells in C . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 021 Clear examples of pervasive transcription are not well defined in any organism because , by definition , the products of pervasive transcription are almost undetectable . We therefore chose to examine the GAL gene cluster ( Figure 8A ) , which is tightly transcriptionally repressed in cells grown in glucose . Under these conditions , antisense ncRNAs are produced from the GAL10 ORF with a known steady-state abundance of one RNA molecule per 14 cells ( Houseley et al . , 2008; Pinskaya et al . , 2009 ) ( arrows in Figure 8B lane 1 ) . Transcription of these ncRNAs is abrogated in a previously described Reb1 binding site mutant ( RBSΔ ) , leaving almost no detectable RNAs from this locus ( Figure 8B lane 3 ) . For reasons that remain unclear , the GAL cluster is slightly de-repressed in the RNAi+ strain ( Figure 8B compare lanes 1 , 2 ) ; nonetheless , the RBSΔ RNAi+ strain ( Figure 8B lane 4 ) only produces very low level heterogeneous transcripts from GAL10 , suggesting that it forms a good model of pervasive transcription . Cloning either wild-type or RBSΔ GAL clusters onto high-copy plasmids substantially increased the levels of detectable ncRNA as expected ( Figure 8B lanes 5 , 7 ) , and these ncRNAs were processed into easily detectable siRNAs ( Figure 8C lanes 6 , 8 ) . Therefore , ncRNAs produced at the level of pervasive transcription are sufficient to mediate extensive siRNA production when the copy number of the transcribing locus is increased . 10 . 7554/eLife . 01581 . 022Figure 8 . RNA interference ( RNAi ) against pervasive transcripts from the repressed GAL cluster . ( A ) Schematic representation of the GAL cluster . ( B ) Non-strand-specific northern blot of non-protein coding RNAs ( ncRNAs ) produced from the GAL locus present at single-copy ( lanes 1–4 ) or high-copy ( lanes 5–8 ) , showing wild-type and RBSΔ mutant . Arrow indicates GAL10 antisense RNA . Strand-specific northern blots for the same RNA are shown in Figure 8—figure supplement 1 . ( C ) GAL10 short interfering RNA ( siRNA ) from the same cells as in B . ( D ) Expression of GAL10 mRNA from a single-copy genomic locus under the control of a Cu2+-responsive promoter in wild-type and RNAi+ strains carrying an empty vector ( lanes 1 , 2 ) , high-copy wild-type GAL cluster ( lanes 3 , 4 ) , or high-copy RBSΔ GAL cluster ( lanes 5 , 6 ) . n = 3 biological replicates , error bars ±1se , ***p<0 . 01 by Student’s t test , y axes in arbitrary units . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 02210 . 7554/eLife . 01581 . 023Figure 8—figure supplement 1 . Strand-specific analysis of GAL10 non-protein coding RNAs ( ncRNAs ) . RNA samples from Figure 8B were separated on two gels and probed with strand-specific probes sense and antisense to the GAL10 optical reading frame ( ORF ) . Known GAL10 antisense ncRNA species ( Houseley et al . , 2008; Pinskaya et al . , 2009 ) are visible in lanes 9 and 10 , and are more prevalent in the RNA interference ( RNAi ) + strain for unknown reasons . These species are highly expressed from the high-copy plasmid ( lanes 13 , 14 ) , along with an additional high molecular weight RNA . No sense RNAs were detected from the single-copy wild-type GAL10 locus ( lanes 1 , 2 ) , although heterogeneous RNA species of 0 . 8–1 . 5 kb are observed in the RNAi+ strain carrying the single-copy RBSΔ GAL10 plasmid ( lane 4 ) . These sense RNAs are more clearly visible from the high-copy RBSΔ GAL10 plasmid in wild-type cells ( lane 7 ) , and are faintly visible in the wild-type GAL10 plasmid ( lane 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 023 The siRNAs produced from the high-copy GAL10 locus are clearly sufficient to degrade the GAL10 ncRNAs in the RNAi+ background ( Figure 8B compare lanes 5 , 7 with lanes 6 , 8 ) ; however , a classical RNAi response should be able to degrade RNA expressed from a separate locus . To test this we introduced the high-copy GAL cluster plasmids into a strain in which the single-copy genomic GAL10 ORF is expressed at high levels from a Cu2+-dependent promoter , allowing expression of the GAL10 mRNA from the single-copy locus while the GAL clusters present on the high-copy plasmids remain fully repressed . As observed for the GAL10 ncRNAs , the GAL10 mRNA was expressed at higher levels in the RNAi+ strain than in the wild-type ( Figure 8D compare lanes 1 , 2 ) but , nonetheless , both wild-type and RBSΔ high-copy GAL cluster plasmids caused highly significant >50% knockdowns of the GAL10 mRNA compared with the empty vector control ( Figure 8D lanes 2 , 4 , 6 ) . This was not an indirect effect of the high-copy GAL clusters alone as , in the wild-type background , GAL10 mRNA levels were slightly increased by the presence of the GAL plasmids ( Figure 8D lanes 1 , 3 , 5 ) . These data demonstrate that pervasive transcription of a high-copy locus is sufficient to instigate an effective RNAi response that can mediate the degradation of a target mRNA in trans .
The ability of the RNAi system to selectively target the products of high-copy sequences such as transposons provides a remarkably efficient genome defense , as well as an effective way to differentiate heterochromatic regions , which are often repetitive , from gene-rich euchromatin . Here we have demonstrated that RNAi has an innate preference for the products of high-copy sequences , probably because RNA from high-copy sequences forms dsRNA more efficiently . It has long been known that cells can recognize and silence high-copy DNA , which would form a basic defense against uncontrolled amplification of transposable elements ( reviewed in Hsieh and Fire , 2000 ) . This of course requires a mechanism to count copy number , or at least differentiate high- and low-copy regions , which has remained mysterious . Our data show that RNAi provides such a mechanism by selectively targeting the products of high-copy loci . The production of siRNA from high-copy DNA , which would be the basis of such a counting mechanism , absolutely requires that all DNA is transcribed; if this does not occur , transposable elements that remain transcriptionally silent would be invisible to the system . Pervasive transcription , the general background of very low level RNA produced across the genome , ensures that the vast majority of the genome is transcribed , and therefore that no region remains completely silent ( Cheng et al . , 2005; Birney et al . , 2007; Kapranov et al . , 2007; Goodman et al . , 2012 ) . The extent to which mammalian genomes are pervasively transcribed has been controversial; however , many of the questions revolve around whether the pervasive transcripts represent defined functional products or whether much of the detected RNA represents random transcriptional noise ( van Bakel et al . , 2010; Clark et al . , 2011 ) . For a general surveillance function , it does not really matter whether pervasive transcription is formed of many discrete transcripts or occurs at random since either process should be sufficient to generate dsRNA . If a large proportion of pervasive transcription does represent random noise , this would be actively advantageous; random transcription would be sequence independent , and therefore transposable elements could not become fully silent by mutating individual promoter sequences . We suggest that the primary function of pervasive transcription lies in ensuring the whole genome is transcribed to allow identification and suppression of transposable elements; this does not conflict with the idea that some proportion of these transcripts may have additional functions . We propose that hybridization kinetics explains the dependence of RNAi on copy number ( shown in Figure 9 ) . The rate of formation of dsRNA from single stranded sense and antisense RNA is proportional to the concentration of each strand of RNA , and so is inversely proportional to the square of the reaction volume . Technically , the reaction volume is the non-excluded volume of the cell; however , this assumes a uniform distribution of RNA throughout the cell . In reality the RNA is far from evenly distributed , so some small volumes may have very high concentrations of RNA and , within these volumes , the rate of dsRNA formation will be dramatically higher than in the bulk of the cell . Single-copy loci cannot simultaneously transcribe sense and antisense RNA ( Hobson et al . , 2012 ) so , although such a locus can give rise to a mixed population of sense and antisense RNA in the cytoplasm over time , in the vicinity of the transcription site only one sense of RNA should ever be present , assuming efficient RNA export . Annealing of sense and antisense RNA must therefore occur in the relatively large non-excluded cytoplasmic volume of the cell , which will be inefficient except for very highly expressed RNA . In contrast , at multi-copy loci the simultaneous production of sense and antisense RNA from many closely spaced sites can lead to high concentrations of sense and antisense RNA around the transcription site , causing efficient duplex formation and hence efficient RNAi . 10 . 7554/eLife . 01581 . 024Figure 9 . Proposed mechanism for RNA interference ( RNAi ) on high-copy loci . The rate of double stranded RNA ( dsRNA ) formation , the required first step in RNAi , is highly dependent on local RNA concentration . Single-copy loci cannot simultaneously transcribe sense and antisense RNA , allowing RNA export to occur before the strands meet and requiring hybridization to occur in the relatively large cytoplasmic volume . In contrast , sense and antisense RNA can be simultaneously transcribed from different parts of a multi-copy locus and , therefore , if the copies are closely juxtaposed in the genome or in 3D space , the local concentration of sense and antisense RNA around the transcription sites should be high , favoring dsRNA formation . DOI: http://dx . doi . org/10 . 7554/eLife . 01581 . 024 This mechanism predicts that dsRNA formation should occur in the nucleus , but we were not able to detect Dicer in the nucleus by immunofluorescence . This reflects the situation in higher eukaryotes where Dicer is largely cytoplasmic , but recent experiments in Drosophila and mammalian cells have detected small quantities of nuclear Dicer , particularly associated with chromatin ( Sinkkonen et al . , 2010; Cernilogar et al . , 2011; Gullerova et al . , 2011; Doyle et al . , 2013 ) . Low but functional levels of Dicer may therefore be present in the nucleus of RNAi+ cells and able to generate siRNA . Alternatively , the dsRNA may be exported and processed in the cytoplasm . To our knowledge , there is no clear evidence for or against export of dsRNA by normal pathways; certainly these would not be too large or too structured to pass through nuclear pores compared , for example , with pre-ribosomes . One notable prediction of this mechanism is that clustering of multi-copy transcription sites would be a particularly efficient way to increase the local density of sense and antisense RNA . All the systems we have described in this paper are clustered: rDNA repeats are arranged in tandem , telomeres are known to cluster together at various points in the cell cycle ( Gotta et al . , 1996 ) , and high-copy 2µ plasmids exist in a discrete focus that is vital for copy number maintenance ( Velmurugan et al . , 2000; Wong et al . , 2002 ) . The comparison of siRNA formation from clustered and dispersed TRP1 loci provides experimental evidence for this effect since , for a given quantity of sense and antisense RNA , the clustered system produces more siRNA . Although the clustered system also produces read-through transcripts , these would not have a higher hybridization rate than the non-read-through RNAs as the hybridization rate depends on the frequency of collisions between molecules . Intriguingly , Tf2 retrotransposons in Schizosaccharomyces pombe are clustered by the action of centromere protein B ( CENP-B ) , which also silences these elements through histone deacetylation ( Cam et al . , 2008 ) . This clustering would allow cells to produce siRNA against the Tf2 elements through pervasive transcription , although multiple mechanisms silence Tf2 retrotransposons ( Yamanaka et al . , 2012 ) , providing an extra defense against retrotransposon activation . Similarly , gypsy retrotransposons in Drosophila are known to cluster ( Gerasimova et al . , 2000 ) , which may again facilitate siRNA production . Hence , the clustering of transposable elements by factors such as CENP-B would facilitate their recognition by RNAi and allow for selective RNA degradation . Mammalian germline cells are replete with small RNAs including endogenous siRNA ( Watanabe et al . , 2006 , 2008; Tam et al . , 2008; Song et al . , 2011 ) , and siRNAs in sperm and oocytes show a pronounced bias towards high-copy sequences that would be effectively explained by the selectivity of the RNAi system towards the products of high-copy loci ( Watanabe et al . , 2006; Song et al . , 2011 ) . However , it remains unclear how some dsRNA precursors of siRNAs are generated , particularly for retrotransposons that are primarily expressed only on the sense strand . We suggest that pervasive transcription would provide sufficient antisense RNA for this role , just as we observed for high-copy GAL cluster sequences in yeast . In comparison to the germline , the response of mammalian somatic cells to dsRNA is distinctly muted . dsRNA could be processed into siRNA , be altered by RNA editing ( Hogg et al . , 2011 ) , or could activate the interferon response leading to apoptosis ( Gantier and Williams , 2007 ) . However , transgenic mice expressing a hairpin dsRNA construct produce minimal siRNAs , little edited RNA , and show no phenotype that might indicate cell death ( Nejepinska et al . , 2012a ) . Nonetheless , siRNAs produced from LINE-1 retrotransposons have been detected in cell culture ( Yang and Kazazian , 2006 ) , and high-copy transfected plasmids expressing a sense-antisense RNA pair do produce detectable siRNA in HEK293 cells ( Nejepinska et al . , 2012b ) . This shows that a basic siRNA response with an apparent bias towards high-copy sequences is functional in mammalian somatic cells .
Yeast deletion strains ( Supplementary file 1 ) were created by standard methods using the oligonucleotides in Supplementary file 1 . Plasmids are described in Supplementary file 1 with construction details . Cells were grown on rich media ( 2% peptone , 1% yeast extract , 2% sugar ) or synthetic media ( 0 . 69% yeast nitrogen base with ammonium sulfate , amino acids , 2% sugar ) for plasmid assays . GAL10 mRNA was induced with 20 µM CuSO4 in Figure 8D . Media components were purchased from Formedium . Cells were grown to mid-log ( OD 0 . 4–0 . 6 ) at 30°C for most experiments or at 25°C for experiments involving trf4Δ mutants . The W303 background strain used here has defects in galactose induction in synthetic media , so strains in Figure 3G , H were diploids of W303xBY4741 that show a normal galactose response . RNA was extracted by three procedures described below . High molecular weight RNA was prepared using the hot phenol method for all experiments except Figures 7B , 3B , 3F and Figure 3—figure supplements 1 , 2 where guanidine thiocyanate ( GTC ) -phenol preparations were used . 5–10 µg glyoxylated RNA was resolved on 1 . 2% gels as described ( Sambrook and Russell , 2001 ) , transferred to Hybond N+ membrane ( GE ) and hybridized with probes listed in Supplementary file 1 using either Church Hyb ( Sambrook and Russell , 2001 ) or UltraHyb ( Life Technologies ) . RNA probes were hybridized at 65°C and washed at 65°C using 0 . 1× SSC , 0 . 1% SDS , DNA probes in Church Hyb were hybridized at 65°C and washed at 65°C with 0 . 5× SSC , 0 . 1% SDS , DNA probes in UltraHyb were hybridized at 42°C and washed at 55°C using 0 . 2× SSC , 0 . 1% SDS . Small RNA enriched fractions were isolated using the mirVANA kit ( Ambion ) . 4–10 µg small RNA was separated on 15% polyacrylamide/8 M urea gels containing 20 mM MOPS or 1× TBE , transferred in 20 mM MOPS or 0 . 5× TBE to Hybond N membrane ( GE ) and chemically cross-linked as described ( Pall and Hamilton , 2008 ) . We observed no difference in cross-linking efficiency between MOPS and TBE gels , but resolution of TBE gels was superior in our hands . siRNAs were detected using random primed probes ( Supplementary file 1 ) hybridized in UltraHyb Oligo ( Life Technologies ) at 42°C and washed with 2× SSC , 0 . 5% SDS at 42°C , U6 control oligonucleotide was labeled using T4 polynucleotide kinase and hybridized in Church Hyb under the same conditions . 10 × 107 cells in 15 ml tubes were re-suspended in 600 µl TES ( 10 mM Tris pH 7 . 5 , 10 mM EDTA , 0 . 5% SDS ) and 600 µl phenol pH 7 . The mixture was incubated at 65°C for 20 min with 30 s vortexing every 5 min , before briefly chilling on ice . Samples were centrifuged for 5 min and the upper phase extracted . This phase ( 5–600 µl ) was extracted twice with phenol:chloroform ( 5:1 ) and once with chloroform before precipitation with 50 µl 3 M sodium acetate ( NaOAc ) pH 5 . 2 and 1 . 1 ml ethanol . The pellet was washed with 70% ethanol and re-suspended in 30 µl water . 2 × 107 cells were lysed by 5 min vortexing at 4°C with 50 µl glass beads and 40 µl GTC-phenol ( 2 . 1 M GTC , 26 . 5 mM Na citrate pH7 , 5 . 3 mM EDTA , 76 mM β-mercaptoethanol , 1 . 06% N-lauryl sarcosine , 50% phenol pH7 ) . 600 µl GTC-phenol was added , mixed , and samples were heated at 65°C for 10 min , then placed on ice for 10 min . 160 µl 100 mM NaOAc pH 5 . 2 and 300 µl chloroform:isoamyl alcohol ( 24:1 ) were added , samples were vortexed and centrifuged at top speed for 5 min . The upper phase was extracted , precipitated with 1 ml ethanol , washed with 70% ethanol and the pellet re-suspended in 6 µl water . 3 µl RNA was analyzed per lane . Small RNAs were isolated using a mirVANA kit ( Life Technologies ) with minor modifications . 35 × 107 cells were thoroughly re-suspended in 100 µl lysis/binding buffer , 200 µl glass beads were added , and the samples were vortexed for 5 min at 4°C . 500 µl lysis/binding buffer were added and the samples were mixed before proceeding with the isolation as per the manufacturer’s instructions , starting with addition of the miRNA homogenate additive . After isolation the samples were generally re-precipitated and re-suspended in 20 µl water . 20 × 107 cells were harvested and split into two aliquots , then re-suspended in 600 µl 10 mM Tris pH 7 . 5 , 10 mM EDTA on ice . Cells were lysed with glass beads ( 10 cycles of 30 s vortex , 60 s on ice ) , and 5 µg RNase A was added to one aliquot followed by 30 min incubation on ice . After centrifugation for 10 min at 4500×g , 600 µl lysate was extracted , SDS added to 0 . 5% , and RNA extracted by the hot phenol method as above . RNase A treated samples were re-suspended in 12 µl water . Cells from 2 ml saturated culture were washed with 50 mM EDTA , then spheroplasted with 250 µl 0 . 34 U/ml lyticase ( Sigma L4025 ) in 1 . 2 M sorbitol , 50 mM EDTA , 10 mM DTT . After centrifuging at 1000×g , the cells were gently resuspended in 400 µl of 0 . 3% SDS , 50 mM EDTA , 100 µg/ml RNase A and incubated at 37°C for 30 min . 4 µl of 20 mg/ml proteinase K was added and the samples were mixed by inversion and heated to 65°C for 30 min . 160 µl 5 M potassium acetate was added after cooling to room temperature , the samples were mixed by inversion and then chilled on ice for 30 min . After 10 min centrifuging at 10 , 000×g , the supernatant was poured into a new tube containing 500 µl phenol:chloroform pH 8 and the samples were mixed on a wheel for 15 min . The samples were centrifuged for 10 min at 10 , 000×g and the upper phase was extracted using cut tips and precipitated with 400 µl isopropanol . Pellets were washed with 70% ethanol , air-dried and left overnight at 4°C to dissolve in 30 µl TE . After gentle mixing , 10 µl of each sample was digested with 40 U of EcoRV , ethanol precipitated and separated on a 25 cm 1% TBE gel at 90 V overnight . The gel was washed in 0 . 25 N HCl for 15 min , 0 . 5 N NaOH for 45 min , and twice in 1 . 5 M NaCl 0 . 5 M Tris pH 7 . 5 for 20 min before transfer to HyBond N+ membrane in 6× SSC . The membrane was probed for URA3 and GAL7 in Church Hyb at 65°C and washed with 0 . 5× SSC , 0 . 1% SDS at 65°C . Cells were grown to OD 0 . 5 in YPD and the cultures mixed as required , 4 ml per coverslip . The cells were fixed with 440 µl 37% formaldehyde ( Merck , microscopy grade ) for 15 min at room temperature , then centrifuged for 2 min at 4600 rpm . The cells were washed three times with 1 ml of buffer B ( 0 . 1 M sodium phosphate pH 7 . 5 , 1 . 2 M sorbitol ) , then re-suspended in 100 µl buffer B containing 3 µl 17 U/µl lyticase ( Sigma L2524 ) and 10 mM DTT for 15 min . The cells were centrifuged for 2 min at 1000×g , then washed with 1 ml buffer B . The cells were re-suspended in 40 µl buffer B , applied to a poly-L-lysine coated coverslip ( Zeiss 18 × 18 × 0 . 170 ± 0 . 005 mm ) and left for 20 min before washing twice with buffer B . Coverslips were treated with −20°C methanol for 6 min , then dipped in −20°C acetone for 10 s , followed by two washes with PBS . Coverslips were blocked for 30 min with 5% milk 0 . 3% Triton-X100 in PBS , washed with PBS , then incubated overnight at 4°C with primary antibodies in 50 µl 1% BSA 0 . 3% Triton-X100 in PBS . Coverslips were washed three times with PBS and incubated for 30 min at room temperature with secondary antibodies 1:1000 in same buffer as primaries . After washing three times with PBS , the samples were dehydrated with 70% , 90% , and 100% ethanol and mounted in Pro-long Gold with DAPI ( Life Technologies ) . Antibodies were rabbit anti-GFP ( Life Technologies A11122 ) at 1:500 and mouse anti-dsRNA J2 ( ESC 10010200 ) at 1:1000 . Images were acquired using a Nikon N-SIM microscope comprising a Nikon Ti-E microscope , Nikon 100× 1 . 49 NA lens , Nikon SIM illumination module , and Andor iXon 897 EM-CCD camera . SIM data were acquired in ‘3D-SIM’ mode using five phases and three rotations . DAPI , Alexa Fluor 488 , and Alexa Fluor 594 dyes were excited using 405 , 488 , and 561 nm laser light , respectively . Super-resolution images were reconstructed using Nikon Elements software . Equivalent wide-field images were reconstructed in FIJI ( ImageJ , NIH ) by summing the phase shifts from one grid rotation . Gels and phosphor screens were imaged using FLA 3000 ( Fuji ) or FLA 7000 ( GE ) systems . Quantification was performed using AIDA ( Fuji ) or ImageQuant ( GE ) , depending on the scanner used . Images were prepared for publication with ImageJ by smoothing and minimal contrast enhancement . Images from the FLA3000 had a Gamma Shift of 3 applied . 5′ RACE was performed with an ExactSTART Eukaryotic mRNA 5′- and 3′-RACE Kit ( Epicentre ) as per manufacturer’s instructions , except that reverse transcription was primed from random hexamers . 50 ml of cells at 0 . 7 × 107 cells/ml were harvested , washed , and frozen on nitrogen . Cells were thawed , washed twice in 1 ml lysis buffer ( 50 mM HEPES pH7 . 5 , 50 mM KCl , 5 mM MgCl2 , 1 mM DTT , 1× complete protease inhibitors ( Roche ) ) and transferred to 2 ml tubes , then re-suspended in 300 µl lysis buffer . 300 µl zirconium beads were added and cells were lysed by vortexing 10 × 30 s with 30 s on ice between cycles . The lysate was clarified by centrifuging twice for 5 min at 14 , 000 rpm , a 12 µl aliquot was removed for total RNA and the remaining lysate was split in half and 2 . 5 µl mouse anti-dsRNA J2 ( ESC 10010200 ) added to one aliquot . Antibody was bound for 2 hr at 4°C , then 20 µl GammaBind beads ( GE ) pre-blocked overnight with 1% BSA were added and incubation continued on a wheel for 2 hr at 4°C . The beads were washed 5× for 10 min with 1 ml wash buffer ( 10 mM Tris pH 7 . 5 , 120 mM NaCl , 5 mM MgCl2 , 0 . 1% NP-40 , 1 mM DTT ) . RNA was extracted from the beads and total lysate samples using Tri-reagent ( Sigma ) according to the manufacturer’s instructions . 1 µg total lysate and whole immunoprecipitation samples were treated with RQ1 DNase ( Promega ) , purified by phenol:chloroform and ethanol precipitation , then reverse transcribed from random hexamers using Superscript III ( Life Technologies ) . Quantitative PCR was performed using Maxima SYBR qPCR mix ( Fermentas ) . Multi-copy loci are problematic for standard high-throughput sequencing mapping pipelines and are commonly discarded . Reads mapping to a non-unique genome sequence are usually assigned at random to one copy in the genome , therefore the total reads are divided evenly amongst the copies and the apparent abundance of RNA matching each copy is effectively divided by the copy number . In order to assess total RNA abundance ( as in Figure 4A , B ) , we multiplied all total RNA read counts by the copy number to obtain the real total RNA abundance . However , we decided that the siRNA abundance should be analyzed per producing locus because each copy in the genome was analyzed separately for comparison with single-copy loci ( i . e . , we quantified how many siRNA reads an individual copy of a multi-copy locus produced , not how many the combined copies produced ) . We therefore did not multiply the siRNA read counts by the copy number . Such normalizations clearly have the potential to introduce systematic biases , and we therefore repeated the analysis in Figure 4 either with no copy number normalization or with both total and siRNA read counts multiplied by copy number ( Figure 4—figure supplement 2 ) . We found that although the distributions changed somewhat , the majority of total RNA abundance categories had higher siRNA levels for multi-copy loci irrespective of the copy number normalization applied . We note that the alternative copy number analysis in Figure 4C , D did not require any normalization; indeed , any copy number normalization would simply cancel out in the calculation of the siRNA:total RNA ratio , therefore any systematic bias that might be introduced by copy number normalization would have no effect . | Genes contain the codes that are needed to make the proteins used by cells . This code is transcribed to make a messenger RNA molecule that is then translated to make a protein . However , other types of RNA called non-coding RNA molecules can disrupt this process by binding to messenger RNA molecules , with matching sequences , before translation begins . This phenomenon , which is known as RNA interference , involves enzymes called Dicer and Argonaute . Many cells contain large numbers of non-coding RNA molecules—so called because they are not translated to produce proteins—and many of these are capable of starting the process of RNA interference . However , most do not , and the reasons for this are not understood . Now , work by Cruz and Houseley has provided new insight into this phenomenon by showing that it is related to the number of copies of the gene encoding such RNAs in the genome . Yeast cells normally do not have the genes for RNA interference , but Cruz and Houseley used genetically engineered yeast cells containing Dicer and Argonaute . Although most of the messenger RNA molecules in these cells showed no change , the expression of some genes with high ‘copy numbers’ was reduced . Further experiments that involved adding more and more copies of other genes showed that RNA interference could selectively target messenger RNA molecules produced from genes with an increased copy number—particularly if the copies of the genes were clustered in one location in the genome . RNA interference is also used to defend against DNA sequences that invade and multiply within a genome , such as viruses and other ‘genetic parasites’ . As such , the effect observed by Cruz and Houseley could explain why entire genomes are often continuously copied to RNA at low levels . This activity would allow the monitoring of the genome for the invasion of any genetic parasites that had multiplied to high numbers . Following on from this work , the next challenge will be to understand how gene copy number and location are balanced to achieve a selective RNA interference system . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"genetics",
"and",
"genomics"
] | 2014 | Endogenous RNA interference is driven by copy number |
Variants in IFIH1 , a gene coding the cytoplasmatic RNA sensor MDA5 , regulate the response to viral infections . We hypothesized that IFIH1 rs199076 variants would modulate host response and outcome after severe COVID-19 . Patients admitted to an intensive care unit ( ICU ) with confirmed COVID-19 were prospectively studied and rs1990760 variants determined . Peripheral blood gene expression , cell populations , and immune mediators were measured . Peripheral blood mononuclear cells from healthy volunteers were exposed to an MDA5 agonist and dexamethasone ex-vivo , and changes in gene expression assessed . ICU discharge and hospital death were modeled using rs1990760 variants and dexamethasone as factors in this cohort and in-silico clinical trials . About 227 patients were studied . Patients with the IFIH1 rs1990760 TT variant showed a lower expression of inflammation-related pathways , an anti-inflammatory cell profile , and lower concentrations of pro-inflammatory mediators . Cells with TT variant exposed to an MDA5 agonist showed an increase in IL6 expression after dexamethasone treatment . All patients with the TT variant not treated with steroids survived their ICU stay ( hazard ratio [HR]: 2 . 49 , 95% confidence interval [CI]: 1 . 29–4 . 79 ) . Patients with a TT variant treated with dexamethasone showed an increased hospital mortality ( HR: 2 . 19 , 95% CI: 1 . 01–4 . 87 ) and serum IL-6 . In-silico clinical trials supported these findings . COVID-19 patients with the IFIH1 rs1990760 TT variant show an attenuated inflammatory response and better outcomes . Dexamethasone may reverse this anti-inflammatory phenotype . Centro de Investigación Biomédica en Red ( CB17/06/00021 ) , Instituto de Salud Carlos III ( PI19/00184 and PI20/01360 ) , and Fundació La Marató de TV3 ( 413/C/2021 ) .
The spectrum of disease after infection by SARS-CoV-2 ( COVID-19 ) may range from mild respiratory symptoms to a severe form of lung injury fulfilling acute respiratory distress syndrome ( ARDS ) criteria ( Grasselli et al . , 2020 ) . In these severe cases , systemic response to infection may be associated with multiorgan failure and death ( Du et al . , 2020 ) . Therefore , outcomes in critically ill COVID-19 patients are related not only to viral clearance , but also to preservation of homeostasis . The mechanisms responsible for the development of severe forms of COVID-19 have not been fully elucidated , but non-adaptative inflammatory responses play a central role . The only strategies that have decreased mortality in this population , steroids ( WHO Rapid Evidence Appraisal for COVID-19 Therapies ( REACT ) Working Group et al . , 2020 ) and blockade of the IL-6 pathway ( REMAP-CAP Investigators et al . , 2021 ) , aim to limit this exacerbated immune response to prevent organ dysfunction . The human gene IFIH1 , located in the reverse strand of chromosome 2 , encodes MDA5 , a helicase that acts as a cytoplasmatic virus receptor . After binding to a viral RNA strand , MDA5 interacts with a mitochondrial adapter ( MAVS , mitochondrial antiviral signaling protein ) , triggering the transcription of type-1 interferon genes and ultimately the systemic inflammatory response . In humans , the rs1990760 polymorphism encodes a variant of the IFIH1 gene ( NM_022168: c . 2836G > A [p . Ala946Thr] ) that has been related to different susceptibility to viral infections and autoimmune disorders ( Gorman et al . , 2017 ) . By regulation of IFN-dependent pathways , IFIH1 participates in a feedback loop that ultimately modulates viral clearance and host inflammatory responses . In experimental models of Coxsackie virus infection , TT variants in rs1990760 result in lower pro-inflammatory cytokine levels without a major reduction in viral clearance ( Domsgen et al . , 2016 ) . Although the role of this variant in Coronavirus infections has not been explored , it has been proposed that the rs1990760 TT variant could confer resistance to SARS-CoV-2 infection and that differences in allelic frequencies could explain the epidemiological features of the pandemic in different countries ( Maiti , 2020 ) . We hypothesized that , once infection is established , the inflammatory response in severe COVID-19 patients could be conditioned by IFIH1 variants . To test this hypothesis , we prospectively followed a cohort of critically ill patients with confirmed infection by SARS-CoV-2 , in which peripheral blood gene expression , cell populations , concentrations of immune mediators , and clinical outcomes were studied and related to rs1990760 variants .
This single-center prospective , observational study was approved ( ref . 2020/188 ) by the Clinical Research Ethics Committee of Principado de Asturias ( Spain ) . Informed consent was obtained from each participant or next of kin . Given the exploratory nature of the study objective and the absence of previous data , no formal sample size calculations were performed . The study started after approval from the ethics committee and finished in December 2020 , after the second pandemic wave . All patients with confirmed SARS-CoV-2 infection ( Escudero et al . , 2020 ) and meeting the Kigali modification ( Riviello et al . , 2016 ) of ARDS criteria ( to include patients without mechanical ventilation ) from March 16 , 2020 to December 10 , 2020 were included in the study . Exclusion criteria were age <18 , any condition that could explain the respiratory failure other than COVID-19 , do not resuscitate orders or terminal status , or refusal to participate . Patients were followed until hospital discharge , and clinical and analytical data collected . The main outcome was intensive care unit ( ICU ) discharge alive and spontaneously breathing . Secondary outcome was hospital discharge . All biochemical analyses ( cytokine measurements , RNAseq , estimation of cell populations… , see below ) were performed by researchers blinded to genotype and outcome . The presence of SARS-CoV-2 was analyzed by detecting viral genome using a multiple quantitative retrotranscriptase ( RT ) -PCR . Nucleic acids were purified by MagNa Pure 96 System ( Roche , Geneva , Switzerland ) from the swabs transport medium . The extracts were subjected to an amplification reaction using TaqMan Fast 1-Step Master Mix ( Life Technologies , Carlsbad , CA ) supplemented with a mixture of primers ( Thermo Fisher Scientific , Walthan , MA ) and TaqMan MGB probes ( Applied Biosystems , Foster City , CA ) directed against ORF1ab and N genes ( Supplementary file 1a ) . Amplifications and subsequent analysis were carried out using the Applied Biosystems 7500 Real-Time PCR System ( Escudero et al . , 2020 ) . Amplification of viral genes with a Ct number lower than 35 was considered positive . Viral load was normalized by the number of cells and expressed as copies/1000 cells , as previously described ( Gómez-Novo et al . , 2018 ) . Clearance of SARS-CoV-2 was evaluated by fitting viral load in tracheobronchial samples after its peak value over time using an exponential decay function and calculation of viral clearance half-life ( λViral Clearance ) . DNA was extracted from total blood leukocytes ( 1 ml ) with an automated equipment ( Promega Maxwell ) . IFIH1 rs1990760 C/T polymorphism determined by real-time PCR with TaqMan genotyping master mix ( Life Technologies ) and TaqMan probes ( Thermo Fisher Scientific , assay C_2780299_30 ) in an ABI-7500 device . The genotyping strategy was validated by Sanger sequencing of selected individuals from the three genotypes . Based on previous data showing that patients with a TT genotype have a different immune response compared to CC and CT ( Wawrusiewicz-Kurylonek et al . , 2020; Zhang et al . , 2018 ) , patients with these two variants were grouped and compared against the TT group . Two blood samples were taken in the first 24 hr after ICU admission . No patient had received steroids at the time of sampling . About 3 ml of blood was collected in Tempus tubes ( Thermo Fisher Scientific ) for RNA isolation and immediately stored at –80°C . Additional 5 ml was collected in a Vacutainer serum tube ( BD Biosciences ) , and isolated by centrifugatio , and stored at –80°C until analysis . After thawing , blood from Tempus tubes was diluted in phosphate-buffered saline ( PBS ) ( 1:3 v/v ) and centrifuged at 3000 rpm for 30 min . Supernatant was discarded and the pellet resuspended in TRIzol ( Sigma-Aldrich , Poole , UK ) and precipitated overnight with isopropanol at –20°C . After centrifugation , RNA pellets were washed with 70% ethanol and resuspended in RNAse-free water . RNA quality was assessed using a TapeStation , and only samples with a RIN ( RNA integrity number ) above 8 were analyzed . RNA sequencing was performed using Ion AmpliSeq Transcriptome Human Gene Expression Kit , in an Ion S5 GeneStudio sequencer ( Ion Torrent ) . Briefly , 10 ng of total RNA was retrotranscripted and the obtained cDNA used for library synthesis using Ion AmpliSeq Transcriptome kits to amplify all the canonical human transcripts . After template preparation in an automated Ion Chef Instrument , semiconductor 540 chips were run in an Ion S5 GeneStudio sequencer . Torrent Suite software was used for base calling , alignment , and sequence quality controls . The generated FASTQ files ( available at Gene Expression Omnibus , accession numbers GSE168400 and GSE 177025 ) were mapped against a reference transcriptome ( obtained from http://refgenomes . databio . org ) and transcripts counted using Salmon v1 . 4 software ( Patro et al . , 2017 ) . Raw counts were compared between genotypes using the DESeq2 library ( Love et al . , 2014 ) . The log2-fold change between variants for each gene and the adjusted p-value ( corrected using a false discovery rate of 0 . 05 ) were calculated and analyzed using Ingenuity Pathway Analysis ( Qiagen , USA ) to identify overrepresented gene sets and networks . Over the identified network , in-silico effects of IFIH1 downregulation and addition of exogenous dexamethasone were performed . Circulating cell populations were estimated from gene expression using a previously validated deconvolution algorithm ( Vallania et al . , 2018 ) . Using a reference expression matrix , proportions of 20 different cell lines were calculated . Only cell lines present in more than five samples were considered . As a quality check , we analyzed the correlation between estimated and measured lymphocyte percentages . The obtained correlation coefficient was 0 . 61 ( Figure 3—figure supplement 1 ) . It must be noted , however , that the deconvolution method estimates proportions over transcriptionally active cells , which may not be equivalent to the obtained cell count ( as there may be inactive cells in the latter ) . A panel of inflammatory mediators was studied in serum from patients not receiving steroids during their first ICU day . Serum concentrations of interferons ( IFN ) -β , -γ and -λ , tumor necrosis factor ( TNF ) -α , interleukins ( IL ) -1β and -6 , and chemokines CXCL8 , CXCL9 , CXCL10 , CXCL16 , CCL2 , CCL3 , CCL4 , and CCL7 were measured using a multiplexed assay ( Luminex custom panel ) . Concentrations below the lower limit of detection for a given mediator were replaced with half of that limit . To study the potential interferences between IFIH1 rs1990760 variants and dexamethasone predicted by in-silico analyses , an ex-vivo experiment was designed . Blood samples from healthy volunteers ( genotyped for IFIH1 rs1990760 variants using DNA obtained from buccal swabs ) were collected in EDTA tubes and immediately processed . Peripheral blood mononuclear cells ( PBMCs ) were isolated via density-gradient centrifugation with Lymphoprep ( Axis-Shield PoC AS , Oslo , Norway ) . Cells were washed with red blood cell lysis buffer ( NH4Cl 0 . 1 M , KHCO3 0 . 01 M , EDTA 0 . 1 mM in dH2O , pH 7 . 33 ) and PBS before being resuspended in RPMI-1640 ( Gibco , USA ) + 10% fetal bovine serum ( FBS ) . Cells were seeded in 12-well plates at a final concentration of 1 . 5×106 PBMC/ml and cultured at 37°C and 5% CO2 in presence of medium/FBS , medium/FBS plus a MDA5 ligand ( 1 µg/ml high-molecular weight poly-I:C/LyoVec , Invivogen , USA ) , or medium plus the MDA5 ligand and 10 µM dexamethasone ( Kern Pharma , Spain ) . Although poly-I:C may bind to both RIG-I and MDA-5 , long ( high-molecular weight ) poly-I:C binds specifically to MDA-5 ( Kato et al . , 2008 ) . After 24 hr , cells were collected and homogenized in TRIzol for RNA extraction . 500 ng of total RNA was retrotranscribed into cDNA using an RT-PCR Kit ( High-capacity cDNA rt Kit , Applied Biosystems , USA ) . Expression of STAT1 , STAT3 , FOXO3 , IL6 , and GAPDH was quantified using 5 ng of cDNA per well and in triplicate for each sample . SYBR Green Power up ( Thermo Fisher Scientific ) and 10 µM of the corresponding primers ( Supplementary file 1b ) were used in all the experiments . The relative expression of each gene was calculated as 2−ΔCT ( gene of interest ) – ΔCT ( GAPDH ) . Data from the RECOVERY trial were extracted and used to estimate risk ratios ( RRs ) for each rs1990760 variant , with and without steroids ( see online Appendix for details ) . With these data , a survival model was developed and used to simulate scenarios with different allelic frequencies and baseline risks of death . Hazard ratios ( HRs ) for 28-day mortality were calculated from these simulations . Data are shown as median ( interquartile range ) . All data points and sample sizes represent individual patients/biological replicates . Comparisons between IFIH1 variants were done using Wilcoxon or ANOVA tests , and p-values corrected using a false discovery rate of 0 . 05 . Results from the ex-vivo model were fitted to a mixed-effects model including experimental group and genotype as covariables , and post hoc comparisons evaluated using Holm’s correction . No outliers were discarded . Survival was analyzed using competing risks , Cox regression model , with ICU/hospital discharge alive and spontaneously breathing and death as competing risks , using the Aalen-Johansen estimator , as previously described ( Barbaro et al . , 2020 ) . This competing-risks framework is needed as patients discharged alive have a low probability of death , so censoring at the time of discharge using a standard Kaplan-Meier approach would lead to biased observation , as the probability of death is different than in those still followed ( i . e . , informative censoring ) . Patients with an rs1990760 CC/CT variant not treated with steroids were considered the reference category in all the analyses . All the analyses and plots were performed using the R 4 . 0 . 1 statistical environment ( R Development Core Team , 2020 ) with the packages data . table ( Dowle and Srinivasan , 2021 ) , multcomp ( Hothorn et al . , 2008 ) , survival ( Therneau , 2020 ) , MetaIntegrator ( Haynes et al . , 2017 ) , and ggplot2 ( Wickham , 2016 , p . 2 ) .
About 250 patients were admitted in ICU due to suspected or confirmed COVID-19 . Among these , 227 were included in the study . The study flow chart and reasons for exclusion are shown in Figure 1 . Basic demographic and clinical data are shown in Table 1 . IFIH1 rs1990760 variants in this population met Hardy-Weinberg conditions ( 53 CC , 110 CT , 64 TT , chi-square p=0 . 19 ) . There were no differences in comorbidities or clinical data at admission between genotypes other than a lower PaO2/FiO2 ratio in patients with a TT variant ( Table 1 ) . Gene expression in peripheral blood during the first day of ICU admission was profiled in 42 patients who did not receive steroid therapy at that time ( 11 , 19 , and 12 with rs1990760 TT , CT , and CC genotypes , respectively ) . Expression of IFIH1 was significantly lower in patients with the TT genotype ( Figure 2A and Figure 2—figure supplement 1 ) . Comparison of peripheral blood gene expression between patients with TT and CT/CC variants yielded significant differences in 179 genes ( Figure 2B–C , Supplementary file 2 ) . Visual inspection of the heatmap reveals that differences are quantitative rather than qualitative . Ingenuity pathway analysis revealed several gene networks involved in the regulation of the inflammatory response among the differentially expressed genes ( Figure 2D and Figure 2—figure supplement 2 ) . In-silico predictions suggest that IFIH1 downregulation ( such as in patients with the rs1990760 TT variant ) decreases the expression of pro-inflammatory pathways ( Figure 2—figure supplement 3 ) . Then we assessed the immunological consequences of these differences in gene expression and cell populations . There were no statistically significant differences in inferred percentages of classic CD14+ monocytes , circulating plasma cells , M2 macrophages , and CD56dim NK cells , between genotypes . Patients with the TT genotype showed an increase in hematopoietic precursors and myeloid dendritic cells ( Figure 3 ) . Serum cytokines were measured at ICU admission in 28 patients ( 8 , 10 , and 10 with TT , CT , and CC genotypes , respectively ) . There were no differences in any of the measured interferons , which were below the limit of detection in a large proportion of patients . Patients with the TT genotype showed lower levels of pro-inflammatory mediators , including IL-6 , CXCL10 , CXCL16 , and CCL7 ( Figure 4 ) . There were no significant differences in any of the measured cytokines between patients with CC and CT variants ( Figure 4—figure supplement 1 ) . During the study period , dexamethasone was added to COVID-19 treatment based on results from published trials ( WHO Rapid Evidence Appraisal for COVID-19 Therapies ( REACT ) Working Group et al . , 2020 ) . An in-silico analysis focused on the interactions between IFIH1 and steroids suggested that dexamethasone could disrupt some of the effects caused by IFIH1 downregulation ( Figure 5—figure supplement 1 ) . Specifically , dexamethasone may change expression of several IFIH1-dependent genes , including STAT1 , STAT3 , or FOXO3 among others . To test these predictions , an ex-vivo experiment using PBMCs from healthy volunteers with different IFIH1 rs1990760 variants ( n=5 , 6 , and 7 for CC , CT , and TT variants , respectively ) was performed . Cells were collected and exposed to poly I:C , to mimic SARS-Cov-2 infection , and dexamethasone . Compared to exposure to poly I:C alone , dexamethasone had no effect on STAT1 ( Figure 5A ) or STAT3 ( Figure 5B ) expression . However , the steroid increased the expression of FOXO3 ( Figure 5C ) and IL6 ( Figure 5D ) only in cells with the TT variant . These results suggest that dexamethasone may alter the inflammatory response triggered by MDA5 activation in those patients with the TT variant of the gene . Median follow-up was 25 days ( interquartile range 16–39 days ) and 27 days in survivors ( interquartile range 17–44 days ) . There were no differences in ICU length of stay among groups ( Supplementary file 1c ) . ICU survival and hospital mortality were modeled using IFIH rs1990760 genotype and steroid treatment as interacting covariables . There were no significant differences in the main clinical characteristics among the resulting groups ( Supplementary file 1c ) . Regarding ICU survival ( Figure 6A and Supplementary file 1c ) , 27 out of 35 patients with CC/CT alleles who did not receive steroids were discharged alive and spontaneously breathing from the ICU ( HR: 1 , used as reference ) . In patients with this variant , dexamethasone was not related to a better outcome ( 98 out of 128 patients discharged , HR: 1 . 20 [0 . 78–1 . 38] , p=0 . 41 ) . All patients with the TT allele ( n=14 ) who did not receive steroids were discharged alive ( HR: 2 . 49 [1 . 29–4 . 79] , p=0 . 012 ) . Steroid treatment in patients with the TT variant was related to the loss of this benefit ( 32 out of 50 patients discharged alive , HR: 1 . 03 [0 . 62–1 . 72] , p=0 . 91 ) . Hospital mortality is shown in Figure 6B and Supplementary file 1c . There were nine hospital deaths in patients with the CT/CC genotypes not treated with steroids ( out of a total of 35 patients ) , and 27 out of 128 hospital deaths when treated with steroids , resulting in HR of 1 . 11 [0 . 52–2 . 37] ( p=0 . 80 ) . All patients with the TT allele who were not treated with steroids survived after their hospital stay ( HR 1 . 23×10–7 , confidence intervals , and p-value cannot be computed due to the absence of events ) . Patients with the TT allele and treated with steroids showed a worse outcome ( 19 deaths in 50 patients , HR: 2 . 19 [1 . 01–4 . 87] , p=0 . 05 ) . To investigate the causes responsible for the differences in mortality among groups , we first quantified viral load at diagnosis ( Figure 6C ) and clearance after its peak value ( Figure 6D ) . There were no differences in these parameters related to IFIH1 variants or treatment groups . We also compared serum IL-6 levels at ICU admission and 1 week later . Compared to patients with CC/CT variants not treated with steroids , patients with the same genotype but receiving dexamethasone and patients with the TT variant not receiving steroids showed lower levels of IL-6 at admission ( Figure 6E ) . After 1 week ( Figure 6F ) , serum IL-6 levels remained lower in patients receiving dexamethasone with the CC/CT variant and in those with the TT variant and not treated with this drug . However , IL-6 levels were higher in those with the TT variant and treated with steroids , in line with our ex-vivo findings . To explore the mechanisms behind these differences , peripheral blood gene expression at ICU day 4 was compared between patients treated with or without steroids for each rs1990760 variant . IFIH1 expression decreased in patients with rs1990760 CC/CT variants treated with steroids , but not in those with a TT variant ( Figure 6G ) . In opposite , steroids increased FOXO3 expression only in patients with a TT variant ( Figure 6H ) , resembling the results from the ex-vivo experiments . IL6 gene raw counts at day 4 were below 5 in all the patients , and thus not compared . When the whole transcriptomes were compared , steroids significantly changed the expression of 58 genes in patients with a CC/CT variant and 23 in patients with a TT variant ( Supplementary file 3 ) . Overall , changes in gene expression were qualitatively different between variants , with only three genes in common ( Figure 6I and Figure 6—figure supplement 1 ) . These findings may support the hypothesis that COVID-19 populations with a low proportion of the rs1990760 TT variant would show a better response to dexamethasone . To explore this prediction , data from the RECOVERY trial ( RECOVERY Collaborative Group et al . , 2021 ) were examined . The T-allele frequency in populations from a Black/Asian ancestry is 0 . 13 ( data available at https://grasp . nhlbi . nih . gov/Covid19GWASResults . aspx ) , so only 1 . 7% has a TT variant . In these populations , RR was 0 . 7 ( 0 . 51–0 . 95 ) , whereas in patients from a White ancestry ( with a T allele frequency of 0 . 61 ) , RR increased to 0 . 9 ( 0 . 8–1 . 2 ) . Using these data , RRs related to each rs1990760 variant , with and without steroids , were extracted and a survival model developed ( see online supplement for details ) . A simulation including 500 patients from each rs1990760 variant and treatment arm ( placebo or dexamethasone ) showed a significant reduction in mortality with steroids in patients with a CC/CT variant , but not in those with a TT variant ( Figure 7A ) . In patients not-receiving steroids , HRs related to a TT variant , compared to CC/CT variants , were independent of the allelic frequency ( Figure 7B ) . Simulations of steroid therapy with different allelic frequencies showed that HRs in each specific variant remained constant whereas the effect of steroids in the overall population was dependent on the allele distribution ( Figure 7C and Figure 7—figure supplement 1 ) .
Our results show that critically ill patients with the TT variant in the IFIH1 rs1990760 polymorphism have an attenuated inflammatory response to severe SARS-CoV-2 infection , leading to a decreased mortality . In this selected population , treatment with steroids has no immunomodulatory effects and could be related to worse outcomes . These results confirm the impact of the host response on patients’ outcomes and suggest that patient geno/phenotypes should be taken into account to prescribe steroids in this setting . Host response is a major determinant of outcome in critically ill patients , including those with COVID-19 . Several genetic variants involved in the inflammatory response have been related to SARS-CoV-2 infection and its severity ( Bovijn et al . , 2020; Gómez , 2021; Zhang et al . , 2020 ) . Activation of immune responses causes local and systemic inflammation , aimed to block viral replication . However , exacerbation of these responses can cause organ damage even after virus clearance . Our results recapitulate previous findings in COVID-19 describing the release of pro-inflammatory mediators , NK cell exhaustion , monocyte dysregulation , and emergency hematopoiesis ( Schulte-Schrepping et al . , 2020; Wen et al . , 2020; Wilk et al . , 2020 ) . Some of our findings in patients with the rs1990760 TT variant , including lower levels of circulating pro-inflammatory molecules and a shift toward anti-inflammatory cell populations ( hematopoietic precursors , M2 macrophages , or CD56dim NK cells ) have been linked to better outcomes in other observational studies ( Maucourant et al . , 2020 ) . Recognition of viral RNA by cytosolic receptors leading to the ultimate induction of expression of types I and II IFNs as part of the activation of innate antiviral signaling cascades , may not only relay on MDA5 receptor but also on other retinoid acid-inducible gene ( RIG ) -I-like receptors , or on the endosomal Toll-like receptor 3 ( TLR3 ) ( Takeuchi and Akira , 2010 ) . Interestingly , no differences were found in serum interferons levels . This may be explained by several factors . Most patients included in our study had serum IFNs levels below the inferior detection threshold . This result is similar to the one reported by Galani et al . , 2021 , where IFN-lambda and type I IFN production were both diminished and delayed in moderate-to-severe COVID-19 patients followed up during hospitalization . In our study , IFNs levels were only measured on day 1 of ICU admission , precluding any conclusions regarding time-dependent changes on these mediators’ levels during , for instance , subacute or long-term infection . More importantly , transcriptomic analysis did not reveal significantly different expression levels for IFIH1 gene after correction for multiple comparisons , although single-gene comparison showed a downregulated expression in patients with TT genotype . Although expression levels may be similar , full activation of an antiviral response triggering specific signaling intermediates and transcription factors , may vary between genotypes due to differences in MDA5 molecular architecture and function . The effect of rs1990760 on MDA5 function is related to modifications on the molecular architecture ( Berke and Modis , 2012; Wu et al . , 2013 ) and has been reported by several previous studies ( Looney et al . , 2015; Nejentsev et al . , 2009 ) . Collectively , our findings raise the hypothesis that the TT variant could be related to an attenuated pro-inflammatory response . In line with this observation , a TT variant has been associated with lower pro-inflammatory cytokine levels in patients with lupus ( Robinson et al . , 2011; Zhang et al . , 2018 , p . 1 ) and in experimental viral infections ( Domsgen et al . , 2016 ) . Development of tolerance to viral diseases can be considered a major evolutionary adaptative response ( Schneider and Ayres , 2008 ) , and inhibition of pro-inflammatory responses has been proposed to improve the outcome of severe COVID-19 patients . Notably , pangolins , an intermediate host of coronaviruses ( Xiao et al . , 2020 ) , lack functional IFIH1 . It has been suggested that this deficiency reduces the inflammatory response to coronavirus infections ( Fischer et al . , 2020 ) . IL-6 blockade ( REMAP-CAP Investigators et al . , 2021 ) or steroids ( WHO Rapid Evidence Appraisal for COVID-19 Therapies ( REACT ) Working Group et al . , 2020 ) are the only treatments that have improved the outcome of critically ill COVID-19 patients . However , mortality rates are still around 27–32% ( REMAP-CAP Investigators et al . , 2021; WHO Rapid Evidence Appraisal for COVID-19 Therapies ( REACT ) Working Group et al . , 2020 ) , so new therapeutic approaches are warranted . According to our results , the impact of the treatment with steroids along a given population may depend on intrinsic individual characteristics , allowing a personalized approach based on a specific genomic biomarker . In our study sample , patients with the IFIH1 rs1990760 TT variant constitute a population with a better prognosis , in whom treatment with dexamethasone may be reconsidered as it was associated to higher mortality rates . Indeed , as shown by the in-silico trials , any population enriched for patients with CC/CT variants in rs1990760 will show higher mortality rates and a better response to steroids , due to the low proportion of TT variants . Dexamethasone in patients receiving mechanical ventilation at inclusion in the RECOVERY trial had an HR of 0 . 64 , compared to an HR of 0 . 83 in the whole population . Of note , the proportion of Black/Asian patients ( with a low frequency of a T allele ) in the mechanically ventilated group was 29% , but only 18% in the whole study population . To date , no other genomic markers of personalized therapies in critically ill patients have been identified . The absence of differences in viral clearance and the late increase in IL-6 in these patients suggest that this worse outcome is more related to the disruption of ongoing immunoregulatory mechanisms than to antiviral responses . Our ex-vivo experiments do not fully elucidate the molecular mechanisms behind the interaction between TT variants and steroids , as the effects of FOXO3 upregulation and IL6 expression may be variable ( Joseph et al . , 2016 ) , but clearly illustrate that this specific combination may disrupt the ongoing self-regulation of inflammation . Moreover , steroid-induced changes in gene expression are qualitatively different in patients of each variant . Our findings point toward the therapeutic potential of MDA5 modulation in COVID-19 , either induced by steroids or targeted by other drugs . Amelioration of MDA5-dependent RNA sensing could avoid an exacerbated inflammatory response without impairing viral clearance . In this sense , it has been described that the interferon response triggered by MDA5 is unable to control viral replication ( Rebendenne et al . , 2021 ) . However , genetic ablation of IFIH1 results in increased viral loads and decreased cytokine production ( Yin et al . , 2021 ) , so this approach must be viewed with caution . It is unclear if these findings can be translated to other viral diseases . Steroids may decrease mortality in an unselected ARDS population ( Villar et al . , 2020 ) , but the role of underlying genotypes has not been addressed . An experimental model of Coxsackie virus infection revealed lower IFIH1 and CXCL10 expression in cells with the TT variant ( Domsgen et al . , 2016 ) , with no relevant differences in viral clearance . However , total absence of MDA5 results in impaired clearance of West Nile virus ( Errett et al . , 2013 ) . Our work has several limitations . First , the results must be validated in an independent cohort . Our in-silico simulations reinforce the external validity of our findings , but a pharmacogenomic analysis of patients included in clinical trials is warranted for confirmation . Second , steroid treatment was not randomized , so we cannot discard other underlying factors responsible for the observed differences . Although there could be confounding by indication , there were no baseline differences among groups suggesting a higher severity in steroid-treated patients . Moreover , there are significant differences between genotypes irrespective of the treatment received . In addition , in-silico and ex-vivo experiments are congruent with the observed clinical results , supporting the differential effects of steroid therapy in IFIH1 rs1990760 variants . Third , the favorable outcome of patients with a TT allele without steroids is based on a small sample size . As steroids are now the standard of care for severe COVID-19 , this sample size cannot be increased outside a hypothetical clinical trial focused of personalized steroid therapy according to IFIH1 rs1990760 variants . However , similar reduced numbers have served to identify other variants in the immune response ( Bastard , 2020 ) , and the related finding of increased mortality in this genotype after steroid therapy compared to all other variants is supported by a larger sample . In addition , our sample is representative of European population with a limited racial diversity that may have influenced our results , as the in-silico analyses further suggest . Finally , other studies Pairo-Castineira et al . , 2020; Ellinghaus et al . , 2020 have focused on the genetic variants linked to an increased risk of severe COVID-19 , compared to non-infected populations . However , no genetic markers have been associated to mortality in cohorts of infected patients . According to our data , rs1990760 is linked to the outcome , but no inferences on susceptibility to severe COVID-19 can be extracted . In summary , we have identified a genetic variant of IFIH1 that results in an ameliorated inflammatory response after severe SARS-CoV-2 infection . Patients with the rs1990760 TT genotype show a good outcome . However , this adaptative response was not observed in patients with a TT variant and treated with steroids . These findings suggest that the systemic response to severe COVID-19 is regulated by genetic factors that modulate the response to the infection and the prescribed therapy and , ultimately , may impact the outcome . | Patients with severe COVID-19 often need mechanical ventilation to help them breathe and other types of intensive care . The outcome for many of these patients depends on how their immune system reacts to the infection . If the inflammatory response triggered by the immune system is too strong , this can cause further harm to the patient . One gene that plays an important role in inflammation is IFIH1 which encodes a protein that helps the body to recognize viruses . There are multiple versions of this gene which each produce a slightly different protein . It is possible that this variation impacts how the immune system responds to the virus that causes COVID-19 . To investigate , Amado-Rodríguez , Salgado del Riego et al . analyzed the IFIH1 gene in 227 patients admitted to an intensive care unit in Spain for severe COVID-19 between March and December 2020 . They found that patients with a specific version of the gene called TT experienced less inflammation and were more likely to survive the infection . Physicians typically treat patients with moderate to severe COVID-19 with corticosteroid drugs that reduce the inflammatory response . However , Amado-Rodríguez , Salgado del Riego et al . found that patients with the TT version of the IFIH1 gene were at greater risk of dying if they received corticosteroids . The team then applied the distribution of IFIH1 variants among different ethnic ancestries to data from a previous clinical trial , and simulated the effects of corticosteroid treatment . This ‘mock’ clinical trial supported their findings from the patient-derived data , which were also validated by laboratory experiments on immune cells from individuals with the TT gene . The work by Amado-Rodríguez , Salgado del Riego et al . suggests that while corticosteroids benefit some patients , they may cause harm to others . However , a real-world clinical trial is needed to determine whether patients with the TT version of the IFIH1 gene would do better without steroids . | [
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] | [
"medicine",
"immunology",
"and",
"inflammation"
] | 2022 | Effects of IFIH1 rs1990760 variants on systemic inflammation and outcome in critically ill COVID-19 patients in an observational translational study |
We study the morphogenesis and evolutionary origin of the spectacular erectile ruff of the frilled dragon ( Chlamydosaurus kingii ) . Our comparative developmental analyses of multiple species suggest that the ancestor of Episquamata reptiles developed a neck fold from the hyoid branchial arch by preventing it to fully fuse with posterior arches . We also show that the Chlamydosaurus embryonic neck fold dramatically enlarges and its anterior surface wrinkles , establishing three convex ridges on each lobe of the frill . We suggest that this robust folding pattern is not due to localised increased growth at the positions of the ridges , but emerges from an elastic instability during homogeneous growth of the frill skin frustrated by its attachment to adjacent tissues . Our physical analog experiments and 3D computational simulations , using realistic embryonic tissue growth , thickness and stiffness values , recapitulate the transition from two to three ridges observed during embryonic development of the dragon’s frill .
Lizards can exhibit moveable skin folds at various locations of their body , such as the wings of the flying dragon ( Draco volans ) , the oral display frill of the ‘secret toadhead agama’ ( Phrynocephalus mystaceus ) , and the dewlap of many anole lizard species ( Anolis spp . ) . Here , we investigate the evolutionary developmental origin of the distinctive large erectile ruff ( Figure 1A , B ) of the emblematic Australian/New-Guinean frilled dragon ( Chlamydosaurus kingii ) . This animal spreads its spectacular neck frill for predator deterrence , territorial display and courtship ( Shine , 1990 ) . Figure 1C illustrates that the ventral sides of the ruff are supported by the two ceratobranchial I bones ( CBI ) of the hyoid apparatus ( Beddard , 1905 ) and the dorsal sides are held by the so-called ‘Grey’s cartilages’ ( De Vis , 1883 ) . Erection of the frill is caused by the coordinated movements of the CBI bones and Grey’s cartilages and requires the opening of the mouth . Although their primary function in tetrapods is associated with deglutition , the bones of the hyoid apparatus are also involved in a variety of specialised morphologies and functions such as improved lung ventilation through gular pumping in monitor lizards ( Bels et al . , 1995; Owerkowicz et al . , 1999 ) , extension of the throat in bearded dragons ( Throckmorton et al . , 1985 ) and of the dewlap in Anolis lizards ( Bels , 1990; Font and Rome , 1990 ) , tongue projection in chameleons ( Herrel et al . , 2001 ) , as well as tongue extension and shock absorption in woodpeckers ( Yoon and Park , 2011 ) . In reptiles , the central part of the hyoid apparatus , anteriorly prolonged by an entoglossal process ( EP; Figure 1C–D ) , is associated to three pairs of horns ( Bellairs and Kamal , 1981 ) . The first pair is composed of the hypohyal ( HH ) , ceratohyal ( CH ) and epihyal ( EH ) . The second pair is made of the CBI and epibranchial ( EB ) , while the third pair of horns consists of the ceratobranchial II ( CBII ) . During embryogenesis , the hyoid apparatus develops from the pre-cartilage ( mesenchyme condensation of neural crest origin ) of three branchial arches ( BA ) : the hyoid arch ( i . e . , the second BA = BA2 ) contributes to the development of the central and anterior parts of the hyoid body as well as the first pair of horns , whereas the third and fourth BAs ( BA3 and BA4 ) generate the second and third pairs of horns , respectively ( Bellairs and Kamal , 1981; Creuzet et al . , 2005; Kaufman and Bard , 1999; Köntges and Lumsden , 1996 ) . Contrary to that of the hyoid skeletal elements , the morphogenesis of the frill soft tissues and of the ‘Grey’s cartilage’ are unknown . Here , using computed-tomography and histology approaches , we first show that the highly-developed CBI bones of the frilled dragon are localised into the third ( most dorsal ) skin ridge of the frill and that the ‘Grey’s cartilage’ is not made of cartilage per se , but is a dense connective tissue mainly composed of collagen fibres . Second , our comparative developmental analyses indicate that the existence of a spectacular frill in Chlamydosaurus was made possible by the incomplete fusion of the BA2 with the cardiac eminence and posterior BAs , an evolutionary event that probably occurred at the origin of Episquamata reptiles . This event allowed most members of that lineage to exhibit a conspicuous neck fold ( although it was lost in chameleons , snakes and legless lizards ) that develops from the hyoid BA ( BA2 ) . Hence , the Chlamydosaurus frill is a dramatic outgrowth of the hyoid arch ectoderm . Finally , using 3D reconstruction , analyses of proliferation and computational simulations , we show that the very robust folding pattern of the Chlamydosaurus frill ( all individuals develop three ridges on each of the two lobes of the frill ) is not due to localised increased growth at the position of the ridges , but likely emerges from an elastic instability during the homogeneous growth of the anterior sheet frustrated by the underlying tissues and by its attachment to the neck . This physical ( mechanical ) process also explains the transition from two to three ridges observed during embryonic development of the dragon frill .
The frill of Chamydosaurus is a large and sagitally-symmetric piece of skin attached to the neck and head . The left and right lobes of the frill are connected ventrally by a central crease ( white line if Figure 1A ) . When erected , following the movements of the mandible , hyoid bones and Grey’s cartilages , the frill forms a flat disc positioned in a transversal plane ( Figure 1A ) . At rest , each lobe of the frill always pleats into three convex ridges and two concave folds , in addition to the central concave crease ( Figure 1B ) . We confirmed by computed-tomography ( CT ) that the lower part of the frill is supported by hypertrophied CBI bones ( Beddard , 1905 ) . Unlike in other agamid lizards , such as Pogona spp ( Throckmorton et al . , 1985 ) , where the CBIs are fully enclosed in the throat and do not extend further than the end of the lower jaw , the frilled dragon’s straight and much elongated CBIs have most of their length positioned into the frill ( Figure 1C ) . Skeleton staining and paraffin sections indicate that the frilled dragon CBI bones are ossified whereas other parts of the hyoid apparatus remain cartilageneous ( Figures 1D and 2A–F ) . Anteriorly , the CBIs are attached to the posterior part of the hyoid body within the throat ( Figure 2A ) and are surrounded by muscles . More distally , the CBIs separate from the throat to become incorporated into the third ridge of each lobe of the frill ( Figure 2E–F ) . In its free part , the frill consists of two sheets of skin linked by loose connective tissue ( Figure 2G , H ) , with the anterior sheet being longer than the posterior sheet at the positions of the convex ridges ( Figure 2F , H ) . The Grey’s cartilages ( black arrowhead in Figure 2C–E ) connect the dorsal part of the frill to each side of the head at the vicinity of the tympanic membrane . The proximal end of the Grey’s cartilage is strongly attached to muscles ( Figure 2D ) , the digastric , the attolens chlamydis and the adductor chlamydis , that allow for the movement of the upper part of the frill ( De Vis , 1883 ) . We used staining techniques to investigate further the nature of the Grey’s cartilage , described previously as ‘fibro-cartilagenous' ( De Vis , 1883 ) . As Alcian blue strongly stains the structure in the absence ( Figure 2C–E , I ) , but not in the presence ( Figure 2J ) , of MgCl2 , it is likely to contain weakly-sulphated glycosaminoglycans ( GAGs ) rather than keratan sulphates ( Bancroft , 2002; Scott and Dorling , 1965 ) ; the latter are characteristic components of true cartilage . Elastic staining does not reveal elastic fibres ( Figure 2K ) , while ‘Sirius Red’ staining indicates the presence of collagen fibres ( Figure 2L ) . As , under polarised light , thick and thin collagen fibres appear orange-red and green-yellow , respectively ( Rich and Whittaker , 2005 ) , we could infer that fibres are thick at the surface of the ‘cartilage’ ( near the epidermis , Figure 2M , left panel ) whereas they are heterogeneous in size deeper in the structure ( Figure 2M , right panel ) . We investigated the dragon's frill morphogenesis across pre-hatching development . Around embryonic day 23 ( E23 = 23 days post-oviposition ) , the early frill is visible as a swollen skin outgrowth in the ventral portion of the neck above the heart cavity and shoulders ( Figure 3A ) . Between E24 and E30 , the frill splits into right and left lobes , the central ventral crease becomes visible , and the latero-dorsal side of the outgrowth reaches the tympanic membrane ( Figure 3B ) . At the end of that period ( Figure 3C ) , two ridges become clearly visible on the anterior surface of each lobe while the latero-dorsal part grows towards the back of the head , beyond the tympanic membrane . Between E30 and E40 , a third ridge forms ( Figure 3D ) . At E45 , the basic morphology of the frill is established: the anterior side of each lobe has grown substantially and exhibits three expanded anterior ridges while the lateral side of the frill now extends well beyond the tympanic membrane ( Figure 3E ) . Our dissections of embryos at earlier stages indicate that the skin outgrowth generating the early development of the frill in Chlamydosaurus is present at E15 but is hidden behind the heart cavity . We investigate below this early outgrowth in the context of the branchial arches ( BA ) development . Around E6-E7 , the four first BAs ( BA1 to BA4 ) are visible ( Figure 4A ) on Chlamydosaurus embryos while the BA6 is hidden behind the developing heart and is only discernible on parasagittal sections ( Figure 4B ) . In amniotes , the posterior part of the BA2 ( hyoid arch ) has been called an ‘embryonic opercular flap’ because it grows , expands caudally and covers the posterior BAs . Eventually , the BA2 fuses to the cardiac eminence , causing the internalisation of the BAs 3 to 6 ( Richardson et al . , 2012 ) . In the frilled dragon , the posterior BAs internalise around E11 ( Figure 4C , D ) and are not discernible anymore at E15-E16 ( Figure 4E , F ) . However , here we observe that part of the frilled dragon's BA2 ( arrowhead in Figure 4F and G ) does not fuse to the cardiac eminence , forming the early frill behind the heart ( Figure 4G ) before intensively growing ( Figure 4H , I ) . Hence , we show here that , in Chlamydosaurus , the frill originates from the outgrowth of the BA2 that failed to completely fuse with posterior BAs and the neck . We then studied the internalisation of BAs in other species of squamates . We identify the presence of an incompletely fused BA2 for an extended period of embryonic development in all species investigated ( Figure 5 ) : the leopard gecko ( Eublepharis macularius ) , the ocellated lizard ( Timon lepidus ) , the veiled chameleon ( Chamaeleo calyptratus ) , and the bearded dragon ( Pogona vitticeps ) . The maintenance of the BA2 eminence throughout pre- and post-hatching development in the ocellated lizard and bearded dragon is responsible for the formation of a clearly visible neck fold ( Figure 5F , H and Figure 5—figure supplement 1A–F ) . In the leopard gecko and veiled chameleon , the incompletely fused BA2 will remain visible up to at least developmental stages 34 ( following developmental staging system of Wise et al . , 2009 ) but will eventually fuse with the neck at later stages ( Figure 5—figure supplement 1G–L ) , explaining that these species do not exhibit a conspicuous neck fold ( Figure 5E , G ) . Given that ( i ) the opercular flap ( BA2 ) fuses with the cardiac eminence in non-squamate amniotes such as birds and mammals ( at stages E6 and E10 for the chicken and the mouse , respectively ) , ( ii ) all Squamata lineages we investigated exhibit an unfused opercular flap for some period of their development , and ( iii ) many species of Episquamata exhibit a neck skin fold ( cf . green lineages in Figure 5 ) , the most parsimonious evolutionary scenario ( Figure 5 ) is that the opercular flap forms in all amniotes but it’s complete fusion was abolished at the origin of the Episquamata clade , that is after the divergence of the more ancestral lineages of Gekkota and Scinciformata . This event allowed the development of a neck skin fold in the ancestor of Episquamata , a morphological feature that was secondary lost ( cf . red lineages in Figure 5 ) in chameleons , snakes and various legless lizards . The neck fold was then developed and modified into a spectacular erectile neck ruff during the evolution of the frilled dragon . The folding pattern of the dragon's frill is robust: the left and right lobes of the frill each pleats into three convex ridges and two concave folds ( Figure 1A , B ) . In the two concave folds , the anterior and posterior skin sheets have similar lengths ( Figure 2F , G ) whereas the anterior skin sheet is substantially longer than the posterior skin sheet in the three convex ridges ( Figure 2F , H ) . This morphology indicates that the convex ridges impose the frill to pleat at these positions , inevitably causing the concave folds to also occur . To investigate the origin of the pre-folded pattern , we first tested whether they are generated by local increased proliferation . Indeed , if the anterior surface of the developing frill exhibited six lines of increased growth ( three on the left lobe and three on the right lobe ) superposed to the future location of the ridges , it would explain that the frill folds exactly there because the anterior skin sheet would become locally larger than the posterior skin sheet ( Figure 2F , H ) . Such a pattern of localised growth could be controlled by a corresponding pattern of morphogen gradients generated by a Turing-like ( reaction-diffusion ) mechanism or by unknown positional information . Hence , we used the mitotic marker phospho-Histone H3 ( pH3 ) to quantify proliferation across the frill while and after ridges are formed . Our analyses indicated ( i ) similar cell densities ( of about 0 . 02 cell per μm2 ) and proliferation during and after ridges formation , and ( ii ) no notable difference in proliferation at the location versus in between ridges ( Figure 6 ) . Although these proliferation analyses are limited by the low number of embryos at our disposal , the data generated does not hint at any obvious proliferation spatial patterning . This leaves us with the possibility that the ridges of the dragon's frill are generated mechanically by frustrated homogeneous growth . Indeed , recent analyses have demonstrated the importance of mechanical instabilities in morphogenesis ( Nelson , 2016 ) . For example , uniform growth of the gut at a rate larger than that of the anchoring dorsal mesenteric sheet is sufficient to quantitatively explain the gut looping morphogenesis into the body cavity of vertebrates ( Savin et al . , 2011 ) . Similarly , it has been suggested that the folding of the developing cerebral cortex in mammals is caused by expansion of the grey matter constrained by the more-slowly developing white matter ( Karzbrun et al . , 2018; Richman et al . , 1975; Tallinen et al . , 2016 ) . Using 3D measurements on frilled dragon embryos at various developmental stages , we observe that the linear dimension of the frill surface ( square root of the area ) increases approximately 1 . 3 fold relative to the length of the frill boundary attached to the neck ( Figure 7 ) . This observation is compatible with a physical ( mechanical ) morphogenesis process where homogeneous growth of the frill’s anterior skin sheet , frustrated by a boundary condition ( its attachment to the neck ) , generates an elastic instability that resolves into the formation of three anterior convex ridges on each lobe of the dragon's frill . In other words , contrary to the folding of the brain , which is geometrically constrained all over its surface , the frill might be folding during development because of a ‘curtain-instability process’ , similar to the wrinkling of gravity-induced draping ( Cerda et al . , 2004; Vandeparre et al . , 2011 ) where self-similar wrinkled patterns are generated under boundary confinement . We then construct a more complex model to assess if the folding pattern is additionally influenced by the attachment of the frill’s anterior skin sheet to the underlying loose connective tissue . Below , we measure and implement realistic tissue physical parameters and growth rates into numerical simulations and physical analog experiments to test whether such models can quantitatively predict the robust folding pattern of the Chlamydosaurus dragon's frill . Our 3D numerical growth model , based on a custom finite element method ( see Supplementary Methods ) , assumes that the two lobes of the frill grow independently and are symmetrical with respect to the central crease . We first use a thin semi-cylindrical geometry of thickness T and diameter L ( Figure 8A ) as a simplified model of the embryonic frill anterior surface . We model the frustrated growth of the sheet ( i . e . , the smaller growth rate of the edge attached to the neck relative to the growth rate of the frill ) as follows: the length L of the sheet’s straight edge is maintained constant while tangentially growing the rest of the structure . This simple model indicates that the number of ridges increases with the amount of expansion g ( Figure 8B and Figure 8—figure supplement 1 ) , while it decreases with increasing relative thickness ( T/L ) of the sheet ( Figure 8C and Figure 8—figure supplement 1 ) . To generate a more realistic geometry of the anterior surface of the frill , we apply a curved neck boundary and the presence of a central crease ( Figure 8D ) . Using this geometrical configuration and a quasi-static approach , our simulations recapitulate the transition from two to three ridges observed during embryonic development: for T/L = 0 . 014 ( i . e . , the ratio between the average values of skin thickness and the neck boundary length measured on real embryos; Figure 7 ) , each lobe of the frill exhibits two ridges when g reaches about 1 . 07 and three ridges when it exceeds 1 . 15 ( Figure 8E and Video 1 ) . To further test the single sheet model , we performed a physical analog experiment using a thin semi-cylindrical sheet of polydimethylsiloxane ( PDMS ) gel with a T/L value of 0 . 01 ( Figure 9A ) and constructed an equivalent computer model ( Figure 9B ) . We fixed the PDMS sheet at its straight edge and used hexane to make it swell by about 30% . The resulting geometry ( Figure 9C ) is very similar to that obtained with our numerical simulations ( Figure 9D ) : the swollen PDMS sheet exhibits three wrinkles with a wavelength that increases towards the free edge . Next , to investigate the influence of the substrate ( loose connective tissue ) on the resulting morphology , we generated a more complex model ( Figure 10A ) derived from the actual frill geometry obtained from a high-resolution episcopic microscopy ( HREM ) 3D reconstruction at E23 , that is a stage where the frill appears as a swollen skin outgrowth with no ridge ( Figure 3A ) . As the thickness of the anterior skin sheet exhibits nearly constant values during the period of ridges formation ( Figure 7C ) , we used its average value of 47 µm . By measuring 3D morphological features of the dragon’s frill between E23 ( i . e . , when the anterior skin is smooth ) and E32 ( i . e . , when the third ridge is clearly formed ) , we estimate: ( i ) the linear surface growth , that is the expansion perpendicular to the surface normal of the frill to be g ( s ) ≈1 . 3; ( ii ) the central crease growth ( in the direction C in Figure 10A ) to be g ( c ) ≈2 . 2 , and ( iii ) the outer free edge growth of the frill to be g ( o ) ≈1 . 0 . All these values are relative to the length of the neck boundary ( Figure 7D ) and are used to grow our numerical model ( Figure 10 ) . Note that we assume the simulated materials to be nearly incompressible ( Poisson’s ratio v = 0 . 45 ) , as suggested for biological soft tissues such as skin ( Choi and Zheng , 2005; Hendriks et al . , 2003; Khatyr et al . , 2004 ) . Our simulations show that the frill's surface indeed wrinkles and that the number of convex ridges decreases with the skin-to-substrate stiffness ratio ( Figure 10B ) , as expected from the formula λ=2πT ( μK/3μs ) 1/3 ( Kim et al . , 2011; Wang et al . , 2016 ) where λ is the wrinkling wavelength of a compressed stiff film ( here , the skin of thickness T and shear modulus µk ) on a flat soft substrate ( here , the loose connective tissue of shear modulus µs ) . As our simulations are performed on a curved geometry with fixed inner edge ( attachment to the neck ) , they generate wrinkles that are less regular than those of flat film-substrate systems described by the above equation . For µk/µsvalues ≤ 50 , the wavelength predicted by the equation is larger but similar ( within two standard deviations ) to the mean wavelength generated in the corresponding simulations . The discrepancy increases substantially for larger µk/µs because the fixed inner edge in the simulated model increasingly prevents the wrinkles to ‘repel’ each others . Note also that actual growth and elastic deformations are superposed in our measurements on embryos . The elastic contribution is small: at steady state , we measure a linear surface relative growth of 1 . 27 for an actual growth of 1 . 3 ( using a stiffness ratio of 100 ) . Crucially , as our model generates three ridges for skin-to-substrate stiffness ratios in the range 65–1000 for g ( s ) =1 . 3 , we need to evaluate if the actual value for the developing frill ( i . e . , the ratio between the stiffnesses of the frill's surface skin sheet and of the underlying loose connective tissue ) is within that range . Elastic moduli reported throughout the literature indicate several orders of magnitude variation in stiffness among human tissues ( Cox and Erler , 2011 ) . Hence , we performed depth-dependent nanoindentation measurements ( Figure 11 and Supplementary Methods ) on fresh unfixed Chlamydosaurus embryos and inferred that the frill skin elastic modulus increases from 11’440 ± 200 Pa at E26 to 36’880 ± 20 Pa at E45 ( i . e . , well in the 8–540 kPa range reported for human skin [Iivarinen et al . , 2014] ) , whereas the loose connective tissue modulus remains constant: 220 ± 7 Pa at E26 and 228 ± 14 Pa at E45 . These numbers yield an increase of skin-to-substrate stiffness ratio from 52 to 162 during the development of the ridges . Note that a stiffness ratio of 52 at E26 , that is smaller than the lower bound of the range 65–1000 required to generate three ridges , is not problematic because , at that stage , the frill grew only by g ( s ) =1 . 1 instead of g ( s ) =1 . 3 . Implementing this smaller growth value ( and a corresponding observed g ( c ) =1 . 5; Figure 10C ) in our simulation model generates two ridges , as observed in real embryos . Finally , instead of applying the whole growth at once during numerical simulations , and checking the steady-state result , we also performed a quasi-static approach to progressively grow the HREM-derived realistic 3D model of the frill; skin-to-substrate stiffness ratio was set to 100 for convenience . These new simulations recapitulate the transition from two to three ridges observed during actual frill morphogenesis ( Figure 10C , Video 2 ) . In principle , these semi-quantitative validations of our model could be reinforced by the comparisons of the amplitudes of the folds produced by the model with those observed on the developing embryos . The amplitudes of the middle ridge on the HREM reconstructions ( 104 μm at the end of the second ridge formation and 198 μm when the third ridge is formed ) is larger than the corresponding amplitudes obtained with the numerical model ( 46 μm and 97 μm , respectively ) . This discrepancy remains even after normalisation: the corresponding relative amplitudes are 3 . 3% and 5 . 4% of the inner boundary ( i . e . , the attachment to the neck ) in the HREM reconstructions versus 1 . 7% and 3 . 6% in the simulations . The significance of these differences is difficult to evaluate because the simulations are performed using elasticity parameters evaluated on fresh tissues , whereas 3D measurements were done on HREM reconstructions after fixation and dehydration of the samples . As tissue shrinkage due to dehydration can dissimilarly affect different tissues , the relative amplitudes measured on the fixed samples might not accurately reflect their real values . In addition , we use a strictly elastic model whereas the actual embryonic tissues might experience flow . A full quantitative analysis of folding in the frilled lizard would require live-imaging ( e . g . , with light-sheet microscopy ) , hence , the development of two techniques currently not available in reptiles: ex-ovo incubation ( for imaging ) and transgenesis ( for fluorescent labelling of living tissues ) . Similarly , a better access and availability of Chlamydosaurus embryos would allow to perform controlled local tissue cutting experiments on multiple locations of the frill to measure the distribution of stress prior and during the development of the folds .
The emblematic erectile ruff of the frilled dragon is a large and sagitally-symmetric piece of skin attached to the neck and the head . At rest , the frill pleats into three convex ridges and two concave folds while the animal can spread this structure by the coordinated movements of hyoid-derived hypertrophied CBI bone ( incorporated in the most dorsal ridge ) and the so-called ‘Grey’s cartilage’ that we identify not to be bona fide cartilage . Here , we identify an ancient evolutionary developmental event that paved the way to the much more recent evolution of the spectacular Chlamydosaurus frill . Indeed , by comparing the embryonic development of representatives of the Squamata lineage , we suggest that the ancestor of Episquamata ( Figure 5 ) lost the ability to completely fuse the hyoid branchial arch ( BA2 ) with the cardiac eminence and posterior BAs , allowing for the transformation of this 'embryonic opercular flap' into a conspicuous neck fold . The latter was subsequently lost in chameleons , snakes as well as various legless lizards , while it hypertrophied in Chlamydosaurus . Second , by producing and analysing embryonic series of frill dragons , we show that wrinkles form in the developing frill’s anterior skin , establishing a pattern of three convex ridges that , later in development , allow the structure to robustly fold when rested along the animal's neck . Third , using histological data , analysis of proliferation , physical analogs and computational models , we suggest that the convex ridges are generated by an elastic instability rather than by local increased proliferation patterned by signalling morphogen gradients or positional information . Indeed , we show that homogeneous growth of the embryonic frill's anterior surface is sufficient to robustly produce on each lobe of the frill , first two then three convex ridges when the frill's growth is frustrated by its attachment to the neck . Finally , numerical simulations , implementing ( i ) a more realistic morphology ( inferred from HREM 3D reconstructions ) of the embryonic frill , incorporating the shape of it's skin and connective tissue substrate , ( ii ) measured mechanical parameters of Chlamydosaurus embryonic tissues , and ( iii ) a realistic growth model derived from embryonic series , indicate that the development of two ridges , and the later transition to three ridges , can be explained by a mechanical process that does not require any pre-patterning .
Frilled dragons breed in the wild during the wet season from November to December ( Harlow and Shine , 1999; Shine and Lambeck , 1989 ) . In captivity , we obtained a mean number of 12 . 8 eggs per year per female . The incubation of frilled dragon’s eggs is about 90 days at 29 . 5°C , that is substantially longer than in its close relative bearded dragon ( Pogona vitticeps , 60 days ) . Maintenance of , and experiments on animals were approved by the Geneva Canton ethical regulation authority ( authorisations GE/82/14 , GE/73/16 , and GE/27/19 ) and performed according to Swiss law . These guidelines meet international standards . Computed-tomography scans were performed with a Skyscan-1076 microCT at a resolution of 35 μm ( source: 55 kV , 179 μA ) . Three-dimensional iso-surfaces were created using the Imaris software ( Bitplane , Zurich , Switzerland ) . Skinned heads were dehydrated and stained in 0 . 03% alcian blue in 80% EtOH and 20% acetic acid . The samples were rehydrated and stained in 0 . 01% alizarin red in 1% KOH . The hyoid apparatus was dissected and pictures taken with a Nikon D700 camera . Tissue samples were fixed overnight in 4% PFA , rinsed in 1x phosphate-buffered saline ( PBS ) . Post-embryonic samples were decalcified in Osteosoft ( Merck , 1017281000 ) . Samples were dehydrated before paraffin embedding and sectioned at 7 μm . For hematoxylin/eosin/alcian blue staining , sections were initially treated in a solution of 1% Alcian blue in 3% acetic acid before classical hematoxylin/eosin staining . For ‘critical electrolyte concentration’ applications ( Scott and Dorling , 1965 ) , the initial treatment was replaced by 0 . 05% alcian blue and 0 . 6M MgCl2 in 0 . 2M acetate buffer . Sirius Red staining were performed in 0 . 1% Direct Red 80 ( Sigma , 365548 ) in 1 . 2% picric acid and rinsed in 0 . 5% acetic acid . Elastic staining was performed following the manufacturer’s instructions ( elastic stain kit , Sigma HT25A-1KT ) . Images were acquired with a Pannoramic MIDI Slide scanner ( 3D HISTECH , Budapest , Hungary ) . Polarised images were taken with a Leica DM5500 microscope . Embryos were fixed overnight in 4% PFA , washed in 1x PBS followed by dissection of the head . Tissues were permeabilised in 1x PBS , 0 . 2% Triton X-100 and incubated in proteinase K ( 5 μg/ml ) . The heads were then rinsed in glycine and blocked in a solution containing 0 . 5% goat serum , 0 . 2% BSA , and 0 . 2% Triton X-100 . Anti-Histone H3 ( phospho S10 , ab14955 ) antibodies were diluted by a factor 1500 in blocking solution and samples were incubated overnight at room temperature . Incubation with secondary antibodies ( anti-MS-Alexa Fluor 555; 1/500 dilution ) was performed during 3 hr at room temperature . Finally , nuclei were stained during 1 hr using Nuclear Green ( ab138905 ) at dilution 1/1500 . Z-stacks were acquired using a LSM700 confocal microscope . Nuclei and pH3+ cells were segmented using the Imaris ‘spot’ tool . Areas containing cells were measured using ImageJ ( Schindelin et al . , 2012 ) and cell densities were computed . Cell division rates were calculated as the ratio of pH3+ cells over the total number of nuclei . Embryos were fixed in 4% PFA at +4°C overnight , washed in 1X PBS , dehydrated through a methanol series ( 30% , 50% , 70% , 100%; methanol diluted with 1X PBS ) and stored at −20°C . High-resolution episcopic microscopy ( HREM; Indigo Scientific , Herts , UK ) was performed using the JB4/dye embedding mix ( including eosinB and acridine orange ) following the manufacturer protocol . Embryos were placed in the mix for overnight polymerisation at room temperature followed by baking for 24 hr at 95°C and sectioning at 3 . 5 μm . The pixels defining the skin versus the underlying connective tissue were separated on HREM 2D images by applying the 2-means clustering algorithm in Python and aligned using the Amira software ( ThermoFisher Scientific , Oregon , USA ) . Then , 3D measurements were performed using Meshlab ( Vcli- , 2011 ) and a C ++ based semi-automatic tool ( Milinkovitch et al . , 2013 ) . Fresh embryos were dissected in 1x PBS and fixed and submerged on a home-made support . Nanoindentations were performed using a pre-calibrated FemtoTools FT-S100 Microforce Sensing Probe . Depth-dependant Young’s moduli were calculated using unloading curves ( Oliver and Pharr , 2004 ) and assuming that the surface of the frill consists of a sheet of skin adherent to the underlying loose connective tissue . Multiple models have been suggested for proper extraction of intrinsic material properties of two layers systems ( Menčík et al . , 1997 ) . We use a simple closed-form equation ( Jung et al . , 2004 ) for predicting the elastic moduli of the skin ( Eskin ) and of the substrate ( Esubstrate ) based on the relation between the penetration depth d and the estimated Young’s modulus E . The basic assumption of this approach is that the elastic and plastic responses of the system change progressively from skin-dominated to substrate-dominated as the values of d increases ( Figure 11A ) . More explicitly , we assume that the Young’s modulus E can be written as a simple power law function of the following form:E= Esubstrate ( Eskin/Esubstrate ) LwhereL=1/ ( 1+A ( d/T ) B ) and positive coefficients A and B are calculated using a nonlinear curve-fitting procedure in MATLAB for a range of skin thickness T = 27–72 µm ( Figure 7C ) . Given that E→Eskin when d/T→0 and E→Esubstrate when d/T→∞ , we estimate Eskin = 11'440 ± 200 Pa and Esubstrate = 220 ± 7 Pa at E26 while Eskin = 36'880 ± 20 Pa and Esubstrate = 228 ± 14 Pa at E45 ( Figure 11B ) . This result indicates that the skin modulus is increasing by a factor >3 between E26 and E45 while the substrate’s stiffness remains approximately constant . To test these results , we fix the substrate Young’s modulus to the values estimated above with the empirical model of Jung et al . ( 2004 ) , and use the analytical solution derived by Gao et al . ( 1992 ) to find the best fit ( in terms of mean squared error ) of both the skin thickness and skin Young’s modulus . These analyses yield values of T = 50 µm and Eskin = 17’800 Pa at E26 , and of T = 45 µm and Eskin = 38’550 Pa at E45 . Hence , the analytical model confirms that the skin thickness remains approximately constant , and that the skin modulus increases substantially ( by a factor of about 2 . 2 ) , between E26 and E45 . The somewhat different values of skin stiffness at E26 between empirical and analytical estimates might be due to the low number of data points at small indentation depths . Note also that there is no consensus in the literature on the best approach for estimating the Young’s modulus of biological soft tissues . Indeed , parameters , such as the presence of adhesion forces between the nano-indentation probe and the biological tissue ( Kontomaris , 2018 ) , but also capillary forces at the air-water interface when measuring submerged samples ( Boots et al . , 2019 ) , can bias the experimental estimates . More fundamentally , the potential anisotropy of biological tissues in terms of their viscoelastic and plastic properties can differentially affect specific methods , for example tensile measurements in multiple kinds of soft biological tissues consistently yield larger Young’s modulus estimates than those estimated with indentation methods ( McKee et al . , 2011 ) . Given that all our measurements were performed with the same method in the same conditions , we think that our estimates of the skin-to-substrate stiffness ratio ( i . e . , the most important parameter for our simulations ) are valid . The PDMS elastomer and curing agent ( Sylgard 184 Silicone Elastomer Kit ) were mixed at a 10:1 vol ratio with the addition of a small amount of green paint ( 0 . 03 ml per 100 ml of PDMS ) . The mixture was placed under vacuum ( to avoid bubbles ) in a Petri dish for about 30 min before being cured at 50°C overnight . After cooling , the PDMS sheet was removed from the Petri dish and placed between two wooden planks that were then tightly screwed to each other . The PDMS was swelled by placing the whole system into 98 . 5% Hexane ( Sigma-Aldrich ) for 15 min at room temperature . To study ridges formation under homogeneous growth of the frill constrained at the neck-frill boundary , we constructed a tetrahedral GPU-based numerical model in three dimensions . We assume that the two frill lobes ( attached by the central crease ) grow independently of one another and that the tissue material is neo-Hookean with volumetric strain energy densityW= μ2 [Tr ( FFT ) J−2/3−3]+ K2 ( J−1 ) 2where μ and K=α . μ are the shear and bulk moduli , respectively , F is the deformation gradient , J = det ( F ) , and α can be defined , using the Poisson’s ratio v , by α= ( 2+2ν ) / ( 3−6ν ) . We use a custom finite element method , similar to the one described by Tallinen et al . ( 2016 ) to minimise the elastic energy of the system . Every tetrahedron is defined by a matrix A=[x1x2x3] , where x1 , x2 , and x3 are vectors with the origin in the same vertex of the tetrahedron . Following Jin ( 2014 ) and Rodriguez et al . ( 1994 ) , we define a growth tensor G of a tetrahedron as G (  ) , where  describes the tetrahedron in a stress-free initial configuration . The growth tensor is defined by G=GsGc , where the tensor Gs=gsI+ ( 1−gs ) NNT represents the expansion g ( s ) perpendicular to the surface normal N in the initial configuration , and the tensor Gc=I+ ( gc−1 ) CCT defines the growth of the central crease g ( c ) in the direction C ( Figure 6D and Figure 8A ) . Values gs and gc are calculated using the generalised logistic functiongi=1+ g ( i ) −1 1+0 . 25e−100 ( T−xi ) where T is the thickness of the anterior skin sheet of the frill , and xi is the distance to the surface of the frill for xs ( pink in Figure 8B ) or to the central crease for xc ( purple in Figure 8B ) . This implies that growth can vary from one element to another which might result in configurations that are not physically attainable ( when neighbouring elements do not fit together after growth ) . To make the volume elements compatible , we introduce the elastic deformation tensor F (  ) that defines how the elements change their shapes to insure continuity inside the tissue . The product of the growth and deformation tensors results in the deformation gradient tensor A = FG that maps the stress-free state  before the growth to the stressed state A = FG after growth . To minimise the elastic energy of the system , at each time step , we obtain the Cauchy stress tensorσ=1J∂W∂FFTand the surface traction of each deformed face si=−σni ( i = 1 , 2 , 3 , 4 ) , where ni are normals with lengths proportional to the areas of the deformed faces . Nodal forces are obtained by distributing the traction of each face equally to its three vertices . Finally , the energy of the system is minimised by using damped second-order dynamicsν ( t+Δt ) = ν ( t ) + f ( t ) − γν ( t ) mΔtx ( t+ Δt ) =x ( t ) + ν ( t+Δt ) Δtwhere the time step Δt=0 . 01a/K , and m=a3 represents a node mass , where a is the average distance between neighbouring nodes in the initial configuration . The parameter γ is the viscous damping factor set to 10m in the beginning of the simulation and reduced progressively to speed up convergence . The vectors f , v and x are the force , velocity and position of the node , respectively . The simulations are assumed to be at steady state when the elastic energy of the system stabilises ( i . e . , its value doesn’t change by >0 . 1% during 105 iterations ) and the maximum node displacement is <10−6 . We also implemented a quasi-static approach using a discrete-time formulation . More explicitly , the elastic energy of the system is minimised at each time step ti∈{1 , 2 , … , Tmax} , that is for successive values of growth g ( ti ) corresponding to the measured growth values between time steps ti-1 and ti . This procedure is performed simultaneously for the surface and the central crease . To ensure that the frill is a symmetrical structure , the nodes of the central crease can move only along the vector C in one-layer simulations ( Figure 8D ) , while they are imposed a zero-displacement boundary condition in the normal direction on the central-crease plane in multi-layer simulations ( Figure 10A , central panel ) . | In Jurassic Park , while the computer programmer Dennis Nedry attempts to smuggle dinosaur embryos off the island , he gets attacked and killed by a mid-sized dinosaur that erects a frightening neck frill . This fictional dinosaur is clearly inspired from a real animal known as the ‘frilled dragon’ , that lives today in northern Australia and southern New Guinea . These lizards , also known as Chlamydosaurus kingii , have a large disc of skin that sits around their head and neck . This frill is usually folded back against the body , but can spread in a spectacular fashion to scare off predators and competitors . Folding of the left and right side of the frill occurs at three pre-formed ridges . But , it remains unclear which ancestral structure evolved to become the dragon’s frill , and how the ridges in the frill form during development . Now , Montandon , Fofonjka , and Milinkovitch show that the dragon’s frill , as well as the bone and cartilage that support it , develop from a part of the embryo known as the branchial arches . These are a series of bands of tissue in the embryo that evolved to become the gill supports in fish , and that now give rise to multiple structures in the ear and neck of land vertebrates . In most species , the second branchial arch will eventually fuse with the arches behind it . But in the frilled dragon , this arch instead continues to expand , leading to the formation of the dragon’s spectacular frill . As the frill develops , the front side of the skin forms three successive folds , which make up the pre-formed ridges . Studying the formation of these ridges revealed that they do not emerge from increased growth at the folding sites , but from physical forces – whereby the growth of the frill is constrained by its attachment to the neck . This causes the top layer to buckle , creating the folds of the frill . Montandon , Fofonjka , and Milinkovitch then simulated this mechanism of growth in a computer model and found it could recapitulate how folds develop in the frill of real lizard embryos . These results provide further evidence that physical processes , as well as genetic programs , can shape tissues and organs during an embryo’s development . Furthermore , changes in how the branchial arches develop between lizard species highlights how evolution is able to ‘recycle’ old structures into new shapes with different roles . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"physics",
"of",
"living",
"systems"
] | 2019 | Elastic instability during branchial ectoderm development causes folding of the Chlamydosaurus erectile frill |
Organelles are distributed to daughter cells , via inheritance pathways . However , it is unclear whether there are mechanisms beyond inheritance , which ensure that organelles are present in all cells . Here we present the unexpected finding that the yeast vacuole plays a positive essential role in initiation of the cell-cycle . When inheritance fails , a new vacuole is generated . We show that this occurs prior to the next cell-cycle , and gain insight into this alternative pathway . Moreover , we find that a combination of a defect in inheritance with an acute block in the vacuole biogenesis results in the loss of a functional vacuole and a specific arrest of cells in early G1 phase . Furthermore , this role for the vacuole in cell-cycle progression requires an intact TORC1-SCH9 pathway that can only signal from a mature vacuole . These mechanisms may serve as a checkpoint for the presence of the vacuole/lysosome .
Organelles are essential for cellular functions , and organelle inheritance is likely a major pathway that ensures the presence of organelles in all cells . A requirement for the inheritance of the mammalian Golgi is well established due in part to the elaborate changes in architecture that occur during interphase vs mitosis ( Shorter and Warren , 2002 ) . For other organelles , which are constitutively dispersed , organized mechanisms for their inheritance remain less clear . The budding yeast Saccharomyces cerevisiae provides an excellent model to study the spatial and temporal control of organelle inheritance , in part because its cell division is asymmetric . This asymmetric division requires active organelle transport in each cell-cycle . In budding yeast , most of the organelles are transmitted from mother to daughter cells ( Fagarasanu and Rachubinski , 2007 ) . These include the vacuole/lysosome , mitochondria , the endoplasmic reticulum , peroxisomes , secretory vesicles and late-Golgi elements . Transport of these organelles starts in G1 phase and occurs in coordination with the cell-cycle . However , it is unclear whether there are mechanisms that guarantee the presence of organelles prior to the next round of cell division . Here we present the unexpected finding that the presence of the vacuole is ensured because the vacuole plays an essential role in the initiation of the cell-cycle . During cell division in budding yeast , the daughter cell inherits a vacuole from the mother cell ( Weisman et al . , 1987 ) . The vacuole is transported by a vacuole transport complex , composed of the myosin V motor Myo2 , the vacuole membrane anchored protein Vac8 , and an adaptor protein Vac17 that links Myo2 and Vac8 ( Catlett and Weisman , 1998; Wang et al . , 1998; Ishikawa et al . , 2003; Tang et al . , 2003 ) . Vacuole inheritance is initiated in G1 phase via Cdk1/Cdc28 , which regulates the formation of the vacuole transport complex ( Peng and Weisman , 2008 ) . After formation of the complex , Myo2 moves the vacuole to the daughter cell along actin cables ( Hill et al . , 1996 ) . At the end of the cell-cycle , vacuole transport is terminated by ubiquitylation of Vac17 , which is then degraded by the 26S proteasome ( Yau et al . , 2014 ) . Notably , Myo2 also delivers other cargoes including mitochondria , peroxisomes , secretory vesicles , late-Golgi elements , and astral microtubules . Myo2 binds to each cargo via cargo specific adaptors , which attach to the globular tail domain of Myo2 ( Yin et al . , 2000; Itoh et al . , 2002; Boldogh et al . , 2004; Itoh et al . , 2004; Fagarasanu et al . , 2006; Arai et al . , 2008; Lipatova et al . , 2008; Jin et al . , 2011; Santiago-Tirado et al . , 2011; Eves et al . , 2012; Chernyakov et al . , 2013 ) . Moreover , some of the regulatory pathways for vacuole transport are also utilized by other Myo2 cargoes ( Moore and Miller , 2007; Peng and Weisman , 2008; Fagarasanu et al . , 2009; Jin et al . , 2009; Yau et al . , 2014 ) . Many of the proteins involved in vacuole inheritance are conserved among several species , which suggests that vacuole inheritance confers a selective advantage ( Mast et al . , 2012 ) . These observations suggest that the vacuole plays essential roles . Surprisingly , mutations that block vacuole inheritance do not have a notable impact on cell viability ( Catlett and Weisman , 1998; Ishikawa et al . , 2003 ) . Indeed , previous studies suggest that new vacuole synthesis occurs in the absence of vacuole inheritance ( Weisman et al . , 1990; Gomes De Mesquita et al . , 1997 ) , however at the time of those studies , there were no suitable methods to distinguish an old vacuole from newly formed vacuoles . Moreover the origin of the new vacuole was unknown . Importantly it was not clear how many pathways would need to be blocked in order to prevent vacuole biogenesis . Note that vacuole biogenesis utilizes at least three direct transport pathways: autophagy/Cvt ( from the cytoplasm ) , AP-3/ALP ( from the Golgi ) , and CPY ( from the MVB/endosome ) pathways ( Bryant and Stevens , 1998; Hecht et al . , 2014 ) .
The new mother cells of vac17Δ , which lack FM4-64 ( Figure 1A; open yellow arrowheads ) , had larger vacuoles compared to the newly formed vacuoles in the large buds . These findings suggest that the newly generated vacuoles continue to grow . To establish the relationship between the size of the newly formed vacuoles with the size of unbudded cells or mother cells , we measured vacuole size vs cell size in wild-type and vac17Δ cells . In wild-type cells , the vacuole diameter showed a linear relationship with cell diameter ( Figure 1—figure supplement 1A , B ) . This is consistent with a previous study that demonstrated that vacuole volume correlates with cell volume ( Chan and Marshall , 2014 ) . Notably , we found that in the vac17Δ mutant , the vacuoles grew prior to the generation of a bud ( Figure 1B , black arrow ) . In unbudded new daughters of the vac17Δ mutant , the vacuole diameter relative to cell diameter was smaller than that of wild-type and vac17Δ unbudded old mother cells ( Figure 1B , left panel ) . Notably , the average diameter of the vacuole was only 5 . 7 ( ±1 . 5 ) % of the cell diameter in new unbudded vac17Δ cells . However , after production of a small bud , the average mother cell vacuole diameter was 10 ( ±3 . 8 ) % of the cell diameter . In contrast , there was no significant increase in the relative percent diameter of the vacuoles in wild-type and vac17Δ old mother cells with or without a small bud . These observations show that the vacuoles in the new daughter cells grow to a minimum size prior to producing a bud . This growth occurs either because a minimum vacuole size is required and/or because the vacuole needs to mature prior to the generation of a bud . Note that the vacuoles in vac17Δ new mothers continued to grow , and grew faster than vacuoles in either wild-type or vac17Δ old mother cells ( Figure 1—figure supplement 1C , green line vs the black/pink lines ) . These observations suggest that cells actively synthesize new vacuoles in the absence of inherited vacuoles , and further suggest that there are mechanisms that regulate vacuole size in proportion to cell size . In addition , the average vacuole diameter relative to cell diameter of all the vac17Δ old mother cells was larger than that of wild-type and vac17Δ new mother cells ( Figure 1—figure supplement 1D ) . This suggests that vacuole inheritance is also important for regulating vacuole size in the mother cell . We observed that all vac17Δ mother cells and unbudded cells have a vacuole as defined by the presence of Vph1-GFP . This strongly suggests that the vacuole is required for cell growth and viability . If this were true , then a combination of a vacuole inheritance defect with an additional defect in the synthesis of a new vacuole would render the cell inviable . Similarly , an additional defect in vacuole function ( s ) that are required for bud emergence would result in non-viable cells ( Figure 2A ) . Indeed , a high-throughput screen suggested that over twenty genes might be synthetically lethal with the vac17Δ mutant ( Costanzo et al . , 2010 ) . We individually tested double mutants of vac17Δ with each of the previously proposed candidates that are not essential genes , and found that the double mutants , vac17Δ pep12Δ and vac17Δ vps45Δ displayed synthetic growth defects ( Figure 2B , C ) . Importantly pep12Δ and vps45Δ were also synthetically lethal with additional mutants defective in vacuole inheritance , vac8Δ and myo2-N1304D mutants ( Figure 2—figure supplement 1A–D ) . The corresponding wild-type genes , PEP12 and VPS45 , likely play a critical role in the generation of a new vacuole . 10 . 7554/eLife . 08160 . 005Figure 2 . Pep12 and Vps45 are required for the synthesis of a new vacuole . ( A ) Schematic of pathways predicted to exhibit synthetic lethality with mutations in vacuole inheritance . When vacuole inheritance is defective , the bud generates a new vacuole that is independent of the mother vacuole . If vacuoles play an essential role , then cells with no mechanism to generate a vacuole will not be viable . Furthermore if the new vacuole is defective in its essential function ( s ) , the cell will not be viable . ( B ) The pep12Δ and vps45Δ mutants exhibit a synthetic growth defect with vac17Δ . Results of tetrad dissection . Haploid colonies from tetrads derived from heterozygous diploids of VAC17/vac17Δ PEP12/pep12Δ and VAC17/vac17Δ VPS45/vps45Δ were arrayed vertically on YPD ( rich medium ) plates incubated at 24°C for 3 days . vac17Δ = 17Δ; pep12Δ = 12Δ; vps45Δ = 45Δ; vac17Δ pep12Δ or vac17Δ vps45Δ double mutant = ΔΔ are indicated . ( C ) Quantification of colony size , relative to the average of wild-type colonies . A total of 28 full tetrads and 48 full tetrads were analyzed for vac17Δ pep12Δ and vac17Δ vps45Δ , respectively . Average size in each genotype ( red bar ) . Error bar; SD . ( D ) Both vacuole inheritance and new synthesis are important to maintain functional vacuoles . In the vac17Δ pep12Δ mutant several cells appear to lack a vacuole . Wild-type cells incubated with 10 μM CMAC for 30 min exhibited a blue fluorescent signal in the vacuole lumen . The limiting membrane of the vacuole is indicated by Vph1-GFP expressed from its endogenous locus . Wild-type and vac17Δ cells show normal localization of Vph1-GFP and CMAC . Single pep12Δ cells show abnormal distribution in Vph1-GFP , but not CMAC . The vac17Δ pep12Δ double mutant cells show defects in the localization of Vph1-GFP and CMAC . ( E ) Quantification of cells with a CMAC positive subcellular structure . Any CMAC containing structure with or without Vph1-GFP was scored as a structure . Error bars; SD calculated from four independent experiments with at least 100 cells counted in each strain/experiment . ( F ) New vacuoles are generated from Pep12-positive endosomes . GFP-Pep12/Vph1-CFP expressed in wild-type and vac17Δcells were pulse labeled with FM4-64 . GFP-Pep12 and Vph1-CFP were expressed from the endogenous loci in both strains . Open arrowheads; new vacuoles . ( G ) Quantification of percent daughter cells with Vph1-CFP and/or GFP-Pep12 in vac17Δ cells . Averages from two independent experiments; at least 100 cells counted per experiment . Open circles and triangles indicate each experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 08160 . 00510 . 7554/eLife . 08160 . 006Figure 2—figure supplement 1 . Pep12 and Vps45 are required for the viability of vacuole inheritance mutants . ( A ) A vac8Δ mutant exhibits a synthetic growth defect with either pep12Δ and vps45Δ . Results of tetrad dissections of heterozygous diploids , VAC8/vac8Δ PEP12/pep12Δ and VAC8/vac8Δ VPS45/vps45Δ . vac8Δ = 8Δ; pep12Δ = 12Δ; vps45Δ = 45Δ; vac8Δ pep12Δ or vac8Δ vps45Δ double mutant = ΔΔ are indicated . ( B ) Quantification of colony size , relative to average of wild-type colonies . A total 35 tetrads and 37 tetrads were analyzed for vac8Δ pep12Δ and vac8Δ vps45Δ , respectively . ( C ) The pep12Δ mutant exhibits a synthetic growth defect with myo2-N1304D , which is a vacuole inheritance mutant due to a defect in binding to Vac17 . Survival occurs when wild-type MYO2 is expressed in myo2Δ cells , and in the myo2Δ pep12Δ mutant . In addition , the myo2-N1304D mutant expressed in myo2Δ cells is sufficient for survival . In contrast , the myo2-N1304D mutant expressed in myo2Δ pep12Δ cells is lethal . Plasmids were transformed into a myo2Δ strain and a myo2Δ pep12Δ strain containing YCp50 [URA3] MYO2 . Plasmids tested were pRS413 [HIS3] ( mock ) , pRS413 MYO2 ( pMYO2 ) , or pRS413 myo2-N1304D ( pmyo2-N1304D ) . Transformed colonies were cultured in SC-His-Ura liquid media and serial dilutions spotted onto SC+5-FOA plates to counter select against YCp50 [URA3] MYO2 ( middle panel ) and the same culture was also tested on SC-His-Ura plate ( right panel ) . ( D ) A vps45Δ mutant exhibits a synthetic growth defect with the myo2-N1304D mutant . Wild-type MYO2 expressed in myo2Δ cells , or in the myo2Δ vps45Δ mutant , is sufficient for cell viability . In addition , the myo2-N1304D mutant expressed in myo2Δ cells is sufficient for cell viability . In contrast , the myo2-N1304D mutant expressed in myo2Δ vps45Δ cells is lethal . DOI: http://dx . doi . org/10 . 7554/eLife . 08160 . 006 PEP12 encodes a t-SNARE , and VPS45 encodes a Sec1/Munc18 protein . These proteins function together in the vacuole-protein-sorting pathway from endosomes to the vacuole ( Becherer et al . , 1996; Burd et al . , 1997 ) . To test whether the growth defects of the vac17Δ pep12Δ double mutant are due to a defect in forming new vacuoles , we monitored vacuoles using two markers , Vph1-GFP and CMAC , a small molecule that is taken into the lumen of the vacuole ( Stefan and Blumer , 1999 ) . Wild-type and vac17Δ cells showed normal localization of Vph1-GFP and CMAC ( Figure 2D , E ) . However , single pep12Δ cells , which have a defect in protein sorting to the vacuole , and a partial defect in vacuole inheritance ( Raymond et al . , 1992 ) , showed an abnormal distribution of Vph1 ( Piper et al . , 1997 ) , but not CMAC ( Figure 2D , E ) . The vac17Δ pep12Δ double mutant cells showed defects in the localization of Vph1 and CMAC ( Figure 2D , E ) . These findings suggest that the double mutant does not generate normal vacuoles . PEP12 localization is consistent with its role in the synthesis of a new vacuole . GFP-Pep12 and Vph1-CFP were co-expressed in a vac17Δ mutant labeled with FM4-64 . In this strain , 63% of vac17Δ cells had GFP-Pep12 on the newly synthesized vacuole in the bud , which was Vph1-CFP positive but lacked FM4-64 ( Figure 2F , G ) . In 17% of daughter cells , GFP-Pep12 was present in buds without a vacuole , as indicated by the absence of Vph1-CFP ( vacuole ) . This suggests that a Pep12-positive endosome appears first , and subsequently a Vph1-positive vacuole matures from the Pep12-positive endosome . Note that in only 1 . 6% of cells , Vph1-CFP was present without GFP-Pep12 . That the vacuole in the mother cell must reach a specific size prior to bud emergence ( Figure 1B ) and that a vacuole is required for cell growth ( Figure 2 ) , raised the possibility that the vacuole is required for cell-cycle progression . Thus , we tested whether the vac17 pep12 double mutant arrests at a specific point in the cell-cycle . To perform this analysis , we used the pep12-60tsf mutant , which is temperature sensitive for function ( tsf ) . At elevated temperatures PEP12 function is acutely ablated , but the cells remain viable ( Burd et al . , 1997 ) . Importantly , the vac17Δ pep12-60tsf double mutant , but not pep12-60tsf single mutant , showed a severe growth defect at 37°C ( Figure 3A ) . 10 . 7554/eLife . 08160 . 007Figure 3 . The vacuole is required for cell-cycle progression from early G1 . ( A ) The vac17Δ pep12-60tsf double mutant shows synthetic growth defects at the restrictive temperature , 37°C . Wild-type , vac17Δ , pep12-60tsf and vac17Δ pep12-60tsf strains were cultured in liquid media and serial dilutions were spotted onto YPD plates . The plates were incubated at 24°C , 30°C and 37°C for 2 days . ( B ) The vac17Δ pep12-60tsf double mutant arrests in G1 phase at the restrictive temperature 37°C . Percent cells in G1 phase ( solid lines ) . Yeast strains tested; wild-type , vac17Δ , pep12-60tsf , and vac17Δ pep12-60tsf . Cultures were incubated at 24°C overnight , and then sifted to 37°C for 0 , 2 , 4 , 8 , 12 , or 24 hr . The percentage of G1 cells ( 1N DNA ) was measured using propidium iodide ( PI ) staining and assessed by flow cytometry . The same cultures were analyzed for lethality ( percent dead cells ) ( dashed lines ) . After incubation at 37°C , the number of yeast cells were assessed with a hemocytometer , and their ability to form colonies at 24°C on YPD plates was tested . Lethality was inferred from the number of cells that survived the treatment . Error bars; SD calculated from four independent experiments . *** ( p-value < 1 × 10−3 ) . ( C ) The vac17Δ pep12-60tsf double mutant arrests in early G1 phase at the restrictive temperature 37°C . Cells were scored for the presence of Whi5-3xGFP in the nucleus . Wild-type , vac17Δ , pep12-60tsf , and vac17Δ pep12-60tsf cells , which express Whi5-3xGFP from its endogenous locus , were incubated at 24°C overnight , and then sifted to 37°C for 0 or 4 hr . Error bars; SD calculated from three independent experiments with at least 100 cells counted in each strain/experiment . *** ( p-value < 1 × 10−3 ) . ( D ) Arrested cells that have 1N DNA content are unbudded . Wild-type and vac17Δ pep12-60tsf cells were incubated at 24°C overnight , and then sifted to 37°C for 24 hr . After fixation , yeast were stained with PI , and cells with 1N DNA were sorted by flow cytometry . The sorted cells were observed by microscopy . For both wild-type and the vac17Δ pep12-60tsf double mutant 99% of the cells with 1N DNA were unbudded . Sorted cells from three individual experiments were counted . At least 400 cells were counted for each experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 08160 . 00710 . 7554/eLife . 08160 . 008Figure 3—figure supplement 1 . The vacuole is required for cell-cycle progression from early G1 phase . ( A ) The vac17Δ pep12-60tsf double mutant is arrested in G1 phase at the restrictive temperature , 37°C . Percent cells in G1 ( 1N DNA ) and G2 ( 2N DNA ) phase after incubation at 37°C for 0 , 2 , 4 , 8 , 12 , or 24 hr . Percent cells arrested in G1 are also shown in Figure 3B . Error bars; SD calculated from four independent experiments . ( B ) vac17Δ pep12Δ and vac17Δ vps45Δ double mutants exhibit an accumulation of cells arrested in G1 phase . Flow cytometry analysis of PI staining of yeast strains; wild-type , vac17Δ , pep12Δ , vps45Δ , vac17Δ pep12Δ , and vac17Δ vps45Δ . ( C ) Quantification of percent cells in G1 and G2 phase . Error bars; SD calculated from four independent experiments . ( D ) The vac17Δ pep12-60tsf double mutant arrests in early G1 phase at the restrictive temperature , 37°C . Wild-type , vac17Δ , pep12-60tsf , and vac17Δ pep12-60tsf cells which express Whi5-3xGFP from its endogenous locus , were incubated at 24°C overnight , and then sifted to 37°C for 0 or 4 hr . Images of Whi5-3xGFP localization after cells were incubated at 37°C for 4 hr . Quantification is shown in Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 08160 . 008 To test cell-cycle progression , we labeled DNA with propidium iodide ( PI ) , and measured DNA content via FACS analysis . Wild-type , vac17Δ , pep12-60tsf , and vac17Δ pep12-60tsf cells were incubated overnight at 24°C , then shifted to 37°C . At 24°C ( 0 hr of 37°C ) , the vac17Δ pep12-60tsf double mutant had a normal cell-cycle profile ( Figure 3B and Figure 3—figure supplement 1A ) . After incubation at 37°C for 24 hr , wild-type , vac17Δ and pep12-60tsf showed a similar percent of G1 phase cells ( 1N DNA ) ; 46 ( ±2 ) % , 48 ( ±2 ) % , and 43 ( ±3 ) % , respectively ( Figure 3B and Figure 3—figure supplement 1A ) . In contrast , after incubation at the restrictive temperature for 24 hr , 80 ( ±3 ) % of the vac17Δ pep12-60tsf double mutant cells arrested at G1 phase ( 1N DNA ) . Importantly , at 8 , 12 and 24 hr after the shift to 37°C , the differences between the vac17Δ pep12-60tsf double mutant and the other strains were statistically significant ( all p-values < 1 × 10−3 ) . Consistent with these findings , both the vac17Δ pep12Δ and vac17Δ vps45Δ double mutants exhibited an arrest in G1 phase ( Figure 3—figure supplements 1B , C ) . These results strongly suggest that a functional vacuole is important for cell-cycle progression from G1 phase . The finding that the vac17Δ pep12-60tsf double mutant arrests in G1 phase , suggested that this might be the primary reason for the growth arrest ( Figure 3A ) . To further test this hypothesis , we monitored cell viability at several time points following a shift to 37°C . Cells were counted with a hemocytometer , and the ability of these cells to form colonies at 24°C was assessed ( Figure 3B ) . There was a measurable increase in the lethality of the vac17Δ pep12-60tsf mutant . However the difference between the lethality of the pep12-60tsf single mutant and the vac17Δ pep12-60tsf double mutant was not statistically significant ( p = 0 . 082 ) ( Figure 3B ) . Together , these observations strongly suggest that cells without a functional vacuole arrest at G1 phase , and that this arrest is not an artifact of general cell death . Additional evidence for a specific arrest in G1 phase , came from the finding that when the vac17Δ pep12-60tsf mutant was incubated at the restrictive temperature , there was a striking increase of cells with Whi5 in the nucleus ( Figure 3C and Figure 3—figure supplement 1D ) . Whi5 is a transcriptional repressor of the SBF complex ( Swi4-Swi6 ) , and is localized in the nucleus in early G1 phase ( Costanzo et al . , 2004; de Bruin et al . , 2004 ) . In wild-type cells , Whi5 nuclear localization is transient and released by Cdc28-Cln3 activity , which enables progression to early G1 phase . Further evidence that this is a bona fide G1 arrest , came from the finding that the vac17Δ pep12-60tsf mutant with 1N DNA content , arrests as unbudded cells . We collected 1N DNA cells by flow cytometry , and determined their morphology by microscopy . After incubation at 37°C for 24 hr , 99% of the G1 cells were unbudded , in both the wild-type and vac17Δ pep12-60tsf mutant ( Figure 3D ) . Together these results indicate that the vacuole is required for early G1 progression . The above findings predict that regulation of the cell-cycle requires signaling from the vacuole . Evidence for a candidate signaling pathway came from studies which showed that deletion of Target Of Rapamycin 1 ( TOR1 ) showed synthetic growth defects with vac17Δ and vac8Δ ( Zurita-Martinez et al . , 2007; Costanzo et al . , 2010 ) ( Figure 4—fiugre supplements 1A , B ) . Similarly , another vacuole inheritance mutant , myo2-N1304D , was synthetic lethal with tor1Δ ( Figure 4—figure supplement 1C ) . Notably , the vac17Δ tor1Δ double mutant showed an increase in cells arrested at G1 phase ( Figure 4A , B ) . This arrest in G1 phase was similar to that observed for the vac17Δ pep12-60tsf and vac17Δ pep12Δ mutants ( Figure 3—figure supplements 1A–C ) . This suggests that TORC1 signaling from the vacuole may account at least in part for the G1 arrest observed in the vac17Δ pep12-60tsf and vac17Δ pep12Δ mutants . Interestingly , the vac17Δ tor1Δ mutants generated a new vacuole in the daughter cells ( Figure 4C ) . This suggests that TOR1 functions after the synthesis of the new vacuole . 10 . 7554/eLife . 08160 . 009Figure 4 . TORC1-SCH9 signaling from the new vacuole is required for cell-cycle progression . ( A ) The vac17Δ tor1Δ double mutant exhibits an accumulation of G1 phase cells . Flow cytometry analysis with PI staining of yeast strains; wild-type , vac17Δ , tor1Δ , and vac17Δ tor1Δ . ( B ) Quantification of percent cells in G1 and G2 phase . Error bars; SD calculated from four independent experiments . ( C ) A new vacuole is synthesized in the new daughter cells of the vac17Δ tor1Δ double mutant . Wild-type , vac17Δ , tor1Δ , and vac17Δ tor1Δ cells which express Vph1-GFP from its endogenous locus , were pulse labeled with FM4-64 . Arrowheads; new vacuole in daughter cells . ( D ) The kinase activity of target of rapamycin 1 ( Tor1 ) is required for growth of the vacuole inheritance mutant , vac17Δ . Plasmids were transformed into a vac17Δ tor1Δ mutant containing pRS416 [URA3] TOR1 . Plasmids tested were pRS315 [LEU2] ( mock ) , pRS315 HA-TOR1 , pRS315 HA-tor1-D2275A , or pRS315 HA-tor1-D2294E . Transformed colonies were cultured in liquid media and serial dilutions spotted onto SC+5-FOA or SC-Leu-Ura plates . Plates were incubated at 24°C for 4 days . ( E ) TORC1 signals from the new vacuole via Sch9 . The phospho-mimetic sch9-2D3E mutant partially rescues the growth defect of the vac17Δ tor1Δ mutant . pRS413 ( mock ) , pRS413 VAC17 , pRS413 TOR1 , pVT102-H ( mock ) , pVT102-H SCH9 , pVT102-H sch9-2D3E , or pVT102-H sch9-5A expressed in vac17Δ tor1Δ with pRS416 TOR1 . Transformed colonies were cultured in liquid media and serial dilutions were spotted onto SC-His+5-FOA or SC-His-Ura plates , and incubated at 24°C for 4 days . ( F ) Sch9 signaling requires a functional vacuole . The phospho-mimetic sch9-2D3E mutant does not rescue the growth defect of the vac17Δ pep12Δ mutant . pRS413 ( mock ) , pRS413 VAC17 , pRS413 TOR1 , pVT102-H ( mock ) , pVT102-H SCH9 , pVT102-H sch9-2D3E , or pVT102-H sch9-5A plasmids were expressed in a vac17Δ pep12Δ strain . Transformed colonies were cultured in liquid media and serial dilutions spotted onto an SC-His plate , and incubated at 24°C for 3 to 4 days . DOI: http://dx . doi . org/10 . 7554/eLife . 08160 . 00910 . 7554/eLife . 08160 . 010Figure 4—figure supplement 1 . TORC1-SCH9 is required for the viability of vacuole inheritance mutants . ( A ) The tor1Δ mutant exhibits a synthetic growth defect with vac17Δ and vac8Δ . Results of tetrad dissection of heterozygous diploids , VAC17/vac17Δ TOR1/tor1Δ and VAC8/vac8Δ TOR1/tor1Δ . vac17Δ = 17Δ; vac8Δ = 8Δ; tor1Δ = 1Δ; vac17Δ tor1Δ or vac8Δ tor1Δ double mutants = ΔΔ are indicated . ( B ) Quantification of colony size , relative to average of wild-type colonies . A total of 34 tetrads and 39 tetrads were analyzed for vac17Δ tor1Δ and vac8Δ tor1Δ , respectively . ( C ) The tor1Δ mutant exhibits a synthetic growth defect with the myo2-N1304D mutant . Plasmids were transformed into a myo2Δ and myo2Δ tor1Δ strain containing YCp50 MYO2 . Plasmids tested were pRS413 ( mock ) , pRS413 MYO2 , or pRS413 myo2-N1304D . Transformed colonies were cultured in liquid media and serial dilutions were spotted onto SC+5-FOA or SC-His-Ura plate , and incubated at 24°C for 4 days . ( D ) The vac17Δ kog1-105 mutant showed synthetic growth defects . pRS313 ( mock ) and pRS415 ( mock ) , pRS313 ( mock ) and pRS415 VAC17 , pRS313 KOG1 and pRS415 ( mock ) , pRS313 KOG1 and pRS415 VAC17 , pRS313 kog1-105 and pRS415 VAC17 , or pRS313 kog1-105 and pRS415 ( mock ) expressed in vac17Δ kog1Δ with pRS316 KOG1 . Transformed colonies were cultured in liquid media and serial dilutions were spotted on SC-His-Leu+5-FOA or SC-His-Leu-Ura plates , and incubated at 24°C for 5 days . DOI: http://dx . doi . org/10 . 7554/eLife . 08160 . 010 TOR1 encodes a PIK-related protein kinase ( Alarcon et al . , 1999 ) . In yeast , Tor1 functions in the TORC1 complex , which is composed of Tor1/2 , Kog1 , Lst8 , and Tco89 ( Loewith et al . , 2002 ) . In yeast , TORC1 localizes on the vacuole membrane ( Reinke et al . , 2004; Araki et al . , 2005; Urban et al . , 2007; Sturgill et al . , 2008; Binda et al . , 2009; Jin et al . , 2014 ) , and is a key determinant of nutrient status ( Di Como and Arndt , 1996 ) . We found that the kinase activity of Tor1 was required for growth of vac17Δ ( Figure 4D ) , and that the TORC1 specific subunit KOG1 was also required for growth of vac17Δ ( Figure 4—figure supplement 1D ) . In addition , the lst8-15 temperature sensitive mutant is also synthetically lethal with vac17Δ ( Costanzo et al . , 2010 ) . These results suggest that the kinase activity of the TORC1 complex is required for normal growth of vacuole inheritance mutants . One critical target of TORC1 is the Sch9 kinase , which shares overlapping functions with metazoan S6 kinase ( Ballou et al . , 1991; Oldham et al . , 2000; Christie et al . , 2002 ) . TORC1 directly phosphorylates Sch9 on several serines and threonines , and the phospho-mimetic sch9-2D3E , but not wild-type SCH9 or the Ala-substituted sch9-5A mutant bypasses the TORC1 inhibitor , rapamycin ( Loewith et al . , 2002; Urban et al . , 2007 ) . Notably , the phospho-mimetic sch9-2D3E mutant partially suppressed the growth defect of the vac17Δ tor1Δ mutant ( Figure 4E ) , suggesting that the arrest of this mutant is due in part to defects in TORC1 mediated signaling via SCH9 . It was previously shown that Sch9-2D3E localizes on the vacuole membrane ( Urban et al . , 2007 ) . To test whether bypass of Tor1 by sch9-2D3E requires a functional vacuole , we tested whether sch9-2D3E can suppress the vac17Δ pep12Δ double mutant . Notably sch9-2D3E did not suppress the vac17Δ pep12Δ double mutant ( Figure 4F ) . This suggests that a functional vacuole is required for the roles of the TORC1-SCH9 pathway in cell-cycle progression from G1 phase . While previous studies showed that TORC1 signals from the vacuole/lysosome ( Sancak et al . , 2010 ) , and that Sch9 is activated in that location ( Urban et al . , 2007 ) , it was assumed that once Sch9 is activated , it no longer requires the vacuole for its further downstream functions . However our findings strongly suggest that the vacuole is required for Sch9 function ( s ) after Sch9 is phosphorylated by TORC1 . One possible role of the vacuole in Sch9 function , is that target protein ( s ) of the Sch9 kinase must be present on the vacuole membrane . Alternatively or in addition , the full kinase activity of Sch9 may require other proteins that are on the vacuole membrane . If a cell does not receive a vacuole from the mother cell , the daughter cell generates a new vacuole . The observation that these new vacuoles grow to a specific size prior to generation of a bud , and that TORC1 signaling is also required , raised the possibility that that there are functional differences between newly synthesized vacuoles and inherited vacuoles . As a first approach , we tested the localization of several proteins that are involved in the TORC1 pathway , Tor1 , Kog1 and Sch9 . New vacuoles were defined as Vph1-CFP positive structures that failed to inherit FM4-64 . Notably , in vac17Δ cells , GFP-Sch9 was defective in its localization to the new vacuoles , while the localization of Tor1 and Kog1 were unaffected ( Figure 5A , B ) . To directly address whether Sch9 is eventually recruited to the newly formed vacuole and when this occurs , we correlated the presence of fluorescent signals for Vph1-CFP , Tor1-3xGFP and GFP-Sch9 ( Figure 5C , D ) . These analyses show that Sch9 recruitment to the newly formed vacuole is slower than the recruitment of Vph1 and Tor1 . Specifically , small budded cells did not have fluorescent signals for any of the proteins , which indicates that these small buds do not have a vacuole ( Figure 5C–F ) . As the bud increases in size , in most cases , Vph1-CFP and Tor1-3xGFP appeared simultaneously ( Figure 5C ) . This indicates that Tor1-3xGFP is immediately recruited to the newly formed vacuoles . Moreover in some small budded cells , Tor1-3xGFP was present without Vph1-CFP , which suggests that Tor1 may be present at endosomes in these small budded cells . 10 . 7554/eLife . 08160 . 011Figure 5 . The newly synthesized vacuoles initially lack Sch9 and Fab1 . ( A ) Sch9 does not localize to the newly formed vacuole . Indicated plasmids were transformed into wild-type and vac17Δ strains , which express Vph1-CFP from its endogenous locus: pRS416 GFP-SCH9 , pRS416 TOR1-3xGFP ( D330 ) , or pRS416 KOG1-3xGFP . Transformed cells were pulse labeled with FM4-64 . Open arrowheads indicate a newly formed vacuole ( Vph1-CFP ) that was not inherited ( lack of FM4-64 ) , and is lacking GFP-Sch9 . Closed arrowheads indicate a newly formed vacuole ( Vph1-CFP ) that was not inherited ( lack of FM4-64 ) , and with the GFP fusion protein , either Tor1-3xGFP ( D330 ) or Kog1-3xGFP . ( B ) Quantification of cells with fluorescence ( Vph1-CFP and/or GFP ) in daughter cells , where the mother has both GFP and FM4-64 signals . Averages from two independent experiments , with n = 69 and n = 104 for GFP-Sch9 , n = 166 and n = 110 for Tor1-3xGFP ( D330 ) , and n = 106 and n = 126 cells for Kog1-3xGFP , respectively . Open circles and squares indicate results of each experiment . ( C ) Tor1 is immediately recruited to the newly formed vacuoles . FM4-64 labeled vac17Δ cells that express Vph1-CFP from its endogenous locus , and Tor1-3xGFP expressed from a CEN plasmid with its endogenous promoter were used . Most small budded cells do not have fluorescent signals for any of the proteins , indicating that these small buds do not have a vacuole . In most cases , as the bud increases in size , Vph1-CFP and Tor1-3xGFP appear simultaneously . The middle line in the box plot indicates the median of the data set . The upper edge of the box indicates the 75th percentile of the data set , and the lower edge indicates the 25th percentile . ns; not a significant difference ( p-value > 0 . 10 ) ; *** ( p-value < 1 × 10−6 ) . ( D ) Sch9 recruitment is delayed compared to Tor1 , but eventually occurs . FM4-64 labeled vac17Δ cells expresses Vph1-CFP from its endogenous locus , and GFP-Sch9 expressed from a CEN plasmid with its endogenous promoter were used . In medium sized buds ( 0 . 62 ( ±0 . 14 ) daughter size/mother size ) only Vph1-CFP is present . The average bud size where both Vph1 and Sch9 are present is 0 . 79 ( ±0 . 11 ) . ( E ) Recruitment of Sch9-2D3E to the new vacuole is similar to the recruitment of wild-type Sch9 . FM4-64 labeled vac17Δ cells which express Vph1-CFP from its endogenous locus , and GFP-Sch9-2D3E expressed from a CEN plasmid with its endogenous promoter were used . ( F ) The timing of the recruitment of Fab1 was similar to that observed for Sch9 . FM4-64 labeled vac17Δ cells which express Vph1-CFP from its endogenous locus , and Fab1-3xGFP expressed from a CEN plasmid with its endogenous promoter were used . In medium sized buds ( 0 . 57 ( ±0 . 13 ) daughter size/mother size ) only Vph1-CFP is present . The average bud size where both Vph1 and Fab1 are present is 0 . 76 ( ±0 . 10 ) . ( G ) Model: The vacuole is essential for cell-cycle progression and functions in part through the TORC1 pathway . When the daughter cell receives vacuoles from the mother cell ( 1 ) , the daughter can progress from G1 . If the cell fails to inherit a vacuole ( 2 ) , the cell generates a new vacuole ( 3 ) , which is followed by maturation of the new vacuole prior to G1 progression ( 4 ) . Without a functional vacuole , the daughter cell arrests at G1 phase ( 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08160 . 011 In the vac17Δ mutant , although Tor1-3xGFP and Vph1-CFP are both present on the new vacuoles in medium buds ( approximately 0 . 6 daughter size/mother size ) ( Figure 5C ) , GFP-Sch9 was generally not present until the bud size was larger ( approximately 0 . 8 ) ( Figure 5D ) . Thus Sch9 recruitment is delayed compared to Vph1 and Tor1 , but eventually occurs . Notably the recruitment of the Sch9-2D3E mutant was similar to the recruitment of wild-type Sch9 ( Figure 5E ) , indicating that the growth suppression of the vac17Δ tor1Δ mutant by sch9-2D3E is not due to a faster recruitment to the vacuole . PI ( 3 , 5 ) P2 , an inositol lipid which is present at the vacuole membrane , is required for vacuole membrane targeting of Sch9 ( Jin et al . , 2014 ) . To test whether the delay in Sch9 recruitment might be due to a delay in the generation of PI ( 3 , 5 ) P2 on newly formed vacuoles , we tested the timing of the recruitment of Fab1 , the sole lipid kinase that generates PI ( 3 , 5 ) P2 ( Gary et al . , 1998 ) . Notably the timing of the recruitment of Fab1 was similar to that observed for Sch9 . In a vac17Δ mutant which co-expressed Vph1-CFP and Fab1-3xGFP , only Vph1-CFP is present in medium buds ( approximately 0 . 6 daughter size/mother size ) ( Figure 5F ) . The average bud size where both Vph1 and Fab1 are present was 0 . 76 ( ±0 . 10 ) . Thus Fab1 recruitment is delayed compared to Tor1 , but is similar to that observed for Sch9 . These observations suggest that there are differences between new vacuoles and inherited vacuoles , and that the newly synthesized vacuoles are missing specific components of mature vacuoles that are essential for cell-cycle progression . Together , these observations demonstrate that a functional vacuole is crucial for cell-cycle progression at G1 phase , and that the TORC1-SCH9 pathway is part of this critical function ( Figure 4G ) . TORC1-SCH9 signaling from the vacuole may be involved in G1 progression through its known functions in a ribosome biogenesis and translation ( Barbet et al . , 1996; Jorgensen et al . , 2004; Urban et al . , 2007 ) . Alternatively , the TORC1-SCH9 pathway may signal the cell-cycle machinery that a functional vacuole is present and hence the cell is ready to progress from G1 . Previous studies showed that Sch9 activity is critical for cell size ( Jorgensen et al . , 2004; Urban et al . , 2007 ) , and that cell size correlates with vacuole size ( Chan and Marshall , 2014 ) . These observations together with the current study suggest that TORC1-SCH9 localization and signaling from the vacuole is critical for the regulation of cell size . It is tempting to speculate that an analogous regulation of the cell-cycle occurs from endo-lysosomal membranes in other organisms . In addition , these findings lead to the hypothesis that cells possess novel checkpoint mechanisms that prevent cell-cycle progression at G1 phase in the absence of essential organelles .
Yeast strains used are in Table 1 . Deletion and fusion strains were constructed as described ( Longtine et al . , 1998 ) . A vac17Δ pep12-60tsf double mutant strain was made through mating pep12-60tsf ( CBY9 ) ( Burd et al . , 1997 ) with vac17Δ ( LWY5798 ) ( Ishikawa et al . , 2003 ) . To generate a GFP-PEP12::natNT2 strain , a ClaI-ApaI fragment from pBlueScript SK+ ( pBS ) GFP-PEP12::natNT2 vector was integrated into the PEP12 locus . Yeast cultures were grown at 24°C unless stated otherwise . Yeast extract-peptone-dextrose ( 1% yeast extract , 2% peptone , 2% dextrose; YEPD ) , synthetic complete ( SC ) lacking the appropriate supplement ( s ) , and 5-FOA media were made as described ( Kaiser et al . , 1994 ) . Unless stated otherwise , SC medium contained 2% dextrose . 10 . 7554/eLife . 08160 . 012Table 1 . Yeast strains used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 08160 . 012StrainGenotypeSourceFigureLWY7235MATa , ura3-52 , leu2-3 , -112 , his3-Δ200 , trp1-Δ901 , lys2-801 , suc2-Δ9 ( Bonangelino et al . , 1997 ) –LWY11678MATa , VPH1-GFP::KanMXThis studyFigures 1 , 2 , 4 , Figure 1—figure supplement 1LWY12144MATa , VPH1-GFP::KanMX , vac17Δ::TRP1This studyFigures 1 , 2 , 4 , Figure 1—figure supplement 1LWY15258MATa/α , VAC17/vac17Δ::TRP1 , PEP12/pep12Δ::KanMXThis studyFigure 2LWY15612MATa/α , VAC17/vac17Δ::TRP1 , VPS45/vps45Δ::KanMXThis studyFigure 2LWY14490MATa , VPH1-GFP::KanMX , pep12Δ::KanMXThis studyFigure 2LWY14493MATa , VPH1-GFP::KanMX , vac17Δ::TRP1 , pep12Δ::KanMXThis studyFigure 2LWY15515MATa , GFP-PEP12::natNT2 , VPH1-CFP::KanMXThis studyFigure 2LWY15506MATa , GFP-PEP12::natNT2 , VPH1-CFP::KanMX , vac17Δ::TRP1This studyFigure 2LWY15263 , LWY14462 , LWY12369MATaThis studyFigures 3 , 4 , Figure 3—figure supplement 1LWY5798MATa , vac17Δ::TRP1 ( Tang et al . , 2003 ) –LWY15244 , LWY14468 , LWY12366MATa , vac17Δ::TRP1This studyFigures 3 , 4 , Figure 3—figure supplement 1CBY9MATα , pep12-60tsf , leu2-3 , 112::pBHY11 CPY-Inv LEU2 ( Burd et al . , 1997 ) –LWY15250MATa , pep12-60tsfThis studyFigure 3 , Figure 3—figure supplement 1LWY15249MATa , vac17Δ::TRP1 , pep12-60tsfThis studyFigure 3 , Figure 3—figure supplement 1LWY15799MATa , WHI5-3xGFP::His3MXThis studyFigure 3 , Figure 3—figure supplement 1LWY15791MATa , WHI5-3xGFP::His3MX , vac17Δ::TRP1This studyFigure 3 , Figure 3—figure supplement 1LWY15789MATa , WHI5-3xGFP::His3MX , pep12-60tsfThis studyFigure 3 , Figure 3—figure supplement 1LWY15814MATα , WHI5-3xGFP::His3MX , vac17Δ::TRP1 , pep12-60tsfThis studyFigure 3 , Figure 3—figure supplement 1LWY12364MATa , tor1Δ::KanMXThis studyFigure 4LWY12367MATα , vac17Δ::TRP1 , tor1Δ::KanMXThis studyFigure 4LWY12168MATa , VPH1-GFP::KanMX , tor1Δ::KanMXThis studyFigure 4LWY12193MATa , VPH1-GFP::KanMX , vac17Δ::TRP1 , tor1Δ::KanMXThis studyFigure 4LWY14142MATa , vac17Δ::TRP1 , tor1Δ::KanMX , pRS416 TOR1This studyFigure 4LWY12358 , LWY14487MATa , vac17Δ::TRP1 , pep12Δ::KanMXThis studyFigure 4 , Figure 3—figure supplement 1LWY11657MATa , VPH1-CFP::KanMXThis studyFigure 5LWY13781MATa , VPH1-CFP::KanMX , vac17Δ::TRP1This studyFigure 5LWY15610MATa/α , VAC8/vac8Δ::HIS3 , PEP12/pep12Δ::KanMXThis studyFigure 2—figure supplement 1LWY15614MATa/α , VAC8/vac8Δ::HIS3 , VPS45/vps45Δ::KanMXThis studyFigure 2—figure supplement 1LWY2947MATα , myo2Δ::TRP1 , YCp50-MYO2 ( Catlett and Weisman , 1998 ) Figure 2—figure supplement 1 , Figure 4—figure supplement 1LWY12443MATα , pep12Δ::KanMX , myo2Δ::TRP1 , YCp50-MYO2This studyFigure 2—figure supplement 1LWY15581MATα , vps45Δ::KanMX , myo2Δ::TRP1 , YCp50-MYO2This studyFigure 2—figure supplement 1LWY14497MATa , pep12Δ::KanMXThis studyFigure 3—figure supplement 1LWY14475MATa , vps45Δ::KanMXThis studyFigure 3—figure supplement 1LWY14463MATa , vac17Δ::TRP1 , vps45Δ::KanMXThis studyFigure 3—figure supplement 1LWY1MATa/α , TOR1/tor1Δ::KanMXThis studyFigure 4—figure supplement 1LWY15616MATa/α , VAC8/vac8Δ::HIS3 , TOR1/tor1Δ::KanMXThis studyFigure 4—figure supplement 1LWY12001MATa , tor1Δ::KanMX , myo2Δ::TRP1 , YCp50-MYO2This studyFigure 4—figure supplement 1LWY13595MATa , vac17Δ::TRP1 , kog1Δ::KanMX , pRS316 KOG1This studyFigure 4—figure supplement 1Each above haploid strain is ura3-52 , leu2-3 , -112 , his3-Δ200 , trp1-Δ901 , lys2-801 , suc2-Δ9 , and diploid strain is ura3-52/ura3-52 , leu2-3 , -112/leu2-3 , -112 , his3-Δ200/his3-Δ200 , trp1-Δ901/trp1-Δ901 , lys2-801/lys2-801 , suc2-Δ9/suc2-Δ9 . Vacuoles were labeled in vivo with N- ( 3-triethelammoniumpropyl ) -4- ( 6 ( 4- ( diethylamino ) phenyl ) hexatrienyl ) pyridinium dibromide ( FM4-64 [SynaptoRed C2]; Biotium , Hayward , CA , United States ) essentially as described ( Ishikawa et al . , 2003 ) . In brief , a 2 mg/ml stock solution of FM4-64 in dimethyl sulfoxide was added to early log phase cultures for a final concentration of 80 μM . After 1 hr of labeling , cells were washed and then chased in fresh liquid medium for 3–4 hr . Plasmids used are in Table 2 . To generate an integration vector to express GFP fused to Pep12 from the PEP12 gene locus , pBS GFP-PEP12::natNT2 was made . A 1 . 4 kb ClaI-BstBI fragment of PEP12 was inserted at the ClaI site of pBS . To insert GFP at the N-terminus of Pep12 , an AvrII site was generated at the N-terminus of PEP12 by PCR using primers ( 5′-CAA TAA TTG TGT TGA GAT Gcc tag gTC GGA AGA CGA ATT TTT TGG-3′ ) and ( 5′-CCA AAA AAT TCG TCT TCC GAc cta ggC ATC TCA ACA CAA TTA TTG-3′ ) . The GFP fragment was amplified from pFA6a GFP ( S65T ) -KanMX ( Longtine et al . , 1998 ) by PCR using primers ( 5′-TGA gct agc AGT AAA GGA GAA GAA CTT TTC ACT GG-3′ ) and ( 5′-TGA act agt gtt aat taa ccc ggg gat ccg tcg acc TTT GTA TAG TTC ATC CAT GCC-3′ ) . The NheI-SpeI fragment of GFP was inserted at the AvrII site . The natNT2 maker was amplified from pFA6a natNT2 ( Janke et al . , 2004 ) by PCR using primers ( 5′-CTG tgt aca CAG CGA CAT GGA GGC-3′ ) and ( 5′-TCA tgt aca ACA GGT GTT GTC CTC TGA G-3′ ) . A BsrGI fragment of natNT2 was inserted into the BsrGI site at 3′ UTR of the PEP12 . 10 . 7554/eLife . 08160 . 013Table 2 . Plasmids used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 08160 . 013Plasmid nameDescriptionSourceFigurepBlueScript SK+ GFP-PEP12::natNT2AmpThis studyFigure 2pRS416 TOR1CEN , URA3This studyFigure 4pRS413CEN , HIS3 ( Sikorski and Hieter , 1989 ) Figure 4pRS315 HA-TOR1CEN , HIS3Gift from Dr Robbie LoewithFigure 4pRS315 HA-tor1-D2275ACEN , HIS3This studyFigure 4pRS315 HA-tor1-D2294ECEN , HIS3This studyFigure 4pRS413 VAC17CEN , HIS3This studyFigure 4pRS413 TOR1CEN , HIS3This studyFigure 4pVT102-H2μ , HIS3 ( Vernet et al . , 1987 ) Figure 4pVT102-H SCH92μ , HIS3This studyFigure 4pVT102-H sch9-2D3E2μ , HIS3This studyFigure 4pVT102-H sch9-5A2μ , HIS3This studyFigure 4pRS416 GFP-SCH9CEN , URA3 ( Urban et al . , 2007 ) Figure 5pRS416 GFP-sch9-2D3ECEN , URA3This studyFigure 5pRS416 TOR1-3xGFP ( D330 ) CEN , URA3This studyFigure 5pRS416 KOG1-3xGFPCEN , URA3This studyFigure 5pRS416 FAB1-3xGFPCEN , URA3 ( Jin et al . , 2008 ) Figure 5pRS413CEN , HIS3 ( Sikorski and Hieter , 1989 ) Figure 2—figure supplement 1 , Figure 4—figure supplement 1pRS413 MYO2CEN , HIS3 ( Catlett and Weisman , 1998 ) Figure 2—figure supplement 1 , Figure 4—figure supplement 1pRS413 myo2-N1304DCEN , HIS3 ( Catlett et al . , 2000 ) Figure 2—figure supplement 1 , Figure 4—figure supplement 1pRS416 KOG1CEN , URA3 ( Jin et al . , 2014 ) Figure 4—figure supplement 1pRS313CEN , HIS3 ( Sikorski and Hieter , 1989 ) Figure 4—figure supplement 1pRS415CEN , LEU2 ( Sikorski and Hieter , 1989 ) Figure 4—figure supplement 1pRS415 VAC17CEN , LEU2 ( Jin et al . , 2009 ) Figure 4—figure supplement 1pRS313 KOG1CEN , HIS3 ( Nakashima et al . , 2008 ) Figure 4—figure supplement 1pRS313 kog1-105CEN , HIS3 ( Nakashima et al . , 2008 ) Figure 4—figure supplement 1 pRS416 TOR1 includes 227 bp upstream and 944 bp downstream of the TOR1 gene , the same region as pRS315 HA-TOR1 ( gift from Dr Robbie Loewith ) . For generation of tor1-D2275A , and -D2294E kinase dead mutants ( Zheng et al . , 1995; Alarcon et al . , 1999 ) , the TOR1 gene was mutagenized by site-directed mutagenesis using the following primers: ( D2275A-S ) 5′-GTT ATA TTC TGG GAC TAG GTG cTC GCC ATC CAA GCA ACC TG-3′ , ( D2275A-AS ) 5′-CAG GTT GCT TGG ATG GCG AgC ACC TAG TCC CAG AAT ATA AC-3′ , ( D2294E-S ) 5′-CAC CGG TAA AGT TAT CCA CAT TGA aTT CGG CGA TTG TTT TGA AGC-3′ , ( D2294E-AS ) 5′-GCT TCA AAA CAA TCG CCG AAt TCA ATG TGG ATA ACT TTA CCG GTG-3′ . For generation of pVT102-H SCH9 , SCH9 was amplified by PCR using primers ( 5′-ATA gga tcc ATG ATG AAT TTT TTT ACA TCA AAA TCG-3′ ) and ( 5′-GAG tct aga TAT TTC GAA TCT TCC ACT GAC AAA TTC-3′ ) . A BamHI-XbaI fragment of SCH9 was inserted into the BamHI , XbaI sites of pVT102-H ( Vernet et al . , 1987 ) . For generation of phospho-mimetic sch9-2D3E and non-phospho sch9-5A mutant ( Urban et al . , 2007 ) , the SCH9 gene was mutagenized by site-directed mutagenesis using the following primers: ( T723D/S726D-S ) 5′-CC GAT GAT GAC TGC Tga CCC GCT Aga TCC AGC CAT GCA AGC AAA G-3′ , ( T723D/S726D-AS ) 5′-CTT TGC TTG CAT GGC TGG Atc TAG CGG Gtc AGC AGT CAT CAT CGG-3′ , ( T737E–S ) 5′-CAA GCA AAG TTT GCT GGT TTC gaa TTT GTT GAT GAG TCC GCC ATC-3′ , ( T737E-AS ) 5′-GAT GGC GGA CTC ATC AAC AAA ttc GAA ACC AGC AAA CTT TGC TTG-3′ , ( S758E/S765E–S ) 5′-CCT ACA AAA Cga GTA CTT TAT GGA ACC TGG Tga aTT TAT CCC GGG-3′ , ( S758E/S765E-AS ) 5′-CCC GGG ATA AAt tcA CCA GGT TCC ATA AAG TAC tcG TTT TGT AGG-3′ , ( T723A/S726A-S ) 5′-CCG ATG ATG ACT GCT gCC CCG CTA gCT CCA GCC ATG CAA GCA AAG-3′ , ( T723A/S726A-AS ) 5′-CTT TGC TTG CAT GGC TGG AGc TAG CGG GGc AGC AGT CAT CAT CGG-3′ , ( T737A-S ) 5′-CAA GCA AAG TTT GCT GGT TTC gCC TTT GTT GAT GAG TCC GCC ATC-3′ , ( T737A-AS ) 5′-GAT GGC GGA CTC ATC AAC AAA GGc GAA ACC AGC AAA CTT TGC TTG-3′ , ( S758A/S765A-S ) 5′-CCT ACA AAA CgC GTA CTT TAT GGA ACC TGG TgC CTT TAT CCC GGG-3′ , ( S758A/S765A-AS ) 5′-CCC GGG ATA AAG GcA CCA GGT TCC ATA AAG TAC GcG TTT TGT AGG-3′ . For generation of pRS416 TOR1-3xGFP ( D330 ) , an XbaI site was generated at the D330 position of TOR1 with PCR using primers ( 5′-GTT TAT AAG GAA ATC TTG TTT TTG AAG tct Aga CCC TTT TTG AAT CAA GTG TTC-3′ ) and ( 5′-GAA CAC TTG ATT CAA AAA GGG tcT aga CTT CAA AAA CAA GAT TTC CTT ATA AAC-3′ ) . A 3xGFP fragment was amplified from pFA6a 3xGFP-TRP1 by PCR using primers ( 5′-CGG tct aga GGG TTA ATT AAC GTG AGC AAG GG-3′ ) and ( 5′-AAT CTC GAG gct agc GGG GAT CCG TCG ACC CTT GTA CAG CTC GTC CAT GC-3′ ) . An XbaI-NheI fragment of 3xGFP was inserted at the XbaI site ( Jin et al . , 2014 ) . For generation of pRS416 KOG1-3xGFP , an XbaI site was generated at the C-terminal end of KOG1 by PCR using primers ( 5′-GAG AAT TGA TTA TTT Ttc tag aTA TGT GCC ATT TCT TTT TTT TTC-3′ ) and ( 5′-GAA AAA AAA AGA AAT GGC ACA TAt cta gaA AAA TAA TCA ATT CTC-3′ ) . The 3xGFP fragment was amplified by PCR using primers ( 5′-TCT AGA GGG TTA ATT tct aga AGC AAG GGC GAG GAG C-3′ ) and ( 5′-AAT CTC GAG gct agc GTT AAT TAA CCC GGG GAT CCG TCG ACC-3′ ) . The XbaI-NheI fragment of 3xGFP was inserted at the XbaI site ( Jin et al . , 2014 ) . Quantitation of nuclear DNA was determined as follows: Cells were stained with PI and analyzed by FACS analysis ( MACSQuant 1; Miltenyi Biotec , Germany ) . In most experiments , 10 , 000 cells were examined . Yeast were incubated at 24°C overnight , and then sifted to 37°C for 0 , 2 , 4 , 8 , 12 , or 24 hr . At the start of the experiment , yeast were in log phase growth . 1 . 0 OD600 yeast cultures were collected , washed with 50 mM of Tris-HCl [pH7 . 5] , and fixed with 70% EtOH . Cells were then washed twice with 50 mM of Tris-HCl [pH7 . 5] , followed by sonication . Cells were treated with RNaseA ( Sigma–Aldrich R6513; final 2 mg/ml in 50 mM of Tris-HCl [pH7 . 5] ) at 37°C overnight . Cells were then treated with Pepsin ( Sigma–Aldrich 7000; final 5 mg/ml ) at room temperature for 30 min , and stained with PI ( Sigma–Aldrich 4170 ) 50 mg/ml in 180 mM Tris-HCl [pH7 . 5] , 180 mM NaCl , 70 mM MgCl2 for 1 hr at room temperature . The PI stained cells were analyzed by FACS . | Animals , fungi and other eukaryotes have cells that are divided into sub-compartments that are called organelles . Each type of organelle serves a specific purpose that is essential for the life of the cell . Yeast cells have a large organelle called a vacuole; the inside of the vacuole is acidic and contains enzymes that can break down other molecules . Previous studies have shown that when a budding yeast cell buds to produce a new daughter cell , a process ensures that some of the mother's vacuole is transferred to its daughter . However , yeast mutants that fail to inherit some of their mother's vacuole can still survive . This is because an ‘alternative’ mechanism allows the newly forming daughter to generate its own vacuole from scratch . Jin and Weisman now unexpectedly show that a new daughter cell cannot become a mother cell until its new vacuole is formed . The experiments made use of yeast mutants that were defective in the ‘inheritance’ mechanism , and double mutants that were defective in both the inheritance and alternative mechanisms . The experiments also revealed that a signal from the vacuole is required before the yeast cell's nucleus can start the cycle of events that lead to the cell dividing . Jin and Weisman suggest that this newly identified communication between the vacuole and the nucleus may help to ensure that critical organelles are present in all cells . Though it remains unclear why the yeast vacuole is critical for a cell to divide , these findings suggest that the mammalian lysosome ( which is similar to the yeast vacuole ) may perform a similar critical role in mammals . If this is the case , then understanding how these organelles communicate with the nucleus may provide new insights into how to prevent the uncontrolled growth of tumors and cancer . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2015 | The vacuole/lysosome is required for cell-cycle progression |
Neurons show diverse timescales , so that different parts of a network respond with disparate temporal dynamics . Such diversity is observed both when comparing timescales across brain areas and among cells within local populations; the underlying circuit mechanism remains unknown . We examine conditions under which spatially local connectivity can produce such diverse temporal behavior . In a linear network , timescales are segregated if the eigenvectors of the connectivity matrix are localized to different parts of the network . We develop a framework to predict the shapes of localized eigenvectors . Notably , local connectivity alone is insufficient for separate timescales . However , localization of timescales can be realized by heterogeneity in the connectivity profile , and we demonstrate two classes of network architecture that allow such localization . Our results suggest a framework to relate structural heterogeneity to functional diversity and , beyond neural dynamics , are generally applicable to the relationship between structure and dynamics in biological networks .
A major challenge in the study of neural circuits , and complex networks more generally , is understanding the relationship between network structure and patterns of activity or possible functions this structure can subserve ( Strogatz , 2001; Newman , 2003; Honey et al . , 2010; Sporns , 2011 ) . A number of neural networks show a diversity of time constants , namely different nodes ( single neurons or local neural groups ) in the network display dynamical activity that changes on different timescales . For instance , in the mammalian brain , long integrative timescales of neurons in the frontal cortex ( Romo et al . , 1999; Wang , 2001; Wang , 2010 ) are in striking contrast with rapid transient responses of neurons in a primary sensory area ( Benucci et al . , 2009 ) . Furthermore , even within a local circuit , a diversity of timescales may coexist across a heterogeneous neural population . Notable recent examples include the timescales of reward integration in the macaque cortex ( Bernacchia et al . , 2011 ) , and the decay of neural firing rates in the zebrafish ( Miri et al . , 2011 ) and macaque oculomotor integrators ( Joshua et al . , 2013 ) . While several models have been proposed , general structural principles that enable a network to show a diversity of timescales are lacking . Studies of the cortex have revealed that neural connectivity decays rapidly with distance ( Holmgren et al . , 2003; Markov et al . , 2011; Perin et al . , 2011; Levy and Reyes , 2012; Markov et al . , 2014; Ercsey-Ravasz et al . , 2013 ) as does the magnitude of correlations in neural activity ( Constantinidis and Goldman-Rakic , 2002; Smith and Kohn , 2008; Komiyama et al . , 2010 ) . This characteristic is apparent on multiple scales: in the cerebral cortex of the macaque monkey , both the number of connections between neurons in a given area and those between neurons across different brain areas decay rapidly with distance ( Markov et al . , 2011 , 2014 ) . Intuitively , local connectivity may suggest that the timescales of network activity are localized , by which we mean that nodes that respond with a certain timescale are contained within a particular region of the network . Such a network would show patterns of activity with different temporal dynamics in disparate regions . Surprisingly , this is not always true and , as we show , additional conditions are required for localized structure to translate into localized temporal dynamics . We study this structure–function relationship for linear networks of interacting nodes . Linear networks are used to model a variety of physical and biological networks , especially those where inter-node interactions are weighted ( Newman , 2010 ) . Most dynamical systems can be linearized around a point of interest , and so linear networks generically emerge when studying the response of nonlinear networks to small perturbations ( Strogatz , 1994; Newman , 2010 ) . Moreover , for many neurons the dependence of firing rate on input is approximately threshold-linear over a wide range ( Ahmed et al . , 1998; Ermentrout , 1998; Wang , 1998; Chance et al . , 2002 ) , and linear networks are common models for the dynamics of neural circuits ( Dayan and Abbott , 2001; Shriki et al . , 2003; Vogels et al . , 2005; Rajan and Abbott , 2006; Ganguli et al . , 2008; Ganguli et al . , 2008; Murphy and Miller , 2009; Miri et al . , 2011 ) . The activity of a linear network is determined by a set of characteristic patterns , called eigenvectors ( Rugh , 1995 ) . Each eigenvector specifies the relative activation of the various nodes . For example , in one eigenvector the first node could show twice as much activity as the second node and four times as much activity as the third node , and so on . The activity of the network is the weighted sum of contributions from the eigenvectors . The weight ( or amplitude ) of each eigenvector changes over time with a timescale determined by the eigenvalue corresponding to the eigenvector . The network architecture determines the eigenvectors and eigenvalues , while the input sets the amplitudes with which the various eigenvectors are activated . In Figure 1 , we illustrate this decomposition in a simple schematic network with three eigenvectors whose amplitudes change on a fast , intermediate and slow timescale respectively . 10 . 7554/eLife . 01239 . 003Figure 1 . The activity of a linear network can be decomposed into contributions from a set of eigenvectors . On the right is shown a sample network along with the activity of two nodes ( cyan and yellow ) . The activity of this network is the combination of a set of eigenvectors whose spatial distributions are shown in blue , green and red on the left . The nodes are colored according to the contributions of the various eigenvectors . Each eigenvector has an amplitude that varies in time with a single timescale given by the corresponding eigenvalue; here the blue , green and red eigenvectors have a fast , intermediate and slow timescale , respectively . The cyan node is primarily a combination of the blue and green eigenvectors; hence its activity is dominated by a combination of the blue and green amplitudes and it shows a fast and an intermediate timescale . Similarly , the yellow node has large components in the green and red eigenvectors , therefore its activity reflects the corresponding amplitudes and intermediate and slow timescales . DOI: http://dx . doi . org/10 . 7554/eLife . 01239 . 003 In general , the eigenvectors are poorly segregated from each other: each node participates significantly in multiple eigenvectors and each eigenvector is spread out across multiple nodes ( Trefethen and Embree , 2005 ) . Consequently , timescales are not segregated , and a large number of timescales are shared across nodes . Furthermore , if the timescales have largely different values , certain eigenvectors are more persistent than others and dominate the nodes at which they are present . If these slow timescales are spread across multiple nodes , they dominate the network activity and the nodes will show very similar temporal dynamics . This further limits the diversity of network computation . In this paper , we begin by observing that rapidly-decaying connectivity by itself is insufficient to give rise to localized eigenvectors . We then examine conditions on the network-coupling matrix that allow localized eigenvectors to emerge and build a framework to calculate their shapes . We illustrate our methods with simple examples of neural dynamics . Our examples are drawn from Neuroscience , but our results should be more broadly applicable for understanding network dynamics and the relationship between the structure and function of complex systems .
Our first model architecture is motivated by observations that as one progresses from sensory to prefrontal areas in the primate brain , neurons receive an increasing number of excitatory connections from their neighbors ( Wang , 2001; Elston , 2007; Wang , 2008 ) . We model a chain of nodes ( i . e . , neurons , networks of neurons or cortical areas ) with connectivity that decays exponentially with distance . In addition , we introduce a gradient of excitatory self-couplings along the chain to account for the increase in local excitation . The network is shown in Figure 3A and the coupling matrix W is given by ( 3 ) W ( j , k ) ={μ0+Δrjfor j=k ( self-coupling ) μfe− ( j−k ) /lcfor j>k ( feedforward connections ) μbe ( j−k ) /lcfor j<k ( feedback connections ) . 10 . 7554/eLife . 01239 . 005Figure 3 . Localized eigenvectors in a network with a gradient of local connectivity . ( A ) The network is a chain of 100 nodes . Network topology is shown as a schematic with a subset of nodes and only nearest-neighbor connections . The plot above the chain shows the connectivity profile , highlighting the exponential decay and the asymmetry between feedforward and feedback connections . Self-coupling increases along the chain , as shown by the grayscale gradient . ( B ) Sample eigenvectors ( filled circles ) in a network with a weak gradient of self-coupling , so that localized and delocalized eigenvectors coexist . Localized eigenvectors are described by Gaussians , and predictions from Equation 4 are shown as solid lines . Eigenvectors are normalized by maximum value . The network is described by Equation 3 , with μ0 = −1 . 9 , Δr = 0 . 0015 , μf = 0 . 2 , μb= 0 . 1 and lc = 4 . ( C ) Sample eigenvectors ( filled circles ) along with predictions ( solid lines ) in a network with a strong gradient , so that all eigenvectors are localized . Network parameters are the same as B , except Δr = 0 . 01 . ( D ) Heat map of eigenvectors from network in ( C ) on logarithmic scale . Eigenvectors are along rows , arranged by increasing decay time . All are localized , and eigenvectors with longer timescales are localized further down in the chain . Edge effects cause the Gaussian shape to break down at the end of the chain , but eigenvectors are still localized at the boundary . ( E ) Dynamical response of the network in ( C ) to an input pulse . Nodes early in the chain show responses that decay away rapidly , while those further in the chain show more persistent responses . DOI: http://dx . doi . org/10 . 7554/eLife . 01239 . 00510 . 7554/eLife . 01239 . 006Figure 3—figure supplement 1 . Co-existence of localized and delocalized eigenvectors in a network with a weak gradient of local connectivity . ( A ) Left panel: eigenvalues of the network ( filled circles ) along with the region of the complex plane in which α2 > 0 ( gray shaded region ) . Eigenvectors corresponding to eigenvalues within this region are predicted to be localized . ( B ) Eigenvectors corresponding to the colored eigenvalues in panel A . Eigenvalues within the gray region correspond to localized eigenvectors . Eigenvectors outside the gray region are progressively more delocalized . Eigenvectors are shown as solid lines for ease of visualization . ( C ) Heat map of eigenvectors on logarithmic scale . Eigenvectors are along rows , arranged by increasing decay time . The network is described by Equation 3 in the main text , with μ0 = − 1 . 9 , Δr = 0 . 0015 , μf = 0 . 2 , μb = 0 . 1 , and lc = 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 01239 . 006 The self-coupling includes a leakage term ( μ0 < 0 ) and a recurrent excitation term that increases along the chain with a slope Δr . Nodes higher in the network thus have stronger self-coupling . Connection strengths have a decay length lc . μf scales the overall strength of feedforward connections ( i . e . , connections from early to late nodes in the chain ) while μb scales the strength of feedback connections . In general we set μf > μb . If the gradient of self-coupling ( Δr ) is strong enough , some of the eigenvectors of the network will be localized . As the gradient becomes steeper this region of localization expands . Our theory predicts which eigenvectors will be localized and how this region expands as the gradient becomes steeper ( Figure 3—figure supplement 1 ) . By applying the theory sketched in the previous section ( and developed in detail in the appendix [Supplementary file 1] ) , we find that the value of the eigenvector width for the localized eigenvectors ( α in Equation 2 ) is equal to ( see Section 3 of Supplementary file 1 ) ( 4 ) α2=μf−μb2Δr ( 1+cosh ( 1lc ) ) . This equation asserts that α2 is inversely proportional to the gradient of local connectivity , Δr , so that a steeper gradient leads to sharper localization , and α2 increases with increasing connectivity decay length , lc . Note that in this case the eigenvector width is independent of the location of the eigenvector ( or the particular timescale ) . In Figure 3B , we plot sample eigenvectors for a network with a weak gradient , where localized and delocalized eigenvectors coexist . We also plot the analytical prediction for the localized eigenvectors , which fits well with the numerical simulation results . For more details on this network see Figure 3—figure supplement 1 . In Figure 3C , we plot sample eigenvectors for a network with a strong enough gradient that all eigenvectors are localized . As shown in Figure 3D , all the remaining eigenvectors of this network are localized . In Figure 3E , we plot the decay of this network’s activity from a uniform initial condition; as predicted from the structure of the eigenvectors , decay time constants increase up the chain . With a strong gradient of self-coupling , Equation 4 holds for all eigenvectors except those at the end of the chain , where edge effects change the shape of the eigenvectors . These eigenvectors are still localized , at the boundary , but are no longer Gaussian and appear to be better described as modulated exponentials . Equation 4 also predicts that eigenvectors become more localized as feedforward and feedback connection strengths approach each other . This is counter-intuitive , since increasing feedback strength should couple nodes more tightly . Numerically , this prediction is confirmed only when μf − μb is not close to 0 . As seen in Figure 4 , when μf − μb is small , the eigenvector is no longer Gaussian and instead shows multiple peaks . Strengthening the feedback connections leads to the emergence of ripples in the slower modes that modulate the activity of the earlier , faster nodes . While the first-order approximation of the shape of vλ breaks down in this regime , Equation 4 is locally valid in that the largest peak sharpens with increasing symmetry , as seen in Figure 4B . 10 . 7554/eLife . 01239 . 007Figure 4 . Second-order expansion for partially-delocalized eigenvectors . Same model with a gradient of local connectivity as in Figure 3 . ( A ) Schematic of the predicted shape . Eigenvectors ( black ) are the product of an exponential ( blue ) and an Airy function ( red ) . The constant in the exponential depends on the asymmetry between feedback ( μb ) and feedforward ( μf ) strengths . In the left panel , μf − μb is large and the product is well described by a Gaussian . In the right panel , μf − μb is small and the exponential is shallow enough that the product is somewhat delocalized . ( B ) Analytically predicted eigenvector shapes ( solid lines ) compared to numerical simulations ( filled circles ) for four values of μb . For each value of μb one representative eigenvector is shown . As μb approaches μf , eigenvectors start to delocalize but , as per Equation 4 , the maximum peak is sharper . β2 is the steepness of the exponential ( Equation 5 ) . The network is described by Equation 3 with μ0 = −1 . 9 , Δr = 0 . 01 , μf = 0 . 2 , and lc = 4 . μb = 0 . 125 , 0 . 15 , 0 . 175 , and 0 . 19 . DOI: http://dx . doi . org/10 . 7554/eLife . 01239 . 007 We extend our expansion to second-order in vλ ( appendix [Supplementary file 1] , Sections 5 & 6 ) to predict that the eigenvector is given by ( 5 ) vλ ( j ) =eβ2 ( j−j0 ) Ai ( β1 ( j−j0 ) +β22β12/3 ) eiωjwith ( 6 ) β1=Δrcsch ( 12lc ) 4sinh ( 1lc ) 3 ( μf+μb ) and β2= ( μf−μb ) coth ( 12lc ) ( μf+μb ) where , Ai is the first Airy function ( Olver , 2010 ) . The eigenvector is the product of an exponential and an Airy function and this product is localized when the exponential is steep ( Figure 4A ) . The steepness of the exponential depends on μf − μb . When this difference is small the exponential is shallow and the trailing edge of the product is poorly localized . Figure 4B shows that this functional form accurately predicts the results from numerical simulations , except when the eigenvector is almost completely delocalized . These results reveal that an asymmetry in the strength of feedforward and feedback projections can play an important role in segregation of timescales in biological systems . The second-order expansion demonstrates that the approach is general and can be extended as needed . While the first-order expansion in vλ generically gives rise to modulated Gaussians , the functional form of the eigenvectors from a second-order expansion depends on the connectivity ( appendix [Supplementary file 1] , Section 5 ) and , in general , the asymptotic decay is slower than that of a Gaussian . The previous architecture was a chain of nodes with identical inter-node connectivity but varying local connectivity . We now consider a contrasting architecture: a chain with no self-coupling but with a location-dependent bias in inter-node connectivity . We build this model motivated by the intuitive notion that nodes near the input end of a network send mostly feedforward projections , while nodes near the output send mostly feedback projections . The network architecture is shown in Figure 5A . 10 . 7554/eLife . 01239 . 008Figure 5 . Localized eigenvectors in a network with a gradient of connectivity range . ( A ) The network consists of a chain of 50 identical nodes , shown here by a schematic . Spatial length of feedforward connections ( from earlier to later nodes ) decreases along the chain while the spatial length of feedback connections ( from later to earlier nodes ) increases along the chain . The network is described by Equation 7 , with μ0 = −1 . 05 , μf = 5 , μb = 0 . 5 , f0 = 0 . 2 , f1 = 0 . 12 , b0 = 6 , b1 = 0 . 11 . Normally-distributed randomness of standard deviation σ = 10−5 is added to all connections . ( B ) Five sample eigenvectors , with numerical simulations ( filled circles ) well fitted by the analytical predictions ( solid lines ) . Note the effect of added randomness on the rightmost eigenvector . ( C ) Heat map of eigenvectors on logarithmic scale . Rows correspond to eigenvectors , arranged by increasing decay time . All eigenvectors are localized , but timescales are not monotonically related to eigenvector position . ( D ) Dynamical response of the network to an input pulse . Long timescales are localized to nodes early in the network while nodes later in the network show intermediate timescales . DOI: http://dx . doi . org/10 . 7554/eLife . 01239 . 008 Connectivity decays exponentially , as in the previous example , but the decay length depends on position . Moving along the chain , feedforward decay length decreases while feedback decay length increases: ( 7 ) W ( j , k ) ={μ0for j=k ( self-coupling ) μfe− ( f0+f1k ) ( j−k ) for j>k ( feedforward connections ) μbe ( b0−b1k ) ( j−k ) for j<k ( feedback connections ) The parameters f0 , f1 , b0 , and b1 control the location-dependence in decay length , μ0 is the leakage term , and μf and μb set the maximum strength of feedforward and feedback projections . We also add a small amount of randomness to the connection strengths . As before we calculate the eigenvector width , α . In this case , for a wide range of the parameters in Equation 7 , α2 is positive and approximately constant for all eigenvectors . Therefore , all eigenvectors are localized and have approximately the same width ( appendix [Supplementary file 1] , Section 4 ) . Four eigenvectors are plotted in Figure 5B along with theoretical predictions . Figure 5C shows all of the eigenvectors on a heat map and demonstrates that all are localized . The fastest and slowest timescales are localized to the earlier nodes while the intermediate timescales are localized towards the end of the chain . The earlier nodes thus show a combination of very fast and very slow time courses , whereas the later nodes display dynamics with an intermediate range of timescales . Such dynamics present a salient feature of networks with opposing gradients in their connectivity profile . In Figure 5D , we plot the decay of network activity from a uniform initial condition; note the contrast between nodes early and late in the chain . While the eigenvectors are all localized , different eigenvectors tend to cluster their centers near similar locations . Near those locations , nodes may participate in multiple eigenvectors , implying that time constants are not well segregated . This is a consequence of the architecture: nodes towards the edges of the chain project most strongly towards the center , so that small perturbations at either end of the chain are strongly propagated inward . The narrow spread of centers ( the overlap of multiple eigenvectors ) reduces the segregation of timescales that is one benefit of localization . We find that adding a small amount of randomness to the system spreads out the eigenvector centers without significantly changing the shape . This approach is more robust than fine-tuning parameters to maximally spread the centers , and seems reasonable in light of the heterogeneity intrinsic to biological systems ( Raser and O’Shea , 2005; Barbour et al . , 2007 ) . Upon adding randomness , most eigenvectors remain Gaussian while a minority are localized but lose their Gaussian shape . The significant overlap of the eigenvectors means that the eigenvectors are far from orthogonal to each other . Such matrices , called non-normal matrices , can show a number of interesting transient effects ( Trefethen and Embree , 2005; Goldman , 2009; Murphy and Miller , 2009 ) . In particular we note that the dynamics of our example network show significant initial growth before decaying , as visible in the scale of Figure 5D . As observed in the last section , the heterogeneity intrinsic to biological systems can play a beneficial role in computation . Indeed , sufficient randomness in local node properties has been shown to give localized eigenvectors in models of physical systems with nearest-neighbor connectivity , and the transition from delocalized to localized eigenvectors has been suggested as a model of the transition from a conducting to an insulating medium ( Anderson , 1958; Abou-Chacra et al . , 1973; Lee , 1985 ) . A similar mechanism should apply in biological systems . We numerically explore eigenvector localization in a network with exponentially-decaying connectivity and randomly distributed self-couplings . The network connection matrix is given by ( 8 ) W ( j , k ) ={μ0+N ( 0 , σ2 ) for j=kμce−|j−k|/lcfor j≠k , where , N ( 0 , σ2 ) is drawn from a normal distribution with mean zero and variance σ2 . As σ2 increases , the network shows a transition to localization . This transition is increasingly sharp and occurs at lower values of σ as the network gets larger . Figure 6 shows a network with sufficient randomness for the eigenvectors to localize , with sample eigenvectors shown in Figure 6B . These show a variety of shapes and are no longer well described by Gaussians . Importantly , there is no longer a relationship between the location of an eigenvector and the timescale it corresponds to ( Figure 6C ) . Thus while each timescale is localized , a variety of timescales are present in each region of the network , and each node will show a random mixture of timescales . This is in contrast to our previous examples , which have a spatially continuous distribution of time constants . The random distribution of time constants is also observed in the decay from a uniform initial conditions , as shown in Figure 6D . 10 . 7554/eLife . 01239 . 009Figure 6 . Localized eigenvectors in a network with random self-coupling . ( A ) The network consists of 100 nodes arranged in a chain . The plot above the chain shows the connectivity profile . Self-coupling is random , as indicated by the shading . The network is described by Equation 8 with μ0 = −1 , μc = 0 . 05 , lc = 4 , σ = 0 . 33 . ( B ) Four eigenvectors are shown , localized to different parts of the network . Note the diversity of profiles . ( C ) Heat map of eigenvectors on logarithmic scale . Rows correspond to eigenvectors , arranged by increasing decay time . All eigenvectors are localized , though the extent of localization ( the eigenvector width ) varies; and there is no relationship between the timescale of an eigenvector and its spatial location in the network . ( D ) Dynamical response of the network to an input pulse . Note that the diversity of dynamical responses is more limited , and bears no relationship to spatial location . DOI: http://dx . doi . org/10 . 7554/eLife . 01239 . 009
Local connectivity is insufficient to create localized temporal patterns of activity in linear networks . A network with sharply localized but translationally invariant connectivity has delocalized eigenvectors . This implies that distant nodes in the network have similar temporal activity , since they share the timescales of their dynamics . Breaking the invariance can give rise to localized eigenvectors , and we study conditions that allow this . We develop a theory to predict the shapes of localized eigenvectors and our theory generalizes to describe eigenvectors that are only partially localized and show multiple peaks . A major finding of this study is the identification of two network architectures , with either a gradient of local connectivity or a gradient of long-distance connection length , that give rise to activity patterns with localized timescales . Our approach to eigenvector localization is partly based on Trefethen and Embree ( 2005 ) ; Trefethen and Chapman ( 2004 ) . The authors study perturbations of translationally invariant matrices and determine conditions under which eigenvectors are localized in the large-N limit . We additionally assume that the connectivity is local , since we are interested in matrices that describe connectivity of biological networks . This allows us to calculate explicit functional forms for the eigenvectors . We stress that the temporal aspect of the network dynamics should not be confused with selectivity across space in a neural network . Even if temporal patterns are localized , a large proportion of network nodes may be active in response to a given input , albeit with distinct temporal dynamics . Conversely , even if temporal patterns are delocalized , nodes show similar dynamics yet may still be highly selective to different inputs and any stimulus could primarily activate only a small fraction of nodes in the network . Our results are particularly relevant to understanding networks that need to perform computations requiring a wide spread of timescales . In general , input along a fast eigenvector decays exponentially faster than input along a slow eigenvector . To see this , consider a network with a fast and a slow timescale ( 1/|λfast| and 1/|λslow| ) , and having initial condition with components afast and aslow along the fast and the slow eigenvectors respectively . As shown in Equation 11 , the network activity will evolve as afaste−|λfast|t+aslowe−|λslow|t . For a node to show a significant fast timescale in the presence of a slower , more persistent timescale , the contribution of this slow timescale to the node must be small . This can happen in two ways , corresponding to the terms of Equation 10 . If the input contributes little to the slower eigenvectors then their amplitudes will be small at all nodes . This requires fine-tuned input ( exponentially smaller along the slow eigenvectors ) and means that the slow timescales do not contribute significantly to any node . Alternately , as in the architectures we propose , the slow eigenvectors could be exponentially smaller at certain nodes; these nodes will then show fast timescales for most inputs , with a small slow component . The architecture with a gradient of local connectivity ( Figure 3 ) may explain some observations in the larval zebrafish oculomotor system ( Miri et al . , 2011 ) . The authors observed a wide variation in the time constants of decay of firing activity across neurons , with more distant neurons showing a greater difference in time constants . They proposed a model characterized by a chain of nodes with linearly-decaying connectivity and a gradient of connection strengths , and found that different nodes in the model showed different timescales . Furthermore , the introduction of asymmetry to connectivity ( with feedback connections weaker than feedforward connections ) enhanced the diversity of timescales . This effect of asymmetry was also seen in an extension of the model to the macaque monkey oculomotor integrator ( Joshua et al . , 2013 ) . Our work explains why such architectures allow for a diversity of timescales , and we predict that such gradients and asymmetry should be seen experimentally . With a gradient of local connections , time constants increase monotonically along the network chain . By contrast , with a gradient of connectivity length ( Figure 5 ) , the relationship between timescales and eigenvector position is lawful but non-monotonic , as a consequence of the existence of two gradients ( feedforward connectivity decreases while feedback increases along the chain ) . The small amount of randomness added to this system helps segregate the timescales across the network , while only mildly affecting the continuous dependence of eigenvector position on timescale . This suggests that randomness may contribute to a diversity of timescales . The connection between structural randomness and localization is well known in physical systems ( Anderson , 1958; Abou-Chacra et al . , 1973; Lee , 1985 ) . We applied this idea to a biological context ( Figure 6 ) , and showed that localization can indeed emerge from sufficiently random node properties . However , in this case nearby eigenvectors do not correspond to similar timescales . A given timescale is localized to a particular region of the network but a similar timescale could be localized at a distant region and , conversely , a much shorter or longer timescale could be localized in the same part of the network . Thus , the timescales shown by a particular node are a random sample of the timescales of the network . Chemical gradients are common in biological systems , especially during development ( Wolpert , 2011 ) , and structural randomness and local heterogeneity are ubiquitous . We predict that biological systems could show localized activity patterns due to either of these mechanisms or a combination of the two . Furthermore , local randomness can enhance localization that emerges from gradients or long-range spatial fluctuations in local properties . We have focused on localization that yields a smooth relationship between timescale and eigenvector position; such networks are well-placed to integrate information at different timescales . However , it seems plausible that biological networks have evolved to take advantage of randomness-induced localization , and it would be interesting to explore the computational implications of such localization . It could also be fruitful to explore localization from spatially correlated randomness . An influential view of complexity is that a complex network combines segregation and integration: individual nodes and clusters of nodes show different behaviors and subserve different functions; these behaviors , however , emerge from network interactions and the computations depend on the flow of information through the network ( Tononi and Edelman , 1998 ) . The localized activity patterns we find are one way to construct such a network . Each node participates strongly in a few timescales and weakly in the others , but the shape and timescales of the activity patterns emerge from the network topology as a whole and information can flow from one node to another . Moreover , as shown in Figure 7 , adding a small number of long-range strong links to local connectivity , as in small-world networks ( Watts and Strogatz , 1998 ) , causes a few eigenvectors to delocalize while leaving most localized . This is a possible mechanism to integrate computations while preserving segregated activity , and is an interesting direction for future research . 10 . 7554/eLife . 01239 . 010Figure 7 . Strong long-range connections can delocalize a subset of eigenvectors . ( A ) Left panel: connectivity of the network in Figure 3 with long-range connections of strength 0 . 05 added between 10% of the nodes . The gradient of self-coupling is shown along the diagonal on another scale , for clarity . Right panel: eigenvectors shown as in panel C of Figure 3 . ( B ) Left panel: connectivity of the network in Figure 5 with long-range connections of strength 0 . 05 added between 10% of the nodes . Right panel: eigenvectors shown as in panel C of Figure 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 01239 . 010
We rewrite the connectivity matrix in terms of a relative coordinate , p = j−k , as ( 12 ) W ( j , k ) =c ( j , j−k ) . Thus , c ( j , 2 ) = W ( j , j − 2 ) indexes feedforward projections that span two nodes , and c ( 5 , p ) = W ( 5 , 5 − k ) indexes projections to node 5 . Note that in the translation-invariant case , c ( j , p ) would be independent of j ( appendix [Supplementary file 1] , Section 1 ) , while the requirement of local connectivity means that c ( j , p ) is small away from p = 0 . For any fixed j , c ( j , p ) is defined from p = j − N to p = j − 1 . We extend the definition of c ( j , p ) to values outside this range by defining c ( j , p ) to be periodic in p , with the period equal to the size of the network . This is purely a formal convenience to simplify the limits in certain sums and does not constrain the connectivity between the nodes of the network . Consider the candidate eigenvector vλ ( j ) = gλ ( j ) eiωj . The dependence of gλ on j allows the magnitude of the eigenvector to depend on position; setting this function equal to a constant returns us to the translation-independent case ( see appendix [Supplementary file 1] , Section 1 ) . Moreover , note that gλ ( j ) depends on λ , meaning that eigenvectors corresponding to different eigenvalues ( timescales ) can have different shapes . For example , different eigenvectors can be localized to different degrees , and localized and delocalized eigenvectors can coexist ( see Figure 3—figure supplement 1 for an illustration ) . ω allows the eigenvector to oscillate across nodes; it varies between eigenvectors and so depends on λ . Applying W to vλ yields ( 13 ) [Wvλ] ( j ) =∑k=1NW ( j , k ) gλ ( k ) eiωk=∑k=1Nc ( j , j−k ) gλ ( k ) eiωk ( 14 ) = ( ∑p=j−Nj−1c ( j , p ) gλ ( j−p ) e−iωp ) eiωj , here , the term in brackets is no longer independent of j . So far we have made no use of the requirement of local connectivity and , given that gλ is an arbitrary function of position and can be different for different timescales , we have placed no constraints on the shape of the eigenvectors . By including an oscillatory term ( eiωj ) in our ansatz , we ensure that gλ ( j ) is constant when connectivity is translation-invariant; this will simplify the analysis . We now approximate both c ( j , p ) and gλ ( j − p ) to first-order ( i . e . , linearly ) :c ( j , p ) ≈c ( j0 , p ) +∂c∂j|j0 , p ( j−j0 ) ( 15 ) gλ ( j−p ) ≈gλ ( j ) −g'λ ( j ) p , where , j0 is a putative center of the eigenvector . Substituting Equation 15 into Equation 14 we get ( 16 ) [Wvλ] ( j ) = ( ∑p=j−Nj−1[c ( j0 , p ) +∂c∂j|j0 , p ( j−j0 ) ][gλ ( j ) −g'λ ( j ) p]e−iωp ) eiωj We expect these approximations to be valid only locally . However , if connectivity is local then the major contribution to the sum comes from small values of p . For large values of p , gλ ( j − p ) is multiplied by connectivity strengths close to 0 and so we only need to approximate gλ for p close to 0 . Similarly , in approximating c ( j , p ) around j = j0 , we expect our approximation to be good in the vicinity of j = j0 . However , if our eigenvector is indeed localized around j0 , then gλ ( k ) is small when |k−j0| is large . For small p , large values of |k−j0| approximately correspond to large values of |j−j0| , and so c ( j , p ) makes a contribution to the sum only when j ≈ j0 . The zeroth-order term in Equation 16 is ( ∑p=j−Nj−1c ( j0 , p ) e−iωp ) gλ ( j ) eiωj=λ ( j0 , ω ) vλ ( j ) The function in parentheses is periodic in p with period N ( recall that c ( j , p ) was extended to be periodic in p ) . Thus to zeroth-order vλ is an eigenvector with eigenvalue ( 17 ) λ ( j0 , ω ) =∑p=1Nc ( j0 , p ) e−iωp . For λ to be an exact eigenvalue in Equation 16 , the higher-order terms should vanish . By setting the first-order term in this equation to 0 , we obtain a differential equation for gλ ( j ) : ( 18 ) −α ( j0 , ω ) 2g'λ ( j ) = ( j−j0 ) gλ ( j ) where , ( 19 ) α ( j0 , ω ) 2=−∑ppc ( j0 , p ) e−iωp∑p∂c∂j|j0 , pe−iωp . Thus α2 is a ratio of discrete Fourier transforms at the frequency ω . Note that the denominator is a weighted measure of network heterogeneity at the location j0 . Also note that α2 can be written in terms of λ as ( compare the twist condition of Trefethen and Embree , 2005 ) : ( 20 ) α2 ( j0 , ω ) =−i∂λ∂ω∂λ∂j0 . Solving for gλ in Equation 18 yieldsgλ ( j ) =C1e− ( j−j0 ) 22α ( j0 , ω ) 2 , where , C1 is a constant . Thus , to first-order , the eigenvector is given by the modulated Gaussian function ( 21 ) vλ ( j ) =e− ( j−j0 ) 22α ( j0 , ω ) 2+iωj . In general , α can be complex . In order for vλ to be localized , Re ( α2 ) must be positive for the corresponding values of j0 and ω , and we only accept an eigenvector as a valid solution if this is the case . Thus the approach is self-consistent: we assumed that there existed a localized eigenvector , combined this with the requirement of local connectivity to solve for its putative shape , and then restricted ourselves to solutions that did indeed conform to our initial assumption . For an expanded version of this analysis along with further discussion of what the analysis provides , see the appendix ( Supplementary file 1 ) , Section 2 . | Many biological systems can be thought of as networks in which a large number of elements , called ‘nodes’ , are connected to each other . The brain , for example , is a network of interconnected neurons , and the changing activity patterns of this network underlie our experience of the world around us . Within the brain , different parts can process information at different speeds: sensory areas of the brain respond rapidly to the current environment , while the cognitive areas of the brain , involved in complex thought processes , are able to gather information over longer periods of time . However , it has been largely unknown what properties of a network allow different regions to process information over different timescales , and how variations in structural properties translate into differences in the timescales over which parts of a network can operate . Now Chaudhuri et al . have addressed these issues using a simple but ubiquitous class of networks called linear networks . The activity of a linear network can be broken down into simpler patterns called eigenvectors that can be combined to predict the responses of the whole network . If these eigenvectors ‘map’ to different parts of the network , this could explain how distinct regions process information on different timescales . Chaudhuri et al . developed a mathematical theory to predict what properties would cause such eigenvectors to be separated from each other and applied it to networks with architectures that resemble the wiring of the brain . This revealed that gradients in the connectivity across the network , such that nodes share more properties with neighboring nodes than distant nodes , combined with random differences in the strength of inter-node connections , are general motifs that give rise to such separated activity patterns . Intriguingly , such gradients and randomness are both common features of biological systems . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] | [
"neuroscience"
] | 2014 | A diversity of localized timescales in network activity |
The controversy surrounding 'gain-of-function' experiments on high-consequence avian influenza viruses has highlighted the role of ferret transmission experiments in studying the transmission potential of novel influenza strains . However , the mapping between influenza transmission in ferrets and in humans is unsubstantiated . We address this gap by compiling and analyzing 240 estimates of influenza transmission in ferrets and humans . We demonstrate that estimates of ferret secondary attack rate ( SAR ) explain 66% of the variation in human SAR estimates at the subtype level . Further analysis shows that ferret transmission experiments have potential to identify influenza viruses of concern for epidemic spread in humans , though small sample sizes and biological uncertainties prevent definitive classification of human transmissibility . Thus , ferret transmission experiments provide valid predictions of pandemic potential of novel influenza strains , though results should continue to be corroborated by targeted virological and epidemiological research .
The emergence of deadly animal-origin influenza viruses in human populations , such as influenza A ( H5N1 ) ( Chan , 2002; Li et al . , 2004 ) , and influenza A ( H7N9 ) ( Gao et al . , 2013; Gong et al . , 2014; Li et al . , 2014 ) , has underscored the need to rapidly determine the pandemic potential of novel strains found in humans or in zoonotic reservoirs . Although characterizing human transmissibility of emerging influenza viruses is a perpetual challenge , animal models are often used to characterize transmission among mammals , which can be viewed implicitly as a preliminary screen for pandemic potential in humans . Ferrets are the preferred animal model for influenza transmission studies because clinical signs , pathogenesis and sialic acid distribution are similar in ferrets and humans ( Maher and DeStefano , 2004; Shinya et al . , 2006; Bouvier and Lowen , 2010 ) . Consequently , the ferret model has been used to assess numerous aspects of influenza transmission potential including: phenotypic traits associated with transmission ( Belser et al . , 2008 , 2013; Song et al . , 2009; van Doremalen et al . , 2011; Blumenkrantz et al . , 2013 ) , transmission under antiviral prophylaxis ( Oh et al . , 2014 ) , and the relative transmissibility of drug resistant ( Herlocher et al . , 2004; Hurt et al . , 2010; Kiso et al . , 2010; Seibert et al . , 2010; Duan et al . , 2011; Hamelin et al . , 2011 ) , emerging ( Maines et al . , 2006; Itoh et al . , 2009; Belser et al . , 2013; The SJCEIRS Working Group , 2013; Watanabe et al . , 2013; Zhu et al . , 2013; Xu et al . , 2014 ) , or lab-created isolates ( Herfst et al . , 2012; Imai et al . , 2012; Sutton et al . , 2014 ) . Despite the widespread use of ferrets to assess transmission of influenza , the suitability of ferrets to assess pandemic potential in humans remains unknown , because the relationship between transmission in ferrets and in humans has never been assessed quantitatively ( Palese and Wang , 2012; Casadevall and Imperiale , 2014; Lipsitch , 2014 ) . In fact , conspicuous differences in ferret and human transmissibility for influenza A ( H7N9 ) have cast doubt on the validity of the ferret model for assessing transmission in humans ( Lipsitch , 2013 ) . As a consequence , ferret studies can only be interpreted , strictly , in terms of general mammalian transmissibility ( Herfst et al . , 2012; Imai et al . , 2012; Casadevall and Imperiale , 2014; Casadevall et al . , 2014 ) . Furthermore , the recent controversy surrounding ‘gain-of-function’ ( GOF ) experiments on highly pathogenic avian influenza A ( H5N1 ) in ferrets ( Herfst et al . , 2012; Imai et al . , 2012 ) and proposed GOF experiments on A ( H7N9 ) viruses ( Fouchier et al . , 2013 ) has led to ethical questions about influenza GOF experiments and scientific questions about the use of ferrets to assess transmission ( Morens et al . , 2012; Casadevall and Imperiale , 2014; Casadevall et al . , 2014; Lipsitch , 2014; Lipsitch and Galvani , 2014; Russell et al . , 2014 ) . With the U . S . government halting funding and calling for a voluntary moratorium and period of review on such experiments as of October 2014 ( White House Office of Science and Technology Policy , 2014 ) , groups on all sides of the debate have issued renewed calls for studies on the link between influenza transmissibility in ferrets and in humans ( Morens et al . , 2012; Lipsitch , 2013 , 2014; Casadevall and Imperiale , 2014 ) . Here , we address this gap by compiling ferret transmission studies and comparing their results to estimates of influenza transmission in humans .
To assess the quantitative relationship between influenza transmission in ferrets and in humans , we assembled data from all published ferret transmission studies that met our inclusion criteria , including ferret experiments designed to test transmission in the presence of direct contact ( co-housing ) or by respiratory droplets ( adjacent housing allowing air exchange ) . For each experiment , we calculated the secondary attack rate ( SAR ) , which is defined as the probability of infection for a susceptible individual following known contact with an infectious individual ( Halloran , 2005 ) . To match the close contact found in ferret studies , we reviewed estimates of SAR in humans obtained from household contact data ( Figure 1 ) . 10 . 7554/eLife . 07969 . 003Figure 1 . Boxplots of influenza SAR estimates by subtype . ( A ) Human SAR , ( B ) ferret respiratory droplet SAR , and ( C ) ferret direct contact SAR . Solid , black lines represent the subtype medians . Boxes give the inter-quartile range with whiskers extending out up to 1 . 5 times this range . Points represent extreme values . The number of estimated SARs for each subtype is given above each box-and-whisker plot ( n ) . Subtypes were ordered according to the mean human SAR value in all panels . Shading depicts the known human transmission pattern of the subtypes ( red—supercritical; blue—subcritical ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 00310 . 7554/eLife . 07969 . 004Figure 1—source data 1 . Estimates of human household SAR . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 00410 . 7554/eLife . 07969 . 005Figure 1—source data 2 . Ferret influenza transmission studies via respiratory droplets using human isolates . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 00510 . 7554/eLife . 07969 . 006Figure 1—source data 3 . Ferret influenza transmission studies via direct contact using human isolates . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 00610 . 7554/eLife . 07969 . 007Figure 1—source data 4 . Ferret influenza transmission studies via respiratory droplets and direct contact using avian isolates . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 00710 . 7554/eLife . 07969 . 008Figure 1—figure supplement 1 . Analysis supporting inclusion of SAR estimates from isolates generating using reverse genetics . Comparison of ferret SAR for wild-type influenza isolates and their counterparts engineered using reverse genetics under ( A ) respiratory droplet and ( B ) direct contact transmission . Because experiments were not paired , SAR estimates for a wild-type isolate were plotted against the mean SAR for the reverse genetic derived isolate and vice-versa . Estimates from the same isolate are joined by a line with the isolate name given . Both supercritical ( red ) and subcritical ( blue ) isolates are shown . The dashed gray line denotes a one-to-one relationship between the two . Note that some points are jittered for clarity ( see Figure 1—source data 2 , 3 for full data ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 008 When comparing estimates of human and ferret SAR across subtypes , we found that , as expected ( Lakdawala and Subbarao , 2012 ) , ferret SAR estimates from current experimental designs do not quantitatively align with human SAR estimates—ferret SAR estimates are typically higher than the corresponding human estimate . However , ferret and human SAR estimates are correlated . For respiratory droplet experiments , the ordering of subtypes by ferret SAR was similar to that in human SAR ( Figure 1A , B ) , and mean ferret respiratory droplet SAR explained 66% of the variation in mean human SAR estimates across subtypes ( p = 0 . 003 , Figure 2A ) . Direct contact transmission in ferrets was not significantly related to human SAR at the subtype level ( p = 0 . 14 , Figure 2A ) , suggesting that for estimates of human-to-human transmissibility , direct contact experiments may have less value than respiratory droplet experiments . 10 . 7554/eLife . 07969 . 009Figure 2 . Analysis of subtype-specific SAR . ( A ) Comparison of human SAR and ferret SAR for ferret respiratory droplet ( black squares ) and direct contact ( red circles ) . Data points are the mean human SAR by subtype vs the weighted mean ferret SAR by subtype , where weights are determined by the number of ferrets used in each experiment . Lines give the best fit weighted linear regression models with weights given by the number of human SAR estimates . The solid line indicates a significant relationship between ferret respiratory droplet SAR and human SAR described by the given equation ( significant terms are bolded; p = 0 . 003 ) , while the dashed line indicates a non-significant relationship ( p = 0 . 14 ) for ferret direct contact transmission . ( B ) The degree of overlap in the distributions of ferret respiratory droplet SAR estimates for each subtype . Dark purple indicates subtypes with complete overlap , while white indicates no overlap . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 00910 . 7554/eLife . 07969 . 010Figure 2—figure supplement 1 . Analysis of subtype-specific SAR including avian isolates for H5N1 and H7N9 . ( A ) Comparison of human SAR and ferret SAR for ferret respiratory droplet ( black squares ) and direct contact ( red circles ) . Lines give the best fit weighted linear regression models with weights given by the number of human SAR estimates . The solid line indicates a significant relationship between ferret respiratory droplet SAR and human SAR described by the given equation ( significant terms are bolded; p = 0 . 004 ) , while the dashed line indicates a non-significant relationship ( p = 0 . 18 ) for ferret direct contact transmission . ( B ) The degree of overlap in the distributions of ferret respiratory droplet SAR estimates for each subtype . Dark purple indicates subtypes with complete overlap , while white indicates no overlap . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 010 Despite the strong relationship observed between mean ferret and human SAR estimates ( Figure 2A ) , distributions of ferret SAR estimates for each subtype overlapped substantially ( Figure 2B ) . These overlaps prevent the result from any given ferret experiment ( e . g . , on a novel , uncharacterized strain ) from being unambiguously aligned with the transmission potential exhibited by any particular , previously-characterized subtype . To improve the power to assess pandemic potential , we specified two clusters of subtypes with distinct transmission patterns in humans: subtypes with sustained human-to-human transmission ( i . e . supercritical; H1N1 , H3N2 , H2N2 and pH1N1 ) and subtypes without sustained human-to-human transmission ( i . e . subcritical; H7N9 , H5N1 , H7N7 , H7N2 , H7N3 and H9N2 ) . Using logistic regression , we identified ranges of ferret SAR that characterize supercritical and subcritical influenza viruses ( Figure 3 ) . Ferret respiratory droplet SAR was a significant predictor of the probability that a virus is supercritical or subcritical in humans ( p < 0 . 0001; Figure 3A , Table 1 ) . By accounting for the uncertainty in this relationship , we identified ranges of ferret SAR that indicate a high probability of strains being identified as supercritical or subcritical ( Figure 3A ) . However , a range of intermediate ferret SAR values yielded equivocal results ( i . e . the 95% confidence interval for classification included a classification probability of 0 . 5 ) . Direct contact transmission was also a significant predictor of supercritical or subcritical transmission in humans ( p = 0 . 01; Figure 3B , Table 1 ) . Information theoretic model comparisons showed marginal support for a bivariate model using both respiratory droplet and direct contact transmission data ( Table 1 ) . Considering the bivariate distribution of SAR estimates , however , it is clear that respiratory droplet SAR has the potential for greater specificity in predicting supercritical transmission ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 07969 . 011Figure 3 . Weighted logistic regression predicting the probability of a supercritical classification based on ferret SAR . ( A ) Respiratory droplet SAR and ( B ) direct contact SAR . Solid black line gives the fit of the weighted logistic regression , where model weights are given by the number of ferrets in each experiment . Dashed black lines give the 95% confidence interval on the model predictions . Shading in the prediction interval represents values of SAR for which the 95% confidence intervals for predicted model fit do not overlap a probability of 0 . 5 ( the dashed red line ) indicating a high probability of being supercritical ( red shading ) or subcritical ( blue shading ) . The gray shading represents SAR values where the 95% CI on the prediction overlaps 0 . 5 , providing equivocal classification . Circles show the individual ferret SAR estimates ( See Figure 1—source data 2 , 3 ) for supercritical ( top in red ) and subcritical viruses ( bottom in blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 01110 . 7554/eLife . 07969 . 012Figure 3—figure supplement 1 . Comparison of ferret SAR via respiratory droplet and direct contact transmission for single influenza isolates . Each point represents a single set of experiments that tested an isolate for transmission in ferrets under both respiratory droplet and direct contact transmission with other experimental protocols held fixed . Isolates belonging to subcritical subtypes are depicted by blue squares , and supercritical subtypes are depicted by red circles . Note that some points are jittered for clarity ( see Figure 1—source data 2 , 3 for full data ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 01210 . 7554/eLife . 07969 . 013Figure 3—figure supplement 2 . Effect of uncertainty in ferret SAR on its relationship with the probability of being classified as supercritical . ( A ) Respiratory droplet SAR and ( B ) direct contact SAR . To assess the impact of binomial uncertainty in ferret SAR estimates , we simulated 1000 datasets by taking binomial samples from each data point . Here , the binomial probability for each was given by the observed SAR and the number of trials was the number of ferrets used . To introduce binomial uncertainty into those experiments with an SAR of 0 or 1 , we set the binomial probability to 0 . 1 or 0 . 9 , respectively . The solid line is the average model fit to all of the simulated datasets and is nearly identical to that in Figure 3 . Dashed lines give the 97 . 5 and 0 . 025 percentiles of the upper and lower bounds , respectively , of the 95% confidence intervals on model predictions from each of the simulated datasets . These indicate much more uncertainty in model predictions across datasets that generates a larger equivocal region of ferret SARs than observed in Figure 3 . However , values of ferret SAR indicative of subcritical and supercritical strains still exist , indicating that our qualitative results are robust to binomial uncertainty . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 01310 . 7554/eLife . 07969 . 014Figure 3—figure supplement 3 . ROC curves for classifying pandemic potential using different definitions of transmission and transmission routes . Receiver operating characteristic ( ROC ) curves and area under the curve ( AUC ) using ( A ) seroconversion and/or viral isolation or ( B ) viral isolation alone as evidence for transmission in ferrets when classifying influenza isolates as either supercritical or subcritical in humans . Lines indicate ferret respiratory droplet SAR ( red ) or ferret direct contact SAR ( black ) . Curves were calculated from raw data shown in Figure 3 , using a range of SAR classification thresholds from 0 to 1 . Numbers indicate the threshold values for which the true positive rate ( i . e . the sensitivity ) and false positive rate ( i . e . the complement of the specificity ) changed . Threshold values intermediate to any of those depicted have true positive and false positive rates identical to that of the next lowest value shown . The dashed gray line is the one-to-one line corresponding to random classification . AUC values are shown in the figure legend with higher values corresponding to higher predictive power . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 01410 . 7554/eLife . 07969 . 015Table 1 . Parameter estimates for the weighted logistic regression relating human transmission class to ferret SARDOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 015DataModelβ0βRDβDCΔAICFull dataDirect contact−4 . 39-6 . 30-Respiratory droplet−3 . 526 . 10--Restricted dataRespiratory droplet + direct contact−1 . 768 . 72−3 . 760Respiratory droplet−3 . 776 . 42-3 . 623Direct contact−3 . 07-3 . 7457 . 348Bolded estimates are significant at the α = 0 . 05 level . Due to differing data between ferret respiratory droplet and direct contact transmission experiments , no model selection was done on the full data . Instead , model selection was done only for studies where authors performed respiratory droplet and direct contact transmission experiments on the same isolate . The classification thresholds we identified for likely supercritical or subcritical subtypes account for uncertainties arising from the structure of our model , but not for uncertainties arising from the experimental data used to inform the model . Binomial uncertainties in ferret SAR data can be substantial , as ethical and logistic considerations limit sample sizes in these experiments ( Nishiura et al . , 2013 ) . By re-fitting our logistic regression model to 1000 simulated datasets generated by binomial re-sampling of each data point , we found that the relationship between ferret SAR and a supercritical classification is quite robust to this uncertainty ( Figure 3—figure supplement 2 ) . However , while our analysis was fairly insensitive to binomial uncertainty within the aggregate data , attempts to classify SAR estimates from any individual experiment will be more sensitive to binomial uncertainty . For example , we applied our model to the most transmissible strains from two recent GOF studies on H5N1 avian influenza ( Imai et al . , 2012; Herfst et al . , 2012; Figure 4—source data 1 ) . All three strains had a ferret SAR that fell into the supercritical range , but the confidence intervals for the SAR estimates overlapped with the subcritical and/or equivocal ranges , preventing definitive classification ( Figure 4A ) . Similarly , we found that studies on 1918 pandemic H1N1 , a known pandemic strain , had ferret SAR estimates indicative of supercritical transmission , but again wide confidence intervals overlapped the subcritical and equivocal ranges ( Figure 4B ) . SAR estimates for H7N9 , known to be subcritical in humans , spanned the supercritical , subcritical , and equivocal ranges ( Figure 4C ) . Even if results across all ferret respiratory droplet trials for H7N9 were aggregated into a single SAR estimate ( representing 42 ferrets in all ) , we found an equivocal classification of human transmission pattern ( Figure 4C ) . Consequently , care must be taken to avoid over-interpreting the results of ferret transmission studies . 10 . 7554/eLife . 07969 . 016Figure 4 . Predictions of the transmission pattern for current and historical isolates of concern . ( A ) Gain-of-function experiments with H5N1 avian influenza ( Herfst et al . , 2012; Imai et al . , 2012 ) , ( B ) the reconstructed 1918 pandemic H1N1 strain ( Tumpey et al . , 2007; Imai et al . , 2012 ) , and ( C ) H7N9 avian influenza . Solid black curves and shading represent the logistic regression fit and likely transmission pattern , respectively , as depicted in Figure 2 . Horizontal lines give the 95% Wilson-score interval for each binomial estimate . In all panels , transmission is defined using seroconversion and viral isolation in nasal washes . In ( C ) , green triangles represent individual experiments , while the green square is the aggregated data across all twelve H7N9 transmission experiments in ferrets . Notice that 6 data points are represented at a SAR of 0 . 33 and 3 at a SAR of 1 . See Figure 1—source data 2 and Figure 4—source data 1 for full data . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 01610 . 7554/eLife . 07969 . 017Figure 4—source data 1 . Ferret influenza transmission studies via respiratory droplets using strains from gain-of-function experiments with H5N1 avian influenza and the reconstructed 1918 pandemic H1N1 strain . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 01710 . 7554/eLife . 07969 . 018Figure 4—figure supplement 1 . Sample size calculations . Sample size estimates to achieve 80% power at a significance level of 0 . 05 when testing whether a ferret SAR estimate is greater than the lower limit of the identified supercritical window ( 0 . 643 , Figure 3A ) . Sample sizes were calculated using a one-sided binomial exact test . DOI: http://dx . doi . org/10 . 7554/eLife . 07969 . 018
For the first time , we have demonstrated a quantitative link between estimates of transmission efficiency of influenza among ferrets and among humans , at the subtype level . However , there is little power to resolve human SAR using ferret SAR estimates from single experiments . Instead , we observed ranges of ferret SAR distinguishing supercritical from subcritical subtypes that may be useful in identifying influenza viruses that pose greater or lesser risk of pandemic spread—especially for viruses with very high or low ferret SAR . In all analyses , including comparisons of sensitivity and false positive rate of various classification thresholds ( Figure 3—figure supplement 3 ) , we found that respiratory droplet transmission in ferrets was a better indicator of transmission in humans than direct contact transmission . However , direct contact experiments used in conjunction with respiratory droplet experiments can provide additional information on transmission in humans . Sample size is a serious challenge to operational use of the results shown here . The largest sample size we found in our review of transmission studies was twelve ferrets ( Herlocher et al . , 2002 ) . Even for a supercritical strain with an assumed ferret SAR of 1 , 8 ferrets must be tested to classify that strain as supercritical with 80% power at a significance level of 0 . 05 ( Figure 4—figure supplement 1 ) . For an assumed ferret SAR of 0 . 8—more in line with zoonotic strains of interest ( Figure 4 ) , but closer to the lower end of the supercritical range—achieving the same power would require more than 30 ferrets ( Figure 4—figure supplement 1 ) . Such a sample size is obviously prohibitive . It is important to note , though , that data from future experiments should refine the relationship in Figure 3 , expanding the ranges corresponding to subcritical and supercritical transmission , and hence lowering the sample size requirements somewhat . Other design changes could also enhance the value of ferret transmission experiments for informing risk assessments . In particular , it is vital to standardize experimental design in order to reduce noise and strengthen inference , beginning with establishing standard definitions of transmission for ferret experiments ( i . e . viral titers in nasal washes vs serologic evidence ) . Discord between viral isolation and antibody data within a single experiment highlights this need and shows that serologic data is often a more sensitive metric of pathogen exposure ( Figure 1—source data 2 , 3 ) . It has been questioned whether seroconversion always reflects a productive viral infection , but recent imaging studies indicate that seroconversion can detect infections that manifest deep in the respiratory tract , which would be missed by nasal wash measurements ( Karlsson et al . , 2015 ) . Although all of our results were robust to the choice of transmission definition ( ‘Results’ not shown ) , defining transmission by viral isolation alone slightly increased predictive power for direct contact experiments , and slightly decreased predictive power for respiratory droplet experiments ( Figure 3—figure supplement 3 ) . Ultimately , this suggests that transmission should be assessed using both serological and viral data to aid in comparisons across experiment types , while allowing for exploration of exposure vs active infection . Dosing protocols can also vary widely across and within studies , in terms of viral titer and volume and even incompatible units . Standardized dosing protocols could reduce variability in ferret SAR estimates substantially . Additional data on time to infection , clinical signs , and mechanistic insights such as receptor binding affinities , none of which are systematically collected under standard protocols , could add value to ferret studies by giving additional power to differentiate among influenza viruses and subtypes with similar transmission outcomes . Despite these challenges , ferret transmission experiments can contribute distinctive insights into the pandemic potential of novel influenza isolates . Our results show that ferret experiments provide a tool with relatively high sensitivity and specificity for identifying strains that may be supercritical in humans ( Figure 3—figure supplement 3 ) . Based on current scientific knowledge , risk screening might also incorporate high-throughput virologic and genetic screens used to identify isolates of concern by looking for genetic changes associated with altered binding affinities and other markers of transmission in mammals ( Russell et al . , 2012 ) . Ultimately , however , human transmission is a complex and partially understood phenotype that is difficult to predict using these initial screens ( Russell et al . , 2014 ) . Ferrets can provide a potential link between underlying virologic and genetic changes and potential transmissibility in humans . Future analyses should attempt to simultaneously incorporate data on the presence of specific mutations ( e . g . , PB2-K627E , Van Hoeven et al . , 2009 ) and virologic factors ( e . g . , binding to α2-6 sialic acid glycans , Belser et al . , 2008 ) into the present analysis of ferret transmissibility to determine if these genetic and virologic screens provide additional information on human transmission not captured by ferrets alone . The resolution of our analysis was limited to the subtype level , because human transmissibility data are not available for more specific strains . Some caution is needed when assessing transmission of novel isolates within a subtype , as out-of-sample predictions under this scenario are likely to be particularly hazardous . As a first assessment of the effect of within-subtype variation , we re-ran our analyses in Figure 2 for a broader dataset including H5N1 and H7N9 strains isolated from avian hosts ( Figure 1—source data 4; Figure 2—figure supplement 1 ) . The results were consistent with our main findings , giving some confidence that our results are robust to such within-subtype variation ( Figure 1—source data 4; Figure 2—figure supplement 1 ) . Additional data on consensus viral sequences within human outbreaks are needed to relate human SAR estimates more specifically to isolates tested in ferrets and clarify the effect of within-subtype variation on predicted human transmission behavior . In the absence of these data , our analysis represents a new null model against which deviations within subtypes can be measured to identify strains that can provide additional information on the molecular features associated with transmissible phenotypes in ferrets and/or humans . Recently , the obvious disparity between highly efficient H7N9 transmission in some ferret experiments and inefficient H7N9 transmission in humans ( see Figure 2B , Figure 4C ) led to questions about the general validity of the ferret transmission model ( Lipsitch , 2013 ) . Our results at least partially assuage these concerns . In spite of the substantial variation we observed within H5N1 and H7N9 subtypes , our results show that , statistically , isolates more transmissible in ferrets are more likely to be capable of sustained transmission in humans . Yet our data also demonstrate that the ferret transmission model is fallible: for H7N9 , an emerging virus of great concern , ferret transmission experiments sometimes yield results that obviously contradict observed patterns in humans . These results are anomalous within the general mapping of ferret transmissibility to human transmissibility and thus , as mentioned previously , may present an opportunity to gain new insight into the molecular drivers of this complex phenotype . However , when screening emerging influenza viruses for pandemic potential , both false negatives and false positives have important consequences for health policy decisions . The deviations of H7N9 from the general correlation between human and ferret transmissibility underscore the importance of corroborating transmission estimates from the ferret model with other lines of evidence . The ultimate evidence to corroborate human transmission comes from epidemiological patterns of infection in humans . For a true pandemic influenza virus , however , such data are likely to come too late , highlighting the need for reliable methods to provide early warning on strains with pandemic potential . Here we have put forward the first guidelines for translating the results of ferret experiments into a measure of pandemic potential in humans . Given the continued use of ferrets in other areas of influenza research ( e . g . , vaccine development ) , this finding enhances the broad value of ferret experiments . However , given pragmatic limitations on sample sizes in ferret studies , uncertainties in ferret SAR estimates are likely to limit the operational utility of these guidelines . This coupled with the biological complexities underlying transmissibility suggests that , at this time , ferret transmission data provide a valuable but imperfect correlate of human transmissibility , and further evidence is needed to assess whether other lines of evidence can improve this predictive capacity .
Most ferret transmission studies report the number of secondary infections amongst a specified number of naïve ferrets that are exposed to single inoculated individuals . This enables calculation of the SAR , which is the probability of infection for a susceptible individual following a known contact with an infectious individual ( Halloran , 2005 ) and establishes a metric of transmissibility in ferrets that is directly comparable to household SAR in humans . We obtained estimates of SAR in humans from household contact data using two methods . Ad hoc SAR estimates are obtained by taking the ratio of infected household contacts over total household contacts . This method is widely used , but may overestimate SAR , as it assumes each household experiences only one disease introduction ( the index case ) and ignores the possibility of multiple household exposures to an exogenous reservoir ( Longini et al . , 1982 ) . Meanwhile , maximum likelihood procedures for SAR estimation use statistical models to simultaneously estimate the probability of secondary transmission within a household ( SAR ) and the probability of infection from the community ( or other source ) . Thus , these estimates attempt to correct for the possibility of multiple introductions from an exogenous source ( Longini and Koopman , 1982; Longini et al . , 1982 ) . However , even these estimates can be strongly skewed by the inclusion or exclusion of specific clusters , especially early in an outbreak when data is limited ( Aditama et al . , 2012 ) . Furthermore , variation in existing , population-level immunity to specific strains , and the use of different case ascertainment methods in specific studies also inevitably skew estimates made using either procedure . Because each method has unique biases and limitations , we used published estimates of SAR based on either method , or calculated an ad hoc SAR estimate ourselves from data on the total and infected number of household contacts in an outbreak . Human SAR estimates are only considered in our initial regression analysis ( Figure 2A ) , so they do not influence our classification model ( Figure 3 ) . To assess the relationship between human and ferret transmissibility of influenza , we reviewed existing estimates of subtype-specific SAR in humans and ferrets . We searched PubMed and Web Of Science [v5 . 15] databases using the following queries: ( influenza AND household AND transmission AND H#N# ) and ( influenza AND ‘secondary attack rate’ OR SAR AND human AND H#N# ) for human studies and ( influenza AND transmission AND ferret* AND H#N# ) for ferret studies . We repeated searches for subtypes H1N1 , H7N9 , H3N2 , H7N7 , H7N9 , H7N2 , H9N2 , H5N1 , H7N3 , and H2N2 . To ensure comprehensive coverage , additional studies were identified using reference lists from search results and additional spot searches were also conducted . We excluded isolates that represented outliers from identified subtypes ( i . e . 1918 pandemic H1N1 ( Tumpey et al . , 2007 ) and novel swine-origin H3N2 in 2009–10 ( Pearce et al . , 2012 ) ) . Searches were completed on 20 July 2015 . Although the transmission potential of unique isolates within a subtype may vary , SAR in humans was reported only at the subtype level , preventing us from analyzing isolate-specific transmission potential . Overall , we found data for all three measures ( ferret direct contact , ferret respiratory droplet , and human SARs ) for ten influenza A subtypes: H7N3 , H9N2 , H7N7 , H7N2 , H5N1 , H7N9 , H2N2 , pH1N1 ( i . e . influenza A ( H1N1 ) pdm09 virus ) , H3N2 , and seasonal H1N1 ( Figure 1 ) . We excluded ferret transmission studies that included serial passage of human isolates in ferrets prior to transmission experiments . To maintain consistency in transmission mechanisms , we excluded studies that inoculated ferrets by routes other than intranasal with a liquid inoculum ( e . g . , ocular inoculation or aerosol inhalation ) and that inoculated ferrets with a lower viral dose than was typical for ferret transmission studies ( <103 50% egg infectious dose [EID50] ) . We excluded studies where naive ferrets were not exposed to inoculated ferrets at 1 day post-inoculation , as was standard , and studies where the duration of contact was restricted . We also excluded trials in which ferrets were vaccinated or administered antiviral drugs for treatment or prophylaxis . If transmission of more than one subtype and/or isolate was tested in a single study ( using different sets of immunologically naive ferrets for each isolate ) , we treated each subtype/isolate-specific data point separately . However , for some analyses , we grouped data from isolates belonging to the same subtype—the one exception being separation of 2009 pandemic H1N1 isolates ( pH1N1 ) and pre-2009 H1N1 isolates ( H1N1 ) . We distinguished between direct contact transmission experiments ( in which sentinel ferrets were co-housed with the donor ferret ) and respiratory droplet transmission experiments ( in which ferrets were housed in adjacent cages designed to allow for airborne exchange , but in which direct or indirect contact between sentinels and donors is not possible ) . Transmission amongst ferrets was determined in each study using either a viral titer in nasal washes or a positive serologic test ( i . e . hemagglutination inhibition assay ) or by a combination of both tests . We noted any discrepancies between the two transmission mechanisms ( Figure 1—source data 2 , Figure 1—source data 3 ) and conducted analyses that showed our results were relatively robust to the transmission definition used ( Figure 3—figure supplement 3 ) . To promote quality of comparison between ferret and human studies , we only included data from ferret studies that tested one or more wild-type human isolates . While avian and other animal isolates maintain close sequence homology with human isolates ( Claas et al . , 1998 ) , the transmission of animal isolates into humans is associated with genetic bottlenecks ( Zaraket et al . , 2015 ) and considerable within-host adaptation ( Linster et al . , 2014 ) . These evolutionary barriers lead to avian precursors that have lower mortality in mice , less morbidity in ferrets , and lower viral titers in human epithelial cells ( Belser et al . , 2013; Watanabe et al . , 2014; Zaraket et al . , 2015 ) . Thus , these cross-species and within-host barriers have the potential to obscure the relationship between transmission in ferrets and transmission in humans , and we excluded avian and other animal strains from the main analysis as a result . We did , however , compile a database of ferret transmission experiments using avian isolates from subtypes H5N1 and H7N9 ( Figure 1—source data 4 ) to test the validity of this exclusion . Avian isolates in these subtypes have the benefit of contemporary sampling in both space and time with their human counterparts . Supplementary analyses including these avian isolates showed that our results were robust to the exclusion of non-human isolates ( Figure 2—figure supplement 1 ) . We also included wild-type isolates from humans generated using reverse genetics techniques . Although viral isolates rescued through reverse genetic techniques are often assumed to have lower transmissibility , analyses with and without these rescued isolates yielded indistinguishable results . Indeed , for the small number of isolates for which we could make direct comparisons , isolates generated using reverse genetics exhibited similar transmissibility to their wild-type counterparts ( Figure 1—figure supplement 1 ) . Thus , our data set contained a total of 81 respiratory droplet ( Figure 1B; Figure 1—source data 2 ) and 76 direct contact transmission trials ( Figure 1C; Figure 1—source data 3 ) . Because we considered only household SAR , we excluded studies with non-standard household definitions ( e . g . , dormitories , health care centers , summer camps ) , and studies where household contacts could not be distinguished from broader community contacts . We also excluded data from studies of zoonotic strains where prior contact with potential livestock or wildlife reservoirs was noted for multiple contacts , thus hindering the distinction between primary and secondary cases . In order to represent a broad range of human SAR estimates , we included both prospective and retrospective household studies that either provided an explicit SAR estimate or reported data sufficient to calculate a SAR . This yielded a total of 83 estimates of human SAR ( Figure 1A; Figure 1—source data 1 ) . Quantitative comparison of SAR in ferrets and SAR in humans was performed using linear regression ( Figure 2A ) . Because human SAR estimates are not typically made for individual isolates , the comparison was done at the subtype level using the mean value of all estimates belonging to a subtype . For ferret experiments , we used a weighted mean by subtype , where the weights were given by the number of ferrets used in each experiment; for human estimates , we used the simple mean by subtype . The potential uncertainty in subtype mean SAR was large , especially for human SAR , where several emerging subtypes ( i . e . H7N3 , H9N2 , H7N7 , and H7N2 ) only had one or two estimates ( Figure 1A ) . To allow for this uncertainty , we used a weighted linear regression with model weights given by the number of human SAR estimates . To create Figure 2B , we developed an empirical measure for the overlap between distributions of ferret SAR estimates for pairs of subtypes that was a simple variant of other overlap indices used in ecology ( Ricklefs and Lau , 1980 ) . This was calculated by comparing the more transmissible and less transmissible of each of the subtypes , taking the minimum SAR estimate for the more transmissible subtype and the maximum SAR estimate for the less transmissible subtype , and counting the number of estimates for both subtypes that fell within this range of overlap ( normalized by the total number of estimates for both subtypes ) . This yielded a measure between 0 and 1 , where zero indicated that the ranges of observed SAR estimates for two subtypes were completely distinct and one indicated that the ranges completely overlapped , rendering the subtypes indistinguishable on the basis of SAR . Examination of Figure 2B revealed two distinct clusters of subtypes whose distributions of SAR estimates overlapped almost completely . H1N1 , H3N2 , pH1N1 , and H2N2 are supercritical subtypes with sustained transmission among humans; H7N2 , H7N7 , H9N2 , and H7N3 are subcritical subtypes with weak transmission among humans . This grouping suggests there may be potential to use ferret SAR estimates for broader functional classification of viruses with or without pandemic potential . However , two subtypes of concern , H7N9 and H5N1 ( both known to be subcritical in humans ) , were anomalies within the natural clusters we observed in Figure 2B: H7N9 clustered with supercritical subtypes , while H5N1 was weakly associated with both groups . We interpreted this as important biological variation within the group of subcritical subtypes , but considering the overarching interest in predicting whether particular subtypes might have pandemic potential , for our further analyses we chose to group subtypes according to their observed transmission pattern in humans ( i . e . supercritical—H1N1 , H3N2 , pH1N1 , H2N2 and subcritical—H7N3 , H9N2 , H7N7 , H7N2 , H5N1 , H7N9 ) . To determine how ferret SAR was related to supercritical and subcritical classifications , we used a weighted logistic regression ( Figure 3 ) . Here , model weights are based on the number of ferrets used in each experiment , thus allowing for more confidence in estimates with larger numbers of ferrets . Ferret SAR estimates that corresponded to a high probability of an isolate being classified as supercritical or subcritical ( i . e . low probability of being supercritical ) were determined by calculating 95% confidence intervals for the predicted model fit and identifying the ranges where these confidence intervals were either wholly above ( supercritical ) or wholly below ( subcritical ) a value of 0 . 5 ( representing a random guess of supercritical or not ) . The sensitivity and specificity of various thresholds in ferret SAR were also assessed ( Figure 3—figure supplement 3 ) . All analyses were done using R Statistical Software version 3 . 1 . 2 ( R Development Core Team , 2014 ) . | Every year , thousands of people develop influenza ( flu ) . After being infected by the influenza virus , the immune systems of most people adapt to fight off the virus if it is encountered again . However , there are many different strains of influenza , and new strains constantly evolve . Therefore , although someone may have developed resistance to one previously encountered strain , they can still become ill if another strain infects them . Different strains of the influenza virus have different abilities to spread between people and make them ill . One way that scientists assess whether a particular strain of influenza is a threat to people is by studying ferrets , which develop many of the same flu symptoms as humans . However , questions have been raised over how accurately ferret studies reflect whether a particular virus strain will spread between humans . Controversy has also arisen over experiments in which ferrets are infected with genetically engineered strains of influenza that mimic how a strain that has evolved in birds could adapt to cause a pandemic in humans . In 2014 , the United States government suggested that such research should be temporarily stopped until more is known about the risks and usefulness of these studies . Now , Buhnerkempe , Gostic et al . have compared the results of 240 ferret and human studies that aimed to assess how easily strains of influenza spread . Specifically , the studies looked at how often a healthy ferret or human became ill when exposed to an animal or human infected with a particular strain of influenza . The results of the ferret transmission studies matched well with transmission patterns observed in human studies . Ferret studies that assessed how the influenza virus is transmitted through the air via sneezes and coughs were particularly good at predicting how the virus spreads in humans . But Buhnerkempe , Gostic et al . caution that ferret studies are not always accurate , partly because they involve small numbers of animals , which can skew the results . There also needs to be more effort to standardize the procedures and measurements used in ferret studies . Still , the analysis suggests that overall , ferret studies are a useful tool for making an initial prediction of which influenza strains may cause a pandemic in humans , which can then be verified using other methods . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"microbiology",
"and",
"infectious",
"disease"
] | 2015 | Mapping influenza transmission in the ferret model to transmission in humans |
In the neocortex , critical periods ( CPs ) of plasticity are closed following the accumulation of perineuronal nets ( PNNs ) around parvalbumin ( PV ) -positive inhibitory interneurons . However , how PNNs tune cortical function and plasticity is unknown . We found that PNNs modulated the gain of visual responses and γ-oscillations in the adult mouse visual cortex in vivo , consistent with increased interneuron function . Removal of PNNs in adult V1 did not affect GABAergic neurotransmission from PV cells , nor neuronal excitability in layer 4 . Importantly , PNN degradation coupled to sensory input potentiated glutamatergic thalamic synapses selectively onto PV cells . In the absence of PNNs , increased thalamic PV-cell recruitment modulated feed-forward inhibition differently on PV cells and pyramidal neurons . These effects depended on visual input , as they were strongly attenuated by monocular deprivation in PNN-depleted adult mice . Thus , PNNs control visual processing and plasticity by selectively setting the strength of thalamic recruitment of PV cells .
During postnatal maturation , sensory processing goes through a critical period ( CP ) , a developmental interval , in which neural circuits are shaped by sensory experience . After this time window , plasticity declines significantly , and learning becomes more difficult ( Hensch , 2005; Espinosa and Stryker , 2012 ) . In the visual cortex , the closure of the CP is paralleled by the structural maturation of the extracellular matrix , and , in particular , of perineuronal nets ( PNNs ) . These are composed by a conglomeration of chondroitin sulphate proteoglycans , extracellular matrix and cell-adhesion molecules that , in the neocortex , accumulates selectively around fast-spiking , PV basket cells ( Pizzorusso et al . , 2002; Berardi et al . , 2004; Bernard and Prochiantz , 2016 ) . Importantly , chemical breakdown of PNNs reactivates ocular dominance plasticity in the adult visual cortex ( Pizzorusso et al . , 2002 ) , promotes juvenile forms of extinction of fear memories in the amygdala ( Gogolla et al . , 2009 ) and functional recovery after brain injury ( Bradbury et al . , 2002; Gherardini et al . , 2015 ) . Therefore , PNNs were proposed to act as a structural brake to experience-dependent plasticity , restricting the extent to which a neural circuit can change during late postnatal development ( Berardi et al . , 2004 ) . PV cells represent a major subtype of cortical GABAergic interneurons , specialized in providing fast and reliable perisomatic inhibition to principal neurons ( PNs ) , thereby controlling their output spiking properties and driving network oscillations in the β-γ-frequency range ( Freund and Katona , 2007; Isaacson and Scanziani , 2011; Buzsáki and Wang , 2012; Tremblay et al . , 2016 ) . In addition to controlling cortical circuit activity ( Hensch , 2005; Buzsáki and Wang , 2012 ) , PV cells shape sensory plasticity ( Fagiolini et al . , 2004; Hensch , 2005; Donato et al . , 2013; Toyoizumi et al . , 2013; Kuhlman et al . , 2013; Gogolla et al . , 2014; Lensjø et al . , 2017; Takesian et al . , 2018 ) . In particular , the strength of inhibition from PV cells was proposed to define the temporal window of the CP of cortical plasticity: increasing GABAergic neurotransmission accelerates the onset of the CP , whereas a reduction of inhibition delays the onset of plasticity ( Fagiolini et al . , 2004; Hensch , 2005; Hensch and Fagiolini , 2005 ) . Despite the mechanisms underlying the CP have been extensively studied ( Hensch , 2005; Hübener and Bonhoeffer , 2014 ) , very little is known about how PNN accumulation around PV cells changes the cellular and synaptic properties of these interneurons , thus affecting cortical circuits and limiting plasticity . In this context , it is crucial to pinpoint the functional mechanisms linking PNN accumulation around PV cells to its modulation of activity-dependent plasticity . Indeed , accumulating evidence indicates that dysfunctions of cortical circuits involving PV cells as well as PNN maturation are implicated in several psychiatric diseases including autism and schizophrenia ( Marín , 2012; Sorg et al . , 2016 ) . Here , we describe how PNN removal in adult mice altered the gain of visual processing and the power of γ-oscillations in vivo; we reveal the underlying synaptic circuitry and its sensitivity to sensory plasticity . In particular , we found that PNNs set the strength of thalamic inputs to PV cells selectively , leaving neuronal excitability and unitary synaptic GABAergic transmission from these interneurons intact . This resulted in a strong and differential modulation of feed-forward inhibition onto PNs and other PV cells . Importantly , plasticity induced by short monocular deprivation ( MD ) strongly attenuated these effects , indicating that PNN-mediated modulation of thalamic input onto PV cells depends on visual activity . These results reveal the synaptic and circuit mechanisms by which PNNs restrict sensory plasticity in the adult visual cortex .
To test the effects of in vivo PNN removal on adult cortical circuit function , we stereotaxically injected the primary visual cortex ( V1 ) of adult mice ( >P70 ) with the bacterial enzyme chondroitinase ABC ( ChABC ) , 2–3 days prior to electrophysiological experiments ( see Materials and methods ) . This is a standard procedure to effectively and locally disrupt PNNs , revealed by the absence of Wisteria floribunda agglutinin staining ( WFA , Figure 1A , B ) ( Pizzorusso et al . , 2002; Lensjø et al . , 2017 ) . Importantly , this approach was shown to re-open adult cortical plasticity ( Pizzorusso et al . , 2002;de Vivo et al . , 2013 ) . We first measured gain adaptation of contrast perception , which is a fundamental computation performed by the primary visual cortex ( Carandini and Ferster , 1997; Anderson et al . , 2000; Atallah et al . , 2012 ) . We recorded visually-evoked extracellular potentials ( VEPs ) in V1 of adult mice in response to an alternating checkerboard of varying contrast presented to the contralateral eye ( Figure 1C , D ) . We found an enhancement of adaptation in ChABC-injected animals , measured as a significant decrease of the slope of the transfer function ( Figure 1E ) . Moreover , the spectral power of the local field potential during both resting state and visual activity was increased by ChABC treatment in the high γ-frequency band ( 40–80 Hz; Figure 1F–I ) consistent with a previous report ( Lensjø et al . , 2017 ) . It should be pointed out that the power increment associated to the visual stimulation has a large bandwidth , consistent with the fact that the checkerboard reversal is an intrinsically transient stimulus that produces a response of brief duration . More prolonged stimuli , like drifting gratings , cause a longer response associated to a stricter bandwidth of the oscillations ( Welle and Contreras , 2016; Veit et al . , 2017 ) . Importantly , PV cells are known to strongly modulate the gain of contrast sensitivity ( Atallah et al . , 2012 ) , and improve network synchrony during γ-oscillations ( Cardin et al . , 2009; Sohal et al . , 2009; Isaacson and Scanziani , 2011; Buzsáki and Wang , 2012 ) . Therefore , our results suggest that the enzymatic disruption of PNNs results in increased activity of inhibitory interneurons during visual stimulation . Increased perisomatic inhibitory activity in vivo could be explained by one or a combination of the following causes: i ) increased intrinsic excitability and/or spiking activity of inhibitory interneurons , ii ) alterations of the excitatory and/or inhibitory drive onto specific elements of the cortical networks favoring the recruitment of local GABAergic interneurons . We tested all these possibilities in L4 of V1 , which is a prominent target of the visual thalamus and shows a stronger PNN enrichment around PV cells , as compared to other cortical layers ( Figure 2—figure supplement 1A , B , C ) . Moreover , changes in the strength of thalamocortical connections were proposed to underlie visual cortical plasticity ( Coleman et al . , 2010; Jaepel et al . , 2017 ) . Indeed , we found that the vast majority of layer 4 PV cells in the adult is enwrapped by PNNs to some degree ( Figure 2—figure supplement 1A , B , C ) , although in superficial and deep cortical layers ( layer 2/3 , deep layer 5 and layer 6 ) the amount of PNNs is much lower ( Figure 2—figure supplement 1A , B , C ) . We conclude that the probability of recording from PNN-free PV cells in layer 4 of sham-treated animals is very low . We recorded from PV cells and PNs in acute brain slices from adult ( >P70 ) mice , which underwent PNN digestion in vivo , in the presence and absence of adult cortical plasticity , induced by short ( 2–3 days ) monocular deprivation ( MD; Materials and Methods and Figure 2—figure supplement 1D ) . Interestingly , action potential waveform , firing dynamics and passive membrane properties of both PV cells and PNs were unaffected by PNN digestion , in the presence and absence of MD ( Figure 2—figure supplement 2; Figure 2—figure supplement 3; Tables S1-2 in Supplementary file 1 ) . These results are in contrast with previous studies showing altered firing after acute PNN disruption in vitro ( Dityatev et al . , 2007; Balmer , 2016 ) , or when the PNN protein brevican was knocked out ( Favuzzi et al . , 2017 ) . We then analyzed glutamatergic synaptic transmission on PV cells and PNs . Enzymatic disruption of PNNs significantly increased amplitude and frequency of spontaneous excitatory postsynaptic currents ( sEPSCs ) onto PV cells ( Figure 2A , B; Table S3 in Supplementary file 1 ) . Conversely , glutamatergic transmission onto PNs was unaffected by ChABC treatment ( Figure 2—figure supplement 4; Table S3 in Supplementary file 1 ) . Increased neurotransmission onto PV cells following PNN removal was due to quantal synaptic transmission , and not increased slice excitability as revealed by increased miniature ( m ) EPSC frequency in the presence of 1 µM tetrodotoxin ( TTX; Figure 2C , D; Table S3 in Supplementary file 1 ) . Plasticity induced by MD ( Pizzorusso et al . , 2002; Berardi et al . , 2004; de Vivo et al . , 2013 ) significantly counteracted the strong increase of sEPSC amplitude and mEPSC frequency on PV cells in PNN-depleted mice ( Figure 2E–H; Table S3 in Supplementary file 1; for ChABC-mediated effects on sEPSC amplitude and mEPSC frequency: p < 0 . 01 in control vs . p > 0 . 05 in MD ) , although sEPSC frequency remained potentiated . These results indicate that the increased synaptic recruitment of PV cells , induced by PNN degradation , is sensitive to visual input . We then investigated if a specific glutamatergic pathway was involved . First , we studied intracortical circuitry but , surprisingly , we found a very low yield of connected intracortical PN-PV cell pairs in L4 of adult mice ( 5% , n = 104 ) , as opposed to young animals ( 40% , n = 20; Figure 2—figure supplement 5 ) , likely because of re-routing of PN axons to L2/3 in adult mice . Therefore , we focused our attention on the thalamocortical pathway , which carries sensory information . Using adeno-associated viruses ( AAVs ) , we expressed the light-sensitive opsin channelrhodospin 2 ( ChR2 ) in the dorsolateral geniculate visual thalamic nucleus ( dLGN ) of adult mice ( Figure 2I; Figure 2—figure supplement 6; see Materials and Methods ) . After 10–12 days , we injected either sham solution or ChABC in V1 of the same mice ( Figure 2J; Figure 2—figure supplement 6A ) , and , in some cases , MD was performed at the time of ChABC/sham injection ( Figure 2—figure supplement 6A ) . In the absence of PNNs , PV cells exhibited larger light-evoked monosynaptic thalamocortical responses ( in the presence of TTX and the K+ channel blocker 4-aminopyridine , 4-AP; see Materials and Methods ) than in control mice ( Figure 2K , Table S4 in Supplementary file 1 ) . We analyzed threshold responses to reduce the risk of misinterpreting our results due to variable expression of ChR2 in different mice . Threshold responses were obtained by setting the power of light stimuli yielding alternating synaptic failures and responses ( see Materials and Methods ) . Of note , light stimulation parameters used to evoke threshold responses were overall similar across different animal groups in all conditions ( Table S4 in Supplementary file 1; p > 0 . 05 ) . Importantly , plasticity induced by MD significantly attenuated the potentiated recruitment of PV cells by thalamic afferents in the absence of PNNs ( Figure 2L; Table S4 in Supplementary file 1; size effect Glass’ delta: 2 . 11 vs . 1 . 75; control vs . MD , respectively , see Materials and Methods ) . Similarly to spontaneous glutamatergic neurotransmission , PNN digestion did not affect monosynaptic thalamocortical recruitment of PNs , both in the absence and presence of MD ( Figure 2M , N; Table S4 in Supplementary file 1 ) . AAVs can produce anterograde infection ( Zingg et al . , 2017 ) . However , in our hands , we never detected mCherry- ( and thus ChR2- ) positive cell bodies in the neocortex ( Figure 2—figure supplement 6C , D ) , thus excluding a possible contamination of intracortical responses in our experiments . These results indicate that PNN accumulation controls the strength of thalamic glutamatergic synapses onto PV cells selectively , and this is forcibly modulated by visual activity . Cortical plasticity is strongly modulated by inhibition ( Fagiolini et al . , 2004; Hensch , 2005; Toyoizumi et al . , 2013; Kuhlman et al . , 2013 ) . Accordingly , we found that removal of PNNs is consistent with a more strongly inhibited cortical network ( Figure 1 ) ( Sohal et al . , 2009; Cardin et al . , 2009; Atallah et al . , 2012 ) , and enhanced recruitment of PV cells ( Figure 2 ) . To test directly if PNN disruption in adult mice changed inhibitory synapses onto PV cells and PNs , we pharmacologically isolated spontaneous and miniature inhibitory postsynaptic currents ( s- and mIPSCs ) , using a high-chloride intracellular solution , in the continuous presence of glutamate receptor antagonists ( see Materials and Methods ) . We found that sIPSC amplitude and frequency onto PV cells were increased upon removal of PNNs ( Figure 3A , B; Table S5 in Supplementary file 1 ) . Interestingly , however , AP-independent quantal mIPSC transmission was not affected by PNN removal ( Figure 3C , D; Table S5 in Supplementary file 1 ) . These results indicate that spontaneous GABAergic transmission onto PV cells was due to a network effect , rather than a direct synaptic alteration . Plasticity induced by sensory deprivation prevented the ChABC-mediated increase of sIPSC amplitudes , whereas frequency was still increased in PNN-depleted mice ( Figure 3E–H; Table S5 in Supplementary file 1 ) . Importantly , we did not find any change in sIPSCs in PNs following ChABC injection , both in the absence and presence of MD ( Figure 2—figure supplement 4E–H; Table S5 in Supplementary file 1 ) . To directly test whether local GABAergic synapses to and from PV cells were not affected by PNN removal , we performed simultaneous recordings in PV-PV and PV-PN connected pairs , in the presence and absence of MD . In addition , PV-PV inhibitory connections were also examined as autaptic self-inhibiting responses , which are highly common in neocortical PV cells ( Deleuze et al . , 2014 ) ( Figure 2—figure supplement 5 ) . Overall , we found that PNN disruption did not alter unitary GABAergic transmission from PV cells onto themselves , other PV cells and PNs , in terms of magnitude and short-term plasticity , both in the absence and presence of MD ( Figure 3I–T; Table S6 in Supplementary file 1 ) . Altogether , these results indicate that GABAergic synapses from PV cells onto PNs and other PV cells are not altered by PNN removal in the adult mouse visual cortex . Rather , increased GABAergic spontaneous activity results from increased AP-dependent activity of interneurons . The results of Figures 2 and 3 suggest a selective and experience-dependent increase of thalamic excitatory neurotransmission onto PV cells in response to PNN removal , with no direct effect on quantal and unitary GABAergic responses . Thalamic activation of PV cells is very potent in neocortical L4 ( Gabernet et al . , 2005; Sun et al . , 2006; Cruikshank et al . , 2010; Bagnall et al . , 2011 ) , generating strong feed-forward inhibition ( FFI ) , which is responsible for sharpening contrast sensitivity and controlling the temporal resolution of sensory integration ( Gabernet et al . , 2005 ) . Importantly , PNN degradation affected contrast sensitivity ( Figure 1 ) and AP-dependent sIPSCs onto PV cells ( Figure 3 ) . We therefore tested whether increased thalamic recruitment of PV cells affects disynaptic FFI in cortical L4 . We isolated thalamic-induced FFI in both PV cells and PNs by activating ChR2-positive fibers at the reversal potential for glutamate-mediated responses , and in the absence of TTX and 4-AP ( Figure 4A , B; Table S7 in Supplementary file 1; see Materials and Methods ) . We found that ChABC treatment strongly increased FFI in PV cells , elicited by threshold stimulations . Again , MD strongly attenuated this effect ( Figure 4A , B , shaded area; size effect Glass’ delta: 6 . 89 vs . 2 . 30; control vs . MD , respectively ) . Surprisingly , we did not detect any change of FFI on PNs , when measured at threshold ( Figure 4C , D ) . Importantly , however , at a higher stimulus intensity ( 1 . 5 x threshold ) , FFI was significantly increased by ChABC treatment also in PNs , and sensory deprivation prevented FFI potentiation onto PNs ( Figure 4E , F ) . Altogether , these results indicate a preferential gating of disynaptic , feed-forward inhibition in L4 . Indeed , compared to PNs , FFI on PV cells was more sensitive to modulation by PNN accumulation . In addition , the boost of FFI was strongly dependent on visual input , similarly to s- , m-EPSCs and thalamocortical glutamatergic activation of PV cells . PNN accumulation in the visual system co-occurs with the end of the CP ( P30-35 ) ( Pizzorusso et al . , 2002 ) . Importantly , PNN disruption in V1 of adult mice re-opens cortical plasticity ( Pizzorusso et al . , 2002 ) , and is associated to increased glutamatergic synaptic transmission selectively in PV cells ( Figure 2 ) . We therefore measured glutamatergic and GABAergic neurotransmission on PV cells and PNs before ( <P20 ) , at the peak ( P25-32 ) and after ( P40-60 and >P70 ) the CP , to test if post-CP accumulation of PNNs is associated to changes of the overall strength of glutamatergic neurotransmission onto PV cells . Interestingly , we found that the maturation of cortical circuits after the CP is accompanied by a decrease of glutamatergic neurotransmission onto PV cells ( both sEPSC amplitudes and frequency; Figure 5A–C , Table S8 in Supplementary file 1 ) , whereas GABAergic inhibition on PV cells was unchanged throughout development ( Figure 5D–F , Table S8 in Supplementary file 1 ) . This developmental decrease of glutamatergic strength was selective for PV cells , as both glutamatergic and GABAergic neurotransmissions on PNs were stable across pre- and post-CP stages ( Figure 5G–L , Table S8 in Supplementary file 1 ) . These results suggest that the accumulation of PNNs around PV cells during post-CP development determines a change in the excitation-to-inhibition ratio in this interneuron type , and ChABC-mediated disruption of PNNs in adult animals ( Figure 2 ) recapitulates some juvenile features of visual cortical circuits .
In this study , we demonstrate that PNNs modulate the gain of contrast sensitivity and network synchrony during cortical γ-oscillations . This is associated to a selective increase of thalamic glutamatergic recruitment of PV interneurons in the absence of PNNs , without altering their excitability and the quantal properties of their GABAergic synapses . Increased thalamic recruitment of PV cells in the absence of PNNs strongly affects FFI differentially in PV cells and PNs ( Figure 6A ) . All effects depended on the presence of normal visual input after ChABC in vivo treatment , as they were reduced or prevented by sensory plasticity induced by MD ( Figure 6B ) . Interestingly , selective decrease of glutamatergic neurotransmission onto PV cells is present during the post-CP development , which is paralleled by accumulation of PNNs around these interneurons . ChABC degrades chondroitin sulfate proteoglycans ( CSPGs ) , the major PNN component , down to their disaccharide building blocks . However , CSPGs are also diffusely present in the entire extracellular environment . Thus , ChABC disrupts the entire extracellular matrix beyond PNNs , possibly leading to compounding effects on PV cell circuits , unrelated to PNNs . For these reasons , in all our experiments ( both in control and MD ) , we have recorded from both PV cells and PNs . We did not observe any change in PNs in any parameter we investigated ( AP waveform , firing dynamics , sEPSCs , sIPSCS , uIPSCs and thalamic glutamatergic activation ) , as opposed to PV cells . We thus conclude that the effects that we report in the manuscript are due to PNN accumulation around these interneurons . Importantly , our approach is known to re-open visual plasticity in adult animals ( Pizzorusso et al . , 2002 ) . Accordingly , it has been recently demonstrated that genetic ablation of aggrecan ( and thus PNN structure ) in adult mice results in the re-opening of cortical plasticity ( Rowlands et al . , 2018 ) . Reduced gain of contrast sensitivity and increased power of γ-oscillations are consistent with increased activity of PV interneurons . Indeed , optogenetic activation of PV cells was shown to have a strong effect on the gain of visual contrast sensitivity ( Atallah et al . , 2012 ) , and determine the level of γ-power and information transfer in the neocortex ( Cardin et al . , 2009; Sohal et al . , 2009 ) . Therefore , the PNN-dependent effects shown here are consistent with either increased PV-cell excitability and firing , and/or altered synaptic transmission to and from these cells . Indeed , it has been suggested that removal of PNNs in the adult results in alterations of AP waveform and firing ( Dityatev et al . , 2007; Balmer , 2016 ) . In contrast , here we show that intrinsic excitability of PV cells and PNs is not affected by in vivo enzymatic digestion of PNNs in the adult visual cortex . This discrepancy could be due to different approaches used to remove or diminish the expression of PNNs: in vivo vs . in vitro digestion , or genetic downregulation . Indeed , acute in vitro digestion of PNNs might not entirely recapitulate the actual remodeling of cortical networks induced by visual input . In addition , genetic downregulation of brevican ( Favuzzi et al . , 2017 ) , a key component of the extracellular matrix , could be prone to developmental and/or homeostatic processes that we did not induce here , due to the relatively short depletion of PNNs in adult animals , known to re-open visual plasticity in adult animals ( Pizzorusso et al . , 2002 ) . The enzymatic disruption of PNNs is advantageous , because it relies on an acute degradation ( 2–3 days ) at a mature developmental stage when cortical circuits are fully developed . Compensatory effects are minimal during such a short time of PNN degradation , thus allowing dissecting the exact mechanism underlying the re-opening of cortical plasticity , without other developmental effects present when PNNs are removed genetically ( Favuzzi et al . , 2017 ) . These developmental studies are extremely important , but they address a different question , centered on the development of cortical circuits in the absence of PNNs . Interestingly , recent evidence indicates a similar re-opening of cortical plasticity in a specific mouse model , in which PNNs were genetically knocked out in PV cells from adult mice ( Rowlands et al . , 2018 ) . Another important mechanism by which PNNs might limit cortical plasticity could be attributed to decreased inhibition from PV cells . A recent study suggest that reduced cortical GABAergic neurotransmission is responsible for unlocking juvenile plasticity in adult animals , although this was not tested directly , but inferred by spiking activity ( Lensjø et al . , 2017 ) . In fact , here we show that network-induced FFI and AP-dependent spontaneous synaptic inhibitory neurotransmission were rather enhanced by PNN removal , although quantal and unitary GABAergic neurotransmission were not affected . These results strongly suggest that increased thalamic glutamatergic strength onto PV cells increases network-dependent FFI ( Figure 6A ) . The enhanced thalamic recruitment of PV cells and the consequent network-driven increase of GABAergic transmission ( Figure 4G–H ) explain the reduction of the gain of contrast adaptation and the enhancement of γ-oscillations measured in vivo ( Figure 1 ) , as they are both consistent with increased activity of PV cells ( Sohal et al . , 2009; Cardin et al . , 2009; Atallah et al . , 2012 ) . Increased FFI onto PV cells and PNs was most likely due to a ChABC-mediated effect on PV-cell recruitment and not on unitary GABAergic responses , as they were not affected by PNN depletion , as shown in Figure 3 . Interestingly , FFI on PV interneurons was much more sensitive to the presence of PNNs than FFI on PNs . This well agrees with the PV cell-specific modulation of sIPSCs by PNNs . The stronger sensitivity of FFI onto PV cells , could be due to the significantly stronger unitary PV-PV connections , as compared to PV-PN synapses ( p < 0 . 05 ) . The preferential PNN-mediated alteration of FFI onto PV cells might favor L4 circuit disinhibition ( and thus paradoxical excitation ) for visual stimulations at low intensities , whereas FFI on PNs becomes more prominent at stronger visual stimuli , thus reducing the gain of the visual adaptation curve ( Figure 1 ) . Moreover , PNN-dependent modulation of FFI in L4 might affect spike-timing precision of both PV cells and PNs , and change the integration window during the initial steps of sensory processing ( Gabernet et al . , 2005 ) . The increased glutamatergic recruitment of PV cells induced by PNN degradation could result from alterations induced in the pre- or postsynaptic site ( or both ) . Our optogenetic approach does not allow dissecting the precise synaptic site affected by PNN degradation . This is especially true for the protocol used to isolate thalamocortical inputs onto PV cells and PNs , which is nonetheless useful to prevent unwanted multi-synaptic activity , typical of these powerful synapses ( Petreanu et al . , 2009 ) , but it compromises the analysis of presynaptic release probability . The increase of mEPSC frequency but not amplitude on PV cells , induced by PNN degradation , suggests a presynaptic modulation . Yet , future experiments will be necessary to unequivocally determine whether the absence of PNNs alters thalamic synapses via pre- or postsynaptic mechanisms . Our MD results demonstrate that the potentiated thalamic recruitment in the absence of PNNs depends on visual activity . A small , albeit significant , reduction of PV-cell and PN recruitment induced by sensory deprivation in adult mice is present also in control conditions ( p < 0 . 05; not shown ) , suggesting that 2–3 days of MD are sufficient to slightly downregulate contralateral thalamic input in the binocular cortex . However , in the absence of PNNs , MD effects are more pronounced . Therefore , we conclude that the MD-dependent reduction of PV-cell recruitment is likely the mechanism that is responsible for re-opening cortical plasticity following PNN degradation in adult animals ( Pizzorusso et al . , 2002; Rowlands et al . , 2018 ) . Interestingly , this effect is similar to that occurring in young animals during the CP , in which sensory deprivation reduces the firing rate of PV cells due to a decrease of synaptic excitatory drive onto these interneurons ( Kuhlman et al . , 2013 ) . Accordingly , we found a significant reduction of glutamatergic synaptic strength selectively on PV cells , during a post-CP maturation window that is paralleled by enrichment of PNNs around PV cells . The lack of changes of the inhibitory strength during the post-CP development suggest that the accumulation of PNNs is associated to a change in the excitatory-to-inhibitory ratio only onto PV cells , and that might be responsible for the closure of cortical plasticity in the adult . Importantly , visual cortical plasticity during CP results from overall decreased inhibition ( Hensch , 2005 ) . Accordingly , we also detect a decrease of action potential-dependent inhibition ( both on PV cells and PNs ) , when we induce plasticity with MD . In order to induce plasticity in the adult , however , glutamatergic recruitment of PV cells ( and consequently their inhibitory activity ) has to increase ( as it happens during the CP; Figure 5 and Kuhlman et al . , 2013 ) . When PNNs are removed , glutamatergic synapses are enhanced , AP-dependent inhibition too , and this boost is subject to sensory-dependent plasticity that we measured as a decreased level of inhibition . We conclude that PNN accumulation during post-CP development might exert a protective role , selectively dampening thalamic excitation of PV cells ( and thus excessive cortical circuit inhibition ) at the expense of reducing cortical plasticity . Age-dependent reduction of plasticity of thalamic synapses onto PV cells might be thus instrumental for correct mature sensory representation . Accordingly , deficits in PNN formation during development have been associated with brain diseases involving altered sensory perception , such as schizophrenia and autism ( Sorg et al . , 2016 ) .
Experimental procedures followed National and European guidelines , and have been approved by the authors' institutional review boards ( French Ministry of Research and Innovation and Italian Ministry of Health ) . In order to identify PV interneurons we used PvalbCre mice ( Jackson Laboratory Stock Number 008069 ) . To selectively express EGFP in PV-positive cells , we bred PvalbCre mice with mice harboring the R26R CAG-boosted EGFP ( RCE ) reporter allele with a loxP-flanked STOP cassette upstream of the enhanced green fluorescent protein ( EGFP ) gene ( RCE mice , kindly provided by Gordon Fishell , New York University ) , obtaining PvalbCre::RCE mice . Male mice of different postnatal age groups were used , recapitulating developmental stages and accumulation of PNNs around PV cells: <P20 ( before the CP ) , P25-P32 ( CP ) , P40-P60 ( maturation of PNNs ) and >P70 ( adult ) . In vivo experiments were performed on adult C57BL/6J mice older than P70 ( Jackson Laboratory stock number 000664 ) . All mice used in the study were reared in a 12 hr light/dark cycle with food ad libitum . To disrupt PNNs locally in V1 , adult mice underwent a stereotaxic injection of the bacterial enzyme chondroitinase ABC ( ChABC ) from Proteus vulgaris ( Sigma ) or of the phosphate-buffered saline solution ( PBS - control ) . ChABC was prepared beforehand: the powder was reconstituted in 0 . 01% bovine serum albumin aqueous solution for a final concentration of 100 mU/mL . Before each injection , reconstituted ChABC was diluted in a second buffer containing 50 mM Tris , 60 mM sodium acetate and 0 . 02% bovine serum albumin ( pH = 8 . 0 ) in order to obtain a final concentration of 40 U/mL . Adult mice were placed in an anesthesia induction cage ( 3% isoflurane Iso-Vet; 250 mL air ) until insensitive to nociceptive stimuli ( tail pinch ) and then fixed on a stereotaxic apparatus with a mouth mask constantly delivering isoflurane ( 2–2 . 5% isoflurane; 200 mL air ) . The analgesic buprenorphine ( 0 . 1 mg/kg - Buprecare ) was intraperitoneally injected and an ophthalmic ointment was applied on the eyes . Body temperature was constantly controlled and maintained to 37 . 5° using a heating pad . An incision was done in the skin ( the local anesthetic bupivacaine was applied before the incision; 0 . 25% in NaCl 0 . 9% ) and a small hole was drilled in one hemisphere at 2 . 9 mm lateral from Lambda . Small glass capillaries ( external diameter of 40 µm; internal diameter of 60 µm ) , beveled in order to ensure a better penetration into the tissue and therefore produce less damages , were filled with 1 µL ChABC or PBS . Two injections of 350 nL each ( with a rate of 100 nL/min ) were realized at a depth of 800 µm and then 400 µm , with 5 min of interval . The skin was sutured with a non-absorbable 3/0 filament ( Ethicon ) , an antiseptic ( betadine ) was applied on the skin and the mouse gently removed from the frame and kept at 37°C in a heated chamber until full recovery . In vivo experiments or brain slices for electrophysiology were prepared 2–3 days post-injection . The thalamocortical ( TC ) pathway was studied by an optogenetic approach: the light-sensitive opsin channelrodopsin-2 ( ChR2 ) was expressed on the membrane of glutamatergic neurons in thalamic dorsolateral geniculate nucleus ( dLGN ) of adult mice . ChR2 was transduced by stereotaxic injections of an adeno-associated ( AAV ) virus , expressing ChR2 under the promoter of the calcium/calmodulin dependent protein kinase II ( CaMKII ) ( AAV9 . CaMKIIa . hChR2 ( H134R ) -mCherry . WPRE . hGH; Addgene#: 20297 , Penn Vector Core , University of Pennsylvania ) . Viral particles were injected in the hemisphere , in which PNNs were subsequently degraded . The procedure was similar to the ChABC/PBS injections ( see section above ) except for the following points: i ) we used a rigid needle ( Hamilton , 33-gauge , 13 mm , pst4-20° ) , which is more appropriate to target deep structures such as the dLGN; ii ) the coordinates of injection site were 2 . 06 mm posterior to Bregma – 2 mm lateral to midline – 3 . 2 mm deep from the surface of the skull; iii ) one 50 nL injection was performed at a rate of 50 nL/min ( viral titer: 2 . 5 × 1013 particles/mL , diluted at a factor five in fresh PBS ) . After 10–12 days , sufficient for an adequate expression of ChR2 , mice were treated with the ChABC or PBS injections . In some cases , mice were monocularly deprived as described below . AAVs can produce anterograde infection ( Zingg et al . , 2017 ) . However , in our hands , we never detected mCherry- ( and thus ChR2 ) -positive cell bodies in the neocortex ( Figure 2 , Figure 2—figure supplement 6 ) , thus excluding a possible contamination of intracortical responses in our experiments . Occasionally , AAVs spread to the lateral posterior ( LP ) nucleus of the thalamus , which also projects to V1 . However , the axons originating from this thalamic nucleus do not innervate cortical layer 4 , but mainly layers 1 and 5b ( Roth et al . , 2016 ) . Therefore , it is unlikely that we activated axons from LP in our experiments . In some experiments , following the injection of ChABC/PBS , the eyelid of the left eye ( contralateral to the injected hemisphere ) was sutured shut . The anti-inflammatory Diprosone ( 0 . 05% ointment ) was applied on the eye and the superior and inferior eyelids were gently removed with fine scissors . Four stitches were realized with non-absorbable 6/0 filament ( Ethicon ) . 1–2 drops of the anti-inflammatory Tobradex were put in the sutured-eye and the mouse was removed from the frame and kept at 37°C in a heated chamber until full recovery . Mice were killed when signs of infection were observed or if the sutured eye re-opened . Brain slices for electrophysiology were prepared 48 to 72 hr post treatment and surgery . During surgery and recordings , body temperature was maintained constant through a heating pad and respiration and heartbeat were monitored ( heart rate range 420–580 bpm ) . Oxygen-enriched air was administered through all procedures . All necessary efforts were made to minimize the stress of the animals . Mice were anesthetized by intraperitoneal injection of urethane ( 0 . 8 ml/kg in 0 . 9% NaCl; Sigma ) and head restrained during the duration of the recordings . The depth of anesthesia was evaluated by monitoring the pinch withdrawal reflex and other physical signs ( respiratory and heart rate ) . Additional doses ( 10% of initial dose ) were intraperitoneally administered to maintain the level of anesthesia if necessary . As an additional indication , we carefully monitored the electrophysiological signature of urethane induced deep sleep: up/down states frequency , duration and amplitude were similar in the two experimental groups during the recordings ( data not shown ) . A portion of the skull overlying the visual cortex ( 0 . 0 mm anteroposterior and 2 . 9 mm lateral to the lambda suture ) was drilled and the dura mater was left intact . A chamber was created with a thin layer of a dental cement around the edges of the craniotomy . Cortex was maintained constantly wet with ACSF containing ( in mM ) : 120 NaCl , 3 . 2 KCl , 2 CaCl2 , 1 MgCl2 , 1 K2HPO4 , 10 HEPES , 26NaHCO3 , ( pH = 7 . 4 ) . Animals deeply anesthetized under urethane were sacrificed by cervical dislocation without regaining consciousness at the end of the experiment . Local field potentials ( LFPs ) and visually evoked potentials ( VEPs ) were recorded by a glass micropipette ( impedance ~ 2 MΩ , filled with ACSF solution ) positioned into the visual cortex at a depth of 250–300 µm . A common reference Ag-AgCl electrode was placed on the cortical surface in the ACSF bath . Electrophysiological signals were amplified 1000-fold ( EXT-02F , NPI ) , band pass filtered ( 0 . 1–1000 Hz ) , and sampled at 2 kHz . Visual stimuli were generated on a LCD display ( mean luminance at maximum contrast , 3 cd/m2 ) by a MATLAB custom program that exploits the Psychophysics Toolbox , and the luminance of the stimuli was calibrated by means of a radiometer ( Konica Minolta ) . Transient VEPs were recorded in response to the reversal of a checkerboard every 2 s ( spatial frequency 0 . 04 c/deg ) . The response to a blank stimulus ( 0% contrast ) was also recorded to estimate noise . In order to record intrinsic and synaptic properties of L4 neurons of V1 , we prepared acute cortical slices from mice at different postnatal ( P ) ages ( <P20; P25-P32; P40-P60 and >P70 ) , and adult ( >P70 ) mice previously injected with either PBS ( sham ) or ChABC . For these experiments , we used slices cut in the sagittal plane ( 350 μm thick ) . In experiments from deprived-animals as well as in which thalamocortical neurons expressed ChR2 , we cut slices in the coronal plane ( 350 μm thick ) , to localize the binocular zone of V1 ( V1b ) . Animals older than 25 days were subject to intracardial perfusion of ice-cold cutting solution ( see below ) before extracting the brain . This procedure improved the quality of slices and preserved the integrity of the tissue significantly . Animals were deeply anesthetized with pentobarbital ( 50 mg/kg - Euthasol Vet ) and 100 µL of Choay heparine was injected in the left ventricle of the heart before perfusion . Animals were then perfused through the heart with a choline-based cutting solution containing the following ( in mM ) : 126 choline chloride , 16 glucose , 26 NaHCO3 , 2 . 5 KCl , 1 . 25 NaH2PO4 , 7 MgSO4 , 0 . 5 CaCl2 , cooled to 4°C and equilibrated with 95% O2/5% CO2 . The brain was then quickly removed ( for groups of mice aged <P20 , this procedure started right after deep anesthesia ) and immersed in the same cutting choline-based solution ( 4°C , equilibrated with 95% O2/5% CO2 ) . Slices were cut with a vibratome ( Leica VT1200S ) in cutting solution and then incubated in oxygenated artificial cerebrospinal fluid ( aSCF ) composed of ( in mM ) : 126 NaCl , 20 glucose , 26 NaHCO3 , 2 . 5 KCl , 1 . 25 NaH2PO4 , 1 MgSO4 , 2 CaCl2 ( pH 7 . 35 , 310-320mOsm/L ) , initially at 34°C for 30 min , and subsequently at room temperature , before being transferred to the recording chamber where recordings were obtained at 30–32°C . Whole-cell patch-clamp recordings were performed in L4 of the primary visual cortex neurons V1 . In MD animals , cells were patched in the binocular portion V1b . Inhibitory PV-expressing interneurons , labeled with GFP , were identified using LED illumination ( OptoLED , blue , λ = 470 nm , Cairn Research , Faversham , UK ) and by their typical fast-spiking firing behavior in response to depolarizing DC current steps . Excitatory principal neurons ( PNs ) were visually identified using infrared video microscopy by their relatively small size round cell body and no apical dendrites . Accordingly , when depolarized with DC current pulses PNs exhibited a typical firing pattern of regular-spiking cells . We used different intracellular solutions depending on the type of experiment and the nature of the responses we wanted to assess . To study intrinsic excitability , AP waveform and glutamatergic spontaneous transmission , electrodes were filled with an intracellular solution containing ( in mM ) : 127 K-gluconate , 6 KCl , 10 Hepes , 1 EGTA , 2 MgCl2 , 4 Mg-ATP , 0 . 3 Na-GTP; pH adjusted to 7 . 3 with KOH; 290–300 mOsm . The estimated reversal potential for chloride ( ECl ) was approximately −69 mV based on the Nernst equation . To measure GABAergic currents ( both sIPSCs and uIPSCs in paired recordings ) , neurons were recorded using an intracellular solution containing ( in mM ) : 65 K-gluconate , 70 KCl , 10 Hepes , 1 EGTA , 2 MgCl2 , 4 Mg-ATP , 0 . 3 Na-GTP; pH adjusted to 7 . 3 with KOH; 290–300 mOsm . The estimated ECl was approximately −16 mV based on the Nernst equation . Under these recording conditions , activation of GABAA receptors resulted in inward currents at a holding potential ( Vh ) of −70 mV . Experiments using optical stimulation of ChR2-positive thalamocortical fibers were done with a cesium-based intracellular solution containing ( in mM ) : 125 CsMeSO3 , 3 CsCl , 10 Hepes , 5 EGTA , 2 MgCl2 , 4 Mg-ATP , 0 . 3 Na-GTP , 5 QX314-Cl; pH adjusted to 7 . 3 with CsOH; 290–300 mOsm . This solution allowed voltage-clamping neurons at various membrane potentials . ECl- was approximately −63 mV based on the Nernst equation . Voltage values were not corrected for liquid junction potential . Patch electrodes were pulled from borosilicate glass capillaries and had a typical tip resistance of 2–3 MΩ . Signals were amplified with a Multiclamp 700B patch-clamp amplifier ( Molecular Devices ) , sampled at 20–50 KHz and filtered at 4 KHz ( for voltage-clamp experiments ) and 10 KHz ( for current-clamp experiments ) . Signals were digitized with a Digidata 1440A and acquired , using the pClamp 10 software package ( Molecular Devices ) . For intrinsic excitability experiments , neurons were recorded in current-clamp mode . In order to avoid any contribution of differences and variations in the membrane resistance ( Rm ) on the frequency-current curves , the injected current was adjusted in each cell as a function of Rm . This value was determined by the Ohm’s law ( I= ΔV/Rm ) : we injected an amount of current ( I ) to obtain a ΔV of ~10 mV , depending on the actual Rm of each cell , and increasing the amount of depolarizing current to obtain a ΔV of 5 mV , for a total of 15 current steps . Single AP were obtained by injecting brief ( 2 ms ) current steps of increasing amplitude from a Vm of ~ −70 mV in order to determine the minimal current intensity required to elicit a spike in each cell . This current was then injected 20 times and we averaged the trials for each cell from which we calculated the first derivative of the Vm and constructed planar phase plots to extract AP threshold values . Synaptic events were recorded in voltage clamp mode for at least 2–3 min . EPSCs ( spontaneous and miniatures ) were isolated by clamping the cells at −70 mV , using an intracellular solution containing [Cl-] yielding a ECl- ~ −69 mV . In some experiments , we applied the glutamate receptor antagonist DNQX at the end of the recording and we could not detect any residual response ( not shown ) . GABAAR-mediated currents where pharmacologically isolated by applying 10 µM of DNQX while recorded neurons at −70 mV , using an intracellular solution with a [Cl-] yielding a calculated ECl- of ~ −16 mV . For paired recordings , unitary synaptic responses were elicited in voltage-clamp mode by brief somatic depolarizing steps evoking action currents in presynaptic cells . We used a high-chloride intracellular solution ( ECl ~ −16 mV ) , which allowed us measuring glutamatergic ( PN-PV ) and GABAergic synaptic responses ( PV-PN , PV-PV and autapses ) simultaneously . Neurons were held at −80 mV and a train of 5 presynaptic spikes at 50 Hz was applied to infer short-term plasticity of synaptic responses . Optical stimulation: ChR2 activation was obtained by brief ( 0 . 3 and 1 . 0 ms ) light flashes on cortical slices , using a blue LED ( λ = 470 nm; Thorlab ) collimated and coupled to the epifluorescence path of an Olympus BX51 microscope mounting a 60 X water immersion objective ( 1 . 0 NA ) . Light intensity was controlled by the analog output of an A/D card ( Digidata 1440A ) via a power supply ( Thorlabs , LEDD1 ) , and calibrated with a photodiode and a power meter . Light power ranged between 0 . 053 and 1 . 12 mW , over a spot of 0 . 28 mm of diameter . Although thalamocortical axons innervating cortical L4 were severed from their cell bodies , activation of ChR2-expressing fibers generated robust responses onto postsynaptic neurons ( Kloc and Maffei , 2014 ) . Light-evoked responses were recorded in voltage clamp mode in L4 PV cells and PNs . Direct recruitment of cortical neurons was examined in ACSF containing 1 µM of TTX , to remove polysynaptic activity , and 100 µM of the K+-channel antagonist 4-aminopyridine ( 4-AP ) , to enhance axonal depolarization . This approach ensures monosynaptic transmission from thalamocortical afferents selectively , without contamination of polysynaptic activity ( Petreanu et al . , 2009 ) . Yet , in the presence of TTX and 4AP , the depolarization of presynaptic terminals differs from an action potential , as it is likely slower and broader . Consequently , glutamate released by ChR2-positive thalamic fibers is temporally dispersed due to a strong asynchronous release component . We therefore measured charge transfer of thalamocortical synaptic responses on both PV cells and PNs . Disynaptic inhibition was measured in regular ACSF and IPSCs were isolated by holding neurons at the reversal potential for glutamate-mediated responses ( between + 10 and+15 mV , taking into account the liquid junction potential and series resistance ) . In order to minimize response variability due to potential different level of expression of ChR2 across animals and slices , we performed optical stimulations at a light intensity inducing detectable responses with occasional failures . This light intensity was refereed as threshold stimulation ( Gabernet et al . , 2005; Bagnall et al . , 2011 ) . The duration of light stimulations were 0 . 3 ms for feed-forward ( FFI ) , and 1 ms for thalamocortical glutamatergic activation . The stimulus duration was longer in the latter case , due to the presence of TTX and 4-AP . With these constant pulse durations and by varying illumination intensity , the threshold stimulation was determined for each cell . For all experiments , neurons were discarded from the analysis if the access resistance was >30 MΩ . All drugs were obtained from Tocris Cookson ( Bristol , UK ) or Sigma-Aldrich ( St-Louis , USA ) . Thick slices used for electrophysiology experiments ( 350 µm ) were fixed overnight in 4% paraformaldehyde in phosphate buffer ( PB , pH 7 . 4 ) at 4°C . Slices were then rinsed three times at room temperature ( 10 min each time ) in PBS and pre-incubated 1 hr at room temperature in a blocking solution of PBS with 0 . 3% Triton and 10% bovine serum albumin . Slices were then incubated 3 . 5 hr at room temperature in PBS with 0 . 3% Triton and Fluorescein Wisteria floribunda lectin ( WFA-FITC; Vector Laboratories ) which binds to the N-acetylgalactosamime of PNNs . Slices were then rinsed three times in PBS ( 10 min each ) at room temperature , coverslipped in mounting medium and stored at 4°C . Immunofluorescence was then observed with an epifluorescence macroscope ( Nikon AZ100 ) and images were acquired . This post-hoc staining was used to check PNN degradation . Experiments were discarded if a clear disruption of the extracellular matrix was not evident . Parvalbumin staining was realized on 40 µm-thick slices . Slices were fixed overnight in 4% paraformaldehyde in phosphate buffer ( PB , pH 7 . 4 ) at 4°C and rinsed ( 10 min each time ) in PBS . A pre-incubation in a blocking solution of PBT with 0 . 2% Triton and 3% bovine serum albumin was done at room temperature for 1 hr . Slices were incubated overnight ( 4°C ) in the same blocking solution containing the primary rabbit ant-PV antibody ( 1:1000; Thermo Scientific ) . Slices were then rinsed three times in PBS ( 10 min each ) at room temperature and incubated with Cy-2-anti-rabbit antibody ( 1:400; Jackson IR ) for 3 . 5 hr at room temperature . Slices were then rinsed three times in PBS ( 10 min each ) at room temperature and coverslipped in mounting medium . Immunofluorescence was then observed with a confocal microscope ( Olympus , FV-1000 ) or a slide scanner ( Zeiss , Axio Scan . Z1 ) and images were acquired . The in vivo recordings were analyzed by a custom program written in MATLAB ( Source Code 1 ) . Experiments on firing dynamics and unitary paired recordings in slice were analyzed with Clampfit ( Molecular Devices ) , Origin ( Microcal ) and custom-made scripts in MATLAB ( Mathworks; Source Code 2 ) . Firing frequencies were averaged across three trials . Failures of unitary synaptic responses were included in the analysis . Spontaneous and miniatures synaptic events were detected using custom written software ( Wdetecta , courtesy J . R . Huguenard , Stanford University; https://hlab . stanford . edu/wdetecta . php ) based on an algorithm that calculate the derivative of the current trace to find events that cross a certain defined threshold . Amplitude and frequencies of the events were then binned and sorted , using other custom-written routines ( courtesy J . R . Huguenard , Stanford University; https://huguenard-lab . stanford . edu/public/; Ulrich and Huguenard , 1996; Manseau et al . , 2010 ) . AP waveforms were investigated using a phase plot analysis based on a routine developed with MATLAB ( courtesy J . Simonnet; Source Code 2 ) to measure AP threshold , peak and width . Passive properties as well as optical stimulation experiments were analyzed with Clampfit . Light-induced EPSCs were averaged across at least 20 trials and failures were removed from the analysis ( threshold stimulation ) . Data were acquired as a long continuous session together with a synch signal generated by the visual stimulator . Average VEPs were obtained by phase locking the trial averaging on the synch signal . Spectral analysis was performed within the same software package using the Chronux toolbox for multitaper spectral analysis . Different treatments ( ChABC vs . sham ) or experimental conditions ( MD vs . control ) or age groups were allocated randomly across mice . All statistical analysis were performed in Prism ( GraphPad Software , Inc . ) . Normality of the data was systematically assessed ( D'Agostino and Pearson omnibus normality test ) . Normal distributions were statistically compared using paired t test two-tailed or One-way ANOVA followed by Bonferroni’s Multiple Comparison post hoc test for more than two independent groups . When data distributions were not normal or n was small , non-parametric tests were performed ( Mann Whitney test and Kruskal-Wallis test followed by Dunn's multiple comparison test for more than two groups , respectively ) . For the comparison of firing dynamics and short-term plasticity , Two-way repeated-measures ANOVAs were used followed by post-hoc Holm Sidak and Bonferroni’s multiple comparison tests for in vivo and in vitro experiments , respectively . Differences were considered significant if p < 0 . 05 ( *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 , ****p < 0 . 0001 ) . Values are presented as mean ± SEM of n experiments . To measure if MD effectively changed ChABC-mediated effects on synaptic transmission , we used a variation of Cohen's d ( Cohen , 1988; Lakens , 2013 ) , Glass' Δ , which uses only the standard deviation of the control group ( Glass et al . , 1981; Lakens , 2013 ) , when each group has a different standard deviation . Glass′Δ= ( M2−M1 ) /SD1 Where M1 and M2 are the mean values for control and MD animals , and SD1 is the standard deviation of the control animals . | Our brains continue to develop after we are born . As sights , sounds and smells flood our senses , networks of neurons go through periods of rapid rewiring . Known as ‘postnatal critical periods’ , the brain uses these periods to adapt to the signals supplied by our senses . For example , a postnatal critical period exists where infants develop the ability to process what they can see . If their vision is blocked until after the end of the critical period , they may not ever fully gain normal vision . In the outer layer of the brain , known as the cortex , neurons called parvalbumin basket cells appear to help to regulate critical periods . The basket cells synchronize the activity of groups of neurons , creating rhythmic patterns of neural impulses . In the visual cortex these patterns are the brain's way of representing incoming information from the eyes . When a critical period ends , dense nets of protein and sugar start to form around the basket cells in the neural circuit . Dissolving the nets in adult animals re-activates the ability of the circuit to rewire its connections . How the nets limit this rewiring in the first place was not known . Faini et al . have now investigated the role of the nets on the visual cortex of adult mice . Monitoring the activity of neurons revealed that the nets around basket cells ‘muffle’ an important circuit that forms part of the visual pathway . The nets reduce the strength of incoming signals from the eyes before they reach the basket cells . Disrupting the nets allows the visual signals to get through and enables the connections between neurons to respond in a similar way to their behaviour during the postnatal critical period . However , these changes in neural activity were much reduced in mice that had been prevented from seeing out of one eye . This emphasizes the importance of sensory input for rewiring neural circuits . Faini et al . propose that the build-up of nets helps to protect basket cells in the visual cortex from being over-activated by sensory circuits . But this comes at the cost of reducing the ability of the neurons to form new connections , hence making learning and acquiring new skills more difficult . The brains of individuals with psychiatric conditions such as schizophrenia and some forms of autism show disrupted nets around basket cells . Investigating the roles of these nets in more detail could therefore help researchers to develop new treatments for such conditions . More widely , understanding precisely how cortical circuits lose their ability to rewire themselves improves our knowledge of how we learn and store memories . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Perineuronal nets control visual input via thalamic recruitment of cortical PV interneurons |
Common garden experiments that inoculate a standardised growth medium with synthetic microbial communities ( i . e . constructed from individual isolates or using dilution cultures ) suggest that the ability of the community to resist invasions by additional microbial taxa can be predicted by the overall community productivity ( broadly defined as cumulative cell density and/or growth rate ) . However , to the best of our knowledge , no common garden study has yet investigated the relationship between microbial community composition and invasion resistance in microcosms whose compositional differences reflect natural , rather than laboratory-designed , variation . We conducted experimental invasions of two bacterial strains ( Pseudomonas fluorescens and Pseudomonas putida ) into laboratory microcosms inoculated with 680 different mixtures of bacteria derived from naturally occurring microbial communities collected in the field . Using 16S rRNA gene amplicon sequencing to characterise microcosm starting composition , and high-throughput assays of community phenotypes including productivity and invader survival , we determined that productivity is a key predictor of invasion resistance in natural microbial communities , substantially mediating the effect of composition on invasion resistance . The results suggest that similar general principles govern invasion in artificial and natural communities , and that factors affecting resident community productivity should be a focal point for future microbial invasion experiments .
Microbial communities of all types are challenged by the arrival of dispersing microbes , which may displace resident taxa and alter ecosystem functioning ( Amalfitano et al . , 2015; Fernandez-Gonzalez et al . , 2021; Kinnunen et al . , 2016; Litchman , 2010; Mallon et al . , 2015a; Thakur et al . , 2019 ) . A primary way in which established communities defend against such invasion attempts is by achieving a high level of productivity ( strictly defined as community members’ growth rate but often approximated by standing biomass ) in their home environment before invaders arrive ( Huston , 2004 ) . High levels of productivity are concomitant with the depletion of resources , the establishment large populations and the occupation of physical space – and consequently , invaders’ ( in ) ability to survive in that environment ( Ghoul and Mitri , 2016; Huston , 2004; Stubbendieck et al . , 2016 ) . Studies with simplified , synthetic microbial communities have suggested that community productivity is of such importance to a microbial community’s invasion resistance that it is often the main explanation for the effect that the composition of the community inoculated into the microcosm has on invasion resistance ( Eisenhauer et al . , 2013; Hodgson et al . , 2002; van Elsas et al . , 2012; Yang et al . , 2018 ) . In these studies , microcosms containing a sterile culture medium ( e . g . lab broth , autoclaved soil ) are inoculated with artificial microbial communities constructed from different combinations of culturable strains and/or dilutions of natural communities ( typically 90% at each dilutions step ) . Each of these communities – differing in their composition – is then invaded with the same population of microbes to assess the relationship between community composition , productivity and invasion resistance . Such experiments have shown that the presence of individual and/or combinations of resident species ( composition ) with the highest cell densities and/or growth rates ( productivity ) explain most of the variation in invasibility between communities ( Eisenhauer et al . , 2013; Hodgson et al . , 2002; van Elsas et al . , 2012; Yang et al . , 2018 ) . Furthermore , since higher diversity communities are generally more productive ( Bell et al . , 2005b ) , there is often a negative relationship between a resident community’s diversity and invasibility ( e . g . Hodgson et al . , 2002 ) . However , the relationship between a microbial community’s composition , productivity and its invasion resistance is far from clear-cut . For example , Hodgson et al . , 2002 noted that whilst the effects of composition on invasion resistance mainly manifested as productivity differences , there was also evidence of productivity-independent effects of composition ( e . g . niche complementarity , antagonism , facilitation ) . Similarly , De Roy et al . , 2013 found no clear relationship between composition and productivity in artificial microbial communities grown in complex media , but nonetheless a clear , strong relationship between resident community composition and invasion resistance . These results from experiments with artificial communities suggest that despite the predominance of productivity , there is scope for productivity-independent effects of composition on invasion resistance . Such direct effects of composition might become more apparent in natural communities , where niche space is likely to be more highly packed and productivity differences less pronounced ( Eisenhauer et al . , 2012 ) . However , the relationships between microbial community composition and invasion resistance have not been studied in communities whose compositional differences reflect naturally occurring ( rather than lab-engineered ) compositional differences . Here , we use natural microbial communities to test the hypotheses that ( A ) resident community productivity is a primary factor underpinning invasion resistance and ( B ) the effect of resident community composition on invasion resistance is mostly mediated by productivity . We conducted experimental invasions into 680 bacterial communities collected from rainwater pools ( water-filled beech tree holes/phytotelmata ) . These types of model communities have been extensively used in previous microcosm experiments ( Bell et al . , 2010; Bell et al . , 2005b; Fiegna et al . , 2015a; Fiegna et al . , 2015b; Foster and Bell , 2012; Glücksman et al . , 2010; Jones et al . , 2017; Lawrence et al . , 2012; Rivett et al . , 2016; Rivett and Bell , 2018; Scheuerl et al . , 2020 ) and are particularly useful for conducting biodiversity-ecosystem functioning type experiments because diversity is likely to be determined largely by habitat size in nature ( Bell et al . , 2005a; Woodcock et al . , 2007 ) . We grew and invaded these communities in a common garden , complex medium reflecting their natural growth medium ( a beech leaf-based ‘tea’ ) , assessing the survival of two invaders ( Pseudomonas fluorescens and Pseudomonas putida ) at 24 , 96 , and 168 hr ( 1 , 4 , and 7 days ) post-invasion ( Figure 1A ) . Such invasion conditions are similar to that used by van Elsas et al . , 2012 , in which an established community was inoculated with an approximately equal density of invading cells . This scenario is probably relatively common in microbial communities ( Mallon et al . , 2018; Rillig et al . , 2015 ) where an established community may frequently face the invasion of an entire microbial community ( e . g . during faecal deposition on soil , leaf fall into soil or water , flooding of terrestrial ecosystems by aquatic ecosystems ) . For example , as noted by Mallon et al . , 2018 , deposition of faeces into soil may introduce as much as 109 E . coli cells per gram ( Tenaillon et al . , 2010 ) . For our tree hole system in which autumn leaf fall is a main input ( beech tree holes were often found to be full or overflowing with leaves during fieldwork ) , there is evidence that in nature there are 106–109 bacteria per gram of beech leaves ( Holm and Jensen , 1972 ) falling into phytotelma containing approximately 106 bacteria per ml ( Walker et al . , 1991 ) . Thus , we believe our conditions are a somewhat realistic representation of the high-magnitude , low-frequency autumn invasion of bacteria into tree hole microbial communities that have established over the previous year . Following previous experiments , we analysed the relationship between invader survival and the composition of the community added to each microcosm at the start of the experiment . Unlike previous experiments using synthetic microbial assemblages , the inoculated communities were naturally occurring bacterial assemblages . We quantified the starting community composition of the microcosms using amplicon sequencing ( 16S rRNA locus ) to estimate the sequence abundances of OTUs inoculated into the microcosms ( Day 0 ) . As well as estimating the starting community composition in this way ( i . e . 16S genotypes present ) , we also took various phenotypic measurements of the communities at 7 and 14 days of growth in the laboratory microcosms ( the latter being the day of invasion ) . Community productivity was the primary community-level phenotype of interest to us , quantified by measuring cell density and respiration before invasion . Additionally , we took phenotypic measurements related to metabolic activity as an alternative hypothesis to productivity , and to further contextualise any results related to composition and productivity . These measurements were the potential metabolic activity ( ATP levels ) and capacity of the communities to degrade a set of four specific substrates that we expected to be important components of these environments ( cellulose , chitin , xylose , and phosphate ) . The experiment allowed us to identify the main components of invasion resistance in natural communities , and assess whether the effects of resident community composition on invasion resistance primarily manifest as effects on productivity .
To select the components of resident community starting composition and realised community phenotype that best predicted the two invaders’ survival in the microcosms across the three sampling points , we compared explanatory variable importance using random forest regressions ( Materials and methods: Statistical techniques ) . Random forest regressions included all explanatory variables relating to starting composition and realised phenotype . The best representation of composition was selected by testing different dimensionality reduction techniques and selecting the best compromise between variance explained and number of variables . We selected a network approach reducing composition to key species' communities , hereafter termed functional groups ( see Materials and methods: Computational techniques ) . Explanatory power of the random forest regressions ranged from 8 . 14% to 59 . 37% ( pseudo R2 ) , with substantially better performance predicting invasions at 24 compared to 96 and 168 hr , and for P . fluorescens compared to P . putida invasions . This correlated with the number of invasions falling below the detection limit of the invasion assays; invader survival declined with time since invasion , and was also lower for P . putida than for P . fluorescens ( Figure 1B; Materials and methods: Laboratory techniques ) . Note that the random forest method somewhat accounts for the difficulty of differentiating noise from true signal below the detection limit by iteratively sub-sampling which samples are included in each of its constituent regression trees . However , where there are larger numbers of invasions falling below the detection limit , this becomes more difficult and hence explanatory power drops . Nonetheless the method was sufficient to identify the most important explanatory variables for downstream analysis , where we account for the detection limit problem with a sensitivity analysis ( Materials and methods: Statistical techniques ) . Across random forests , phenotypic measures related to community productivity before invasion were by far the strongest and most consistent individual predictors of invader survival ( Figure 2 ) . Measures of community cell yield and respiration consistently had the highest variable importance values – quantified as the relative increase in the Mean Square Error ( % IncMSE ) obtained when the data associated to the variable under analysis is absent from a regression ( Figure 1 ) . The variable importance of cell yield and respiration was generally higher than that of other variables relating to either phenotype or the starting composition of the microcosms . However , the abundance of Functional Group 18 ( containing 12 OTUs assigned to Cedecea spp . , Citrobacter werkmanii , Erwinia persicina , Erwinia rhapontici , Escherichia shigella spp . , Klebsiella pneumoniae , Pantoea agglomerans , Pantoea vagens , Serratia fonticola , Serratia liquefaciens , Serratia quinivorans , and Trabulsiella spp . ) was sometimes of comparable variable importance to the productivity variables . All these variables were approximately linearly and negatively correlated with invader survival; increasing cell yield , respiration and abundance of Functional Group 18 was correlated with lower invader success ( Figure 3 , Figure 3—figure supplement 1 ) . Other variables relating to starting composition were weaker individual predictors of invader survival . As well as Functional Group 18 , other functional groups including Functional Group 5 ( Acinetobacter genomospecies 3 , Acinetobacter towneri , Novispirillum itersonii and Ralstonia pickettii ) and 20 ( Acidovorax spp . , Acinetobacter calcoaceticus , Acinetobacter johnsonii , Aquabacterium spp . , Brevundimonas aurantiaca , Caenimonas spp . , Delftia lacustris , Herbaspirillum rubrisubalbicans , Herbaspirillum spp . , Leptothrix spp . , Massilia timonae , Paucimonas spp . , Phenylobacterium spp , Pseudomonas balearica , Pseudomonas pseudoalcaligenes and Stenotrophomonas maltophilia ) also appeared to be somewhat important for invader survival . There was some evidence of a weak negative relationship between Simpson’s diversity and invader survival ( Figures 2 , 3 Figure 3—figure supplement 1 ) – although the abundance of key functional groups was a more reliable indicator of invader survival across time-points and invaders ( Figure 2 ) . There was little evidence for a strong effect of phylogenetic diversity ( Rao’s quadratic entropy; a phylogenetic equivalent to Simpson’s diversity index ) or phylogenetic distance of the community from the invader ( although these results are subject to the reliability of 16S phylogenetic tree; see Discussion: Unexplained variation ) . Overlaying the abundance of the most important functional group ( Functional Group 18 ) on the above described relationships ( Figure 3 , Figure 3—figure supplement 1 ) suggested that the presence of particular 16S genotypes determined overall community phenotype ( in terms of productivity and invasion resistance realised in laboratory microcosms ) . Highly invadable communities with a low cell density and respiration were also those with a low abundance of Functional Group 18 OTUs and similarly , communities with a low abundance of Functional Group 18 had a lower level of respiration ( blue colours in Figure 3 and Figure 3—figure supplement 1 ) . There was little evidence for strong effects of more specific measures of community phenotype , with ATP activity and the potential to degrade specific substrates having a weak effect overall ( Figure 2; Figure 3—figure supplements 2 and 3 ) . Having identified community productivity as the most important predictor of invasion resistance and identified the abundance of functional groups inoculated into the microcosms as the putative cause of this , we aimed to estimate the extent to which the effect of starting composition on invasion resistance was mediated by productivity . We used structural equation models ( SEMs ) to identify whether starting community composition was associated with lower invader survival solely because it determined the productivity of communities . Having specified the latent variables using the most important explanatory variables ( the functional group abundances and the two measures of cell density; see Materials and methods: Statistical techniques ) , we sought to understand the mediation of composition by productivity across our experiments by using three different SEMs in which these latent ( dependent ) variables Composition and Productivity influence the latent ( independent ) variable Invasion following one of these hypothesis: Model comparison strongly supported the idea that the effect of starting composition on invader survival was strongly mediated by realised productivity , with the No Mediation model being easily rejected ( ΔAIC = 166 . 98 ) . The Partial Mediation model was the best fitting of the three models and was significantly better than the Complete model ( likelihood ratio test , ΔChi2 = 7 . 81 , pval = 0 . 0052 ) - though this should be interpreted with caution as the quality of the models is not optimal ( CFI ~ 0 . 63 ) . Based on the estimated coefficients of the best , Partial Mediation model ( Figure 4 ) , we estimated that a minimum of 48% of the effect starting composition on invader survival ( the product of coefficients C→P x P→I ) was mediated by productivity ( 0 . 47 x −0 . 63 = −0 . 30; which represents a proportion −0 . 30/–0 . 63 = 0 . 48 of the total effect ) . Furthermore , comparing the total ( direct + indirect ) effects of productivity ( P→I + C|P→I ) and composition ( C→I + C|P→I ) revealed that the total effect of productivity ( −0 . 33 + −0 . 3 = −0 . 63 ) was much stronger than that of composition ( −0 . 13 + −0 . 30 = −0 . 43 ) . Given the relatively small ( but significant ) difference between the Partial and Complete mediation model fits , the strong effects of productivity in the Partial model , and the sub-optimal quality of the models , we conducted an additional sensitivity analysis ( see Materials and methods: Statistical methods ) . This sensitivity analysis tested how sensitive model selection was to the invader survival values fell below the strict detection limit of the invasion assay ( see Materials and methods: Laboratory techniques ) . The sensitivity analysis supported the Complete ( the best model in 97 . 4% of permutations ) over the Partial mediation model ( the best model in 2 . 6% of permutations ) , highlighting the marginality of the model selection result . Taking everything into account , we believe the most conservative interpretation of these results , therefore , is that most of the effects of composition occurred through productivity .
There is ample evidence that resident community productivity reduces the potential for the growth of invading species in microbes and non-microbes alike ( Crawley and Heard , 1999; Hodgson et al . , 2002; Jousset et al . , 2011; Kinnunen et al . , 2016 ) . Cell density and respiration were likely important predictors of invasion resistance in our system because they were good proxies for how much the resident community had used the available resources that could otherwise have been used by the invader i . e . resource limitation . Interestingly , productivity at 7 days before ( rather than immediately prior to ) invasion had the most explanatory power in the models – suggesting that resource limitation may have operated in complex ways . One possibility is that those communities that could degrade the resource base most rapidly could also do so most efficiently , leaving less resource for the invader at 14 days . Alternatively , communities that have been at carrying capacity for longer periods may exhibit behaviours that select against invasion . For example , limited resources and/or metabolic stress can select for antibiotic production ( Craney et al . , 2013 ) or the production of inhibitory secondary metabolites ( Watrous et al . , 2013 ) . Structural equation modelling revealed that productivity strongly mediates the effect of starting composition on invasion success - further emphasising the extent to which invasion resistance is mainly a side-effect of productivity . This result is similar to that of previous studies with much more artificial microbial microcosms . Hodgson et al . , 2002 demonstrated a strong negative correlation between P . fluorescens SBW25 invasion success and resident P . fluorescens SBW25 community diversity ( R2 = 0 . 7 ) in 2–5 strain communities grown in a defined laboratory medium , driven by the presence of the most productive P . fluorescens SBW25 strain type . Another more recent study with five-strain communities of Ralstonia grown in defined laboratory media , showed that resistance to invasion by another Ralstonia strain was determined by the presence of one or two strains with the fastest growth rate ( Yang et al . , 2018 ) . What is remarkable about our results is not that we found a similar result to these studies per se , but that we found a similar result in microcosms containing a complex growth medium ( beech leaf tea ) inoculated with communities with richnesses of 66–236 OTUs . In our system as in more artificial microcosm experiments , it appeared that the starting abundance of particular group of bacteria – in our case , a putative ‘functional group’ containing several Enterobacterales ( Functional Group 18 ) – determined the level of invasion resistance realised by the community ( i . e . a dominance/selection effect ) . The most parsimonious explanation for this effect in our system was simply that this was the most productive functional group , and a higher starting abundance gave this group of bacteria a better chance of outcompeting other bacteria and raising the level of growth in the community ( i . e . a priority effect ) . Alternatively , the starting abundance of particular functional groups in a microcosm may determine community productivity in more complex ways – for example , because the most abundant early colonisers cause local environmental changes that affect the growth potential of other colonising bacteria ( e . g . through cross-feeding and competition ) . In dental biofilms it has been demonstrated that a full biofilm can only be achieved with a certain order of colonisation , with metabolically-similar groups of early colonisers causing local environmental changes that allow for subsequent bacteria to colonise , allowing the full development of the biofilm ( Mazumdar et al . , 2013 ) . Similarly , in a study system more similar to our own , it has been shown that the ability bacterial species to colonise a new phyllosphere environment depends on the local density ( and by implication , identity ) of other neighbouring bacteria colonising the environment at the same time – likely because more intense competition for resources reduces the average growth success of the population ( Remus-Emsermann et al . , 2012 ) . Such species interactions/succession-type dynamics are likely to be at least partly driving productivity-mediated effects of composition in our system also – especially as our previous work has shown our ‘functional groups’ correspond to distinct metabolic profiles inferred from predicted metagenomes ( Pascual-García and Bell , 2020a ) . There was also some evidence that community composition impacted invasion resistance independently of the impacts of community productivity , although in a more minor way – as has been seen in a more limited sense in previous studies with artificial communities ( De Roy et al . , 2013; Hodgson et al . , 2002 ) . One explanation for this is that groups of bacteria which were abundant in some communities but not others were functionally distinct; specialising on resource-poor or resource-rich environments ( Pascual-García and Bell , 2020a ) and degrading harder-to-access and/or rarer components of the resource pool . Previous work using the same study system has also shown that the starting composition of communities affects the extent to which recalcitrant substances are able to be used during subsequent community growth ( Rivett et al . , 2016 ) . Another possible explanation is that particular resident species acted as ecosystem engineers ( Pascual-García et al . , 2020 ) . We have already mentioned the possibility of bacteria-driven environmental changes setting limits on the community productivity and leaving certain communities more vulnerable to invasion . Additionally , there may be other ways in which particular species engineer the environment to affect invader survival more directly - such as by modifying the pH of the environment or producing reactive oxygen species . Regarding pH modification , bacterial acidification has been shown to be an important factor in recent microbial invasion experiments in soil microcosms , which showed that Pseudomonas may be prevented from invading by the presence of species that alter the growth medium towards more acidic pH values ( Amor et al . , 2020 ) . Regarding the production of reactive oxygen species , this has recently been demonstrated to be important in in-host studies of invasion resistance; E . faecalis has been shown to protect C . elegans nematodes from Staphylococcus aureus invasion by the producing reactive oxygen species which act as an antimicrobial to kill the invaders ( Ford and King , 2021 ) . However , whilst these more specialised mechanisms of invasion resistance are likely to be fairly common , we emphasise again that they are likely to play a more minor role c more general factors such as community growth rate . We found some evidence for the expected negative relationship between the diversity ( Simpson’s diversity of OTUs ) of the inoculated community and the invaders’ survival ( Figure 3 , Figure 3—figure supplement 1 ) in this semi-natural system , although this was relatively weak ( R2 = 0 . 09 and 0 . 06 for P . fluorescens and P . putida , respectively ) . Although it is hard to make a direct comparison , superficially at least , this weak relationship contrasted with previous experiments using artificial communities constructed from isolates or created by dilution-to-extinction of natural communities , where the relationship was clearer ( De Roy et al . , 2013; Eisenhauer et al . , 2013; van Elsas et al . , 2012 ) . Nonetheless , given that previous experiments often found that the negative diversity-invasion relationship plateaued at higher diversities ( Bonanomi et al . , 2014; Eisenhauer et al . , 2013; van Elsas et al . , 2012 ) , the weak negative slope observed in our high-diversity experiment may simply be because our communities all had relatively high levels of diversity . This suggests that species-like diversity is likely to be a less important predictor of invasion resistance in natural rather than artificial communities . Phylogenetic diversity metrics had even poorer explanatory power , though this result should be interpreted with caution as gene trees based solely on the 16S gene have substantial limitations and estimating the phylogenetic diversity of natural communities is still a challenge ( Rajendhran and Gunasekaran , 2011 ) . Regardless , some preliminary conclusions about the importance of 16S phylogenetic diversity can be made . First , a phylogenetic equivalent of Simpson’s diversity ( Rao’s quadratic entropy ) was not related to invader survival , suggesting 16S phylogenetic alpha diversity is not predictive of invader survival . Secondly , the 16S-based phylogenetic distance of the communities from the invader ( estimated number of mutations to between per nucleotide on their 16S gene ) and the invaders’ success only had a very weak positive relationship with invader survival ( Figure 2 ) . Darwin’s ‘naturalisation hypothesis’ posits that invaders are more likely to establish when invading communities composed of resident species that are , on average , more phylogenetically distant from them , because of greater supposed niche differences ( Darwin , 1859 ) . However , evidence from previous bacterial microcosm experiments with artificial communities to support this hypothesis is mixed ( Gu et al . , 2019; Jiang et al . , 2010; Kinnunen et al . , 2018; Li et al . , 2019 ) , and we did not find strong evidence to support this hypothesis in natural communities . Somewhat related to the phylogenetic hypothesis , by using two different invaders in separate experiments , we also explored whether resident communities were better at resisting an invader when they had a high abundance of conspecific species ( as was the case for P . putida; see Materials and methods: Laboratory techniques ) . Qualitatively , these results corroborate the idea that invaders with conspecific residents ( P . putida ) are less successful – although the general mechanisms of invasion resistance appear similar in both P . fluorescens and P . putida experiment ( distribution of the variable importance values in Figure 2 , shape of relationships between invader survival and the explanatory variables in Figure 3 ) . A more dedicated study with more invader treatments would be needed to study this sub-hypothesis properly . More specific measures of community phenotype related to bacterial metabolism were not informative . Again , this suggested that invasion resistance is primarily the result of generic rather than specific mechanisms . Finally , although we successfully explained much of the variation in invasion using the measured variables , a large component of the variation remained unexplained across the six sub-experiments . Given the predominant role of composition and productivity , we expect that some of this unexplained variation is due to the limitations of our techniques in capturing these key components of the resident communities . Regarding composition , firstly , 16S amplicon sequencing was likely inadequate at capturing the microbial composition of the resident community inocula , and metagenomic/shotgun sequencing might have better characterised the compositional and functional diversity of communities and correlated with invasion . Secondly , although we tried to ensure that only bacteria were present in our communities ( through filtering , passage and anti-fungal treatment , visual and flow-cytometric observation; see Materials and methods: Field techniques ) , it is possible that some non-bacterial microorganisms such as fungi or phage survived and could have affected invader survival . This is difficult to avoid and so future studies might improve explanatory power by keeping the inoculated communities even more intact , characterising compositional effects on invasion at multiple taxonomic and trophic levels simultaneously ( again using metagenomic sequencing ) . Thirdly , additionally characterising community composition nearer to the point of invasion may have improved explanatory power and better contextualised the results . Our unpublished work with a subset of these communities suggests that despite our lab acclimation period ( see Materials and methods: Field techniques ) , some more minor compositional changes still occur in these communities after inoculation ( as should be expected ) , with communities converging towards a similar composition but maintaining a high level of diversity . Characterising these changes and accounting for them in models of future experiments might therefore help confirm , for example , whether differences in communities’ invasion resistance are the result of the rate at which certain functional groups of bacteria take over the community during growth . Regarding productivity , the strong predictive power of productivity at 7 days ( i . e . 7 days prior to invasion ) implies that the temporal dynamics of the community growth is an important component of invasion resistance . Increasing the temporal resolution of growth measurements may therefore have improved the predictive power , as would methods to better distinguish active vs dormant proportions of the population . Aside from productivity and composition , measuring additional variables in the growth period including pH ( Amor et al . , 2020 ) and better inferring functional performance using metagenomics and/or metabolomics could have also helped identify the role of environmental modification in preventing invasions . We also do not exclude the possibility that the unexplained variation may result from the stochastic nature of invasions , since stochastic processes can partly govern invader survival in microbial communities ( Amor et al . , 2020; Kinnunen et al . , 2018 ) . Experiments with artificial communities have suggested that the composition of microbial communities mostly affects invasion resistance by affecting their productivity ( Bonanomi et al . , 2014; Eisenhauer et al . , 2013; Hodgson et al . , 2002; van Elsas et al . , 2012; Yang et al . , 2018 ) . Our experiment lends support to the extension of this claim to natural communities; semi-natural microcosms of bacteria mostly achieve invasion resistance primarily ( though not wholly ) as the result of their productivity , which is the result of the identity and abundance of the bacteria with which they are inoculated . A new generation of microbial invasion ecology experiments with more natural microbial communities ( Bell , 2019 ) might therefore benefit from placing a greater emphasis on productivity where diversity has historically been the focus of experimental designs . Productivity is arguably likely to be more variable than diversity among natural communities , because natural communities are already saturated with species and the population density of a community is more easily affected by environmental disturbances ( e . g . dilution by rainfall , environmental pollution ) . Several lines of evidence in experimental microbial ecology exploring the effect of disturbances on invasion already suggest that disturbances primarily affect invasion success by affecting the population size of the resident community , for example ( e . g . Lear et al . , 2020; Mallon et al . , 2015b ) . Equally , work outside microbial invasion ecology with pathogens frequently suggests that these ‘invaders’ frequently gain entry into host microbiomes by reducing the density of resident competitors ( e . g . Brown et al . , 2008; Wei et al . , 2018 ) . Efforts to characterise resident community growth trajectories before invasion – and what drives differences and disturbances to them - should thus be continued and extended in order to better understand microbial invasion resistance . We believe that taking resident community productivity as a null hypothesis for invasion resistance in this way will reveal more clearly the way invasion resistance emerges and is disrupted in natural microbial communities of all types ( Amalfitano et al . , 2015; Fernandez-Gonzalez et al . , 2021; Kinnunen et al . , 2016; Litchman , 2010; Mallon et al . , 2015a; Thakur et al . , 2019 ) .
All statistical analyses were performed in the R programming language ( R Development Core Team , 2017 ) . Before carrying out analyses , we visually inspected plots of the two-way relationships between invader survival and each of the explanatory variables under various transformations to identify appropriate transformations . The response variable , invader survival , was log10 transformed after adding a pseudocount ( +1 ) to each of the density/ml estimates , in order to deal with zero values . All the explanatory variables except for the three diversity indices ( Simpson’s diversity and phylogenetic distance from the invader ) were log10 transformed . Other R packages used include ‘rstudioapi’ for setting working directory ( Ushey et al . , 2020 ) , ‘tibble’ , ‘stringr’ and ‘plyr’ for data wrangling ( Müller and Wickham , 2020; Wickham , 2019; Wickham , 2011 ) , ‘plotrix’ for standard error calculation ( Lemon , 2006 ) , ‘caret for extracting variable importance values from random forests ( Kuhn , 2020 ) , RColorBrewer for figure production ( Neuwirth , 2014 ) , ape for phylogenetic tree wrangling ( Paradis and Schliep , 2019 ) and ‘grateful’ for generating a bibliography of the packages used ( Rodríguez-Sánchez and Hutchins , 2020 ) . Data and R code is available via OSF and Github , copy archived at swh:1:rev:476dcf8bdf672b9126ccdb8a3b463125721aea89 ( Jones , 2021 ) . | Much like animals and plants , microorganisms such as bacteria and fungi naturally live in communities , where different species exist together and share the same resources . These communities can be quite stable over time and resist the invasion of new species – for example , by collectively and rapidly consuming all the available resources before invaders arrive . The gut microbiome is one example of such a microbial community , but there are many others . There have been many studies of how artificial microbial communities created in the lab resist invasion , but it remains unclear how naturally-occurring microbial communities do so , because they are harder to study in the lab . A leading theory is that certain combinations of microbes ( i . e . communities ) grow and consume resources faster than other combinations – this is known as achieving high productivity . Jones et al . conducted invasion experiments across hundreds of naturally-occurring microbial communities collected from woodland puddles that form in the exposed roots of beech trees . Each community contained different combinations of bacteria , but they all largely survived by breaking down leaf litter , so Jones et al . created a tea from beech leaves in which to grow these natural communities in the lab . The relationships between community composition , productivity and invasion resistance were then assessed using a combination of DNA sequencing , measurements of community growth and measurements of invader survival . Jones et al . found that natural combinations of bacteria that grew well together drove invasion resistance in these communities , mirroring results seen in much more artificial communities grown in the lab . These results suggest that productivity is a key factor underpinning invasion resistance in naturally-occurring microbial communities . This is a useful insight that could shape thinking about how the long-term stability of beneficial microbial communities – such as healthy gut microbiomes – might be improved , and how harmful communities – such as dental plaques – could be destabilised . The next step will be to conduct similar experiments in other natural microbe communities to see how generally applicable these results are . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"microbiology",
"and",
"infectious",
"disease"
] | 2021 | Relationships between community composition, productivity and invasion resistance in semi-natural bacterial microcosms |
Methadone maintenance treatment ( MMT ) can alleviate opioid dependence . However , MMT possibly increases the risk of motor vehicle collisions . The current study investigated preliminary estimation of motor vehicle collision incidence rates . Furthermore , in this population-based retrospective cohort study with frequency-matched controls , opiate adults receiving MMT ( cases ) and those not receiving MMT ( controls ) were identified at a 1:2 ratio by linking data from several nationwide administrative registry databases . From 2009 to 2016 , the crude incidence rate of motor vehicle collisions was the lowest in the general adult population , followed by that in opiate adults , and it was the highest in adults receiving MMT . The incidence rates of motor vehicle collisions were significantly higher in opiate users receiving MMT than in those not receiving MMT . Kaplan–Meier curves of the incidence of motor vehicle collisions differed significantly between groups , with a significant increased risk during the first 90 days of follow-up . In conclusion , drivers receiving MMT have higher motor vehicle collision risk than those not receiving MMT in opiate users , and it is worthy of noticing road safety in such drivers , particularly during the first 90 days of MMT .
Approximately 58 million people worldwide had opioid user in 2019 , with 30 million accounting for opiate users ( United Nations Office on Drugs and Crime , 2020 ) . Iatrogenic opioid dependence has become an epidemic in many developed countries , particularly the United States ( Anderson , 2017 ) ; in other countries , the number of people using heroin is steadily increasing ( United Nations Office on Drugs and Crime , 2018 ) . Opioid dependence clearly presents a public health challenge worldwide . Methadone maintenance treatment ( MMT ) is a primary medication-assisted treatment for opioid dependence ( United Nations Office on Drugs and Crime , 2020; Darke et al . , 2006; Hall et al . , 2000; Kleber , 2008 ) . MMT can reduce opioid and heroin dependence , opioid overdose incidence , criminal activity , all-cause mortality , and HIV and hepatitis C virus transmission ( Degenhardt et al . , 2011; Mathers et al . , 2010; Hickman et al . , 2018; Kamarulzaman et al . , 2016; Chen et al . , 2012 ) . Methadone , a full µ-opioid receptor agonist , can alleviate opioid dependence , both ( methadone and opioid ) of which can influence psychomotor performance and cognitive functioning in healthy volunteers ( Zacny , 1995; Zacny , 1996; Rothenberg et al . , 1977; Rothenberg et al . , 1980; Rothenberg et al . , 1980 ) . Hence , older research described little or no difference in cognitive functioning between MMT patients and healthy controls ( Gordon , 1970; Gritz et al . , 1975; Appel and Gordon , 1976; Grevert et al . , 1977; Appel , 1982; Robinson and Moskowitz , 1985 ) . Given such trends , it is likely that when compared with a higher dose of methadone , MMT patients using a stable dose non-significantly impaired ( Moskowitz and Robinson , 1985; Kelley et al . , 1978 ) . On the contrary , increasing recent evidence has supported that MMT patients using a stable dose may be impaired on a broad set of neuropsychological tests that related to psychomotor speed , decision-making , working memory , and meta-memory ( Mintzer and Stitzer , 2002 ) , information processing , attention , short-term verbal and visual memory , long-term verbal memory and problem-solving ( Darke et al . , 2000 ) and cognitive functioning ( Mintzer et al . , 2005; Verdejo et al . , 2005; Prosser et al . , 2006; Rapeli et al . , 2007; Soyka et al . , 2008; Prosser et al . , 2009 ) , decision-making ( Rotheram-Fuller et al . , 2004; Ersche et al . , 2006 ) , and driving aptitude ( Schindler et al . , 2004; Baewert et al . , 2007 ) . However , MMT itself may elevate the motor vehicle collision risk ( Bramness et al . , 2012; Corsenac et al . , 2012; Leveille et al . , 1994 ) . Thus far , few observational epidemiological studies ( Bramness et al . , 2012; Corsenac et al . , 2012; Leveille et al . , 1994; Babst et al . , 1973; Blomberg and Preusser , 1974; Maddux et al . , 1977 ) have been published on the relationship between motor vehicle collisions and MMT; of these , three were performed in the 1970s and used a case-comparison design to investigate drivers receiving MMT in the United States in small-to-medium-sized cohorts ( Babst et al . , 1973; Blomberg and Preusser , 1974; Maddux et al . , 1977 ) . Their results indicated no significant differences in the rate of motor vehicle collisions between drivers receiving MMT and healthy controls . By contrast , three more recent studies ( Bramness et al . , 2012; Corsenac et al . , 2012; Leveille et al . , 1994 ) found that patients receiving buprenorphine maintenance treatment or MMT had a significantly increased incidence of motor vehicle collisions . Although these studies used medium-to-large-sized cohorts , they neglected some potential risk factors for motor vehicle collisions among drivers receiving MMT , particularly opiate use . Most patients receiving MMT in the aforementioned studies had a history of opioid or heroin dependence . Analgesic opioid users have a 1 . 8 times higher motor vehicle collision risk than do nonusers ( Leveille et al . , 1994; Gibson et al . , 2009 ) . Similarly , heroin users have higher motor vehicle collision risk than do healthy people ( Edwards and Quartaro , 1978 ) . Thus , opiate use history must be identified when estimating motor vehicle collision risks related to MMT use . Few studies have proposed effective measures for motor vehicle collision prevention in drivers receiving MMT . To investigate whether drivers receiving MMT have an increased motor vehicle collision risk , we analyzed nationwide motor vehicle collision incidence rate in three groups as preliminary data: general adult population , adult opiate users , and adults receiving MMT . Furthermore , by using these data , we created population-based matched retrospective cohorts of new opiate users receiving and not receiving MMT . Moreover , we should provide some suggestions for early prevention of motor vehicle collisions in opiate users receiving MMT .
Data were retrieved from the Taiwan National Health Insurance Research Database ( NHIRD ) ( Wu and Lee , 2016; Hu et al . , 2019 ) and six Taiwanese population-based administrative registries , namely the management information system of substitution maintenance therapy , Ministry of Health and Welfare , the road accident registry of injurious crashes , National Police Agency , Ministry of the Interior , and four independent management information systems at the Ministry of Justice , Republic of China ( Taiwan ) . The four information systems were the case management system of Drug Prevention and Control Center , processing system of criminal records , criminal case system of drug case prosecutor briefed the transfer of information , and punitive administrative system for the use of Category 3 or 4 Narcotics ( illicit drugs ) . Preliminary data were independently collected by the aforementioned governmental departments and managed by the Health and Welfare Data Science Center , Ministry of Health and Welfare . Data from different systems were linked using the unique national identification numbers assigned to each citizen in Taiwan . For the consideration of privacy protection , all of the personal identifications were recorded , only authorized researchers were permitted to process databases in a separated designate area , and only statistical results were allowed to be carried out for publications . Personal identifiers were removed after the linkage and before the analysis . We combined and organized the four registry databases from the four independent management information systems at the Ministry of Justice ( Figure 1 ) . The total number of opiate users in the registry from 1956 to 2016 was 107 , 213 . From the four registry databases , we identified new opiate users between 2010 and 2016 ( n = 15 , 996 ) , who were defined the first detection by law enforcement . The new opiate users were excluded if they ( 1 ) were registered at <20 years of age; ( 2 ) used opiate before 2010; and ( 3 ) had incomplete information on age , sex , education status , income , residing in area , etc . Of the new opiate users , we selected those receiving MMT as the ( MMT ) exposed group , who were regular methadone users . The date of first MMT administration was defined as the index date . Other new opiate users not receiving MMT were randomly selected as the ( MMT ) unexposed group after they were frequency-matched to the exposed group at a ratio of 1:2 according to age , sex , and opiate use duration . Thereafter , the index date of two matched unexposed users was the same day as that of the exposed user . The included participants had not been in jail after their index date . Figure 1 depicts the flow of patient selection in the present study . Potential covariates , including history of motor vehicle collisions , driving under the influence ( DUI ) , antidepressant use , and BZD ( including Z-drug ) use before the index date , were included in the analysis ( Bramness et al . , 2012; Corsenac et al . , 2012; World Health Organization , 2018; Chang et al . , 2013; Engeland et al . , 2007; World Health Organization , 2004 ) . The road accident registry of injurious crashes provided information regarding motor vehicle collisions involving personal death , injury , or vehicle damage on Taiwanese roads; the subjects of the current study were focused on the drivers of the road accidents in the database . The main outcome was the incidence of motor vehicle collisions after the index date . All participants were followed until motor vehicle collision after the index date , death , end of follow-up in registry records , or the end of 2016 . Data regarding methadone treatment were extracted from the 2007–2016 management information system of substitution maintenance therapy , which includes information on all prescriptions issued by at least two psychiatrists in Taiwan . This registry omits drug administration information of individuals who were hospitalized or received medications dispensed by outpatient departments . Methadone is dispensed to individuals who meet the criteria for opioid use ( dependence or abuse ) as defined by the International Classification of Disease , Ninth Revision , Clinical Modification ( ICD-9-CM 304 . 00–304 . 03 , 304 . 70–304 . 83 , and 305 . 50–305 . 53 ) . Patients receiving MMT must comply with daily witnessed ingestion under the supervision of a pharmacist or psychiatric nurse and are forbidden to take medication away from treatment sites . Methadone dosing and treatment duration are individualized , varying according to patient tolerance and clinical response across treatment stages ( induction , titration , and stabilization ) according to the Regulations for MMT Guidelines in Taiwan Centers for Disease Control , 2007; Department of Health , Executive Yuan of Taiwan , 2006 . All statistical analyses were performed using SAS ( version 9 . 1 , SAS Institute , Cary , NC , USA ) . Participants were stratified on the basis of age , sex , duration of opiate use , education status , income level , urbanity , history of motor vehicle collisions , DUI , antidepressant use , and BZD ( Z-drug ) use by using the Pearson χ2 test . To determine the independent effect of MMT on motor vehicle collision risk in a previous opiate user , we used a proportional hazard model after adjusting for motor vehicle collisions , as stated previously . The Kaplan–Meier method and log-rank test were used to estimate and compare the incidence of motor vehicle collisions between participants receiving and not receiving MMT , assuming a two-tailed alpha level of statistical significance of 0 . 05 .
Figure 2 shows the crude incidence rates ( CIRs ) of motor vehicle collisions in Taiwan during 2009 to 2016 among the general adult population , adult opiate users , and adults receiving MMT . The CIRs of motor vehicle collisions in the general adult population slightly increased from 19 . 2 per 1000 person-years in 2009 to a peak of 30 . 6 per 1000 person-years in 2014 but steadily decreased to 30 . 3 per 1000 person years in 2016 . Over the follow-up period , the CIRs of motor vehicle collisions in adultopiate users followed a similar trend––from 28 . 2 to 46 . 8 per 1000 person-years . In the general adult population receiving MMT , the highest CIR of motor vehicle collisions was noted in 2012 ( 58 . 7 per 1000 person-years ) , with a wide range of 37 . 5 to 58 . 7 per 1000 person-years . Overall , the CIRs of motor vehicle collisions from 2009 to 2016 were the lowest in the general adult population , followed by those in adult opiate users , and they were the highest in adults receiving MMT . During the study period , the incidence rates of motor vehicle collisions were 6 . 5 and 2 . 2% in the MMT group and controls , respectively . The mean time interval from the index date to a motor vehicle collision was similar between the groups ( 234 . 8 ± 295 . 0 and 226 . 7 ± 287 . 7 days , respectively; p=0 . 886; Table 1 ) . Figure 3 shows the Kaplan–Meier analysis for motor vehicle collision-free survival between opiate users receiving MMT and those not receiving in long-term and short-term follow-up . From the survival curve of 7-year follow-up , the upper panel of Figure 3 shows that both MMT and control groups had similar patterns of motor vehicle collision-free survival . From the further analysis during 100-day follow-up , it began to have no differences between both groups in the first 30 days of MMT intervention , but , after the first 30 days , differences were noted . Notably , a rapid descending curve in the MMT group was discovered in the first 90 days; later , the curve descended steadily . Overall , descending rate of motor vehicle collision-free survival in the MMT group ( 6 . 5%; range 5 . 23–7 . 08%; 95% confidence interval [95% CI] , 5 . 23–7 . 08 ) was significantly higher than that in the control group ( 2 . 2%; range 1 . 90–2 . 41%; log-rank test p<0 . 001 ) ( Figure 3 ) . Univariate analysis results indicated that compared with the controls , opiate users receiving MMT had an increased risk of motor vehicle collisions ( crude hazard ratio [HR] 3 . 00 [95% CI , 2 . 05–4 . 38]; log-rank test p<0 . 001 ) . The risk of motor vehicle collisions was the highest in participants with a history of DUI ( crude HR 2 . 26 [95% CI , 1 . 26–4 . 01]; p=0 . 006 ) , followed by that in those with antidepressant exposure ( crude HR 2 . 25 [95% CI , 1 . 26–3 . 93]; p=0 . 005 ) . Factors predictive of motor vehicle collisions were history of motor vehicle collisions ( crude HR 1 . 70 [95% CR , 1 . 17–2 . 48]; p=0 . 006 ) , BZD ( Z-drug ) use ( crude HR 1 . 62 [95% CI , 1 . 09–2 . 42]; p=0 . 018 ) , and rural location ( crude HR 1 . 68 [95% CI , 1 . 14–2 . 46]; p=0 . 009; Table 2 ) . Multivariate analysis showed that after adjustments for income level , urbanity , education status , history of motor vehicle collisions , DUI , BZD ( Z-drug ) use , and antidepressant use , the adjusted HR for motor vehicle collisions in the MMT group was 2 . 75 ( 95% HR , 1 . 87–4 . 04; p<0 . 001 ) , indicating that opiate users receiving MMT had a significantly increased motor vehicle collision risk . In addition , participants residing in urban areas were 1 . 56 times more likely to encounter motor vehicle collisions ( 95% CI , 1 . 05–2 . 32; p=0 . 027 ) than were controls ( after adjustment; Table 2 ) . No differences were observed in the incidence of motor vehicle collisions between groups with respect to history of motor vehicle collisions ( adjusted HR 1 . 41 [95% CI , 0 . 94–2 . 10] ) , DUI ( adjusted HR 0 . 61 [95% CI , 0 . 33–1 . 13] ) , antidepressant use ( adjusted HR 1 . 70 [95% CI , 0 . 95–3 . 05] ) , or BZD ( Z-drug ) use ( adjusted HR 1 . 25 [95% CI , 0 . 82–1 . 91] ) .
The major strength of our study is the use of a population-based sample of nationwide motor vehicle collision incidences and a case-comparison cohort for motor vehicle collisions . The findings may provide the basis for designing measures to prevent motor vehicle collisions during MTT , particularly in ethnic Chinese populations . Most observational studies showed different risks of motor vehicle collisions between two populations ( the general population and the MMT population ) . We not only added the third population ( the opiate-using population ) in the preliminary study but also controlled for the potentially contributing factor of opioid use in the cohort study of opiate users receiving MMT and those not receiving . Furthermore , the data are not subject to reporting or recall bias because we used high-quality data from six administrative registries and the NHIRD . Nevertheless , we acknowledge several limitations . First , the administrative data was the lack of information on road exposure ( driven hours or distances in a given period ) because unobservable differences between the road exposure may confound the relationship of interest . Motor vehicle collisions in the study were assessed using the police registry . Hence , unreported or minor motor vehicle collisions may not be included in our analysis . Second , the number of opiate users and duration of opiate use may be underestimated . Opiate use is often underreported because it is illegal in Taiwan . Moreover , opioid doses were recognized much higher risks among MMT users than among non-MMT opioid users; hence , opiate dose was uncaptured because it was unavailable in our database . The data about enrollment in a nongovernmental organization institution for rehabilitation and severity of opiate use were lacking . Third , data from drivers’ licenses were lacking in our registry . Although motor vehicle collision risk may be underestimated in the study , we excluded participants aged <20 years to diminish underestimation . Fourth , the degree of exposure to unfavorable road conditions and information regarding unsafe road infrastructure , inadequate traffic laws , car speed , driver fatigue , talking on cell phones , and unsafe vehicles was unavailable in the present study ( World Health Organization , 2018; Chang et al . , 2013; Engeland et al . , 2007; World Health Organization , 2004 ) . Finally , the NHIRD only provides information regarding the dispensing of prescribed medications . Because nonadherence is considered a potential confounder , caution should be exercised when comparing our findings with the results reported by other groups in which data are collected from clinical settings ( Chang et al . , 2013; Engeland et al . , 2007; World Health Organization , 2004; Babst et al . , 1973; Blomberg and Preusser , 1974; Maddux et al . , 1977; Gibson et al . , 2009; Edwards and Quartaro , 1978 ) . In line with previous findings , we provided compelling evidence that opiate users on MTT have a significantly increased motor vehicle collision risk . Individuals receiving MMT should be informed of this risk , so that they can take appropriate measures to prevent motor vehicle collisions , particularly during the first 90 days of MMT and if living in rural areas . | In 2019 , 58 million people were estimated to use opioids – a group of substances that include drugs like heroin and morphine . Dependence on opioids can be managed using a prescribed dose of an opioid called methadone , which is administered through a controlled treatment plan . This so-called methadone maintenance treatment manages withdrawal symptoms in opioid-dependent individuals and can reduce the occurrences of overdose , criminal activity and transmission of diseases such as HIV . However , methadone acts on the same brain receptors as other opioids , and individuals receiving methadone may experience impaired motoric and cognitive functioning , including reduced driving ability . It is therefore important to know whether methadone maintenance treatment may increase an individual’s risk to cause road accidents . To assess motor vehicle collision risk associated with individuals receiving methadone maintenance treatment , Yang et al . analysed data from the Taiwan National Health Insurance Research Database and six Taiwanese administrative registries , including the ministries of health and welfare , interior and justice , and registries in substitution maintenance therapy , road accidents and the National Police Agency . Initial analyses found that individuals receiving treatment had a higher risk to be involved in car accidents than the general adult population or those without methadone maintenance treatment . Further tests showed that individuals receiving treatment were at three times higher risk of collisions than individuals not receiving treatment , particularly in the first 90 days . These findings may help individuals undergoing methadone maintenance treatment manage their risk of motor vehicle collisions . Further investigation is needed to reveal the underlying mechanisms of methadone-related impairment of driving ability . | [
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] | [
"medicine"
] | 2021 | Risk of motor vehicle collisions after methadone use |
We determined whether the mutations found in ovarian cancers could be identified in the patients' ovarian cyst fluids . Tumor-specific mutations were detectable in the cyst fluids of 19 of 23 ( 83% ) borderline tumors , 10 of 13 ( 77% ) type I cancers , and 18 of 18 ( 100% ) type II cancers . In contrast , no mutations were found in the cyst fluids of 18 patients with benign tumors or non-neoplastic cysts . Though large , prospective studies are needed to demonstrate the safety and clinical utility of this approach , our results suggest that the genetic evaluation of cyst fluids might be able to inform the management of the large number of women with these lesions .
Ovarian cancer is the most lethal gynecologic malignancy , with 21 , 290 estimated new cases and 14 , 180 estimated deaths in the United States in 2015 . Approximately 1 . 3% of women will be diagnosed with ovarian cancer during their lifetime ( Howlader et al . , 2014 ) . These cancers commonly present as an adnexal mass with cystic components , but are not associated with specific symptoms . As a result , two-thirds of ovarian cancers are diagnosed at late stage ( Stage III or IV ) , when the 5-year survival is less than 30% ( Howlader et al . , 2014 ) Complicating the diagnosis of ovarian cancer is the fact that ovarian cysts are common in women of all ages , with a prevalence of 35% and 17% in pre- and post-menopausal women , respectively ( Pavlik et al . , 2013 ) . These cysts are frequently benign and found incidentally on routine imaging ( Pavlik et al . , 2013 ) . Though malignancy is an unusual cause of the cysts , 30% of the cysts exhibit radiographic features suspicious for malignancy , such as solid areas or mass ( Pavlik et al . , 2013 ) . In addition to the anxiety that such findings provoke , many women undergo unnecessary surgery for cysts that are not malignant and may not be responsible for the symptoms they have . For example , only 5% of 570 women in a large ovarian cancer screening randomized trial who underwent surgical evaluation actually had a malignancy ( Buys et al . , 2005 ) . Similarly in another study of symptomatic women , only 4% of 197 women who had concerning features on transvaginal ultrasound were ultimately diagnosed with ovarian cancer ( Gilbert et al . , 2012 ) . Compounding this issue is the fact that surgery for ovarian cysts requires general anesthesia and is associated with significant morbidity , causing complications in 15% of women ( Buys et al . , 2011 ) . These complications include damage to nerves and ureters , bleeding , infection , and perforation of adjacent viscera . Furthermore , the procedure often results in hormonal and fertility loss ( in the case of bilateral oophorectomy ) ( Buys et al . , 2011 ) . Even minimal procedures such as ovarian cystectomy can affect fertility in premenopausal women by decreasing follicular response and oocyte number ( Loh et al . , 1999; Demirol et al . , 2006 ) . If a preoperative test could help determine whether the cystic lesion was benign or malignant , unnecessary surgery and its associated complications could be avoided in many patients . This would be particularly helpful for women of reprodutive age who wish to preserve their fertility , as well as women whose medical comorbidities or functional status makes anesthesia and surgery hazardous . Ovarian cysts and tumors are classified as non-neoplastic , benign , borderline , or malignant based on microscopic examination after surgical removal ( Figure 1 ) . Non-neoplastic cysts are by far the most common class of ovarian cysts . They are frequently found in pre-menopausal women , arising when an egg is not released properly from either the follicle or corpus luteum and usually resolve spontaneously within several months ( Christensen et al . , 2002 ) . Benign cystic tumors , such as cystadenomas and cystadenofibromas , rarely progress to malignancy ( Cheng et al . , 2004; Levine et al . , 2010 ) . No genetic alterations have yet been identified in either non-neoplastic cysts or in benign cystic tumors ( Cheng et al . , 2004 ) . Neither of these cyst types requires surgery unless they are symptomatic or large ( Levine et al . , 2010 ) . These cysts can be easily sampled with ultrasound-guided fine-needle aspiration within minutes in an outpatient setting without the need for anesthesia ( Duke et al . , 2006 ) . 10 . 7554/eLife . 15175 . 003Figure 1 . Schematic showing classes of ovarian cysts and the diagnostic potential of the cyst fluid . Ovarian cysts and tumors are currently classified according to microscopic evaluation after surgical removal . The majority of ovarian cysts are non-neoplastic ( often 'functional' in premenopausal women ) . Ovarian tumors with combined cystic and solid components are either benign tumors , borderline tumors , or malignant cancers ( type I or II ) . Only cysts associated with borderline tumors and cancers require surgical excision . We show here that the DNA purified from cyst fluid can be analyzed for somatic mutations commonly found in their associated tumors . The type of mutation detected not only could indicate the type of tumor present but also could potentially inform management . DOI: http://dx . doi . org/10 . 7554/eLife . 15175 . 003 At the other end of the spectrum are epithelial ovarian cancers , which are potentially lethal and unequivocally require surgery . A dualistic model has been proposed to classify these neoplasms ( Kurman and Shih , 2010 ) . Type I tumors are composed of low-grade serous , low-grade endometrioid , clear cell , and mucinous carcinomas . They are generally indolent , frequently diagnosed at early stage ( Stage I or II ) , and develop from well-established precursor lesions ( 'borderline' or 'atypical proliferative' tumors , as described below ) ( Kurman and Shih , 2011 ) . Type I cancers commonly exhibit mutations in KRAS , BRAF , CTNNB1 , PIK3CA , PTEN , ARID1A , or PPP2R1A ( Kurman and Shih , 2010 ) . In contrast , type II tumors are generally high-grade serous carcinomas . They are highly aggressive , most often diagnosed in late stage ( Stage III or IV ) , and have suggested origins from the distal fallopian tube ( Lee et al . , 2007 ) . Type II cancers are the most clinically important group of ovarian cancers , comprising 75% of all ovarian carcinomas and responsible for 90% of ovarian cancer deaths ( Kurman and Shih , 2011 ) . They almost always harbor TP53 mutations ( Cancer Genome Atlas Research Network . 2011 ) . Also unlike type I cancers , which are relatively chemo-resistant and more often treated only with surgical excision , type II cancers respond to conventional chemotherapy , particularly after maximal debulking to reduce tumor burden ( Bristow et al . , 2002; Schmeler et al . , 2008 ) . 'Borderline' or 'atypical proliferative' tumors lie in the middle of this spectrum , between the malignant cancers and the generally harmless non-neoplastic or benign lesions . They are distinguished from carcinomas by the absence of stromal invasion and are precursors of type I cancers . In light of their potential for malignancy , the standard of care for borderline tumors is surgical excision . Following surgery , the prognosis is excellent compared to ovarian cancers , with 10-year survival rates over 94% ( Sherman et al . , 2004 ) . A minor but significant portion of borderline tumors recur after surgery , however , and a subset of the recurrences are found to have advanced to type I cancers ( Shih et al . , 2011 ) . This progression is consistent with molecular findings: serous borderline tumors typically exhibit mutations in BRAF or KRAS , like their malignant counterparts ( low-grade serous carcinoma ) ( Mayr et al . , 2006; Jones et al . , 2012 ) . The presence of a BRAF mutation in a borderline tumor is associated with better prognosis and a low probability of progression to carcinoma ( Grisham et al . , 2013 ) . In contrast , KRAS mutations are associated with the progression to type I cancers ( Tsang et al . , 2013 ) . The examination of fluids from pancreatic , renal , and thyroid cysts is routinely used in clinical management ( Frossard et al . , 2003; Lin et al . , 2005; Volpe et al . , 2007 ) . The fluids have historically been studied by cytology to identify malignant cysts . Ovarian cysts share many features with these other types of cysts , in that they are common , often diagnosed incidentally , and are nearly always benign . However , aspiration of ovarian cyst fluid for cytology is not standard-of-care . From a historical perspective , the difference in diagnostic management partly lies in the fact that cytology has not proven to be very informative for ovarian cysts , particularly for distinguishing benign vs . borderline tumors ( Moran et al . , 1993; Martínez-Onsurbe et al . , 2001 ) . There have also been concerns raised about the safety of ovarian cyst aspiration ( see Discussion ) , and these concerns have not often been raised for other types of cysts . More recently , genetic analysis of specific types of cyst fluids has been considered as an aid to cytology , given that conventional cytology often has limited sensitivity and specificity ( Wu et al . , 2011b ) . Based on the emerging success of the molecular genetic evaluation of other types of cysts , we reasoned that a similar approach could be applied to ovarian cysts . Evaluation of DNA from cells and cell fragments shed into the cyst fluid would presumably allow the identification of tumor-specific mutations . Unlike other , conventional markers of neoplasia such as CA-125 , cancer gene mutations are exquisitely specific indicators of a neoplastic lesion ( Vogelstein et al . , 2013 ) . Moreover , the type of mutation can in some cases indicate the type of neoplastic lesion present ( Wu et al . , 2011a ) . Yamada et al . have demonstrated that mutations can be detected in the cystic fluid of ovarian tumors by querying exons 4 to 9 of TP53 , achieving sensitivities of 12 . 5% and 10% , for borderline and malignant tumors , respectively ( Yamada et al . , 2013 ) . Recently developed , extremely sensitive methods for mutation detection , capable of identifying one mutant template allele among thousands of normal templates in a panel of genes , could potentially increase this sensitivity ( Kinde et al . , 2011; Murtaza et al . , 2013; Newman et al . , 2014 ) . In this study , we have applied one of these technologies to assess mutations in ovarian cyst fluids and to inform the development of tests that could eventually be applied to patients .
DNA was isolated from surgically excised ovarian cysts of 77 women . Ten of them had non-neoplastic cysts , 12 had benign tumors , 24 had borderline tumors , and 31 had cancers ( 13 Type I and 18 Type II ) . Age , histopathologic diagnosis , stage , and other clinical information are provided in Supplementary file 1 . The median amount of DNA recovered from the cysts was 222 ng ( interquartile range ( IQR ) of 53 to 3120 ng ) ( Supplementary file 2 ) . There was no significant difference in the amounts of DNA between borderline tumors and type I or type II cancers . However , the borderline tumors and cancers contained significantly more DNA than the non-neoplastic cysts or benign tumors ( 4453 ± 6428 ng vs . 62 ± 64 ng; p<0 . 001 , Wilcoxon rank-sum test ) . We designed a multiplex PCR-based assay that could simultaneously assess the regions of 17 genes frequently mutated in ovarian tumors . The amount of DNA shed from neoplastic cells was expected to be a minor fraction of the total DNA in the cyst fluid , with most DNA emanating from normal cells . We therefore used a sensitive detection method , called Safe-SeqS ( Safe-Sequencing System ) , to identify mutations in cyst fluid samples ( Kinde et al . , 2011 ) . In brief , primers were designed to amplify 133 regions , covering 9054 distinct nucleotide positions within the 17 genes of interest ( Supplementary file 3 ) . Three multiplex PCR reactions , each containing non-overlapping amplicons , were then performed on each sample . One primer in each pair included a unique identifier ( UID ) for each template molecule , thereby drastically minimizing the error rates associated with PCR and sequencing , as described previously ( Kinde et al . , 2011 ) . Under the conditions used in the current experiments , mutations present in >0 . 1% of template molecules could generally be reliably determined ( Kinde et al . , 2011; 2013; Bettegowda et al . , 2014 ) . We could not perform sequencing on five cysts ( four non-neoplastic cysts and one cyst associated with a borderline tumor ) because there was insufficient DNA ( <3 ng recovered ) , and these were scored as uninterpretable . When this assay was applied to cyst fluid samples with sufficient DNA , no mutations were identified in the 18 cysts obtained from patients with simple cysts ( n = 6 ) or benign tumors ( n = 12 ) ( Table 1 ) . This was in stark contrast to the fluids obtained from the 18 patients with type II cancers , all of which were found to contain a mutation ( Table 1 ) . Ten ( 77% ) of the 13 cyst fluids from patients with type I cancers and 19 ( 83% ) of the 23 cyst fluids from patients with borderline tumors contained at least one detectable mutation . When categorized by the need for surgery ( i . e . , presence of a borderline tumor or a type I or type II cancer ) , the sensitivity of this assay was 87% ( 47 of 54 cysts; 95% confidence interval of 75% to 95% ) and its specificity was 100% ( 95% confidence interval of 74% to 100%; Table 1 ) . 10 . 7554/eLife . 15175 . 004Table 1 . Detection of tumor-specific mutations in cyst fluid . The fraction of samples detected and the median fraction of mutant alleles are indicated , grouped by cyst type , cancer stage , and the need for surgery . DOI: http://dx . doi . org/10 . 7554/eLife . 15175 . 004Fraction of samples detected ( 95% confidence interval ) Median fraction of mutant alleles ( IQR ) Total # of samplesTypeNon-neoplastic0% ( 0–46% ) 0% ( 0–0% ) 6Benign tumor0% ( 0–26% ) 0% ( 0–0% ) 12Borderline tumor83% ( 61–95% ) 2 . 4% ( 1 . 5–10 . 8% ) 23Type I cancer77% ( 46–95% ) 7 . 8% ( 3 . 3–28 . 7% ) 13Type II cancer100% ( 81–100% ) 60 . 3% ( 31 . 3–70 . 8% ) 18Cancer stageEarly ( I and II ) 82% ( 48–97% ) 7 . 4% ( 3 . 0–30 . 9% ) 11Late ( III and IV ) 95% ( 75–100% ) 51 . 2% ( 30 . 2–69 . 5% ) 20Cysts requiring surgeryNo0% ( 0–26% ) 0% ( 0–0% ) 18Yes87% ( 75–95% ) 12 . 6% ( 2 . 7–40 . 2% ) 54 Ovarian cancers are generally detected only late in the course of disease , perhaps explaining the poor prognosis of patients . Accordingly , only 11 of the 31 cysts associated with cancers in our study had early ( Stage I or II ) disease ( Supplementary file 1 ) . As expected , most of these were type I carcinomas ( n = 8 ) . Nevertheless , it was encouraging that mutant DNA could be detected in nine ( 82% ) of these 11 patients ( Table 1 ) . Mutations could be detected in 95% of the 20 patients with Stage III or IV cancers ( Table 1 ) . A variety of control experiments were performed to confirm the integrity of these results . One informative positive control was provided by the analysis of DNA from the tumors , using the identical method used to analyze DNA from the cyst fluids . Fifty-three of the 55 borderline and malignant cases had tumors available for this purpose . Every mutation identified in a tumor was found in its cyst fluid , and vice versa . As expected , the mutant allele frequencies in the tumors were often , but not always , higher than in the cyst fluids ( Supplementary file 2 ) . As another positive control , we used an independent PCR and sequencing reaction to confirm each of the cyst fluid mutations listed in Supplementary file 2 . This validated not only the presence of a mutation , but also confirmed its fractional representation . The median relative difference between the fractions of mutant alleles in replicate experiments was 7 . 0% ( IQR of 3 . 5% to 8 . 9% ) . Finally , four patients were found to have two independent mutations ( Supplementary file 2 ) . For example , the cyst fluid of patient OVCYST 081 , who had a high-grade endometrioid carcinoma , harbored a missense mutation ( R280K ) in TP53 plus an in-frame deletion of PIK3R1 at codons 458 and 459 . The TP53 mutation was found in 3 . 0% of alleles while the PIK3R1 mutation was found in 3 . 7% of the alleles analyzed . Similar mutant allele frequencies among completely different mutations in the cyst fluid of three other patients provided further indicators of reproducibility . All genetic assays were performed in a blinded manner , with the operator unaware of the diagnoses of the patients from whom the cyst fluids were obtained . In addition to DNA from normal individuals used as controls , additional negative controls were provided by the simple cysts and benign tumors . Using the identical assay , none of the DNA from their cyst fluids contained detectable mutations . A final control was provided by the borderline and malignant tumors themselves . In general , only one or two of the 9054 base-pairs ( bp ) queried were mutated in any one tumor ( Supplementary file 2 ) . The other ~9000 bp could then be independently queried in the corresponding cyst fluid , and none of these positions were found to be mutated . The mutant allele fractions in the cyst fluids tended to be higher in the type II cancers ( median of 60 . 3% ) than the type I cancers ( median of 7 . 8% ) or borderline tumors ( median of 2 . 4% ) , though there was considerable overlap ( Table 1; Figure 2A ) . With respect to stage , the DNA from cyst fluids of late-stage cancers had higher median mutant allele fractions ( 51 . 2% ) than those of early-stage cancers ( 7 . 4% ) or borderline tumors ( 2 . 4% ) , but with considerable overlap ( Table 1; Figure 2B ) . 10 . 7554/eLife . 15175 . 005Figure 2 . Mutant allele fractions . ( A ) Classification by tumor type . No mutations were found in the DNA of non-neoplastic or benign cysts . Of the cysts that required surgery , the median mutant allele fraction was higher in the cyst fluids associated with type II cancer ( 60 . 3% ) than type I ( 7 . 8% ) or borderline tumors ( 2 . 4% ) . ( B ) Classification by tumor stage . The DNA from cyst fluids of late-stage cancers had a higher median mutant allele fraction ( 51 . 2% ) than those of early-stage cancers ( 7 . 4% ) or borderline tumors ( 2 . 4% ) . Percent mutant allele is depicted on a logarithmic scale . Horizontal bars depict median and IQR . DOI: http://dx . doi . org/10 . 7554/eLife . 15175 . 005 On the other hand , the type of mutation varied considerably among these cysts ( Figure 3 ) . In type I tumors , the genes mutated were BRAF ( n = 1 ) , KRAS ( n = 5 ) , NRAS ( n = 1 ) , PIK3R1 ( n = 1 ) , PPP2R1A ( n = 1 ) , PTEN ( n = 1 ) , or TP53 ( n = 3 ) . Two distinct mutations were found per sample in three type I cancers . The BRAF mutation ( V600_S605 > D ) was unusual that it resulted from an in-frame deletion/insertion rather than the base substitution ( V600E ) characteristic of the vast majority of BRAF mutations reported in the literature . This mutation has been observed in a papillary thyroid cancer and a cutaneous melanoma ( Cruz et al . , 2003; Barollo et al . , 2014 ) . The deletion results in loss of a phosphorylation site in the activation loop of BRAF , while the insertion of an aspartic acid has been suggested to increase BRAF kinase activity by mimicking an activating phosphorylation ( Davies et al . , 2002 ) . In contrast , all but one type II cancers ( 94% of 18 ) had mutations in TP53; the only exception was OVCYST 073 , a high-grade endometrioid carcinoma . The borderline tumors were distinguished by yet a different pattern from that of the either type I or type II cancers . Of the 19 mutations in borderline tumors , 12 ( 63% ) were BRAF V600E , never observed in type I or type II cancers , and the remainder were at KRAS codon 12 or 61 ( Supplementary file 2 ) . 10 . 7554/eLife . 15175 . 006Figure 3 . Mutated genes found in the cyst fluid samples . Yellow boxes represent mutations with mutant allele frequency ( MAF ) between 0 . 1% and 1%; orange boxes represent mutations with MAF between 1 and 10%; red boxes represent mutations with MAF greater than 10% ( * indicates patients with insufficient DNA for analysis; ** indicates patients with two detected mutations ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15175 . 006 A multivariate analysis was used to identify the most informative molecular features of cyst fluids and to compare them to the commonly used serum biomarkers for ovarian cancer , HE4 ( human epididymis protein 4 ) and CA-125 ( Bast et al . , 1983; Hellström et al . , 2003 ) ( Table 2 ) . We defined 'informative' as indicating a need for surgery ( i . e . , borderline tumors or type I or II cancers ) . The amount of DNA in cyst fluids was generally , but not significantly , higher in the cysts requiring surgery ( p=0 . 69 , Table 2 ) , though there were many cysts not requiring surgery that had higher DNA levels than cysts requiring surgery ( Figure 4A ) . Similarly , the serum CA-125 levels were significantly higher in cysts requiring surgery ( p=0 . 01 , Table 2 ) , but there were many cysts not requiring surgery that had higher levels than those requiring surgery ( Figure 4B ) . Serum HE4 levels were not correlated with cyst type ( p=0 . 92 , Table 2; Figure 4C ) . On the other hand , the presence of a mutation was highly informative for the presence of a cyst requiring surgery in the multivariate analysis , as no mutations were found in cysts not requiring surgery ( p<0 . 001 , Table 2 ) . 10 . 7554/eLife . 15175 . 007Table 2 . Multivariate analysis for markers associated with need for surgery . The presence of a mutation , cyst DNA amount , and common serum biomarkers for ovarian cancer were analyzed for association with cysts that require surgical removal ( Firth’s penalized likelihood logistic regression ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15175 . 007Criteriap valueMutation present<0 . 001Serum CA-125 elevated0 . 01HE4 elevated0 . 92Cyst DNA amount0 . 6910 . 7554/eLife . 15175 . 008Figure 4 . Markers associated with the need for surgery . Cyst DNA amount and levels of commonly used ovarian cancer serum biomarkers are plotted according to the cyst type and need for surgery . ( A ) The amounts of DNA in cyst fluids was generally higher in cysts requiring surgery ( blue ) than those that do not ( red ) , but no significant correlation was found ( p=0 . 69 ) . ( B ) CA-125 levels was significantly higher in cysts that required surgery than those that do not ( p=0 . 01 ) . ( C ) Serum HE4 levels was not correlated with the need for surgery ( p=0 . 92 ) . All values are depicted on a logarithmic scale . P-values were calculated using Firth’s penalized likelihood logistic regression in a multivariate model ( See Materials and methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15175 . 008
Ovarian cancer is the most lethal gynecologic cancer in women . However screening is not recommended by the U . S . Preventive Services Task Force using current diagnostic approaches , which too frequently lead to “important harms , including major surgical interventions in women who do not have cancer” ( Moyer and Force , 2012 ) . Our study was driven by the underlying principle that clonal cancer driver gene mutations are causative agents of neoplasia and absent in non-neoplastic conditions ( Vogelstein et al . , 2013 ) . We have demonstrated here that driver mutations in ovarian tumors are also present in their associated cyst fluids . Moreover , the mutant allele frequencies in the cyst fluids are relatively high ( median 12 . 6% , IQR of 2 . 7% to 40 . 2% ) , facilitating their detection . There were no mutations detected in the cyst fluids that were not also present in the tumors , and vice versa . Also importantly , no mutation was identified in non-neoplastic cysts or cysts associated with benign tumors . Overall , mutations were detected in a major fraction ( 87% ) of cysts requiring surgery but not in any cyst that did not require surgery ( Table 1 ) . Although most ( 87% ) of the 54 cysts requiring surgery had detectable mutations in their fluidic compartment , 7 did not . All of these seven cysts occurred in borderline tumors or type I cancers , while mutations were always ( 100% ) detectable in type II cancers ( Table 1 ) . There are two potential explanations for our failure to detect mutations in these seven cysts . First , it is possible that the mutant DNA concentration in these cysts was below the level of technical sensitivity of our assay ( ~0 . 1% mutant allele fraction ) . We excluded this possibility by evaluating the tumors themselves: no mutations were detected in any of the tumors from these 7 patients . The second , and therefore more likely explanation , is that our panel of 133 amplicons , containing regions of 17 genes , was not adequate to capture the mutations that were present . Unlike type II cancers , which nearly always contain TP53 mutations ( 94% of the type II cancers we studied , for example ) , the genomic landscapes of type I cancers and borderline tumors are more heterogeneous and not as well studied ( Kurman and Shih , 2010 ) . Further genetic evaluation of these tumors should facilitate the incorporation of additional amplicons in the panel to reach higher sensitivities . Nevertheless , the 100% sensitivity for type II cancers in our study is highly encouraging , given that these cancers account for over 90% of ovarian cancer deaths . One limitation of our study is the number of patients evaluated . Though excision of ovarian cysts is one of the most commonly performed surgical procedures , banking of cyst fluids is not common , even in academic centers . Thus , we only had relatively small numbers ( n = 22 ) of non-neoplastic cysts and benign tumors available for study . Even so , the differences in genetic alterations among the various cyst types were striking ( Table 1 ) . Our study will hopefully stimulate collection and analyses of ovarian cyst fluids that will be able to establish smaller confidence limits around the sensitivities and specificities reported in the current study . A potential clinical limitation of our approach is the concern by gynecologists that needle puncture of a malignant ovarian cyst leads to seeding of the peritoneum . This concern is based on inconclusive evidence about the dangers of cyst rupture during surgery and is , at best , controversial . A study of 235 patients who had pre-operative ovarian cyst aspirates reported no recurrence in all 7 malignant cases with a mean follow-up of two years ( Mulvany , 1996 ) . Furthermore , a meta-analysis found no difference in progression-free survival for 2382 early-stage ovarian cancer patients who had experienced intraoperative cyst rupture vs . no rupture ( Kim et al . , 2013 ) . We acknowledge that , under current recommendations , a surgical spill could upstage a localized tumor ( i . e . stage 1A to IC ) and subject the patient to chemotherapy with its associated morbidity . However , the idea that malignant cysts might shed cancer cells if needle-punctured seems incongruent with the widespread practice of laparoscopic removal of ovarian cysts ( Hilger et al . , 2006 ) . Laparoscopic removal of a cyst carries a significant risk of cyst rupture , conceivably higher than when a tiny needle is inserted under ultrasound-guidance ( Havrilesky et al . , 2003 ) . Finally , malignant pancreatic cysts might be viewed to be as dangerous as malignant ovarian cysts ( Howlader et al . , 2014 ) , yet the standard-of-care for pancreatic cysts involves repeated sampling of cyst fluid through endoscopic ultrasound over many years ( Chang et al . , 1997; Eloubeidi et al . , 2003 ) . Though pancreatic cysts and ovarian cysts lie in different anatomical compartments , it is encouraging that aspiration of pancreatic cysts is not associated with an increased risk of disease recurrence or mortality in patients with pancreatic cancer ( Beane et al . , 2011; Ngamruengphong et al . , 2013; Kudo et al . , 2014; Ngamruengphong et al . , 2015 ) . Recent advancements in methods of plugging biopsy tracts to prevent tumor cell dissemination , as well as the increased use of intraperitoneal chemotherapy , further lessen the concern of tumor cell dissemination following ovarian cyst aspiration ( Armstrong et al . , 2006; Tran et al . , 2014; Tsang et al . , 2014; Tewari et al . , 2015 ) . On the basis of these observations and recent developments , we believe that ultrasound-guided aspiration of ovarian cyst fluids would likely be a safe and well-tolerated procedure . As noted in the introduction , as many as twenty patients with benign ovarian cyst lesions undergo surgery for each case of ovarian cancer found ( Buys et al . , 2005 ) . In addition to the psychological impact a potential diagnosis of cancer has on patients , surgery for benign lesions entails considerable cost and morbidity . With the ever increasing sensitivity and use of imaging modalities , the number of patients with incidentally found ovarian cysts is expected to rise . This urgently calls for a diagnostic method that reliably differentiates between the harmless cysts that can be managed expectantly and malignant cysts that require surgical resection . OVA-1 is the only FDA-cleared test to date that aims to distinguish benign versus malignant adenxal mass . It measures levels of five serum markers ( CA-125 , β-2 microglobulin , apolipoprotein A1 , Prealbumin , and transferrin ) and is used to stratify patients who should consult a gynecologic oncologist rather than a general gynecologist for surgery . However the test has a specificity of 43% for ovarian cancer , which is even lower than that of CA-125 alone ( Ueland et al . , 2011 ) . While the test might encourage patients with suspected ovarian cancer to seek specialized care , it would not decrease the number of unnecessary surgeries for women with benign adnexal masses . This study was driven by the need for a biomarker that would help distinguish malignant ovarian tumors from benign lesions and thereby reduce the number of unnecessary surgeries . Such distinction is often difficult based on symptoms and conventional diagnostic criteria . For example , in a large study of 48 , 053 asymptomatic postmenopausal women who underwent ultrasound examination by skilled sonographers , 8 ( 17% ) of the 47 ovarian cancers that were identified occurred in women with persistently normal sonographic findings ( Sharma et al . , 2012 ) . All eight cases were type II cancers , highlighting the potential utility of an additional assay to detect this highly lethal and aggressive type of ovarian cancer . On the other hand , of the 4367 women with abnormal sonographic findings , less than 1% of cases proved to have malignancy upon surgery . Furthermore , of the 32 women with borderline or Type I cancers diagnosed , 22 ( 69% ) had a serum CA-125 level within the clinically accepted normal range ( ≤35 units/mL ) . In our study , 18 of 18 ( 100% ) type II cancers were detectable by virtue of the mutations found in cyst fluid DNA while none of the 18 benign or non-neoplastic cyst fluid contained such mutations . It is also important to note that the readout of our assay is quantitative and not dependent on the skill level of the reader ( in contrast to sonography ) . Finally , the procedure can be performed minimally invasively in an outpatient setting . The goal of our test is not to replace clinical , radiologic , or sonographic evaluation but to augment them with molecular genetic markers . Our study , though only proof-of-principle , illustrates one route to improve the management of patients with ovarian cysts . Genetic analysis is not the only such route; proteomics could also provide clues to the correct diagnosis ( Bandiera et al . , 2013; Kristjansdottir et al . , 2013 ) . One can easily imagine how such additional information could be used to inform clinical practice in conjunction with current diagnostic approaches . For example , if a cyst contained low amounts of DNA , no detectable mutations , and if the patient had low CA-125 levels , our data suggest that it is very unlikely to be a borderline tumor or malignant lesion . Either no surgery , or laparoscopic rather than open surgery , could be recommended for that patient , even if there were some solid component upon imaging . The option to avoid surgery would be particularly valuable for pre-menopausal women who generally have a low risk of ovarian cancer and might wish to preserve their fertility , as well as patients who are poor surgical candidates . However , our assay in its current format cannot completely rule out malignancy because a fraction of early-stage cancer patients did not have detectable mutations in their cysts . Therefore , patients whose clinical and functional status allows them to undergo surgery and anesthesia might still choose to have a surgical procedure . On the other hand , a minimally invasive test that provides additional , orthogonal information to patients and surgeons could inform their decision about the advisability of surgery . Our data suggest that a cyst without any solid component upon imaging , and thereby unlikely via conventional criteria to be malignant , should be removed promptly if the cyst fluid contained a TP53 mutation . Radical , rather than conservative , surgery might be appropriate due to the high likelihood of an aggressive type II cancer . In contrast , if a BRAF mutation is identified , the lesion is presumably a borderline or low-grade tumor; thus conservative rather than radical surgery might be sufficient . Furthermore , given that certain types of ovarian cancers ( type II ) tend to respond well to chemotherapy while others ( type I ) are relatively chemo-resistant , knowing the type of cancer present prior to surgery based on the mutation profile could help guide decisions regarding the use of neoadjuvant chemotherapy . Validation of the current data in a large , prospective trial will be required before the approach can be seriously considered for clinical implementation in a non-research setting .
Cyst fluids were collected prospectively from women presenting with a suspected ovarian tumor . Patients were diagnosed by transvaginal sonography or computed tomography and admitted for surgical removal of the cyst due to suspicious imaging findings by gynecologic oncology surgeons at Sahlgrenska University Hospital , Gothenburg , Sweden . The study was approved by the ethical board of Gothenburg University and patients provided written consent . According to the approved protocol , 15 to 20 mL of ovarian cyst fluid was collected after removal of the cyst from the abdomen . All samples were immediately put in 4°C for 15–30 min , centrifuged for 10 min at 500 g , and aliquoted into Eppendorf tubes . The fluids were transferred to −80°C , within 30–60 min after collection . All histology was reviewed by board-certified pathologists ( Supplementary file 1 ) . Ten , 12 , 24 , and 31 cyst fluid samples from patients with non-neoplastic cysts , benign tumors , borderline tumors , and malignant ovarian cancers , respectively , were assayed in this study . Plasma HE4 concentrations were determined using a commercial HE4 EIA assay ( Fujirebio Diagnostics , PA , USA ) and plasma CA-125 levels were measured using the Architect CA 125 II ( Abbott Diagnostics , IL , USA ) . DNA was purified from tumor tissue ( either freshly-frozen , or formalin-fixed and paraffin-embedded ) after microdissection to remove neoplastic components . DNA was purified from tumors and from 1 mL of each cyst fluid sample using an AllPrep DNA kit ( Qiagen , Germany ) according to the manufacturer’s instructions . Purified DNA from all samples was quantified as previously described ( Rago et al . , 2007 ) . A Wilcoxon rank-sum test was used to compare the amount of DNA in the cancers and borderline tumors with the amount of DNA in the simple cysts and benign tumors . The fraction of samples detected by tumor-specific mutations in the cyst fluid , as well as their 95% confidence intervals , was calculated for each tumor type ( Table 1 ) . When the presence of a mutation in the cyst fluid was used to predict the need for surgery , the sensitivity and specificity of the assay , as well as the 95% confidence intervals , were calculated . Firth’s penalized likelihood logistic regression was used to quantify the association between molecular features of cyst fluids and the need for surgery ( Table 2 ) in a multivariate model . The model predictors included the presence of mutation , log10 ( ng ) of cyst DNA and indicators for normal CA-125 and HE4 values . Normal CA-125 values were defined as <35 U/mL and normal HE4 values were defined as <92 pmol/L and <121 pmol/L for pre- and post-menopausal women , respectively , according to the cutoffs used at the Sahlgrenska University Hospital . Statistical analyses were performed using the R statistical package ( version 3 . 1 . 2 ) . Unless noted otherwise , all patient-related values are reported as means ± SD . DNA from either cyst fluids or tumors was used for multiplex PCR , as previously described ( Kinde et al . , 2011 ) with the exceptions noted below . One-hundred-and-thirty-three primer pairs were designed to amplify 110 to 142 bp segments containing regions of interest from the following 17 genes: AKT1 , APC , BRAF , CDKN2A , CTNNB1 , EGFR , FBXW7 , FGFR2 , KRAS , MAPK1 , NRAS , PIK3CA , PIK3R1 , POLE , PPP2R1A , PTEN , and TP53 . Primer sequences are listed in Supplementary file 3 . These primers were used to amplify DNA in 25 μL reactions as previously described except that 15 cycles were used for the inial amplification ( Kinde et al . , 2011 ) . For each sample , three multiplex reactions , each containing non-overlapping amplicons , were performed . Reactions were purified with AMPure XP beads ( Beckman Coulter , PA , USA ) and eluted in 100 μL of Buffer EB ( Qiagen ) . A fraction ( 0 . 25 μL ) of purified PCR products were then amplified in a second round of PCR , as described ( Kinde et al . , 2011 ) . The PCR products were purified with AMPure and sequenced on an Illumina MiSeq instrument . To better distinguish genuine mutations in the samples from artifactual variants arising from sequencing and sample preparation steps , we used Safe-SeqS , an error-reduction technology for detection of low frequency mutations as described ( Kinde et al . , 2011 ) . High quality sequence reads were selected based on quality scores , which were generated by the sequencing instrument to indicate the probability a base was called in error . The minimum quality score requirements were 15 ( base call accuracy of 97% ) and 20 ( base call accuracy of 99% ) for each of the fourteen UID bases and mutant base ( s ) , respectively . The template-specific portion of the reads was matched to reference sequences using custom scripts written in SQL and C# . Reads from a common template molecule were then grouped based on the unique identifier sequences ( UIDs ) that were incorporated as molecular barcodes . Artifactual mutations introduced during the sample preparation or sequencing steps were reduced by requiring a mutation to be present in >90% of reads in each UID family ( i . e . , to be scored as a 'supermutant' ) . Only mutations with mutant allele frequency ( MAF ) above 0 . 1% , the limit of sensitivity of the assay performed , were considered as positive . Silent or intronic mutations ( other than those in the canonical splice sites ) were not considered to be mutations because their significance is unknown . In addition , DNA from the peripheral blood lymphocytes of healthy individuals was used as a control to identify potential false positive mutations ( see main text ) . Only supermutants in samples with frequencies far exceeding their frequencies in control DNA samples ( i . e . , > mean + 5 standard deviations ) were scored as positive . The original sequencing data on all mutations listed in Supplementary file 2 were manually reviewed to confirm that the mutations were correctly called by the Illumina software . Moreover , each of these mutations was validated by an independent PCR and sequencing reaction . | More than a third of women develop ovarian cysts during their lifetimes . The vast majority of these cysts are harmless , but a small number are caused by ovarian cancers . These cancers often produce no symptoms until the disease has spread throughout the abdomen or to other organs , so many women go undiagnosed until their chances of being successfully treated are low . Currently , there is no reliable way to determine whether an ovarian cyst is cancerous without performing surgery . As a result , many women undergo unnecessary , invasive surgeries for harmless ovarian cysts . Tumors shed cells and cell fragments into any fluid that surrounds them . Fluids from cysts in the pancreas , kidney , and thyroid are routinely examined to identify whether they contain cancerous cells . Now , Wang , Sundfeldt et al . show that ovarian cancers also shed DNA into the surrounding cyst fluid . Furthermore , mutations found in this DNA can provide valuable information about whether the cysts are cancerous . The study was performed by extracting DNA from the fluid in ovarian cysts that had been surgically removed from 77 women . Of these cysts , 10 were harmless cysts , 12 were benign tumors , 31 were invasive cancers , and 24 were so-called borderline tumors , which fall somewhere between the benign tumors and invasive cancers . Only cysts associated with the borderline tumors and invasive cancers need to be surgically removed . Here , Wang , Sundfeldt et al . report that DNA mutations that are characteristic of ovarian cancers were found in 87% of the cysts associated with borderline tumors and invasive cancers . In contrast , these mutations were not found in any of the cysts that do not require surgery . Fluid can be extracted from an ovarian cyst with a needle during an outpatient visit . Therefore , the results presented by Wang , Sundfeldt et al . suggest a relatively straightforward way of testing the DNA from ovarian cysts before deciding whether surgery is really necessary . First , however , larger studies that follow women with cysts over time will be necessary to confirm that this type of testing is effective and safe . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"medicine",
"cancer",
"biology"
] | 2016 | Diagnostic potential of tumor DNA from ovarian cyst fluid |
Pioneering work with nonhuman primates and recent human studies established intracortical microstimulation ( ICMS ) in primary somatosensory cortex ( S1 ) as a method of inducing discriminable artificial sensation . However , these artificial sensations do not yet provide the breadth of cutaneous and proprioceptive percepts available through natural stimulation . In a tetraplegic human with two microelectrode arrays implanted in S1 , we report replicable elicitations of sensations in both the cutaneous and proprioceptive modalities localized to the contralateral arm , dependent on both amplitude and frequency of stimulation . Furthermore , we found a subset of electrodes that exhibited multimodal properties , and that proprioceptive percepts on these electrodes were associated with higher amplitudes , irrespective of the frequency . These novel results demonstrate the ability to provide naturalistic percepts through ICMS that can more closely mimic the body’s natural physiological capabilities . Furthermore , delivering both cutaneous and proprioceptive sensations through artificial somatosensory feedback could improve performance and embodiment in brain-machine interfaces .
The absence of somatosensation profoundly diminishes a person’s ability to move and interact within their environment ( Cole and Cole , 1995; Sainburg et al . , 1993 ) . Even with intact vision and hearing , which can provide sensory information about body position , movement , and interaction , basic behaviors such as walking or reach-and-grasp require substantially greater cognitive load without somatosensory feedback . The severity of these deficits underscores how deeply integrated cutaneous and proprioceptive somatosensations are in the neural control of movement , and motivates the problem of restoring sensation when it is missing . However , the complexity of the somatosensory circuit , and the difficulty of writing information into this circuit with sufficient integrity , have posed significant challenges . Recent advances in brain-machine interface ( BMI ) technology have led to renewed efforts in this area , under the hypothesis that providing closed-loop motor-sensory control and feedback pathways could lead to vital increases in performance ( Bensmaia and Miller , 2014 ) . Intracortical microstimulation ( ICMS ) is a promising technique for implementing a return path in which electrical stimuli are written directly into the somatosensory cortex through implanted electrode arrays . Non-human primates ( NHPs ) successfully incorporated ICMS information to perform discrimination , detection tasks ( Romo et al . , 1998; Romo et al . , 2000; Tabot et al . , 2013; Dadarlat et al . , 2015 ) and as sensory feedback for brain control in BMI tasks ( O'Doherty et al . , 2011; Klaes et al . , 2014 ) , and recent human studies have provided insight into the feeling and perception of the sensations produced through ICMS ( Flesher et al . , 2016 ) . However , qualities ascribed by human subjects to these sensations ( e . g . , ‘tingling’ or ‘buzzing’ ) have been mostly artificial in nature ( Johnson et al . , 2013; Flesher et al . , 2016 ) , and it is as yet unclear what range of sensations could be elicited through ICMS . Here , we present novel findings from two experiments: one which tested each electrode over a range of amplitudes with fixed frequency , and one which tested a subset of electrodes over a range of amplitudes and frequencies . We found reliable elicitation of natural cutaneous and proprioceptive sensations spanning a range of stimulus amplitudes and frequencies , obtained from stimulation in S1 of a single human subject ( participant FG , Figure 1; see Materials and methods ) with a C5-level spinal cord injury . We further show that current amplitude , not frequency , of the electrical stimulus differentiates the modality ( i . e . , cutaneous or proprioceptive ) of the elicited percept at some stimulation sites .
In Experiment 1 , over an eight-week period , electrical stimuli were tested across a range of current amplitudes between 20–100 µA , with pulse frequency held constant at 150 Hz ( see Materials and methods ) . Stimulation through 46/96 electrodes ( 48% ) prompted at least one response , and there were in total 381 reported sensations out of 1229 non-catch trials ( see Materials and methods ) . There was weak correlation between the number of electrodes that elicited a sensation and the current amplitude ( r = 0 . 34 , p=0 . 42 , Pearson linear correlation ) . Additionally , we found no correlation between electrode impedance and the likelihood of elicited percepts ( p=0 . 80 , Pearson linear correlation coefficient ) , pooling all electrode responses over all days . Furthermore , there was no significant difference in the aggregate impedances of either electrodes that produced or did not produce percepts ( p=0 . 707 , Kolmogorov-Smirnov two-sample test ) . No false positives were reported in any catch trials , and we found no effect of trial history in the proportion of reported sensations during stimulation ( see Materials and methods ) . The stimulation did not trigger any painful sensations , and no adverse events occurred during any of the sessions . Receptive fields along the upper arm , forearm and hand corresponded to coarse somatotopical organization in the corresponding stimulation sites . Figure 2 shows the most frequently reported receptive field and sensation modality for each electrode across all trials . Of the 46 electrodes with responses , 32 evoked percepts in the upper arm , 18 in the forearm , and two in the hand ( palmar surface of digits and a finger pad ) . In agreement with previous reports , stimulation could produce percepts with variably-sized receptive fields in different electrodes ( Flesher et al . , 2016 ) . For the majority of electrodes ( 24/46 ) , receptive fields were reported in the same body region ( i . e . upper arm or forearm ) or in the same plane ( i . e . anterior or posterior ) across all tested amplitudes . Coarse somatotopy was present between the medial and lateral arrays ( Figure 2B ) ; the medial array was more likely to have reliable receptive fields in the anterior upper arm ( 46% of medial-array receptive fields ) , while stimulation on the lateral array induced sensation more frequently on the posterior forearm ( 51% of lateral-array receptive fields ) . However , there was no clear somatotopical organization within each array as previously reported ( Kim et al . , 2015a; Kaas , 1983; Flesher et al . , 2016 ) . The coarse somatotopy found across arrays but not within arrays , could be due to the small area of cortex sampled by the implants , and the fact that the implants predominantly covered upper arm and forearm , areas with a less established somatotopic map ( Kaas et al . , 1979; Kaas , 1983 ) . Another plausible explanation is that the topography in somatosensory cortex has been remapped after injury ( Kaas et al . , 1983; Florence et al . , 1998; Moore et al . , 2000 ) FG has reported a wealth of qualitative sensations induced by ICMS ( Table 1 ) . Unlike paresthetic sensations experienced post-injury , these naturalistic responses were broadly characterized as cutaneous ( e . g . squeeze ) or proprioceptive ( e . g . rightward movement ) , and as being subjectively similar to sensations experienced prior to injury . At his own discretion , the subject used single-word descriptors to characterize the perceived sensations as accurately as possible . Single-word descriptors have the advantage that they can be compared across large data sets or subjects . However , as experimental advances continue to push the capabilities of ICMS , responses could become more complex and future studies might benefit from more structured descriptors , which take into consideration the complexity of these sensory experiences ( Darie et al . , 2017 ) . We found that 18 electrodes had cutaneous-only responses across all tested current amplitudes , while six electrodes had proprioceptive-only responses; the rest of the electrodes ( 22/46 ) had mixed responses , where the perceived modality ( cutaneous or proprioceptive ) varied as stimulus parameters changed . Of these mixed-response electrodes , 45% evoked mostly cutaneous sensations , 32% evoked mostly proprioceptive sensations , and 23% had an equal number of cutaneous and proprioceptive sensations ( Figure 2B ) . This pattern of cutaneous and proprioceptive evoked sensations complements recent reports of multimodal ( i . e . cutaneous and proprioceptive ) neurons throughout S1 ( Yau et al . , 2016; Kim et al . , 2015b ) . While prior single-unit experiments have defined maps from single neurons to specific unimodal receptive fields ( Kaas et al . , 1979; Kaas , 1983; Friedman et al . , 2004; Romo et al . , 2000 ) , the above results suggest that more than one variable may be represented when mapping with ICMS . This finding may be the product of different mechanisms by which receptive fields are observed through recording versus stimulation , and could be an important topic for future work . We found a significant difference between the amplitudes that elicited cutaneous or proprioceptive responses , with the distribution of proprioceptive responses skewed towards higher amplitudes ( Figure 3A ) , when pooling across all electrodes and amplitudes that produced a sensation ( p=0 . 039 , Kruskal-Wallis nonparametric ANOVA , χ2 ( 1 , 378 ) =4 . 41 , proprioceptive responses N = 79 , cutaneous responses N = 302 ) . To assess consistent current delivery across all electrodes , we measured electrode impedance at the beginning of every session and found no significant difference when comparing proprioceptive or cutaneous responses ( p=0 . 237 , χ2 ( 1 , 378 ) =1 . 39 ) and , furthermore , we found no significant difference between the impedance of proprioceptive- and cutaneous-only ( p=0 . 922 , χ2 ( 1 , 155 ) =0 . 01 ) or mixed-response electrodes ( p=0 . 372 , χ2 ( 1 , 221 ) =0 . 8 ) . To account for potential bias from an uneven distribution of responses across amplitudes , we compared the proportion of proprioceptive and cutaneous responses in a bootstrapped resampling ( N = 10000 ) , in which each repetition drew 15 responses at each amplitude from all data pooled across days ( Figure 3B ) . We observed a clear relationship between the number of proprioceptive and cutaneous responses and stimulation amplitudes , measured through overall positive slopes in the 1st-order polynomial fit at each iteration for proprioceptive responses , and negative slopes for cutaneous responses ( Figure 3C ) . Experiment 2 tested a subset of 5 electrodes with robust responses across all tested amplitudes in Experiment 1 ( Figure 2B , Figure 3D ) . In a pseudorandomly-interleaved fashion , we stimulated each electrode with five amplitudes ( range 20 to 100 μA ) at six different frequencies ( range 50 to 300 Hz ) over the course of three consecutive days ( see Materials and methods ) . We reproduced the effect of amplitude on sensation modality , either proprioceptive or cutaneous , when pooling across all responses ( p=2×10−5 , χ2 ( 1323 ) = 18 . 17 , Figure 3E ) . Similar to the main mapping task , we did not find any significant effect on modality due to electrode impedance ( p=0 . 305 , χ2 ( 1323 ) =0 . 8 ) . Furthermore , there was no significance when testing the effect of frequency in eliciting proprioceptive or cutaneous responses ( p=0 . 22 , χ2 ( 1323 ) = 1 . 48 ) . This amplitude-specific effect on sensation modality is perhaps surprising given the more commonly observed effect of frequency and pulse-width modulation on sensation in the periphery ( Graczyk et al . , 2016 ) . Although there is evidence of tactile and proprioceptive inputs co-modulating S1 firing activity ( Kim et al . , 2015b ) , we are unaware of any reported effect of amplitude or frequency thresholding for different sensory modalities in the CNS . Proprioceptive sensations are commonly thought to derive from activity in areas 2 or 3a , while cutaneous sensations more likely correspond to activity in areas 3b and 1 . From topographical features , we estimate our implants lie in area 1; however , with evidence of interindividual variability in the microstructural organization within S1 ( Geyer et al . , 1999 ) , and the potential for functional reorganization after injury ( Kaas et al . , 1983; Florence et al . , 1998 ) , it is possible that higher current amplitudes could increase the effective range of stimulation to include sensory areas 3a or 2 . Moreover , given the receptive fields activated during stimulation , the two implants are well within the arm and forearm regions of S1 , which might receive a larger ratio of proprioceptive-to-cutaneous signals than hand regions ( McKenna et al . , 1982 ) , making it more likely to activate these different modalities with ICMS . FG also provided subjective measures of sensation intensity and duration . Sensation intensity was ranked from 1 to 10 ( weakest to strongest ) . In Experiment 1 , we found a strong positive correlation between intensity and amplitude ( r = 0 . 2 , p=2 . 1×10−5 , Pearson linear correlation coefficient ) , with an intensity of 2 . 4 ± 1 . 9 a . u . ( mean ±s . d . ) for 20 μA and 4 . 0 ± 2 . 1 a . u . for 100 μA , with a slope of 0 . 02 ( 1st-order polynomial , least squares fitting ) . As subjective measures of intensity are most likely sensitive to day-to-day variability , in post-hoc analysis we also normalized intensity values within each session ( see Materials and methods ) . We measured a negative correlation between the current amplitude and the standard deviation of the intensity ( r = −0 . 6 , p=0 . 12 ) . Duration of the percept was recorded for each response as either short ( sensation lasts only briefly at the onset of stimulation ) , medium ( sensation persists throughout the stimulation but not for the full length of the stimulation ) or long ( sensation lasts the full duration of the stimulation ) . The majority of responses were short ( N = 225 ) , followed by medium ( N = 122 ) with very few long responses ( N = 12 ) . Stimulus duration was not recorded for 22 responses of the 381 responses . For Experiment 2 this trend was replicated ( N = 268 , 55 and 1 . Short , medium and long , respectively ) . There was no relationship between duration of the sensation and either amplitude of stimulation ( p=0 . 1 , χ2 ( 1323 ) =4 . 53 ) or frequency of stimulation ( p=0 . 2 , χ2 ( 1323 ) =2 . 83 ) . To our knowledge , this is the first report in human of replicable , purely naturalistic proprioceptive and cutaneous sensations induced through ICMS . Stimulation over a wide range of amplitudes and frequencies generated qualitatively diverse sensations , although percept modality was strongly linked to variations in amplitude . Pairing these natural sensations with BMIs create a unique opportunity to explore how effectively they can be incorporated in a closed-loop BMI system . For example , the ability to evoke proprioceptive sensations could allow the subject to interpret position- or movement-related information , as previously reported in primate studies ( Tomlinson and Miller , 2016; Dadarlat et al . , 2015 ) , while eliciting cutaneous sensations could improve our ability to deliver richer somatosensory feedback for object manipulation . Together these somatosensory signals have the potential to improve performance and embodiment when using a BMI-controlled external device .
We recruited and consented a 32-year-old male participant ( FG ) with C5-level complete spinal cord injury , 1 . 5 years post-injury , to participate in a clinical trial of a BMI system with intracortical recording and stimulation . The subject has residual sensation in the anterior-radial section of his upper arm , and some residual sensation in the posterior-radial section of his upper arm and forearm , which present as paresthesias . All procedures were approved by the Institutional Review Boards ( IRB ) of the University of Southern California ( USC ) and Rancho Los Amigos National Rehabilitation Hospital ( RLA ) . The implant procedure occurred at Keck Hospital of USC , and study sessions took place at RLA . Surgical planning followed the protocols described in ( Aflalo et al . , 2015 ) , with an additional task for identifying an implant location within somatosensory cortex . In this task , a visual cue prompted the experimenter , who was standing next to the MRI , to reach into the MRI machine with a wooden pole and repeatedly press at one of three points on the subjects right upper limb where he previously reported residual paresthetic sensation; biceps , forearm and thenar eminence . The subject was instructed to attend to any residual sensation he felt at each location and report the number of times the experimenter touched him on the cued location ( Kastner et al . , 1998; Staines et al . , 2002 ) . After functional imaging , three target locations for electrode placement were identified; supramarginal gyrus ( SMG ) , ventral premotor cortex ( PMv ) and primary somatosensory cortex ( S1 ) . One 96-channel , platinum-tipped Neuroport microelectrode recording array ( Blackrock Microsystems , Salt Lake City , UT ) was implanted in each of SMG and PMv . Two 7 × 7 SIROF ( sputtered iridium oxide film ) -tipped microelectrode arrays ( with 48 physically-connected channels each ) were implanted in S1 . The SIROF-tipped electrodes have lower impedance than the platinum-tipped electrodes , and thus are better suited to stimulation . All stimuli consisted of biphasic , charge-balanced , cathodic-leading pulses , with 200 µs width per phase , 53 µs interphase interval , and one-second stimulus duration delivered to a single electrode on the S1 array only . The maximum charge delivered per phase was 20 nC . We selected these parameters , and set electric charge limits according to safe ranges shown in ICMS studies with NHPs ( Kim et al . , 2015a ) . Stimulation was delivered with a Blackrock CereStim device , and stimulation parameters were set and delivered using the CereStim API through MATLAB ( The Mathworks Inc , Natick , MA ) software ( MATLAB code in Source code file 1 ) . Experiment 1: After initial assessment of implant viability , we evaluated the effects of stimulation parameters through a percept-detection task . For this primary mapping task , each of the 96 stimulation electrodes were evaluated at eight amplitudes: 20 , 30 , 40 , 60 , 70 , 80 , 90 , and 100 μA , at 150 Hz . The subject was seated in a wheelchair approximately 1 . 5 meters from a TV screen . The subject was instructed to look at a fixation point in the middle of the screen throughout the experiment . In each trial , after a three-second inter-trial interval , the subject was presented with a large purple circle on the screen indicating that an electric stimulus was being delivered . Then , after a one-second delay , an auditory cue signaled the subject to report whether he felt any sensation . When a sensation was perceived , the subject reported its location on a body and hand map , with anterior and posterior views , by referencing a fine overlaying grid ( Figure 1 ) . The subject also reported qualitative characteristics including the perceived stimulus intensity , the perceived duration of stimulation , and a description of the sensation ( Table 1 ) . Sensations closer in nature to tactile stimuli were classified as cutaneous , and those triggering a feeling of movement or change in position were classified as proprioceptive . To complete the mapping of amplitude , we ran trial blocks where we randomly selected a subset of electrodes . Each block contained three replicates of stimulation per parameter , per electrode . An additional set of trials , numbering 10% of the total trials in a block , were added as ‘catch’ trials , where the visual stimuli on the screen and auditory response cue remained identical but the stimulation did not occur . Catch trials were randomly interleaved among the normal trials . In each block , trials were ordered such that stimulation did not occur to the same or adjacent electrodes concurrently . Experiment 2: For the second mapping task , five electrodes were selected for further evaluation at different amplitudes and frequencies . All the phases of the task and other stimulation parameters were the same as in the previous mapping task . The subset of electrodes selected for this task were those that exhibited the most reliable responses in the first mapping task . We varied the current amplitude ( 20 , 40 , 60 , 80 , 100 μA ) and pulse frequency ( 50 , 100 , 150 , 200 , 250 , 300 Hz ) , and tested each amplitude-frequency combination six times per electrode . The full dataset was obtained over three consecutive days . In each day , each of the five electrodes received two replicates of all possible amplitude and frequency combinations . The order of electrode stimulation was determined pseudorandomly . Throughout the analysis we used the Kruskal-Wallis nonparametric ANOVA statistical test . We calculated correlations between responses using the Pearson linear correlation coefficient . To examine whether response history had a significant effect on the proportion of reported sensations ( de Lafuente and Romo , 2005 ) , we looked at differences between the distribution of reported sensations during stimulation for three conditions: all trials , trials after a reported sensation ( hit ) and trials after no reported sensation ( miss ) . We estimated these distributions for each amplitude in a given experimental session across all tested electrodes , and used Kruskal-Wallis nonparametric ANOVA with Dunn-Sidak multiple comparisons correction to test for significance at each amplitude . Furthermore , we generated a shuffle distribution of probabilities with N = 10 , 000 permutations for hits following a hit or a miss for each amplitude . We found no significant difference between the shuffle distributions and the empirical data , with the actual proportion being within the 5th-95th percentile range of the shuffle distribution . For the bootstrapped resampling of proprioceptive and cutaneous responses in Experiment 1 , we drew 15 samples at each iteration from the total responses at each amplitude ( range 21–93 responses across all amplitudes ) . Where normalized intensity data are reported , we rescaled the raw intensity ( range 1–10 ) to a normalized scale ( range 0–1 ) for each day by subtracting the minimum and then dividing by the maximum . Raw data for all analysis presented in this manuscript can be found as downloadable source data ‘Responses to single-electrode stimulation’ . Specific details can also be found in the first sheet of the raw data file . | Nerves throughout the body send information about touch , temperature , body position and pain through the spinal cord to the brain . A part of the brain called the somatosensory cortex processes this information . Spinal cord injuries disrupt these messages . Even though the somatosensory cortex has not been damaged , sensation is lost for the affected body areas . No treatment exists to repair the spinal cord so the loss of sensation is permanent . Applying electricity to the somatosensory cortex can produce artificial sensations . Scientists are testing this approach to restore a sense of touch for people with spinal cord injury . Early experiments show that using different patterns of electrical stimulation generates unnatural sensations in different body parts . People receiving the stimulation describe it as tingling or shocks . Scientists wonder if they can improve the technique to mimic feelings like touch or body position to make it easier for people with a spinal injury to move or use prostheses . Now , Armenta Salas et al . generated more natural sensations in a person with a spinal cord injury . Instead of taking the usual approach of delivering large currents to the surface of cortex , they inserted small electrodes into the inside of the cortex to stimulate it with small currents . In the experiments , electrodes were implanted in the somatosensory cortex of a volunteer who had lost the use of his limbs and torso because of a spinal injury . Armenta Salas et al . applied different patterns of electrical stimuli and the volunteer reported what they felt like . The patient described sensations like a pinch or squeeze in the forearm or upper arm with certain patterns . In some cases , the patient reported the sensation of the arm moving with stronger electrical currents . The experiments show that electrical stimulation of the brain can recreate some natural sensations . These sensations could help patients using robotic or prosthetic arms become more dexterous . It might also help patients view artificial limbs as part of their bodies , which could improve their sense of wellbeing . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience"
] | 2018 | Proprioceptive and cutaneous sensations in humans elicited by intracortical microstimulation |
The mechanism of transport through the Golgi complex is not completely understood , insofar as no single transport mechanism appears to account for all of the observations . Here , we compare the transport of soluble secretory proteins ( albumin and α1-antitrypsin ) with that of supramolecular cargoes ( e . g . , procollagen ) that are proposed to traverse the Golgi by compartment progression–maturation . We show that these soluble proteins traverse the Golgi much faster than procollagen while moving through the same stack . Moreover , we present kinetic and morphological observations that indicate that albumin transport occurs by diffusion via intercisternal continuities . These data provide evidence for a transport mechanism that applies to a major class of secretory proteins and indicate the co-existence of multiple intra-Golgi trafficking modes .
Nearly one third of the eukaryotic proteins are synthesized at the endoplasmic reticulum ( ER ) and then transported to their cellular destinations through the secretory pathway . Over the years , the general organization of membrane transport along the secretory pathway has been gradually unraveled ( Mellman and Simons , 1992; Mellman and Warren , 2000 ) , and many of the underlying molecular components have been identified ( Rothman , 2002; Schekman , 2002; Emr et al . , 2009 ) . Some key questions , however , remain unresolved ( Pfeffer , 2007; Emr et al . , 2009; Glick and Luini , 2011 ) . A central issue is how cargo proteins traverse the Golgi complex ( Malhotra et al . , 1989; Glick and Malhotra , 1998; Glick and Luini , 2011 ) , a major transport station composed of stacks of flat membranous cisternae . There are three main anterograde transport mechanisms that are in principle possible and might apply to the Golgi: ( a ) transport by compartment progression–maturation; ( b ) transport by dissociative anterograde vesicular carriers and , ( c ) transport via inter-compartment continuities . Among these , the progression–maturation model has gained a degree of consensus as an intra-Golgi traffic mechanism , based on several lines of evidence in mammals ( Bonfanti et al . , 1998; Lanoix et al . , 2001; Martinez-Menarguez et al . , 2001; Mironov et al . , 2001; Rizzo et al . , 2013 ) , yeast ( Losev et al . , 2006; Matsuura-Tokita et al . , 2006; Rivera-Molina and Novick , 2009 ) , algae ( Becker et al . , 1995 ) , and plants ( Donohoe et al . , 2013 ) . Under this model , cargo molecules remain in the lumen of the Golgi cisternae while the cisternae themselves progress through the stack and ‘mature’ through recycling of their resident enzymes . Recently , cisternal progression has been proposed to apply only to the rims ( and not to the core ) of the cisternae in the mammalian Golgi ( Lavieu et al . , 2013 ) . In addition to the Golgi , the progression–maturation principle appears to be involved in the endocytic ( Rink et al . , 2005; Poteryaev et al . , 2010 ) and the phagocytic pathways ( Fairn and Grinstein , 2012 ) in different species . The vesicular transport mechanism , whereby dissociative carriers transport cargoes between successive compartments , operates at many stages of the trafficking pathway and has been proposed to apply also to intra-Golgi trafficking ( Rothman , 2002 ) . Here , however , the evidence is less direct and less conclusive than at other transport segments , with conflicting claims about the presence ( Orci et al . , 2000 ) or absence ( Claude , 1970; Sabesin and Frase , 1977; Severs and Hicks , 1979; Clermont et al . , 1993; Dahan et al . , 1994; Di Lazzaro et al . , 1995; Bonfanti et al . , 1998; Orci et al . , 2000; Martinez-Menarguez et al . , 2001; Mironov et al . , 2001; Gilchrist et al . , 2006 ) of anterograde cargo proteins in the peri-Golgi carriers . Moreover in particular cases , like in microsporidia , intra-Golgi transport appears to occur without COPI vesicles ( Beznoussenko et al . , 2007 ) . Diffusion-based transport via inter-compartment continuities remains the least explored and understood of the traffic mechanisms . Some antecedents , however , are available . Continuity-mediated transport has been observed to occur between endosomes and lysosomes ( Luzio et al . , 2007 ) , and also the exocytic release of cargo from secretory granules ( Rutter and Hill , 2006 ) or synaptic vesicles through transient pores ( kiss-and-run ) ( Rizzoli and Jahn , 2007; Alabi and Tsien , 2013 ) at the plasma membrane can be considered to occur via this modality . For intra-Golgi transport , this mechanism has been discussed several times in the past ( Mellman and Simons , 1992; Weidman , 1995; Mironov et al . , 1997; Marsh et al . , 2004; Trucco et al . , 2004; Mironov et al . , 2005; Beznoussenko et al . , 2007; Glick and Luini , 2011 ) and a few recent intra-Golgi transport models including the mixing–partitioning ( Patterson et al . , 2008 ) , the kiss-and-run ( Mironov and Beznoussenko , 2012; Fusella et al . , 2013; Mironov et al . , 2013 ) and the cisternal progenitor schemes ( Pfeffer , 2010 ) have been proposed that imply transient tubular continuities across cisternae . At the molecular/mechanistic level , Golgi tubule formation has been proposed to be initiated by COPI coatomer-mediated budding ( Yang et al . , 2011 ) , and tubule elongation and fission appear to require the actions of cytosolic phospholipase A2 ( cPLA2 ) and lysophosphatidic acid acyltransferase-γ ( LPAATγ ) , respectively ( San Pietro et al . , 2009 ) ( Yang et al . , 2011 ) . Recent evidence also points to a role for Golgi localized SNAREs and BARS in the dynamics of the intercisternal connections ( Fusella et al . , 2013 ) . Nevertheless , a complete understanding of the molecular players regulating the intra-Golgi connections remains lacking . Altogether , uncertainties remain about the applicability of continuity-based transport to the Golgi . One main reason for this situation has been the long-standing difficulty of demonstrating intercisternal continuities in thin sections for electron microscopy . This obstacle has now been partly overcome by the use of electron tomography and new methods of three dimensional electron microscopy ( Briggman and Bock , 2012 ) , which have revealed the presence of intercisternal tubular continuities under experimental conditions that favor the detection of these tubules , such as the induction of active trafficking ( Marsh et al . , 2004; Trucco et al . , 2004; Vivero-Salmeron et al . , 2008; San Pietro et al . , 2009; Wanner et al . , 2013 ) . The second and main problem , yet to be resolved , is that the mere presence of intercisternal tubules is insufficient to prove a role for these continuities in transport , as these tubules might be too few and unfavorably disposed to support trafficking . To test the continuity-based transport model , it is thus necessary to search for functional evidence of a transport role for these continuities . To this end , we have used soluble secretory proteins as transport markers , as these are globular objects of a few nm in diameter that should easily cross the observed intercisternal tubules and rapidly move from the cis to the trans face of an interconnected stack or ribbon . The transport of some soluble proteins has been studied decades ago using electron microscopic autoradiography ( Caro and Palade , 1964; Jamieson and Palade , 1967; Ashley and Peters , 1969; Castle et al . , 1972 ) and biochemical pulse-chase assays ( Jamieson and Palade , 1967; Lodish et al . , 1983 ) , but their actual mechanism of secretion remains unknown . Comparing the trafficking pattern of prototypic soluble proteins with those of cargoes previously proposed to move by cisternal progression–maturation , we find that soluble proteins cross the Golgi stack at a much faster rate , apparently by diffusion along intercisternal connections; and that this transport mode coexists in the same Golgi complex with the much slower intra-Golgi progression of large , non-diffusible cargo , such as procollagen I ( PC-I ) . Soluble secreted proteins are of great physiological interest because they represent a significant portion ( possibly more than 10% ) of the mammalian proteome and include hormones , growth factors , serum proteins , antibodies , and digestive enzymes . Thus , these results are consistent with a novel mechanism of transport for a major class of secretory proteins , and provide evidence for multiplicity of transport mechanisms that can help to rationalize most of the observed intra-Golgi trafficking patterns .
As prototypes of soluble proteins we used albumin and α1-antitrypsin ( hereinafter termed antitrypsin ) . These are globular , water-soluble proteins roughly 3 nm in diameter that should easily diffuse through the 30–60 nm wide Golgi intercisternal connections ( Trucco et al . , 2004 ) . Albumin is an abundant , non-glycosylated protein , while antitrypsin is N-glycosylated . The trafficking of soluble proteins ( albumin in most experiments ) was characterized and compared with that of PC-I ( Weinstock and Leblond , 1974; Bonfanti et al . , 1998; Mironov et al . , 2001 ) and vesicular stomatitis virus G protein ( VSVG ) ( Bergmann and Singer , 1983; Mironov et al . , 2001; Patterson et al . , 2008 ) , because these cargoes have been extensively characterized and shown to move by cisternal progression ( or rimmal progression [Lavieu et al . , 2013] or compartment progression [Mironov et al . , 2013] . For the sake of brevity , from now onward we will use the term compartment progression to describe the traffic of procollagen and other similar cargo ) . Thus , if albumin moves by diffusion via continuities , it should exhibit transport kinetics and patterns different from VSVG and PC-I . PC-I forms large , stable , non-diffusible aggregates that cannot enter tubules or vesicles and cross the Golgi stack in a gradual fashion by compartment progression ( Bonfanti et al . , 1998; Trucco et al . , 2004 ) ; and VSVG is a large trimeric transmembrane viral protein that shows the same trafficking pattern as PC-I , at least under certain specific conditions ( see below ) . In this study , we only used conditions under which VSVG crosses the Golgi by compartment progression . We first compared the kinetics of intra-Golgi transport of albumin with those of VSVG and PC-I in HepG2 cells , a human hepatoma cell line that secretes both albumin and antitrypsin . To assess traffic rates , we used synchronization techniques by which cargoes can be arrested in the intermediate compartment ( IC ) , and then released , to monitor their synchronous passage through the secretory system . To compare albumin with VSVG , HepG2 cells were infected with VSV and subjected to the following synchronization protocol ( protocol 2 , 'Materials and methods' ) : the secretory pathway was first cleared of cargo by blocking protein synthesis with cycloheximide ( CHX ) ; and then CHX was removed at 15°C . At this temperature albumin and VSVG were re-synthesized relatively efficiently , and were then transported to , and arrested in , the IC ( Mironov et al . , 2001 ) . Finally , the 15°C transport block was removed by shifting the temperature to 32°C , to allow the synchronous passage of albumin and VSVG from the IC to and through the Golgi complex ( Mironov et al . , 2001 ) . Notably , this protocol does not seriously overload/perturb the secretory pathway since , under similar conditions , the Golgi complex has been shown to maintain a normal structure and function ( Trucco et al . , 2004; Mironov et al . , 2001 ) . To monitor cargo passage , we used both immuno-electron microscopy ( immuno-EM ) and immuno-fluorescence . By immuno-EM , albumin was seen at time 0 ( i . e . , at end of the 15°C block ) in the ER and IC at similar levels , with very little in the Golgi stacks ( Figure 1A , B , green arrowheads ) . An earlier study had shown that a soluble protein ( soluble secretory GFP ) concentrates in the IC/Golgi area at 15°C ( Blum et al . , 2000 ) ; however , no EM experiments were carried out to verify the localization . Using immuno-EM , we do not observe any such concentration of albumin in the Golgi after the 15°C block ( time 0 ) . Albumin is clearly restricted to the ER and IC , and absent from the Golgi apparatus ( Figure 1A , B , I , L ) . Within 2 min of release from the 15°C block , albumin entered and filled the entire Golgi , including the trans-Golgi network ( TGN ) , with apparently similar levels throughout ( Figure 1C , D , I ) . After 5 min at 32°C , the distribution of albumin had not changed significantly ( Figure 1E , F , I ) , while at 10 min , albumin was higher in the TGN than in the cis cisternae ( Figure 1G , I ) . Then ( by 20 min ) , albumin began to exit the Golgi , as indicated by its diminishing overall levels in the Golgi stack ( Figure 1H , J , L ) . In sum , albumin spreads through the stack in less than 2 min , then exits the Golgi complex . 10 . 7554/eLife . 02009 . 003Figure 1 . Kinetic patterns of synchronized transport of albumin and VSVG through the Golgi stack . VSV-infected HepG2 cells were synchronized according to the CHX/32-15°C protocol ( 'Materials and methods' ) . Following release of the 15°C block , the cells were examined by immuno-EM ( A–H ) at the indicated times . Panels ( I–K ) show quantification of immuno-EM values as labeling density ( LD ) normalized to the density in the ER , to avoid labeling variability across samples . ( L ) The amount of albumin or VSVG in indicated compartments were normalized to that present at time 0 in the ER and expressed as percentage . Values are mean ± SD from 30 stacks per time point , in three independent experiments for immuno-EM . Bar: 60 nm ( A ) , 50 nm ( B , C , E , G ) , 100 nm ( D , H ) , 80 nm ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 00310 . 7554/eLife . 02009 . 004Figure 1—figure supplement 1 . Kinetic patterns of synchronised transport of albumin and VSVG through the Golgi stack as determined by immunofluorescence . VSV-infected cells were synchronised according to the CHX/32-15°C protocol ( 'Materials and methods' ) . Briefly , the cells were treated with CHX for 3 hr , to clear the secretory pathway of all cargo . Following CHX washout , the cells were incubated at 15°C for 2 hr , to block newly synthesised protein in the IC . The 15°C block was then released by shifting the cells to 32°C for the indicated times ( 0 , 2 , 5 , 10 and 20 min ) . The cells were fixed and labeled for the cargo proteins albumin ( A–J ) and VSVG ( K–T ) and the Golgi markers GM130 and TGN46 as indicated . ( U–W ) Quantification of co-localization of albumin , antitrypsin ( U ) and VSVG ( W ) with Golgi markers as indicated . ( V ) Total fluorescence of albumin and antitrypsin in the Golgi complex , expressed as arbitrary units ( AU ) . Values are mean ±SD of 10 co-localization measurements per time point for immunofluorescence . Bar: 20 μm ( A , K ) , 10 μm ( B , C , L ) , 4 μm ( E–J , M–T ) , 8 μm ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 004 The pattern of VSVG traffic differed from that of albumin . As previously described ( Mironov et al . , 2001; Trucco et al . , 2004 ) , at time 0 , VSVG was depleted in the ER , concentrated in the IC , and nearly absent in the Golgi stacks ( Figure 1A , B , K ) . 2 min after the 15°C block release , VSVG was still mostly in IC elements adjacent to the cis-Golgi ( Figure 1C , D , K ) , and at 5 min it had reached only the first cis-cisterna ( Figure 1E , F , K ) . Later , VSVG gradually reached the medial and then the trans-Golgi ( Figure 1G , K ) . Thus , VSVG moves gradually through the stack in over 15 min , consistent with the compartment progression trafficking mechanism , as expected under these synchronization conditions ( Mironov et al . , 2001; Trucco et al . , 2004 ) . For immunofluorescence experiments ( Figure 1—figure supplement 1 ) , we monitored arrival of both VSVG and albumin at the cis- and trans-Golgi by determining their degree of co-localization with cis- and trans-Golgi markers ( GM130 and TGN46 , respectively ) ( Mironov et al . , 2001; Trucco et al . , 2004 ) . This is feasible because cis- and trans-Golgi markers can be resolved ( to a large though not complete extent ) by confocal microscopy ( Shima et al . , 1997; Trucco et al . , 2004 ) ( 'Materials and methods' ) . Albumin showed a diffuse ER-like distribution at time 0 , with no clear Golgi staining ( Figure 1—figure supplement 1 ) ; then , 2 min after the release of the 15°C block , albumin entered the Golgi stack and co-localized to the same extent with both GM130 and TGN46 ( i . e . , it reached both the cis and trans areas , Figure 1—figure supplement 1 ) , while the ER was still not completely empty . After 5–10 min , albumin had completely left the ER and now localized mostly in the Golgi , where its levels declined in the cis-Golgi , while they remained high in the trans-Golgi ( Figure 1—figure supplement 1 ) , compatible with rapid albumin diffusion through the stack followed by concentration in the TGN . Thus , the export of albumin out of the ER was very efficient , so that by 10 min after the release of the temperature block almost all of the protein had been transported to the Golgi apparatus . Antitrypsin showed very similar distribution and trafficking patterns to albumin ( Figure 1—figure supplement 1 ) . VSVG , instead , showed a punctate ( IC-like ) distribution at 15°C , as previously reported ( Mironov et al . , 2001; Trucco et al . , 2004; Figure 1—figure supplement 1 ) . After the release of the block , VSVG reached the cis-Golgi first ( at 5 min ) ( Figure 1—figure supplement 1 ) , and then later , with a lag of 10–15 min , it arrived at the TGN , as previously described ( Mironov et al . , 2001; Figure 1—figure supplement 1 ) . Again , this is compatible with compartment progression , and is in agreement with the immuno-EM data . Next , we compared albumin and PC-I . We expressed albumin in professional PC-I secretory cells ( human fibroblasts ) by microinjecting albumin cDNA in the nucleus and subjecting the cells to the synchronization protocol 1 ( 'Materials and methods' ) . A limited but sufficient fraction of injected cells expressed albumin . At 15°C ( time 0 ) , albumin was mostly diffuse in the ER ( as seen in HepG2 cells ) , while PC-I was seen in scattered fluorescent ‘spots’ ( presumably PC-I aggregates within the IC ) ( Figure 2A , B; Mironov et al . , 2001; Trucco et al . , 2004 ) . We then increased the temperature to 32°C . In these cells , the PC-I trafficking pattern has been characterized extensively in previous studies: PC-I arrives at the cis-Golgi from the IC in 2–3 min and later progresses to the TGN by compartment progression in 12–15 min ( Bonfanti et al . , 1998; Mironov et al . , 2001 ) . Here , we confirmed that within 3 min after the release of the 15°C block , PC-I aggregates reach the Golgi area but not the TGN ( Figure 2D , E ) ; and by correlative light-immuno-EM ( CLEM ) , we further confirmed that at this time PC-I aggregates reach the cis but not the distal cisternae , well in line with previous reports ( Figure 2F ) ( Bonfanti et al . , 1998; Mironov et al . , 2001 ) . In the same cells , by contrast , albumin filled the Golgi stack rapidly , as in HepG2 cells: at 3 min , it already co-localized with the TGN marker TGN46 ( by immunofluorescence ) ( Figure 2C , E ) and , by EM , it filled the Golgi stacks from cis to trans ( Figure 2G ) . 10 . 7554/eLife . 02009 . 005Figure 2 . Kinetic patterns of synchronized transport of albumin and PC-I through the Golgi stack . Human fibroblast cells were microinjected in the nucleus with cDNA for albumin and incubated for 2 hr before further treatments . Transport was synchronized according to the CHX/32-15°C protocol and the cells were examined by immunofluorescence and immuno-EM . ( A ) Immunofluorescence localization of albumin and PC at the end of the 15°C block . The area in ( A ) indicated by white rectangle is enlarged in ( B ) ( C–E ) . Co-localization between albumin ( C ) or PC ( D ) or of both cargoes ( E ) with TGN46 , 3 min after release of the block . ( F–G ) . Localization of PC ( F ) and albumin ( G ) 3 min after release of the 15°C block by immuno-EM . PC ( indicated by * ) localizes selectively to the cis-cisterna . The cis-side of the Golgi is revealed by the presence of GM130 labeled by immuno-nanogold technique ( indicated by white arrows ) ( F ) . Albumin labeled by immuno-nanogold technique ( black dots ) shows a diffuse localization throughout the Golgi complex ( G ) . Bars: 5 μm ( A ) , 2 μm ( B–E ) , 125 nm ( E and F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 005 Collectively , these results indicate the existence , in the same cells ( and in the same stacks ) , of two different intra-Golgi trafficking patterns for different cargo types , one consistent with gradual compartment progression , for PC and VSVG , and one characterized by the rapid spreading of cargo through the stack , for albumin . A possible limitation of these data is that they were obtained using synchronized traffic waves . Albeit relatively mild ( Mironov et al . , 2001 ) , the traffic synchronization protocols that were applied here might ‘overload’ the secretory pathway . We therefore sought to examine the transport patterns of albumin and PC in cells at steady-state . This can be achieved using GFP-tagged cargoes in living HeLa cells ( 'Materials and methods' ) , which offer controlled expression conditions . GFP-albumin showed steady-state Golgi localization and secretory behavior similar to that of native albumin in HepG2 cells ( Figure 3—figure supplement 1 ) , indicating that this construct can be used as an albumin tracer . Moreover , a characterization of the GFP-albumin dynamics in HeLa cells based on fluorescence recovery after photobleaching ( FRAP ) ( Patterson et al . , 2008 ) , showed that this construct enters and exits the Golgi with half-times of about 3–4 min ( Figure 3—figure supplement 1 ) and diffuses ‘horizontally’ along the Golgi ribbon in seconds , as expected from its soluble nature ( Figure 3—figure supplement 1 ) . We thus proceeded to assess the steady-state transport behavior of GFP-albumin , and to compare it with that of PC . To this end , we bleached the Golgi area ( Figure 3A , B ) and monitored the time for arrival of GFP-albumin from the ER at the cis-Golgi and at the trans-Golgi ( again by quantifying its co-localization with GM130 and TGN46; see above and 'Materials and methods' ) ( Figure 3C–F ) . After 1–2 min ( i . e . , the earliest time at which GFP-albumin had recovered to detectable levels in the Golgi stack ) ( Figure 3C ) , GFP-albumin had reached both the cis-Golgi and the TGN ( Figure 3D–F ) ; in fact , it showed a slightly higher degree of co-localization with TGN46 than with GM130 ( using the unbiased co-localization Method 2 based on automatic thresholding; see 'Materials and methods' ) , indicating that it had already traversed the Golgi stack ( Figure 3D–F , quantification in J ) . Later ( 3 min post-bleaching ) , the GFP-albumin signal became slightly higher in the trans- than the cis-Golgi , and at 12 min ( when recovery was nearly complete ) , it was clearly higher in the trans- than the cis-Golgi ( as seen before bleaching , with a ratio of about 1 . 8 ) ( Figure 3J ) . To control for the possibility that part of the fluorescence signal recovered in the Golgi area might come from the underlying ER , we repeated this experiment using nocodazole-induced ministacks ( Figure 3G–I , K ) , where the cis and trans-Golgi markers are resolved better ( Shima et al . , 1997; Trucco et al . , 2004 ) and the very low background fluorescence of the ER present in the cellular periphery allows a better resolution of Golgi fluorescence . The results were very similar to those obtained with the intact ribbon . Next , to confirm these results by EM we resorted to GFP-photooxidation followed by CLEM experiments . For photooxidation studies , the same experiments as those described above were carried out , and the cells were fixed 2 min after photobleaching , when GFP-albumin fluorescence had recovered in the Golgi . Then , the newly arrived fluorescent protein in the Golgi was excited in the presence of DAB under conditions that favor the photo-oxidation reaction and the formation of DAB electron dense precipitates in the close vicinity of GFP ( Grabenbauer et al . , 2005; Meiblitzer-Ruppitsch et al . , 2008 ) ( 'Materials and methods' ) . The Golgi elements that had been monitored by video microscopy were then examined by CLEM ( Mironov and Beznoussenko , 2013 and 'Materials and methods' ) . The results shown in Figure 3O–S clearly indicated that after 2 min of recovery GFP-albumin was already filling the whole Golgi stack . 10 . 7554/eLife . 02009 . 006Figure 3 . Kinetic patterns of transport of GFP-albumin , VSVG-GFP and PC-III-GFP through the Golgi stack under steady-state conditions . HeLa cells were transfected with GFP-albumin ( A–K ) or PC-III-GFP ( L–N ) . After 16 hr of transfection , the Golgi area was bleached , and entry of these cargoes from the unbleached periphery ( ER ) into the Golgi area was monitored by FRAP . The cells were then fixed at different time points , stained for GM130 and TGN46 , and re-localized for analysis of co-localization of the GFP-tagged cargoes with these Golgi markers . ( A–C ) Bleaching of the Golgi area , as delineated by the dotted line , with post-bleaching recovery for 1 min ( C ) . ( D–F ) Detail of the same Golgi area shown in ( C ) , showing co-localization of GFP-albumin ( green ) with GM130 ( D , red ) , or TGN46 ( E , red ) or both ( F: GM130 , blue; TGN46 , red ) . ( G–I ) Similar experiments carried out on a nocodazole-induced Golgi ministack ( 'Materials and methods' ) , with 1-min post-bleaching co-localization of GFP-albumin ( green ) with GM130 ( G , red ) or TGN46 ( H , red ) or both ( GM130 , blue and TGN46 , red ) ( I ) . ( J ) Quantification of the degree of co-localization of GFP-albumin with GM130 and TGN46 at different time points after bleaching , as illustrated in ( A–F ) . These data are expressed by normalizing the degree of co-localization of GFP-albumin in the TGN46 area to that of albumin in the GM130 area ( set to 1 ) . ( K ) Line scan along the arrow across the Golgi ministack shown in ( I ) . The fluorescence intensities from representative points along the distance were plotted . ( L and M ) Cells were transfected with PC-III-GFP . The Golgi area ( within the dotted line ) was bleached , and the time course of entry of PC-III-GFP to the TGN was monitored . The cells were fixed and stained for TGN46 at 3 min ( L ) and 9 min ( M ) post-bleach , and the overlap between PC-III-GFP with TGN46 was examined . ( N ) Quantification of data in ( L and M ) , expressed as mean ± SD from at least three independent experiments . ( O–S ) To ascertain the earlier observations of rapid filling of the Golgi stack by GFP-albumin ( A–F ) , we resorted to electron microscopy . HeLa cells were transfected with GFP-albumin ( O and R ) or VSVG-GFP ( P ) or PC-III-GFP ( Q ) . The Golgi localized fluorescence was bleached as before ( time 0; O ) and entry of cargo into the Golgi area monitored by FRAP and the cells fixed 2 min after recovery . The GFP fluorescence was then converted to a signal visible at the EM by photooxidation ( see 'Photooxidation' under 'Materials and methods' section ) using Diaminobenzidine ( DAB ) . The DAB product is indicated by arrows . At time 0 the DAB product is present only in the ER with Golgi devoid of staining ( O ) . After 2 min of fluorescence recovery , both VSVG-GFP ( P ) and PC-III-GFP ( Q ) are restricted to the cis-side of the Golgi , while GFP-albumin ( R ) is present throughout the Golgi . In the case of VSVG-GFP , DAB precipitate is visible outside of the Golgi cisternae because GFP is attached to the cytosolic tail of VSVG . In addition , nanogold labeling for Mannosidase II was done in ( P ) that marks the medial-part of the Golgi . The time 0 image shown is from cells expressing GFP-albumin; similar staining was obtained from both VSVG-GFP and PC-III-GFP expressing cells at time 0 . ( S ) The percentage of cells that showed DAB product throughout the Golgi 2 min after recovery was calculated and presented as mean ± SD . Bar: 2 μm ( A–M ) , 220 nm ( O–R ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 00610 . 7554/eLife . 02009 . 007Figure 3—figure supplement 1 . Localization , transport behavior , and dynamics of GFP-albumin at steady-state . ( A and B ) Intra-Golgi distribution of GFP-albumin at steady-state . HeLa cells were transfected with GFP-albumin , kept for 24 hr at 37°C , and then fixed and labeled for immuno-EM with an antibody against GFP ( 10 nm gold ) and TGN 46 ( 5 nm gold; green arrows ) ( A ) . The albumin distribution depicted in ( A ) was quantified and the steady-state GFP-albumin in the cis-Golgi and trans-Golgi is shown as labeling density ( LD ) normalized to that of the ER ( B ) . ( C–E ) Kinetics of secretion of GFP-albumin . HeLa cells expressing GFP-albumin ( lanes 1 , 3 ) and HepG2 cells ( lanes 2 , 4 ) were washed with serum free media and incubated at 37°C for indicated times in serum free media . The cell lysate immediately after wash ( C ) and media after 60 min of incubation ( D ) were resolved by SDS-PAGE and probed with anti-albumin antibody . Quantification of the secreted protein ( as % total ) shows that albumin and GFP-albumin are released into the medium with similar kinetics ( E ) . The minor low-molecular-weight form of GFP-albumin in the intracellular pool ( C , lower band in lane 1 ) is probably a misfolded form of the protein undergoing degradation , and it is not secreted ( D , lane 3 ) . ( F–H ) Dynamics of GFP-albumin at steady-state . HeLa cells were transfected with GFP-albumin and examined using the FRAP approach ( 'Materials and methods' ) . ( F ) Entry of GFP-albumin into the Golgi area . The fluorescence in the whole Golgi area was bleached , and the recovery of fluorescence into the bleached area was monitored . ( G ) Exit of GFP-albumin from the Golgi area . The fluorescence of the whole cell less the Golgi area was bleached , and the loss of GFP-albumin from the Golgi area was monitored . ( H ) Diffusion of GFP-albumin along the Golgi ribbon . The fluorescence of a part of the Golgi area was bleached , and the recovery of fluorescence into the bleached region was monitored . The data expressed are in mean ±S . D from three independent experiments ( E ) or five independent experiments ( F–H ) . Bar: 130 nm ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 00710 . 7554/eLife . 02009 . 008Figure 3—figure supplement 2 . Kinetics of antitrypsin processing by Golgi enzymes reflects its fast kinetics of transport . Transport of antitrypsin ( A ) and VSVG ( B ) along the secretory pathway was monitored by radioactive pulse chase assay . HepG2 cells infected with VSV was pulsed with radioactive aminoacids ( 35S-methionine and cysteine ) for 5 min and then chased for indicated times in cold media . At the end of the chase period the cells were lysed and VSVG or antitrypsin were immunoprecipitated and subjected to Endoglycosidase H digestion as indicated and resolved by SDS-PAGE followed by autoradiography . EndoHr–Endoglycodisase resistant , EndoHs–Endoglycosidase sensitive , immature–EndoHr form of antitrypsin that was not processed by trans-Golgi resident enzymes and mature–EndoHr form of antitrypsin processed by trans-Golgi resident enzymes . It is important to note here that the quantities of antitrypsin and VSVG present are very similar suggesting that the difference in the transport behavior of these proteins is not due to differences in their abundance . In addition , as mentioned in the text , the transport behaviors of antitrypsin and albumin are similar , reiterating further that the differences in the transport behavior between soluble secretory cargoes ( albumin and antitrypsin ) and VSVG/PC is possibly not due to the differences in their abundance . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 008 To monitor the behavior of PC under similar conditions , we used HeLa cells transfected with GFP-tagged PC-III ( a homotrimer that forms large aggregates in the Golgi complex like PC-I , and in general behaves like PC-1; [Perinetti et al . , 2009] and 'Materials and methods' ) . A limited but sufficient number of cells expressed this cargo . We then bleached the whole Golgi area and monitored the rate of entry of PC-III-GFP into the Golgi complex from the ER . PC-III behaved as expected from our previous experiments on PC-I trafficking ( Bonfanti et al . , 1998; Mironov et al . , 2001; Trucco et al . , 2004 ) . At 3 min post-bleaching , some PC-III-GFP aggregates ( in the form of distinct bright puncta ) had already entered the Golgi area , but had not reached the TGN ( i . e . , did not co-localize with TGN46 ) ( Figure 3L , N ) . Later ( at 9 min ) , many more PC-III-GFP aggregates had reached the Golgi stack and some of these co-localized with TGN46 ( Figure 3M , N ) , confirming that PC-III , like PC-I , enters the Golgi and then moves gradually to the TGN , consistent with compartment progression and different from the albumin transport pattern . We also monitored whether the fast transport of soluble cargoes by diffusion is coupled to their processing by Golgi enzymes . To this end , the biochemical maturation of antitrypsin was monitored by the pulse chase assay . Antitrypsin was processed efficiently by Golgi enzymes , as evidenced by increased apparent molecular weight of the protein , with a kinetics reflecting that measured by microscopy-based assays ( Figure 3—figure supplement 2 ) . In the same experiments , VSVG acquired endo-H resistance at a markedly slower rate ( Figure 3—figure supplement 2 ) . Thus , the Golgi residence time of soluble cargo appears to be sufficient for complete glycosylation . Possibly , the high surface/volume ratio of the flat Golgi cisternae maximizes the contact , and hence the efficiency of the reaction , between cargo and enzymes . In summary , extensive kinetic evidence obtained under both traffic-synchronization and steady-state conditions shows the coexistence of two different intra-Golgi trafficking behaviors ( and hence , presumably , different mechanisms ) for different cargo types . PC-I ( also PC-III ) and VSVG enter the Golgi stack and move gradually from cis to trans , consistent with compartment progression , while albumin equilibrates rapidly across the Golgi compartments . The latter behavior is consistent with diffusion via intercisternal continuities . However , purely kinetic data cannot exclude that other mechanisms , such as fast vesicular shuttling , might lead to the same traffic pattern ( Pelham and Rothman , 2000 ) . To distinguish between diffusion- and vesicle-based traffic , we used both morphological and computational approaches . We first examined the Golgi structure in HepG2 cells at steady-state , with a focus on vesicles and tubules , using EM tomography . The Golgi stacks in HepG2 cells comprise 4–6 cisternae that were flanked by vesicles and connected side-by-side by ‘longitudinal’ tubules and fenestrated membranes , as seen in many other cell types . Some longitudinal tubules were Y-shaped and connected heterologous cisternae in neighboring stacks ( not shown ) , in agreement with previous descriptions ( Marsh et al . , 2004 ) . In addition , successive cisternae within individual stacks were sometimes connected by tubules that appeared to be oriented in the cis-trans direction ( ‘vertical’ tubules ) and that were distributed apparently randomly at all levels of the stack , as previously described ( Trucco et al . , 2004 ) . These vertical tubules were often convoluted ( as exemplified in Figure 4A–C and Video 1 ) , and had calibers ranging from 30 to 60 nm . The number of intercisternal connections between heterologous cisternae across the Golgi stack was highly variable . In HepG2 cells , each stack had about 5 ± 2 intercisternal connections ( calculated as described in 'Materials and methods' ) . When a complete tomographic reconstruction of a tubule was possible , it was always found that the tubule was connected to a cisterna ( not shown ) in agreement with previous observations ( Trucco et al . , 2004 ) . A few narrow continuities joining the central areas of adjacent cisternae were also observed ( not shown ) . Similar results were obtained using both chemical fixing and high-pressure freezing ( not shown ) . For comparison , we also examined the Golgi in rat liver . The vertical and Y-shaped tubules were similar to those in HepG2 ( not shown ) . 10 . 7554/eLife . 02009 . 009Figure 4 . EM tomography facilitates the visualization of convoluted intercisternal tubules . HepG2 cells were high-pressure frozen and prepared for EM tomography ( 'Materials and methods' ) . ( A and B ) Tomographic model of a stack from a 200-nm-thick section containing an intercisternal connection . Detail of A shown in B; note the complexity of the convoluted connection ( follow the arrow to identify the continuity ) . ( C ) A gallery of tomographic digital slices ( panels 1–12 ) used to construct the model in ( A and B ) shows a convoluted intercisternal connection , with the small arrow following the connection , and the arrowheads showing the two cisternae that are connected . Note the complexity of the connection , which would be nearly impossible to detect in traditional thin sections . See Video 1 for facilitated visualization of the continuity . Bar: 150 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 00910 . 7554/eLife . 02009 . 010Video 1 . A tomographic reconstruction of the Golgi stack shown in Figure 4 . Scripts used to simulate the transport of albumin across the Golgi stack ( Supplement to Figure 6 ) :DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 010 We then examined the albumin distribution in these Golgi structures by immuno-EM . Albumin was scarce in the ER ( not shown ) and dense in the stacks , and even denser in the TGN ( Figure 5A , C , G , I ) . A similar distribution was also found in rat hepatocytes with a slightly higher concentration of albumin in Golgi and TGN ( not shown ) . The distribution of antitrypsin was indistinguishable from that of albumin ( not shown ) . For comparison , VSVG was markedly more concentrated in the TGN than in the stack ( except in cells expressing high levels of VSVG , where the stacks were filled with this cargo ) ( Figure 5B , D , H , J ) ; and PC-I aggregates ( in human fibroblasts ) were also much more numerous in the TGN than in the Golgi stack ( not shown ) . Thus , a large fraction of VSVG and PC-I in the Golgi area was located in the TGN . Next , we focused on the distribution of cargo in Golgi tubules and vesicles . We first examined the albumin content of Golgi vesicles ( defined as round , 50–60 nm-wide profiles near the Golgi stack; see 'Materials and methods' ) by immuno-EM . The vesicular profiles were markedly depleted of albumin ( Figure 5E , arrows , and 5k for quantification ) . This observation is in agreement with previous in vivo morphological and biochemical observations in animal liver cells ( Dahan et al . , 1994; Gilchrist et al . , 2006 ) but is seemingly at variance with in vitro studies that showed that albumin can be present in vesicles generated by the non-hydrolysable GTP analogue GTPγS from Golgi enriched liver membranes ( Malhotra et al . , 1989 ) ( see also [Caro and Palade , 1964; Jamieson and Palade , 1967] ) . As this difference might be due to the presence of GTPγS ( instead of the natural nucleotide GTP ) in the in vitro experiments ( GTPγS is known to affect cargo sorting into vesicles [Lanoix et al . , 2001] ) , we performed a series of in vitro experiments using GTPγS or GTP , and confirmed that this was indeed the case ( Figure 5—figure supplement 1 ) , explaining the discrepancy between the in vitro and the in vivo data . 10 . 7554/eLife . 02009 . 011Figure 5 . Albumin distribution in Golgi cisternae , vesicles and tubules in HepG2 cells . ( A–D ) Albumin and VSVG distribution in the ER , Golgi stack and TGN ( A and B , immuno-EM; C and D , immunofluorescence ) . White arrows in ( A and B ) indicate TGN46 labeling . For quantification see ( G–J ) . ( E ) Albumin distribution in cisternae , vesicles ( 50–60 nm wide round profiles near the Golgi stack; arrows ) and tubules ( tubular–ovoid profiles; arrowheads ) , by immuno-EM . ( F ) Albumin present in a connected cis-trans vertical tubule , as visualized by the immuno-nanogold technique . Arrowhead indicates albumin and white arrows highlight the intercisternal connection . For quantification see ( K ) . ( G and H ) Quantification of labeling density ( LD ) of albumin and VSVG by immuno-EM , normalized to ER labeling . ( I and J ) Quantification of co-localization ( as described in 'Materials and methods' ) of the cargoes with GM130 ( cis-Golgi ) and TGN46 ( TGN ) markers by immunofluorescence , normalized to their co-localization with GM130 . ( K ) Quantification of distribution of albumin in cisternae , peri-Golgi vesicles and tubules expressed as LD . Bars: 120 nm ( A ) , 210 nm ( B ) , 7 . 5 μm ( C and D ) , 250 nm ( E and F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 01110 . 7554/eLife . 02009 . 012Figure 5—figure supplement 1 . Distribution of cargoes in COPI vesicles . ( A–D ) As noted in the main text , Golgi vesicles appear depleted of albumin in vivo , while vesicles prepared in vitro have been reported to contain albumin ( Malhotra et al . , 1989 ) . We sought to resolve this discrepancy between in vivo and in vitro data by examining the conditions of the in vitro preparation . COPI vesicles were purified exactly as described by Rothman et al . ( Malhotra et al . , 1989 ) . Their study used the non-hydrolysable GTP analogue GTPγS to increase the yield of vesicles in vitro . It was later shown , however , that GTPγS also affects the sorting of cargo into vesicles ( Lanoix et al . , 1999 ) . We thus examined the effects of GTPγS , and we show here that while COPI-coated vesicles that are formed in the presence of GTPγS contained significant levels of albumin , those that are formed in the absence of GTPγS were depleted of albumin , in line with our in vivo data . The exclusion of soluble proteins from vesicles has been reported before ( e . g . , of proinsulin from KDEL-receptor-containing vesicles; [Orci et al . , 1997] ) ; however , the mechanism of exclusion remains unclear . Experimental details: Golgi membranes were isolated from rat liver ( 'Materials and methods' ) and treated with ( A ) or without ( B and C ) GTPγS , as described by Rothman and colleagues ( Malhotra et al . , 1989 ) . The Golgi membranes were then pelleted , and processed for cryoimmuno-EM and labeled with an anti-albumin antibody . After GTPγS treatment , albumin can be seen in COPI vesicular profiles ( round , 50–60 nm in diameter ) ( A , arrows ) , while in the absence of GTPγS , the vesicles ( less numerous ) were not labeled for albumin ( B and C; arrows ) . In both cases , albumin was concentrated in large pleomorphic structures , which were probably cisternal remnants . ( D ) Quantification of the percentage of vesicles containing albumin from ( A ) and ( B ) expressed as mean ± SD ( n = 3 ) . ( E–F ) VSV-infected HepG2 cells at steady-state were labeled with antibody against VSVG according to cryo-immuno EM protocol ( 'Materials and methods' ) . ( E ) Lack of VSVG labeling within elongated tubule-like ( arrow ) and round vesicle-like ( arrowhead ) profiles . ( F ) Morphometric analysis shows that VSVG labeling density ( mean ± SD; n = 30 stacks ) is significantly less in tubules and vesicles than in cisternae . Bar: 120 nm ( A and E ) ; 90 nm ( B and C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 01210 . 7554/eLife . 02009 . 013Figure 5—figure supplement 2 . Gallery of cryo-immuno-gold EM images indicating the presence of albumin in intercisternal tubules . HepG2 cells were labeled for albumin according to the cryo-immuno EM protocol ( 'Materials and methods' ) . Black arrows in all images indicate the convex sides of the intercisternal tubules; black arrowheads indicate the concave sides , and red arrows indicate the luminal continuities generated by the connections . ( A ) The concave part of the bent membrane appears to be clearly continuous across two successive cisternae . Moreover , the position of the albumin-associated gold particles distributed along the bend ( please note that albumin is inside a tubule here , even if part of the convex tubular membrane is not clearly visualized ) , as well as the general disposition of the cisterna are clearly consistent with the presence of a connecting albumin-filled tubule . The external ( convex ) membrane of the tubule ( black arrow ) cannot be easily seen , probably due to its oblique orientation with respect to the plane of the section . ( B ) Here , the tubular loop is clearly visible , but its continuity with the cisternae in the stack is not visible probably due to the brighter areas ( asterisks ) caused by heterogeneous UA-methylcellulose deposition . ( C ) Here , the tubular loop exhibits a convex membrane ( black arrow ) connected to the lower cisterna in the stack . The membrane on the concave side of the tubular bridge ( black arrowhead ) and its connection to the upper cisterna in the stack are less evident , probably due to their oblique orientation with respect to the section plane . ( D ) The bent membrane is well preserved and the ‘connecting’ tubule makes a complete ( 180° ) turn from the level of the ‘upper’ cisterna to the level of the ‘lower’ one . However , after the bend there is a gap between the bent membrane and the rest of the ‘lower’ cisterna . This gap almost certainly represents a perforation of the connected cisterna , as indicated by the following considerations: ( A ) perforations in HepG2 cisternae are very common , and they look very similar to this image; several examples of these perforations can be found in the tomography in Figure 4; ( B ) the bent tubule , if it is not connected to the cisterna , would have to be open ended . These open-ended tubules are nearly never observed in Golgi tomographies . In addition , it is very hard to imagine a tubule turning downward and backward 180° and pointing precisely to the rim of the cisterna . Bars: 75 nm ( A ) , 110 nm ( B ) , 130 nm ( C ) , 100 nm ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 01310 . 7554/eLife . 02009 . 014Figure 5—figure supplement 3 . Presence of albumin in intercisternal tubules revealed by serial sectioning followed by cryo-immuno-gold EM and DAB photooxidation followed by tomography . ( A ) Serial sectioning by cryo-immuno EM indicates the presence of albumin in intercisternal tubules . HepG2 cells were labeled for albumin according to the cryo-immuno EM protocol ( 'Materials and methods' ) . Each image contains the green arrows showing that two cisternae that are not connected in sections 1 and 3 ( A1 and A3 ) are connected in section 2 ( A2 ) . A tubular hairpin ( or loop ) that joins these two cisternae located at different levels of the stack is visible in panel A2 . The gold particles distributed over the tubular connection indicate the presence of albumin in the lumen . Black arrows indicate fiducial marks used for alignment of serial sections . ( B1–B4 ) Gallery of tomographic virtual sections of a Golgi stack with GFP-albumin in intercisternal tubules , as revealed by DAB photooxidation . HeLa cells were transfected with GFP-albumin and 24 hr later subjected to photooxidation procedure , embedded in resin and prepared for EM tomography ( 'Materials and methods' ) . White arrows show the intercisternal connection ( the connection is filled with green arrows ) across two consecutive cisternae . The interpretation here is complicated by the membrane damage induced by the photooxidation procedure , and by the presence of the dark patches of oxidized DAB ( red arrows ) . Nevertheless , the image strongly suggest the presence of a tubule ( white arrow ) connecting cisternae located at different levels of the Golgi stack , as can be visualized in sequential digital slices extracted from the tomogram ( panels B1–B4 ) . Dark patches of oxidized DAB product indicate the presence of GFP-albumin within the tubular bridge and the connected cisternae . Precipitates of DAB are visible also in other cisternae ( black arrow ) . ( B5 ) 3D model of the connected cisternae and of the tubular bridge indicated by red arrow . Bars: 150 nm ( A ) ; 80 nm ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 01410 . 7554/eLife . 02009 . 015Figure 5—figure supplement 4 . Presence of albumin in intercisternal tubules revealed by cryo-immuno EM followed by tomography . ( A–C ) HepG2 cells were labeled for albumin according to the cryo-immuno EM protocol ( 'Materials and methods' ) . Here , thick sections ( 200 nm ) were used instead of the usual thin ( 70 nm ) sections . The Golgi profiles showing albumin labeling were subjected tomographic reconstruction ( 'Materials and methods' ) ( A ) . ( B and C ) Serial sections from the tomogram shown in ( A ) that demonstrates the complicated nature of intercisternal connections that contain albumin . Bar: 90 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 015 We next probed the Golgi tubules . In random , thin sections for immuno-EM , tubules appear as elongated variably oriented tubular–ovoidal profiles in the vicinity of the stack ( 'Materials and methods' ) . These profiles contained albumin at a similar density to that seen in the cisternae ( Figure 5F , arrowheads , quantified in K ) , independently of their orientation ( parallel or perpendicular to the plane of the cisternae ) ( see below ) . For comparison , we also examined VSVG . This cargo has been previously shown to be depleted in Golgi vesicles and , to a lesser extent , in tubules ( Trucco et al . , 2004 ) . We confirmed that VSVG is lower in both vesicular and tubular peri-Golgi profiles than in cisternae ( Figure 5—figure supplement 1 ) . We then attempted to visualize albumin in complete vertical tubular connections . This experiment presented serious difficulties because: ( a ) most connections are convoluted and cannot be included in single thin sections; and , ( b ) even relatively ‘linear’ continuities are very unlikely to be cut through their entire length at random ( here , a reasonable expectation is that less than one connection may be found in hundred sections , assuming six connections per stack [Marsh et al . , 2004; Trucco et al . , 2004] ) . An additional difficulty is that the tubular membranes are often cut obliquely at some point along their length , resulting in defective membrane visualization . We sought to overcome these problems by ( a ) cutting several hundreds of individual thin sections , as well as several serial sections , to find at least a few complete inter-cisternal connections , and , ( b ) combining tomography with albumin labeling by photo-oxidation and cryo-immuno EM . Using the first approach , we succeeded in visualizing a few tubular connections ( Figure 5—figure supplements 2 and 3 ) that showed continuity across heterologous cisternae ( see legend to these figures ) . These connections contained albumin at roughly the same level as in cisternae ( Figure 5F , Figure 5—figure supplements 2 and 3 ) . Moreover , the tomography- and photooxidation-based approaches ( Meiblitzer-Ruppitsch et al . , 2008 ) also indicated that albumin is present in the connecting tubules ( Figure 5—figure supplements 3 and 4 ) . In sum , despite the technical difficulties , the data are consistent with the notion that albumin is depleted in peri-Golgi vesicles and it is present in Golgi intercisternal tubules at levels similar to those seen in cisternae . To further distinguish between continuity-based and vesicle-based albumin transport , we then used two computational models . These models were constructed to assess whether the equilibration of albumin at the observed rates ( within 1–2 min ) through a closed system with a stack-like geometry ( defined using morphological parameters derived from our observations ) could be best explained by a scheme based on simple diffusion of albumin between the cisternae through tubules without biasing forces , or by a scheme based on vesicular transport . Notably , these models are limited to simulating intra-Golgi equilibration of cargoes and do not aim to simulate the entire traffic process through the Golgi including cargo arrival , departure and intra Golgi concentration steps . For diffusion-based transport , we simulated albumin diffusion through either one stack or through Golgi ribbons made of three to five longitudinally connected stacks containing vertical intercisternal tubules ( 'Materials and methods' and Figure 6A , B ) . Different numbers , dispositions and stabilities ( open time ) of these tubules were tested in the simulation ( Table 1; Figure 6C–F ) . We found that the rates of equilibration from cis to trans Golgi are fast , and are easily compatible with our experimental data ( see Figures 1–3 and Figure 1—figure supplement 1 for experimental data ) , even when we simulated infrequent and transient tubular connections ( Figure 6C ) . Also of note , even when connectivity gaps at any level of a stack are simulated , these gaps can be compensated for by the presence of connections at the same level in neighboring stacks longitudinally joined in Golgi ribbons , and/or by the fact that these gaps might be transient , i . e . , that connections might form and disassemble rapidly ( see scheme in Figure 6D–F ) . 10 . 7554/eLife . 02009 . 016Figure 6 . Computational simulations of intra-Golgi transport of albumin by diffusion via intercisternal tubules . ( A ) The Golgi stack was modeled as a system of six circular cisternae connected in series by five ( one per pair of cisternae ) vertical cylindrical tubules . ( B ) The same stack drawn in a ‘distended’ disposition . The size of the cisternae was set to 1 . 5 μm diameter and 30 nm thickness , and the diameter and length of the tubules to 30 nm and 100 nm , respectively . The simulations started with the first cisterna ( cis ) filled with albumin ( black shading ) and all of the others empty ( no shading ) . Albumin was then allowed to diffuse through the connections until it asymptotically reached equilibrium ( gray shading in all cisternae ) . The variations with time in the albumin concentrations in the other cisternae were calculated for the center of the cisternae ( see below ) . For wider tubule diameters of 60 nm and 120 nm , the 90% threshold of equilibrium was reached after 7 . 8 s and 6 . 8 s compared to 14 . 9 s for the 30 nm diameter . For shorter and longer tubules of 30 nm diameter with 50 nm and 150 nm lengths , the 90% threshold of equilibrium was reached after 11 . 4 s and 19 . 1 s . When assuming a system composed only of four circular cisternae , the 90% threshold of equilibrium in the fourth cisternae was reached after 3 . 4 s ( 60 nm diameter , 100 nm length ) and 3 . 0 s ( 120 nm diameter , 100 nm length ) compared to 6 . 4 s for the 30 nm diameter tubules . For shorter and longer tubules of 30 nm diameter with 50 nm and 150 nm lengths , the 90% thresholds of equilibrium were reached after 5 . 0 s and 8 . 4 s . ( C and D ) Time-courses of the equilibration process with different Golgi configurations and with stable or transient ( flickering ) intercisternal tubular connections . ( C ) One stack of six cisternae connected by one stable tubule per pair of adjacent cisternae ( five connections in all ) , or by one transient tubule per pair of adjacent cisternae . The tubules were set to be open for 50% of their time , with equal average open and closed times as indicated . In these simulations , the individual tubules opened and closed randomly . ( D ) Simulation of diffusion based albumin transport in Golgi ribbons of one ( blue ) , three ( red ) , or five ( yellow ) stacks ( each with 6 cisternae ) . The ribbons were completely connected horizontally by tubules joining adjacent cisternae . The total number of connections was five in all cases . For instance , the three-stack ribbon had one or two connections per stack . Nevertheless , complete equilibration was reached in less than 60 s . ( E ) Possible diffusion route of a soluble cargo through the Golgi ribbon with three stacks where the longitudinal tubules connecting the isolated stacks compensate for the vertical connectivity gaps . ( F ) Diffusion of a soluble cargo across a stack through transient intercisternal tubules . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 01610 . 7554/eLife . 02009 . 017Figure 6—figure supplement 1 . Computational simulation of the intra-Golgi equilibration of albumin argues against the classic vesicular transport model . To simulate albumin transport as mediated by vesicles traveling between adjacent cisternae , the geometry of the Golgi cisternae was set to the same parameters as in the standard configuration used for the diffusion model ( see Figure 6 ) . Each stack contained 4 to 6 cisternae , each with a diameter of 1 . 5 microns and a thickness of 30 nm . The vesicles were spherical , with a diameter of 50–70 nm . The simulations started with only the first cisterna filled with albumin ( black ) , and all the others empty ( white ) . Albumin transport is considered to proceed stepwise , where each step is defined as one vesicle detaching from each cisterna and fusing with an adjacent one . The albumin concentration in the vesicles is always assumed to be the same as in the cisterna from where they originated . The albumin concentration in the various cisternae is expressed as a function of the numbers of steps ( per cisterna , as defined above ) . Two scenarios were considered . ( A ) In the first , cargo is exchanged between the neighboring cisternae by anterograde as well as retrograde vesicles . This means that the first and last cisternae exchange one vesicle each at every step , while the other cisternae exchange two vesicles . This is required for the cisternal volumes to be kept constant . In this simulation , 50% and 90% equilibrium were reached after 3948 and 8846 steps per cisterna , respectively ( for comparison , 282 and 933 steps are required for 50% and 90% equilibrium across only two cisternae ) . For larger vesicles with diameters of 60 nm and 70 nm 90% equilibrium is reached with fewer steps ( 5119 and 3224 , respectively ) . However , experimental observations indicate a range of vesicle diameter from 50 nm to 70 nm , so for our calculations described below we considered the vesicle diameter to be 60 nm . ( B ) The second scenario represents the steady-state situation where only anterograde transport vesicles are considered . Here , in each step , one vesicle detaches from each cisterna and fuses with the next one . The volume constancy of the system is maintained by a vesicle coming into the first cisterna at each step ( containing the same albumin concentration as the first cisterna ) , and by an identical vesicle exiting the last cisterna . Here , 50% and 90% equilibrium are reached after 3785 and 6474 steps using 50 nm diameter vesicles ( compared to 563 and 1865 steps for only two cisternae ) . ( C and D ) Changes in albumin concentration in cisternae 2 , 4 , and 6 , expressed as fractions of the concentrations in the same cisterna at equilibrium , as a function of the number of steps , for the scenarios in A and B , respectively . All these simulations were performed using the MATLAB software and the script Skript3_Golgi_vesicular_transport . m is provided as supporting material . Based on the above numbers , the model ( scenario ‘a’ is considered here and the results are similar also with scenario ‘b’ ) predicts that close to 5119 vesicle budding/fusion steps per cisterna are required for 90% equilibration of albumin across the stack , considering 60 nm diameter vesicle , 6 cisternae per stack and the concentration of albumin in vesicles as same as that of the cisterna . Our own observations indicate that the concentration of albumin in the vesicles is 20% of that present in cisterna ( see Figure 5 ) , so approximately 25 , 000 steps per cisterna are needed for 90% equilibration of albumin across the stack . Since such equilibration happens in less than 2 min ( we consider 2 min here , for simplicity ) , each cisterna has to bud ( and fuse with ) close to 12 , 500 vesicles per min or about 200 per s . Based on these numbers one can calculate the turnover of a cisternal rim assuming that: ( A ) the diameter of an ‘ideal’ cisterna in the mammalian Golgi to be 1500 nm , and therefore the perimeter of the cisterna to be approximately 4700 nm , and , ( B ) vesicles are formed from the whole perimeter ( or rim ) of the cisterna , then a maximum of 78 vesicles of 60 nm diameter ( 4700/60 ) can bud from the rim of each cisterna at any given time . Thus , the budding ( and fusing ) of 78 vesicles can be considered one event of turnover of the cisternal rim . Thus , a rim , in order to produce 12 , 500 vesicles per min , has to turnover ( 12 , 500/78 ) 160 times per min i . e . , once every 400 ms , or in other words , the rim turns over 2 . 5 times per second . On varying the diameter of the vesicle from 50 to 70 nm this value ranges from 4 to 2 times per second for a stack with 6 cisternae ( see Table 2 for the results of these calculations ) . Since the typical Golgi of the HeLa cells that we used for live imaging studies contained 4 cisternae , we also simulated vesicle-mediated transport across such a Golgi . Based on the results of this simulation described earlier , the calculated turnover rate of cisternal rims varies between 2 and 1 times per second for a stack with 4 cisternae ( see Table 2 ) . Thus for a range of parameters used for the simulation ( variations in number of cisterna , vesicle diameter and bidirectional or unidirectional vesicle transport ) , the calculated turnover time of a cisternal rim ranges between 250 and 1000 ms ( see Table 2 ) i . e . , turnover of the cisternal rims ranges from 4 to 1 times per second . Regardless of the mode of intra-Golgi transport , if we match these numbers with published observations about the Golgi , we can then ask whether this vesicle-based turnover rates are possible/realistic or else . Actually , there are a few experimental observations , which , in the light of these calculations , indicate that a role of vesicles in albumin transport is not tenable:1 ) One is qualitative and is about the images of budding/fusing vesicles in Golgi stacks observed by electron microscopy . Thus , if each cisternal rim generates 200 vesicles ( and fuses with 200 vesicles ) per second then one would expect to see signs of this enormous budding and fusion activity at the cisternae . This is , however , clearly not the case . The Golgi cisternae look rather 'quiet' , with fairly rare images of budding or fusion . 2 ) Another observation is quantitative and is based on the association-dissociation rate of the Arf/COPI complex at the Golgi in living cells . The life cycle of a COPI-dependent vesicle is schematically comprised of the following steps: exchange of GDP/GTP on Arf , association of Arf-GTP and COPI with the membrane of cisternal rims , budding , fission , and uncoating of the vesicle ( which is initiated by GTP hydrolysis on Arf and the consequent dissociation of Arf from the vesicle membrane ) , and then docking and fusion of the vesicle with the next cisterna . At the uncoating stage , Arf and COPI are shed from the vesicle and become cytosolic . Thus , the on- and off-rate of Arf/COPI at the Golgi ( or at least of the portion of Arf/COPI that is involved in forming vesicles at the rims ) should reflect the overall turnover of the vesicles . Hence , if the rimmal membrane turns over ( i . e . , turns into vesicles and fuses with an equivalent number of vesicles ) 2 . 5 times per second , then the portion of Arf/COPI coat that is located at the rims should cycle , i . e . , detach from ( and attach to ) the rim membrane at the same rate , namely , 2 . 5 times per second . As noted , these considerations are valid only for the fraction of Arf and COPI that participate in vesicle formation at the rim , rather than for the total Golgi Arf and COPI pool; however , the available evidence indicates that a sizable fraction of the Golgi-associated Arf/COPI is located at the cisternal rims ( Oprins et al . , 1993; Weidman , 1995 ) . Hence , a substantial portion of the Golgi-associated Arf/COPI pool would be predicted by the vesicular model of albumin transport to associate with and detach from the rimmal membrane 2 . 5 times per second . Hence , the vesicular model of albumin transport predicts that a substantial portion of Golgi-associated Arf/COPI pool would associate with and detach from the rimmal membrane 2 . 5 times per second . The rate of association/dissociation of Arf and COPI with/from the Golgi complex has been measured in living cells by FRAP and FLIP video microscopy ( Presley et al . , 2002 ) , and its value has been used as a descriptor of the molecular process of Arf and COPI binding to , and dissociating from , the Golgi membrane ( Presley et al . , 2002 ) . It turns out that the Arf/COPI dissociation/association rate at the Golgi is mono-exponential with a half time of 15 s for Arf and 35 s for COPI , and that similar values apply to the dissociation of Arf/COPI in the presence of BFA ( Presley et al . , 2002 ) . These rates are orders of magnitude slower than those predicted by the vesicular model of albumin transport ( notably , the process underlying these slow rates remains unclear [Presley et al . , 2002] ) . Hence , the major fast cycling subpopulation predicted by the vesicular model of albumin transport is not detected . We conclude that this subpopulation of Arf and COPI does not exist , and that the vesicular model of albumin transport is incompatible with the available data . These data and arguments rule out , or at least strongly militate against , the possibility that Golgi vesicles turn over fast enough to mediate the rapid intra-Golgi transport of albumin . A further consideration that argues against a vesicular turnover rate as fast as that predicted by the vesicular model of albumin transport is based on the comparison between this predicted rate and the rate that is experimentally observed for the formation of other coated ( e . g , clathrin- or COPII-coated ) vesicles , in other cellular locations . The rates of formation of these vesicles ( Thor et al . , 2009 ) ( Steinman et al . , 1976 ) and even of synaptic vesicles at one of the fastest brain synapses ( Fernandez-Alfonso and Ryan , 2006 ) , are order of magnitudes lower than those predicted by the vesicular model of albumin transport . Such a difference , while formally possible , would be quite surprising . Given the above considerations , we conclude that the most parsimonious interpretation of the data is that the bulk of the trafficking of albumin through the Golgi is mediated not by the vesicles but by intercisternal tubules . DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 01710 . 7554/eLife . 02009 . 018Figure 6—figure supplement 2 . Computer-simulated model of the formation of a pH gradient between continuous cis- and trans-Golgi cisternae . ( A ) The model is based on three proton-handling components in the stack ( the size and geometry of which are as defined in the legend to Figure 6 ) : proton pumps located in the trans-most cisterna ( cisterna 1: swirls ) , proton channels located in all of the other cisternae ( cisternae 2–6: red circles ) , and a proton buffering system ( made up of proteins and/or lipids; dashes ) that decreases the diffusion coefficient of protons moving through the system . For modeling , the initial luminal pH was set to be identical to the cytosolic pH ( set at 7 . 0 ) . When the pumps start working ( activity of pumps: 10 protons/s in total; modeled on four pumps ) , the proton concentration in the lumen of the trans-most cisterna ( cisterna 1 ) starts increasing locally . Protons then start diffusing towards the cis-Golgi pole ( cisterna 6 ) . Their diffusion is slowed down by the proton buffering system ( proteins and lipids ) . The diffusion coefficient for protons here is set to 10 µm2/s ( for comparison: diffusion coefficient for protons in bulk water is 9 . 3 × 103 µm2/s [Pines et al . , 1988] ) . Moreover , as the protons move , they encounter the proton channels present in cisternae 2 to 6 , through which they can leak out into the extra-luminal space , resulting in a pH gradient across the Golgi stack . The rate of leakage is assumed to be linearly dependent on the difference between the proton concentrations in the lumen and in the cytosol ( zero leakage for an internal pH of 7 . 0 , and a leakage of 1 . 4 protons/s/channel for an internal pH of 6 . 0; the model assumes two channels per cisterna ) . The energy needed to generate the gradient comes from the ATP consumed by the proton pumps . The steepness and shape of the gradient depend on the interplay among these three defined proton-handling mechanisms . ( B ) Time-course of generation of steady-state pH in the cisternae according to this model ( final pH of cisternae 1 to 6: 6 . 00 , 6 . 17 , 6 . 32 , 6 . 45 , 6 . 55 , 6 . 60 ) . ( Assuming a diffusion coefficient of 100 µm2/s would result in a proton pump rate of 100 protons/s and a leakage rate of 14 protons/s at an internal pH of 6 . 0 , and the time scale in the Figure would range from 0 to 6 s . The final pH values would not change . ) DOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 01810 . 7554/eLife . 02009 . 019Table 1 . Simulated intra-Golgi transport of albumin by diffusion , via intercisternal continuities , occurs in the timescale of secondsDOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 019Length of tubule ( nm ) Diameter of tubule ( nm ) Number of cisternaTime needed for 90% equilibration ( s ) 10030614 . 91006067 . 810012066 . 85030611 . 415030619 . 11003046 . 41006043 . 410012043 . 0503045 . 01503048 . 4The size and geometry of the Golgi stack used for the simulations is defined in the legend to Figure 6 . The variable parameters used for the simulation are: length of tubules ( from 50 to 150 nm ) , diameter of tubules ( from 30 to 120 nm ) , and number of cisternae ( between 4 and 6 ) . The tubules here refer to the intercisternal tubules connecting two cisternae of a Golgi stack . The time needed for 90% equilibration of albumin across the Golgi stack under varying combinations of the indicated parameters was computed . As can be seen from the data , equilibration across the cisternae happens in seconds across all the conditions . For vesicle-based models ( Figure 6—figure supplement 1; Table 2 ) , we implemented a quasi-hopping scheme where one ‘event’ includes the loading and unbinding of a vesicle from one cisterna and binding and unloading of the same vesicle to a neighboring cisterna . This scenario is the limiting case allowing for fastest turnover . The considered system had the same geometry and number of cisternae as those used for the diffusion-based model ( Figure 6 ) . We found that in order to mediate cargo equilibration across the Golgi stack in 2 min , each cisterna would need to generate and receive hundreds of vesicles per second ( the exact number varies depending on the variables selected for simulation , e . g . , size of the vesicles or the number of cisternae per stack etc , see Figure 6—figure supplement 1; Table 2 ) . This very high rate of turnover is very difficult to reconcile with several lines of experimental evidence , as discussed in detail in the legend to Figure 6—figure supplement 1 and in the 'Discussion' . 10 . 7554/eLife . 02009 . 020Table 2 . Simulated intra-Golgi transport of albumin mediated by vesicles , predicts extremely fast turnover of cisternal rimsDOI: http://dx . doi . org/10 . 7554/eLife . 02009 . 020Vesicle diameter ( nm ) 506070Number of cisternae464646Steps for 90% equilibration19 , 61744 , 23011 , 35225 , 595714916 , 120Rim turnover time ( s ) 0 . 5760 . 2550 . 8290 . 3681 . 1290 . 500The size and geometry of the Golgi stack and vesicles used for the simulations are defined in the legend to Figure 6—figure supplement 1 . The variable parameters used for the simulation are: diameter of vesicles ( from 50 to 70 nm ) and number of cisternae ( between 4 and 6 ) . Albumin transport is considered to proceed stepwise , where each step is defined as one vesicle detaching from each cisterna and fusing with an adjacent one . For the calculations presented here , the albumin concentration in the vesicles is considered to be 20% of that present in the cisterna ( see Figure 5 and also legend to Figure 6—figure supplement 1 ) . The number of steps required to achieve 90% equilibration of albumin across the stack was computed and the time required for a single turnover event of the cisternal rim ( rim turnover time ) was calculated as described in the legend to Figure 6—figure supplement 1 . The rim turnover time varied from 0 . 25 to 1 . 12 s or in other words , the rim turns over from 4 to 1 times per second , depending on the condition used for the simulation . The results presented here are for the scenario 'a' ( discussed in Figure 6—figure supplement 1 ) and the results are very similar even in the case of scenario 'b' . In conclusion , the observed rate of intra-Golgi albumin transport ( cis-trans equilibration is observed in less than 2 min ) , seems easily compatible with diffusion-based models and inconsistent with vesicle-based quantitative transport models of albumin equilibration through the stack ( see legend to Figure 6—figure supplement 1 and 'Discussion' ) .
The main findings of this study are that soluble cargo proteins such as albumin traverse the Golgi complex rapidly by a mechanism that is different from compartment progression and involves diffusion via intercisternal continuities . This fast diffusion-based transport coexists in the same Golgi stacks with the slower movement of supramolecular cargo such as PC by compartment progression . Three lines of evidence ( kinetic , morphological , and computational ) converge to support the notion of continuity-based transport . The kinetic evidence , which consists of the observation that albumin traverses the stack in less than 2 min , at least fivefold to sixfold faster than PC , is consistent with diffusional transport , a fast process over short distances . It cannot , however , exclude in principle a role for other ( potentially ) rapid transport processes such as intercisternal shuttling by COPI-derived vesicles ( Pelham and Rothman , 2000; Presley et al . , 2002 ) . Thus , further discriminating evidence is required . The morphological experiments show that Golgi vesicles are depleted ( 80% ) of albumin while Golgi intercisternal tubules appear to contain albumin at similar levels as those seen in cisternae . Published biochemical studies concur with this depletion in vesicles ( Dahan et al . , 1994; Sonnichsen et al . , 1996; Gilchrist et al . , 2006 ) indicating that the observed depletion is not an artifact of reduced epitope presentation in the vesicles . These observations therefore favor a role for tubular continuities , over vesicles , in albumin transport . Clearly , the albumin depletion in vesicles is difficult to reconcile with the extreme efficiency of the intra-Golgi albumin transport . Moreover , computational models of intra-Golgi transport by diffusion show that tubules , even when very few and transient , are sufficient to mediate the observed fast rate of albumin transport through the stack . Instead , models of vesicular transport show that Golgi vesicles would have to form and fuse at rates of hundreds of vesicles per cisterna per second , and hence that cisternal rims would have to turnover several times per second , to mediate albumin transport ( see legend to Figure 6—figure supplement 1 for a detailed discussion of this point ) . These computational results appear difficult to reconcile with at least two lines of experimental evidence: ( a ) with such rates of vesicle formation/fusion , there should be morphological signs of such an enormous activity; however , cisternae look quiet , with fairly rare images of budding or fusion; and , ( b ) if cisternal rims turned over a few times per second as a result of vesicle budding , then the vesicular coat protein COPI ( or at least a substantial subpopulation of COPI ) should cycle on and off the Golgi complex at similar rates , i . e . , roughly three times per second; however , the half-time of COPI dissociation from the Golgi complex is roughly 35 s in live cells ( Presley et al . , 2002 ) . Finally , the vesicle turnover rates calculated to be required for albumin transport are orders of magnitude higher than those experimentally observed for any other type of coated vesicles at any cellular location including endocytic vesicles ( about 2 pinocytic vesicles per second per macrophage cell [Steinman et al . , 1976] ) or fast neuronal synapses that form close to 1–15 endocytic vesicles per second per synapse ( Rink et al . , 2005; Fernandez-Alfonso and Ryan , 2006 ) or COPII vesicles ( about 0 . 3 vesicles per exit site per second [Thor et al . , 2009] ) . Collectively , these considerations indicate that the simplest scheme to accommodate all of the available evidence on albumin intra-Golgi transport is one in which this soluble cargo reaches the cis-Golgi via the IC and diffuses rapidly across the Golgi stack in the cis-trans direction through ( most probably transient ) intercisternal tubules and concentrates on the trans-side . At the same time , non-diffusible cargoes that cannot enter intercisternal tubules ( such as PC-I ) traverse the stack slowly , by progression–maturation ( Mironov et al . , 2001 ) . As noted earlier , continuity-mediated transport has been proposed to occur also in the endo-lysosomal pathway ( Luzio et al . , 2007 ) and during the release of synaptic vesicles through flickering pores ( Rizzoli and Jahn , 2007 ) ( Alabi and Tsien , 2013 ) . Among the questions raised by the cargo diffusion model , one concerns the functional significance of its coexistence with maturation . We suggest that each process optimizes the transport of different cargo classes endowed with specific physical properties . For instance , PC-I forms aggregates in the early Golgi ( or earlier ) and is efficiently transported in this condensed state by compartment progression . Albumin and possibly other abundant serum proteins , instead , move from the ER to the Golgi in a soluble state ( Martinez-Menarguez et al . , 1999; Oprins et al . , 2001 ) . Thus , once at the Golgi , these proteins diffuse rapidly via continuities away from the cis-Golgi towards the TGN , where they are concentrated and exported . Existence of different modes of transport possibly also gives the cell/organism flexibility to switch between modes of transport depending on the cargoes being transported in a certain physiological or developmental stage . For example , during the spermatid development there is a clear increase in the proliferation of intracisternal tubules ( Vivero-Salmeron et al . , 2008 ) that we propose may coincide with an increase in the transport of cargoes that depend on diffusional mode of transport . The second question raised by the cargo diffusion model is how the Golgi maintains its compositional polarity in the presence of continuities and does not collapse to form one larger compartment . Several possibilities can be hypothesized . For instance , the collapse of connected cisternae may be prevented by scaffolds that maintain cisternal geometry , or it might actually tend to occur , but at a rate that is too slow compared to the half-life of transient continuities . Regarding domain segregation within continuous compartments , there are several known prior examples . They include the ER , with its rough and smooth domains ( Sitia and Meldolesi , 1992 ) , the neuronal plasma membrane , with different compositions in the soma and dendrites ( Bradke and Dotti , 2000 ) , and even the cytosol , where gradients of calcium and other second messengers are continuously created and maintained across different cytosolic regions ( Zaccolo et al . , 2002 ) . In most cases , the molecular basis for domain segregation is ill-understood , with a few possible exceptions ( Zaccolo et al . , 2002 ) . For the Golgi complex , multiple such domain-generating mechanisms might exist , and they might be based on: ( a ) the small size and the transiency of intercisternal tubules , which might act as filters by remaining open for a time sufficient for the passage of molecules such as albumin that would rapidly diffuse through the tubes , but not for that of larger molecules , such as lipid rafts , or protein clusters , which would be retarded by the small size and curvature of the tubes; ( b ) the arrival at the cis- and trans-Golgi poles of membranes of markedly different compositions and thicknesses from the ER and from endosomes , with the consequent creation of a cis-to-trans lipid compositional gradient through which Golgi resident enzymes may distribute differentially ( Trucco et al . , 2004; Patterson et al . , 2008 ) ; ( c ) the action of cytosolic scaffolds to nucleate different domains in different cisternae; and , ( d ) the action of intralumenal buffering and pumping systems to create gradients of ions , such as calcium and protons , in a fashion similar to that described for cytosolic messengers ( McCarron et al . , 2006 ) ( a quantitative model of this last mechanism and a further discussion of diffusion via continuities are in Figure 6—figure supplement 2 ) . Clearly , much work is needed to clarify whether and how some of these mechanisms apply . Another point of consideration while discussing the cargo diffusion model is the possible relationship between intra-Golgi transport and Golgi export , although a detailed analysis of the mode of albumin export is beyond the scope of this study . Earlier biochemical and our own imaging studies have shown that small soluble cargoes are cleared from the cells quite rapidly and a pH gradient is necessary for such an efficient export ( Yilla et al . , 1993 ) ( our unpublished data ) . Moreover , we show that albumin is concentrated in the TGN ( Figure 5G ) , and that this concentration most likely depends on the low TGN pH , as it is abolished by concanamycin , a specific inhibitor of the vacuolar proton pump that operates in the TGN ( not shown ) . Thus , the pH-dependent concentration of albumin on the trans-side of the stack ( Figure 5G ) appears to be an important driving force for albumin export in that it imparts some sort of directionality to the movement of this cargo by increasing the probability of cargo molecules to localize in the TGN vs the CGN . Based on the above collective considerations and the current results , a hypothetical transport model for albumin might be delineated as follows: albumin reaches rapidly the Golgi from the ER , where it diffuses rapidly across the stack at a rate dependent on the number and transiency of the connections , accumulating in the low pH trans-Golgi compartments . This intra-Golgi equilibration of albumin is fast and hence unlikely to be rate-limiting for export out of the Golgi ( differing in this regard from the slower transport of larger cargo such as PC-I which takes 10–15 min to cross and exit the stack ) . Rather , export is likely to depend on the rate of formation of the export carriers and on the concentration of albumin in these carriers , which , as noted , depends on the TGN pH . A notable consequence of this is that the export efficiency of other soluble cargoes might vary as a function of their propensity to concentrate in the TGN at low pH . The mechanism by which a low pH increases the albumin concentration in the TGN is unknown . On a speculative plane , at low pH , albumin might bind with increased affinity to a TGN protein or lipid , resulting in enhanced concentration and sorting into export carriers . Under these conditions albumin might also tend to self-aggregate , increasing concentration , and sorting efficiency . Interestingly , under the vesicular transport model , concentration in the TGN could be achieved by directional vesicle-mediated transport in a pH-independent way . Thus , the fact that export is pH-dependent is a further indication in favor of the diffusion-based mechanism . Finally , one may ask whether the coexistence of diffusion- and maturation-based mechanisms might help to rationalize the observation of different intra-Golgi trafficking patterns for various cargo proteins and under different conditions . Clearly , the trafficking of large non-diffusible cargoes ( e . g . , PC-I ) and that of soluble proteins ( e . g . , albumin ) is explained in a simple way by the above dual transport scheme . The case of VSVG , a trimeric transmembrane protein , is more complex , in that this cargo has been reported to either progress gradually through the stack like PC-I ( here and Mironov et al . , 2001; Trucco et al . , 2004 ) , or to spread rapidly from cis to trans cisternae like albumin ( Bergmann and Singer , 1983; Patterson et al . , 2008 ) , depending on experimental conditions . Here , a simple explanation would be that this dual behavior might depend on the ability of VSVG to exist in two states , an ‘aggregated’ one , in which it would form large clusters that behave like PC-I ( as proposed by others Griffiths et al . , 1985 ) , and a diffusible one ( mono/oligomers ) that behaves like albumin . Under this assumption , which remains to be verified , also the dual behavior of VSVG could be accommodated by our two-mechanisms scheme . The coexistence of multiple transport principles , involves a loss of simplicity and elegance compared to a single general mechanism . Nevertheless , evidence that multiple cargo transfer strategies are used in the secretory and endo-lysosomal pathways is emerging ( Luzio et al . , 2007 ) and the observation of different cargo transport rates for different cargoes ( Boncompain et al . , 2012 ) concurs well with the notion of multiple transport mechanisms . It is unclear whether the mechanisms so far described represent the full range of the existing transport strategies . The complete scenario will probably emerge through studies of further cargoes and trafficking steps in suitable cell types . Highly secreting cells , including certain cancer lines , might preferentially use one rather than another traffic mechanism . Uncovering the diversity of the trafficking modes will enhance our ability to understand and selectively manipulate different cargo classes , for research or therapy purposes .
HepG2 human hepatoma and HeLa cells ( both from ATCC ) were grown in Minimum Essential Medium ( MEM ) supplemented with 10% foetal calf serum , glutamine and antibiotics . Human fibroblasts were grown and used as described previously ( Mironov et al . , 2001 ) . The following polyclonal antibodies were used: anti-GM130 ( MA De Matteis , TIGEM , Italy ) , anti-TGN46 ( S Ponnambalam , Leeds University , UK ) , anti-albumin and anti-α1-antitrypsin ( DAKO , Denmark ) , and anti-VSVG ( MA De Matteis ) . The following were also used: nanogold-conjugated Fab fragments of anti-rabbit IgG and Gold Enhancer ( Nanoprobes , Yaphank , NY ) ; Protein A conjugated with colloidal gold ( J Slot , Utrecht University , The Netherlands ) ; and anti-rabbit , anti-mouse and anti-sheep antibodies conjugated with Alexa Fluor 488 , Alexa Fluor 546 and Alexa Fluor 633 ( Molecular Probes Europe BV , The Netherlands ) . GTPγS and CHX were from Sigma ( St . Louis , MO ) . The FUGENE6 transfection reagent ( used following the manufacturer instructions ) was from Roche ( Basel , Switzerland ) . To express constructs in human fibroblasts ( which were difficult to transfect ) , microinjection was performed . Albumin was expressed efficiently in 50% of injected cells . Of these , 30% expressed both albumin and PC , while the other 70% expressed only albumin . The other 50% of the injected cells contained PC and very little or no albumin , indicating that these human fibroblasts prefer to express either one or the other of these cargoes . Unless otherwise indicated , all other chemicals and reagents were obtained from previously described sources ( Mironov et al . , 2001 ) . The GFP-albumin construct was prepared by PCR amplification and sequential cloning of the pre-pro-signal region of albumin , the GFP cDNA from the pEGFP-N2 vector ( Clontech laboratories , Takara Bio Europe ) , and the albumin without the signal region , into the pcDNA-4B vector . Specifically , the construct was prepared by subcloning the HindIII-BamHI-digested , PCR-amplified ( forward primer , CCCAAGCTTATGAAGTGGGTAACCTTTATTTCCC; reverse primer CGCGGATCCTCGACGAAACACACCCCTGG ) pre-pro-albumin region into the pcDNA-4B vector , followed by the sub-cloning of the BamHI-EcoRI-digested , PCR-amplified ( forward primer , CGCGGATCCGTGAGCAAGGGCGAGGAGC; reverse primer , CCGGAATTCCTTATACAGCTCGTCCATGCCGAG ) GFP cDNA into the pre-pro-albumin-containing construct , and in turn the sub-cloning of the EcoRI-XhoI-digested , PCR-amplified albumin ( forward primer , CCGGAATTCGATGCACACAAGAGTGACCTTC; reverse primer , CCGCTCGAGTTATAAGCCTAAGGCAGCTTGAC ) into the previous construct . The final construct was verified by direct sequencing . The preparation the PC-III construct was as described previously ( Perinetti et al . , 2009 ) . Confocal and time-lapse images were obtained using a Zeiss LSM510 META confocal system ( Carl Zeiss , Jena , Germany ) . To measure co-localization between cargo and the Golgi marker proteins GM130 and TGN46 , we used approaches modified from published protocols ( Mironov et al . , 2001 ) . Co-localization indicates the intensity of cargo staining within the GM130- or TGN46-stained areas ( cargo localized in other cellular or Golgi regions was not considered ) . Two methods were used: Cryo-immuno-EM , immuno-nanogold-labeling , rapid-freezing cryosubstitution , serial sectioning , 3D reconstructions , and electron tomography were all performed as previously described ( Polishchuk et al . , 1999; Mironov et al . , 2001; Trucco et al . , 2004 ) . Correlative light-electron microscopy was performed as described previously ( Polishchuk et al . , 2000; van Rijnsoever et al . , 2008; Mironov and Beznoussenko , 2013 ) . All samples were analyzed under a Philips Tecnai-12 electron microscope ( FEI/Philips Electron Optics , Eindhoven ) . High-pressure freezing of samples was carried out according to Nicolas et al . ( 1989 ) . For samples from rat liver , CD-COBS Charles River rats were anesthetized with Nembutal and sacrificed . Liver slices were fixed by immersion in 1% gluteraldehyde in 0 . 15 M HEPES ( pH 7 . 2 ) for 1 hr and then post-fixed with 1% reduced OsO4 for 2 hr on ice . For cultured cells , cryo-immuno-EM , cryosections of HepG2 cells were prepared and immunolabeled with antibodies against albumin , antitrypsin , GM130 and TGN46 , and then analyzed as previously described ( Mironov et al . , 2001 ) . Quantification of albumin , antitrypsin and VSVG labeling within the Golgi stacks was performed using the Analysis software ( Soft Imaging Software Corporation; Mironov and Mironov , 1998 ) . In studies of intra-Golgi transport , the number of gold particles were counted per cisterna in 30 stacks per time point ( per experiment , in at least four experiments ) , and normalized to the value in the ER . This normalization was necessary to reduce the labeling variability between different sections and experiments . The TGN was defined as ribosome-less , tubular-reticular membranes adjacent to the trans-Golgi cisternae that were positive for TGN46 labeling . The IC elements were defined as clusters of tubular-vesicular membranes located near the cis side of a stack that were positive for ERGIC53 . Vesicles were defined as round profiles adjacent ( lateral ) to Golgi cisternae and not exceeding 65 nm in diameter , a feature of COPI-derived vesicles . Golgi tubules were defined as elongated membrane profiles with length:width ratios of at least 1 . 5 . For electron tomography , the analysis of intercisternal connections in chemically and cryo-fixed samples was performed on 200-nm-thick sections , passing roughly perpendicularly to the center of the Golgi stack , as described previously ( Trucco et al . , 2004 ) . The numbers of intercisternal connections were counted in single tomograms and varied between 0 and 2 per tomogram . Because the average volume of each Golgi section represented approximately 20% of the total stack volume ( conservatively assuming ‘idealized’ stacks made of round cisternae of about 1 micron in diameter ) , the number of connections detected in each tomogram was multiplied by five to obtain the total number of connections per stack . At least five tomograms were analyzed per experimental condition . Surfaces of Golgi membranes were rendered using the IMOD software ( http://bio3d . colorado . edu/imod/ ) . For photooxidation experiments , HeLa cells transfected with GFP-albumin were fixed with a mixture of 4% formaldehyde and 0 . 05% gluteraldehyde for 5 min and then post-fixed with 2% formaldehyde for 20 min . Next , cells were washed and incubated with 0 . 1% DAB in 0 . 2 M cacodylate buffer ( pH 7 . 4 ) on ice for 30 min . Then , cells immersed into DAB solution ( still kept at 4°C ) were placed under the LSM510 laser scanning confocal microscope ( Zeiss ) and an area of the Golgi complex exhibiting GFP fluorescence was irradiated with laser-derived blue ( 405 nm wave length ) light with maximal intensity . After each scan of the light beam , there was at least a 10 s interval , allowing DAB to diffuse to the site of bleaching . When the fluorescent intensity of GFP emanating from the irradiated area had disappeared , we checked whether the expected brownish DAB precipitate was visible over the irradiated Golgi area . If a modest brownish staining was visible , cells were processed for EM using reduced OsO4 and thiocarbohydrazide . Finally , 200 nm sections were subjected to electron tomography . For photo-oxidation experiments following FRAP studies , the Golgi region of the HeLa cells transfected with GFP tagged constructs was photobleached in the presence of culture medium and then after appropriate recovery time , the cells were fixed and subjected to photooxidation as above . Purified Golgi membranes were prepared from rat liver and incubated in vitro to induce the formation of vesicles , as described by Rothman et al . ( Malhotra et al . , 1989 ) . All of the membranes were then pelleted and prepared for cryo-immuno-EM . The method was essentially as described ( Bonifacino , 2001 ) with small modifications as described below . HepG2 cells infected with VSV were incubated at 32°C for 1h and then kept in methionine/cysteine free DMEM for 30 min . After which the media was substituted with one containing 0 . 2 mCi/ml of radiolabelled ( 35S ) cysteine and methionine for 5 min at 32°C and then incorporation of radioactivity was stopped by substituting with cold media . The cells were then incubated at 32°C for indicated times before lysing in RIPA buffer ( 150 mM NaCl , 20 mM Tris pH 8 . 0 , 0 . 1% SDS , 0 . 5% Sodium deoxycholate and 1% Triton X-100 ) followed by immunoprecipitation with anti-antitrypsin and anti-VSVG antibodies for 6 hr at 4°C . The immunoprecipitate was then subjected to EndoH digestion before being resolved by SDS-PAGE followed by autoradiography . The immunoprecipitates from the radioactive pulse-chase assay were eluted by incubating in the elution buffer ( 0 . 1 M sodium citrate pH 5 . 5 , 0 . 5% SDS and 1% β-mercaptoethanol ) at 90°C for 3–4 min followed by centrifugation at 13000×g for 5 min . The supernatant was then incubated with Endoglycosidase H ( 1000 units/ml ) for 1h at 37°C . The treated samples were then resolved by SDS-PAGE followed by autoradiography . All the simulations were performed using the MATLAB software together with the DIPimage toolbox ( www . diplib . org; Hendriks CLL , Rieger B , van Ginkel M , van Kempen GMP , van Vliet LJ . DIPimage , a scientific image processing toolbox for MATLAB . Delft University of Technology , 1999 . ) . See legends to Figure 6 , Figure 6—figure supplements 1 and 2 and the 3 Matlab scripts included in the supporting material for details of the individual simulations . The simulations of albumin diffusion through the Golgi stack were run both in 2D and 3D . Since the 2D and 3D concentration profiles were in very good agreement with the difference in equilibration time less than 12% , we only show data for diffusion in 2D . For the simulations , we used a mean field approach because we assume that the movement of the proteins in the experiments are mainly driven by diffusion . The diffusion equation ∂c/∂t = D ∇2c , with the albumin concentration c and the diffusion coefficient D , was solved using the finite difference method with the Forward Time , Centered Space ( FTCS ) scheme ( Ames , 1992 ) . Both the time and space derivatives are replaced by finite differences ci , j ( t + Δt ) = ci , j ( t ) + D Δt/Δx2 ( ci−1 , j ( t ) + ci+1 , j ( t ) + ci , j−1 ( t ) + ci , j+1 ( t ) − n ci , j ( t ) ) , where ci , j ( t ) is the concentration at the position i and j in x- and y-direction , respectively . The factor n is the number of nearest neighbor pixels ( n = 4 in the interior and less at the boundary ) . Space and time increments were chosen as 30 nm and 10 µs to ensure the stability criterion D*Δt/Δx2 < 1/2 . The change in the total particle number was monitored by measuring the integrated density ( concentration ) over time . The changes were between 1 and 4 particles out of about 20 , 000 particles over simulation times of 20–30 s . Dirichlet boundary conditions were applied , i . e . the albumin concentration was set to zero outside the cisternae and the tubules . The diffusion coefficient of albumin in the Golgi complex was taken as 10 µm2/s , which is a realistic value for a globular protein of 60 kDa in the cell cytosol ( Hobbie , 1978; Dauty and Verkman , 2004 ) . The system was simulated as a closed system with no flux from the ER and no flux out of the Golgi . Since the intra-Golgi diffusion rate of soluble cargoes is an order of magnitude faster than either entry into or exit out of the Golgi ( Figure 3—figure supplement 1 ) excluding them in the simulations does not affect the conclusions . Additionally the following assumptions are implicit in the model:i . The protein is initially distributed homogeneously in the first cisternae . ii . The initial protein density spreads out everywhere inside the confined volume ( to other cisternae and tubules ) according to the diffusion equation . iii . The diffusion constant for albumin is homogenous throughout the system ( inside the cisternae as well as inside the tubules ) . iv . The protein concentration is high enough so that it can be described by density and not by stochastically moving individual proteins . This is not really a limitation because stochastic random walk motion yields the same diffusion behavior as a continuous particle density . v . No interactions are considered between the diffusing proteins and the system boundaries ( cisternae and tubule walls ) . Of note , the use of modeling here was restricted to assess whether the equilibration rates of albumin across the stack are compatible with the continuity- or the vesicle-based model , or with both . The size of the cisternae was varied and did not significantly change the magnitude of the time to reach equilibrium . Changing the diameter and length of the tubules affected the equilibration times in a way typical for diffusion processes ( see figure legends for results ) . Also , systems with two tubules instead of one connecting adjacent cisternae were simulated . As mentioned above , three Matlab scripts are included in the supporting material that were used by us to produce the data for Figure 6 and Figure 6—figure supplement 1 . All experiments involving immuno-EM and immunofluorescence were performed at least three times on different days , and each treatment was carried out as triplicate samples . For quantification , 15–30 individual measurements were made for each sample ( for instance of the labeling density of albumin in Golgi stacks , or of the degree of co-localization of cargoes with a Golgi marker in a cell ) . For correlative light-electron microscopy ( Figure 3 ) , each experiment was carried out at least three times and at least three cells were examined . Experiments on the live dynamics of GFP-albumin ( Figure 3—figure supplement 2 ) were repeated at least four times . Values are mean ±SD from 30 stacks per time point , in three independent experiments for immuno-EM; and mean ±SD of 10 co-localization measurements per time point for immunofluorescence . | The Golgi is a structure within cells where proteins and other large molecules are modified and prepared for delivery to locations inside or outside of the cell . Each Golgi is made from a stack of flattened sacs called cisternae that are filled with fluid and enclosed by a membrane . Proteins and other molecules are transported to the Golgi by packages called vesicles , which fuse with the outermost cisterna , which is known as the ‘cis-face’ of the Golgi , and unload their contents . From here , the proteins are processed and modified by enzymes as they move through the Golgi towards the ‘trans-face’ on the opposite side . The modified proteins are then re-packaged into vesicles before being sent to their intended destinations . But how do proteins move through the Golgi ? Some researchers have suggested that proteins do not actually move: rather , the stacks of the Golgi move like a conveyer belt as new cisterna are added to the cis-face . However , other researchers have proposed that molecules proceed from one cisterna to the next inside small vesicles . It is also possible that proteins are transported through the Golgi in other ways , or by a combination of two or more methods . Now , Beznoussenko , Parashuraman et al . reveal that some small , soluble , proteins can move through the Golgi by diffusion . These proteins move much quicker than large protein complexes , which suggests that multiple transport mechanisms do co-exist within the Golgi . Furthermore , Beznoussenko , Parashuraman et al . found that these soluble proteins are most likely moving through some narrow tunnel-like connections between the individual cisternae . Following on from the work of Beznoussenko , Parashuraman et al . , the main challenge is to understand how all the different types of proteins that move through the Golgi are transported—which includes roughly a third of all human proteins . As many of these proteins are important for human health , learning to control their transport might create new opportunities to understand and treat disease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2014 | Transport of soluble proteins through the Golgi occurs by diffusion via continuities across cisternae |
Models of cell function that assign a variable to each gene frequently lead to systems of equations with many parameters whose behavior is obscure . Geometric models reduce dynamics to intuitive pictorial elements that provide compact representations for sparse in vivo data and transparent descriptions of developmental transitions . To illustrate , a geometric model fit to vulval development in Caenorhabditis elegans , implies a phase diagram where cell-fate choices are displayed in a plane defined by EGF and Notch signaling levels . This diagram defines allowable and forbidden cell-fate transitions as EGF or Notch levels change , and explains surprising observations previously attributed to context-dependent action of these signals . The diagram also reveals the existence of special points at which minor changes in signal levels lead to strong epistatic interactions between EGF and Notch . Our model correctly predicts experiments near these points and suggests specific timed perturbations in signals that can lead to additional unexpected outcomes .
Development is a dynamical process , so models that purport to be comprehensive must explicitly describe dynamics . Typically models report changes in protein levels and use them to predict phenotypic outcomes . However , the number of parameters involved makes implementation cumbersome and predictions non-intuitive . Classical embryology emerged in the absence of genetics and describes development in terms of overall cell and tissue phenotype . Such studies allow the inference of cell states that must exist even before any overt differentiation or morphogenesis is visible . For example , a cell is competent to respond to signals during a temporal window; it is committed or specified when those signals are no longer required , and determined when other signals cannot deviate it from its normal/assigned fate . Our aim here is to retain the conceptual clarity of classical embryology in models that made novel and quantitative predictions . Developmental states admit an intuitive topographical representation , as proposed by Waddington and later formalized mathematically ( Waddington , 1957; Slack , 1991 ) . The development of a cell is conceived as a downhill path in a shifting landscape controlled by cell signaling . Between two outcomes , or valleys , is always a ridge , and cells poised on the ridge can descend into either valley with equal probability . Once pushed off a ridge , cell fates are determined irrespective of subsequent twists and turns of the valleys . The Waddington picture suggests that cell fate decisions can be separated from the complexity inherent in specification and morphogenesis , which inherently simplifies any model . Vulval development in Caenorhabditis elegans ( Sternberg , 2005 ) is an appealing setting in which to quantify Waddington's landscape metaphor . Here , six vulval precursor cells ( VPCs ) , P3 . p-P8 . p , which are developmentally equivalent ( P3 . p is less competent and ignored in the model ) , receive an EGF signal from the anchor cell ( AC ) , and interact through Notch signaling , to eventually assume one of three different terminal fates ( Sternberg and Horvitz , 1989 ) . Cells P6 . p and P5/7 . p assume the 1° and 2° fates , respectively , and after three divisions form the vulva . The remaining cells , P3/4 . p and P8 . p , are assigned fate 3° . They divide once and fuse with the hypodermis . These basic facts suggested a model in which each cell travels in a landscape with three valleys ( fates ) that we represent in two dimensions to allow EGF and Notch to tilt the landscape independently as development proceeds ( Corson and Siggia , 2012 ) . The movement of a cell in the landscape depends on parameters that quantify the influence of each signal on the direction of motion in the landscape . Values for these parameters were obtained from known terminal VPC fate patterns of animals defective in the two signaling pathways , as well as from limited time-specific perturbations ( ablation of the AC at different stages ) ( Greenwald et al . , 1983; Ferguson and Horvitz , 1985; Sternberg , 1988; Sundaram and Greenwald , 1993; Koga and Ohshima , 1995; Simske et al . , 1995; Shaye and Greenwald , 2002; Félix , 2007; Milloz et al . , 2008; Komatsu et al . , 2008; Hoyos et al . , 2011 ) . Partially penetrant phenotypes are ideal for parameter fitting as they define the locations of ridges . From this data alone , focusing on the competence period ( Ambros , 1999; Wang and Sternberg , 1999 ) , we built a quantitative model for how EGF and Notch signals control fates , without considering the underlying complex genetic networks ( Corson and Siggia , 2012 ) . Our model has no fitting parameters that would couple the EGF and Notch pathways , implying that that if we fit two alleles , one in each pathway , then the outcome of the cross is defined with no additional freedom . Still , the model is sufficient to explain experiments showing epistasis , including non-intuitive interactions that were previously attributed to a context-dependent action of the signals , for example low EGF can promote 2° fate acquisition , or biochemical interactions between the pathways in a single cell . The model can predict context-dependent signals since it applies a nonlinear function to a linear combination of the two vectors representing the pathways . Recent work ( Barkoulas et al . , 2013 ) quantified EGF levels in a series of perturbation lines and examined the cell fate outcomes in multiple combinations with Notch perturbation lines . The new data allows further quantitative refinement of the model , and an extensive test of its predictions . Using the refined model , we now generate a phase diagram that displays the terminal vulval cell fate pattern as a function of the EGF and Notch signal strengths . The diagram predicts the existence of special transition points where epistasis is particularly evident . These special points organize the entire phase diagram and define allowed and forbidden phenotypic transitions in response to continuous changes in signal strength . Verification of these transition points is an important goal of future experiments . Our data demonstrate that taking a more abstract , yet mathematically rigorous , geometric approach can reveal underlying principles of development , and can provide immediate quantitative predictions more difficult to obtain through kinetic modeling of the complex underlying gene and protein networks .
To quantify the Waddington topography yet retain its intuitive appeal requires a geometrical representation of dynamics that we illustrate in two dimensions , Figure 1A , B . The axes of this so-called fate plane are abstract and only acquire meaning when we fit to actual data . Dynamics is represented in the fate plane by arrows ( a flow ) defining the velocity of a point representing the state of a cell . The developmental history of a cell is defined by tracing its trajectory in the fate plane . The very bottom of a valley is defined by a point with only inward pointing arrows . The low point on a ridge separating two valleys , a so-called saddle point , has two inward directions along the ridge and two outward sectors representing flow into the valleys . Turning to C . elegans vulval patterning , we adopt a two-dimensional fate plane to accommodate the two signaling pathways , Figure 1C ( Corson and Siggia , 2012 ) . The VPCs initially form an equivalence group , thus the same model suffices for each cell , and the three possible fates require three distinct points at which cell trajectories stably terminate . Each of the VPCs starts from the same location at the beginning of competence , but their fate planes distort in different ways in response to EGF and Notch signals , yielding divergent trajectories ( Figure 1D and Video 1 ) . The VPCs receive graded EGF levels that are constant in time , consistent with observed mRNA levels in the AC ( Barkoulas et al . , 2013 ) . Notch ligand production , on the other hand , is turned on in cells approaching the 1° fate ( see the response in P7 . p at t = 0 . 5 after P6 . p has reached the dotted line in Figure 1D ) . The model also incorporates fluctuations to account for partially penetrant phenotypes . The states of a VPC for a group of animals is depicted as a cloud of points traveling in the fate plane ( Figure 1E ) . With wild-type parameters , each of these clouds lies well within a valley following competence , corresponding to an invariant fate pattern . New data from Barkoulas et al . ( 2013 ) provide VPC fates for a set of lines with defined EGF levels crossed into strains with perturbed Notch levels to generate a grid of data . This invites comparison with the outcome of the model as signal levels are continuously varied that is , a phase diagram . While our previous model is consistent with experimental outcomes , we can use some of the new data to more finely estimate its parameters . The new fit is fully described under Methods ( see also Tables 1 and 2 , Figure 1—figure supplement 1 ) and its implications explored here . We ran our model for a range of Notch and EGF signaling levels , and determined the collective fate , or pattern , of the VPCs for each condition . The phase diagram in Figure 2 displays the outcome of our simulations and makes some surprising facts intuitive . Its power derives from the empirical observation that VPC fates , hence vulval patterns , are generally discrete and retain their identity under mutations in the signaling pathways . We highlight four notable features that are common to any system with two signals regulating three fates . Triple points evidently require that any pair of fates have a coexistence boundary in the fate plane as is evident in Figure 1E . A cloud of points induced by noise that lands near the point where the three phases meet , will populate all three fates . In the limit of a pure graded induction model for the vulva , Figure 2—figure supplement 1 , the 2° fated domain sits between the 1° and 3° fates . The fate plane is effectively one-dimensional ( flow onto a line with two saddles separating three fates ) and triple points cannot occur . This configuration is ruled out by the anchor cell ablation data ( Corson and Siggia , 2012; Félix , 2007; Milloz et al . , 2008 ) . As the ablation time increases , P6 . p transits from the 3° fate to equal fractions of 1° and 2° , implying boundaries between all three domains . Aspects of the phase diagram are specific to vulva but obvious without a model . Under low Notch , the pattern transitions from 33333 , to 33133 , 31113 , and 111111 as EGF levels increase and become sufficient to induce more distal cells . Under low EGF , strong ectopic Notch activity yields all 2° fates . Some quantitative features are straightforward consequences of the experimental data , Figure 2C . EGF and Notch null mutations are recessive , that is , a half-dosage is sufficient for patterning ( +/− labels ) ; this and an EGF overexpression line with known EGF levels ( JU1107 label; see Table 3 for a list of the experimental lines referred in this study ) delimit the central wild-type region . Similarly , data on animals with impaired Notch function and 1 or 2 ACs position the boundary between 33133 and 31113 . Elsewhere , the structure of the phase diagram is less intuitive and requires explanation . Around 0 . 3 × WT EGF and WT+ 0 . 2 Notch , small changes in EGF and/or Notch lead to five different fate patterns . Removing the molecular noise , Figure 2—figure supplement 2B , reveals that the triple point where P6 . p assumes all three fates lies very close to where the other VPCs convert to fate 2° . This degeneracy is caused by the simultaneous conversion of all five VPCs to 2° fates for EGF = 0 . A second confluence of multiple phases near the JU1100 point ( ~4 × WT EGF ) , as well as the very narrow domain of 33133 for Notch >1 , both are due to how data positions the dashed line in Figure 1E for the onset of Notch ligand production: near to the boundary between fates 1° and 2° . The JU1100 fate proportions ( P4/8 . p partially converted to 1° and 2° fates , and P5/7 . p partially converted to 1° fate ) were used to fit and are represented as a triple point ( P4/8 . p assume all three fates ) nearly coincident with the boundary between WT and 31113 , Figure 2—figure supplement 2A . Similarly , the phenotype of the CB1417 line implies that loss of the 1° and 2° fates occurs around the same EGF level . If we shift the signaling threshold away from the 1° domain , these fate confluences disappear and different patterns become accessible , Figure 2—figure supplements 3 and 4 . ( It is perfectly possible to observe with Notch = 1 the patterns 32223 , 32123 , and 22122 as EGF increases from below one to above , while retaining a large domain for the WT pattern signifying its robustness ) . Importantly , our model makes precise predictions about the activity levels of the EGF and Notch pathways at interesting points in the phase diagram . Figure 2D overlays the new experiments from ( Barkoulas et al . , 2013 ) on our phase diagram . In all but one case ( JU1100 ) , the EGF levels were measured while Notch levels are not and their values constitute a prediction . For the lowest expressing EGF hypomorph , lin-3 ( e1417 ) , ( denoted as CB1417 in ( Barkoulas et al . , 2013 ) ) , our model could not readily accommodate the very low EGF mRNA level in ( Barkoulas et al . , 2013 ) and the reported phenotype ( Table 1 ) . If , however , we use the model to predict the EGF level from the phenotype ( Table 2 ) , it agrees quite well with another measurement of the same allele reported in ( Saffer et al . , 2011 ) ( Table 3 ) . Elsewhere the correspondence between EGF level and VPC fates agrees well with experiments ( Figure 2—figure supplement 5B ) . Taken together , these examples reveal unexpected structure in the phase diagram that could not be gleaned from the available data without a model . The phase diagram in Figure 2A predicts that at certain Notch and EGF signal strengths , strong genetic interactions can be observed , for example the triple point T1 , where P5/7 . p adopt equal proportions of all three fates . Our fit places this point close to WT EGF levels , and 0 . 3 × WT Notch activity , thus a moderate reduction in Notch creates a novel type of ‘sensitized background’ in which modest increases or decreases in EGF both lead to loss of 2° fates ( see path P1 in Figure 3A ) . This prediction is born out by experiment ( Barkoulas et al . , 2013 ) . Whereas animals subjected to RNAi against lin-12/Notch ( fit as 0 . 4 × WT Notch ) show occasional conversion of P5/7 . p to the 3° fate , elevating EGF to only 1 . 25 × WT results in a substantial fraction of P5/7 . p cells adopting a 1° fate , Figure 3B–C and Table 4 . We predict that in the same RNAi background , removing one EGF copy ( 50% EGF activity ) would result in substantial conversion of P5/7 . p to the 3° fates . The different outcomes of Notch reduction under different EGF levels could be taken to suggest a switch between two functions of Notch , induction of the 2° fate and inhibition of the 1° fate ( Barkoulas et al . , 2013 ) . However , our model allows an apparently simpler explanation: EGF and Notch function independently , and because the path P1 cuts through a corner of the WT domain , changes up or down in EGF can result in conversion of 2° to 1° or 3° fates . The confluence of several patterns at the end of the path P2 in Figure 4A and the new data from ( Barkoulas et al . , 2013 ) allow a test of our assertion that the phenotype of a genetic cross can be predicted by fitting only the single mutants . The strain JU2092 ( Figure 4B , C2 , Table 4 ) combines modest ectopic Notch activity ( whose level we fit from an independent experiment , Figure 2C ) , with a defined level of EGF overexpression . In these animals , P4/8 . p cells are partially converted to the 2° fate while leaving P5-7 . p unaffected . This is quite surprising , as increasing EGF leads to a 2° fate without an adjacent 1° fate cell . Our model provides a simple , yet unexpected , explanation ( Figure 4C2 , Video 2 ) . When EGF levels are sufficiently large , we predict that P5/7 . p transit through the 1° fate domain on their way to the 2° fate , and thus transiently signal to P4/8 . p . This transient signal synergizes with ectopic Notch activity to induce 2° fates in P4/8 . p . This prediction depends critically on how we model Notch signaling in the fate plane ( dashed line in Figure 1E ) , a fit obtained from different experiments . The same position of the line for Notch signaling is independently required to explain the outcome of AC ablation experiments ( Corson and Siggia , 2012 ) . Thus , our geometric model remarkably ties together disparate experimental facts . Yet another observation affirming a context-dependent association of signals and fates is the observation that ectopic Notch activity favors the 1° fate in an EGF hypomorph ( Barkoulas et al . , 2013 ) ( path P3 in Figure 5A , B ) . This fact appears in the phase diagram as a slight inclination of the boundary separating the 33333 territory from 33133 , Figure 5A . This counterintuitive behavior of Notch is easily explicable from the trajectories in the fate plane of the model , Figure 5C . The EGF hypomorph leaves P6 . p poised between fates 3° and 1° and stretched out along the line connecting the saddle or decision point to the associated fates . But the Notch signal has no reason to be parallel to the 'ridge' leading into the saddle and it in fact favors fate 1° , explaining the increased 1° fates with ectopic Notch . More generally as a function of increasing Notch , we predict a peak in the fraction of 1° fated P6 . p followed by their conversion to entirely 2° fates for highest Notch as observed ( Barkoulas et al . , 2013 ) , Figure 5B . A converse experiment ( Zand et al . , 2011 ) found that mild EGF signaling could induce the 2° fate in a sensitized background of weak ectopic Notch that alone was not strong enough to induce appreciable 2° fates . The fate plane of the model , Figure 5D , shows that the sensitized state is poised near the saddle point between fates 2° and 3° . Again because the vector representing EGF is not parallel to the flow into the decision point , EGF can tip the balance and favors 2° . We could explain the enhanced fraction of 2° fates induced by EGF observed in Zand et al . ( 2011 ) , merely from the flow in the fate plane around the saddle point , with no special assumptions . By contrast , Ref . ( Zand et al . , 2011 ) related this outcome to a switch between different Ras effectors , suggesting ( if we translate to our language ) that the EGF vector , whose direction remains fixed in our model , rotated and pointed upward toward 2° and away from 1° for low ligand levels . Time-dependent perturbations during the competence period , as exemplified by anchor cell ablation ( Félix , 2007; Milloz et al . , 2008 ) , are more informative for model fitting and validation than end point data . However , other than these ablations , there is relatively little dynamic data to compare with . Timed temperature perturbations in temperature sensitive backgrounds clearly supply dynamic data and our model can predict the conditions under which dramatic effects are expected , Figure 6 , Videos 3–4 . We showed in Figure 5 that our model could explain the rescue of 1° fate induction by Notch in an EGF hypomorph . The predicted effect is very sensitive to timing , and can be realized by a temperature sensitive Notch mutant , Figure 6A . A Notch pulse early in induction diverts P6 . p further away from the ridge between 3° and 1° than a late one , and thus induces more P6 . p cells to the 1° fate . The same holds for the converse effect , induction of 2° fates by weakly activating lin-15 ( ts ) in a background with weak ectopic Notch activity and no AC ( Zand et al . , 2011 ) . The outcome is more dramatic if we take the same integrated level of EGF and concentrate it in the first quarter of the competence period , Figure 6B . This provides a direct test of our interpretation of pathway epistasis vs . Ras effector switching as described in Zand et al . ( 2011 ) . Finally , we predict dramatic effects in the pure lin-15 ( ts ) background , if we include P3 . p in the model and then extend the duration of the heat pulse , Figure 6—figure supplement 1 . Because the induced EGF is uniform across the VPCs , P3 . p gets induced and provides an additional lateral signal to P4 . p . Thus , there is a marked asymmetry between P4/8 . p in the degree to which they are induced to 2° fates . To realize this state may require some tuning of the intensity of the temperature step , but the predicted 22123 pattern is noteworthy since it is induced by EGF and involves adjacent 2° fates . Our fits imply that the EGF pathway is saturated in P6 . p under WT levels . As a result , EGF activity and the rate at which P6 . p progresses toward the 1° fate should not be too sensitive to moderate changes in EGF dosage . This prediction is consistent with the observed dynamics of Notch ligand expression in P6 . p ( an indicator of EGF activity and/or progression toward the 1° fate ) under EGF overexpression , which is similar to WT ( van Zon et al . , 2015 ) . The saturation of the EGF pathway can be tested by ablating the anchor cell in an EGF +/- background . We predict little change from the outcome for WT animals , in contrast with predictions for ablations in Notch perturbation lines ( Figure 6—figure supplement 2 ) . Until this point , we have used a geometric model with no coupling between the two signaling pathways to explain epistasis in a variety of experiments . However , it is well known that Notch signaling inhibits the MAPK pathway and Ras activation inhibits Notch signaling ( Shaye and Greenwald , 2002; Berset et al . , 2001; Yoo et al . , 2004 ) . In what way could the predictions of the model be improved by adding terms to represent the known biochemical interactions within a cell ? A simple way to incorporate one such coupling in the model , with no additional parameters , is to down-regulate Notch signaling in 1° fated cells along the same threshold that defines the production of Notch ligands , cf . ( Shaye and Greenwald , 2002 ) . Such a term could schematically at least represent the down regulation of Notch receptors in 1° cells , and thus capture intra-cell pathway interactions that were not present in the base model that only described the ( nonlinear ) production of Notch ligands in 1° fated cells . Fitting this modified model to the same data , we find only minor adjustments to parameter values ( Table 2 ) , and the predictions are largely unchanged . The resulting phase diagram , Figure 7B , is very similar to Figure 2A . The most noticeable difference is in the regime of reduced Notch and excess EGF: the 21112 domain extends to lower EGF levels , forming a longer boundary with the WT domain . Why the two models differ in this regime is clear from the corresponding cell trajectories , Figure 7C–F . In the absence of coupling , moderate EGF overexpression ( 1 . 5 × WT ) partially converts P5/7 . p to the 1° fate , but the cells are clustered along the 1°−2° boundary and 1° fated cells do not generate sufficient Notch signal to divert P4/8 . p from the 3° fate . Down-regulation of Notch in 1° fated cells restores a saddle point on the 1°−2° boundary that partitions the cells into distinct 1° and 2° groups , allowing the 1° cells to signal strongly to their neighbors . The resulting correlation between 1° fates in P5/7 . p and 2° fates in P4/8 . p is consistent with observations in the JU2113 line , with a slightly lower measured EGF level ( 1 . 25 × WT ) . The same bistability in P5/7 . p implies a direct transition from a WT pattern to 21112 , which is a strong prediction of our geometric modeling scheme , Figure 7B .
It is common in development to find a proliferating pool of progenitor cells that can be diverted into two mutually exclusive states by different signals . An example from early mammalian development is the split of the pluripotent epiblast cells into mesendoderm or neural fates at the time of gastrulation . Thus , our fate plane model with three possible fates is generic . The modeling proceeds from a general mathematical embodiment of the basic principles of development: cell fates are discrete , cells are competent to respond to signals over a limited period , and cells are committed when they assume their normal fates without additional signals . ‘Fates’ in the model are very schematic , they say nothing about the gene networks or morphogenetic events that ultimately define the fates . This simplification is plausible if the competence period ends before overt fate specification or differentiation occurs . The ‘fates’ in the model are then just a mathematical device to describe with minimal parameters all ways that two signals can divert cells towards three fates . Vulval patterning in C . elegans conforms to our assumptions ( Ambros , 1999; Wang and Sternberg , 1999 ) , and adds the simplification that the fates are terminal , so we can ignore subsequent bifurcations of the Waddington valleys . Our fate plane represents the state of a cell . The motion of a cell in the plane will depend on signals impinging on that cell at the moment in question . Logically , the production of Notch ligand will depend only on where the cell is , not on the signals . Thus , we parameterized the onset of Notch ligand production by a line in the fate plane . Its extension far into the 3° fate domain is probably unreasonable , but since there is no situation that samples that region , it’s immaterial . The geometry that accompanies the Waddington landscapes informs the interpretation of genetic experiments . Since the model has three distinct states represented as valleys , there must be ridges between adjacent fates , with the property that with a little noise half the cells will go to each state . The aptly named saddle point is the low point on such a ridge . The topographic analogy makes it very evident that for cells positioned anywhere near the ridge that feeds into the saddle , a small initial perturbation can lead to large changes in outcomes that are generally impossible to predict without a model . A ‘sensitized background’ in genetics is precisely an allele that flows near a saddle point . Thus to infer the activity of a gene by crossing it into a sensitized background may require a model , and conversely such experiments are very informative for model fitting since they expose the boundaries between fates . Figure 5 presented examples for vulva where small increments of Notch or EGF signaling in a sensitized background elicited non-intuitive changes in fates . Our modeling approach yields the most parsimonious parameterization of how two signals can control three fates . The flow diagram is two dimensional to accommodate the two signals , there are three fixed points , one for each fate , and then mathematics requires the presence of two saddles . It is informative to scan the list of 12 parameters in Table 2 that appear in the model . They all plausibly can be independently varied , for example , the time scale , noise level , EGF gradient , and autocrine signal all describe different phenomena . The other parameters define various vectors , also independent . The other models for vulva we discuss below have yet more parameters . Several other models have been constructed for vulval patterning . Reference ( Fisher et al . , 2007 ) constructs a discrete model for the vulva system incorporating virtually all relevant genes . It was shown to be consistent by logical programming techniques . However , from our perspective it does not deal with partial penetrance or treat continuously variable morphogens . A differential equation built from selected genes with Michaelis-Menten dynamics was implemented in Hoyos et al . , ( 2011 ) and required over 40 parameters . Because of the plethora of parameters both papers are reduced to making model inferences by assuming that the relative volumes of parameter space correctly weight embryo phenotypes . Closest to our approach is Reference ( Giurumescu et al . , 2009 ) , who describe a cell with two coordinates ( for Notch and EGF ) . They also do not fit partial penetrance data , or autocrine signaling , ignore the anchor cell ablation data that we found crucial for model selection , and also do not build in multistability . So formally at the end of competence all cells would revert to 3° fate . Nevertheless they use 13 dimensionful parameters , one more than we do . We were able to describe the Notch and EGF signals as two vectors combined by vector addition since they appear inside of nonlinear functions that build in the basic landscape of the VPC dynamics , namely three fixed points and two saddles ( Methods ) . This is why we assert that the inference of interactions is contingent on how we chose to represent the data , and the bulk of this paper is devoted to showing that what appeared as interactions when counting VPC fates , disappear when fit to our model as shown in Figure 3–5 . An analogy with the physics of thermal equilibrium illustrates the point . If the data consists of concentrations , then to discover interaction energies between species one has to fit the logarithm of the concentrations , not the concentrations themselves . While there are no exact results in dynamics analogous to the Arrhenius formula , imposing the basic topology of the flow field on a linear model for the signals is a plausible place to begin . Multiple steps in a signaling pathway are collapsed into one parameter; nevertheless , the model can capture intra-pathway interactions . In ( Corson and Siggia , 2012 ) , we described the cross between a loss-of-function mutant in the MAPK phosphatase , LIP-1 and ectopic EGF from lin-15 , by first adding the lin-15 contribution to the EGF from the anchor cell and then multiplying the sum by a parameter larger than one to account for loss of the phosphatase . The order of composition is dictated by the genetics . Our model that takes a linear combination of the vectors parameterizing each pathway and passes the vector sum through a nonlinear function , fits the genetic data in ( Berset et al . , 2001 ) even though LIP-1 is a Notch target and biochemically inhibits EGF signaling . We do not question the biochemistry , but it is not needed to explain the genetic result . We do not have to introduce a new fitting parameter for the down regulation of the MAPK cascade by the LIP-1 phosphatase to explain the genetic data . In the absence of fitting parameters explicitly coupling the two pathways , our model has reproduced effects previously attributed to pathway epistasis simply by including the basic genetic facts of vulval patterning . The most visible manifestations of intrinsic pathway interactions we found are correlations between the fates of adjacent cells . We were able to account for them schematically in Figure 7 , without additional parameters , merely by assuming that the effects of Notch on 1° fated cells disappears when that cell produces Notch ligands . Mutual pathway inhibition builds determination into our model , that is , a cell that gets close enough to its terminal state cannot be deflected from that state by ectopic signals . To go further , separate parameters would be needed for the down regulation of Notch by EGF and the inhibition of MAPK by Notch signaling ( Shaye and Greenwald , 2002; Berset et al . , 2001; Yoo et al . , 2004 ) , which we have not pursued since the necessary data are still very sparse , and this would have obscured effects that are independent of pathway interactions . Intuitively , the most informative experiments to fit a dynamic model manipulate signaling during the competence period when the cell is poised among its fates . The timed anchor cell ablations were very instructive for the vulva model . With recent advances in microfluidics ( Keil et al . , 2017 ) one can follow larval development for up to three days and image the worms every 8 min . All the vulva divisions can be scored with DIC optics , and it is possible to control signals with temperature and ts-signaling alleles . In Figure 6 , we presented instances where a Notch or EGF pulse delivered by a ts-mutant would have very different outcomes depending on the timing of the pulse within the competence period . These are appealing experiments since the intensity and duration of the pulse ( provided it is much shorter than the competence window ) will be adjusted to induce an appreciable phenotype , and then fit with one parameter . Then one quantifies the outcomes as a function of the phase of the pulse within the competence window with no additional parameters . The competence window itself can be redetermined by applying a temperature step to a ts-allele and the predictions are insensitive to details such as the shape of the pulse due to induction and persistence of the proteins ( van Zon et al . , 2015 ) that are impossible to control; only the relative timing matters . The phase diagram is computed directly from the model , but its general structure can be deduced from qualitative features of the fate plane . Our vulva model in Figure 1E allowed direct contact between any pair of fates . If instead we assumed a fate plane with the 2° fate wedged between and separating the 1° and 3° fates ( as might correspond to a graded action of EGF [Katz et al . , 1995] ) , the WT pattern is preserved but the phase diagram is very different , Figure 2—figure supplement 1 . There are no longer any triple points since three fates can never coexist in the fate plane . But ablation experiments ( Félix , 2007; Milloz et al . , 2008 ) required that the three valleys be adjacent in all combinations ( Corson and Siggia , 2012 ) , implying that phase boundaries meet at triple points . This and a few quantitative phenotypes largely determine the full diagram Figure 2—figure supplements 2–4 . A phase diagram forces the recognition that individual phases have boundaries and that boundaries will in general meet at triple points . A model will assign coordinates to all these features and allow experiments to be targeted to interesting points . Independent of a specific model , consideration of geometry unifies seemingly disparate outcomes . Models that define the fate plane geometrically and then fit to genes can unify what appear to be different signaling modalities . The relative importance of induction of 2° vs . ‘lateral inhibition’ of 1° ( Sternberg , 1988 ) by Notch has been debated . While biochemically the distinction may be very clear ( e . g . when Notch signaling deactivates the MAPK pathway [Berset et al . , 2001; Yoo et al . , 2004] ) , once we insist fates 1° and 2° are distinct ( bistability ) there is no distinction between favoring one or inhibiting the other . While the extensive data in ( Barkoulas et al . , 2013 ) requires the authors to admit context-dependent signals ( e . g . lateral inhibition vs . induction depends on EGF ) , our model accounts for all contexts with one set of parameters . Similarly , vulval patterning is often conceptualized as an interplay of ‘graded’ ( EGF level defines 1° vs . 2° fates [Katz et al . , 1995; Sternberg and Horvitz , 1986] ) and 'sequential' ( Notch induces 2° fates [Koga and Ohshima , 1995; Simske et al . , 1995] ) signaling , which are proposed to be partially redundant ( Kenyon , 1995 ) . Our quantitative model suggests a specific instance of this: EGF reaching P5/7 . p is not required for 2° fate induction but reduces loss of 2° fates under reduced Notch activity ( Corson and Siggia , 2012 ) . Genes and their interactions are the ultimate building blocks of development , so how can a phenomenological model that eschews any one to one correspondence with genes contribute to our knowledge of development ? For any signaling pathway , there are upwards of five components required from receptor to transcription and already 10’s of parameters . There are solid grounds for believing most of these parameters do not matter for the behavior of the system ( http://www . lassp . cornell . edu/sethna/Sloppy/ ) . Signal transmission between cells is less understood , particularly for hydrophobic ligands such as Hedgehog . Though pathway components and their biochemical properties have been elucidated in cell culture systems , it is far from evident that actual parameters fit to those systems have any relevance to the embryo . If the goal is to be more quantitative about how an egg turns into an embryo , then it is not clear to us , how denser lists of genes and their interactions will help . Rather pieces of the problem will be parameterized and the components joined in a structure defined by phenomenology . Evolution has built many levels of redundancy into development that insure a robust outcome but complicate efforts to reconstruct the process from its genetic components . Geometric models take a step back from genetic reductionism and impose basic embryological phenomenology ( competence , commitment , determination etc . ) . They begin from a parsimonious representation of how signals define fates that largely eliminates redundancy among parameters , and leads to sharper fits and more believable predictions . Geometric models should be the method of choice when confronted with sparse in-vivo data in developmental biology ( Corson et al . , 2017 ) .
The model ( Corson and Siggia , 2012 ) describes the state of each VPC by a vector r→ in two-dimensional space and its dynamics by the stochastic differential equation ( 1 ) dr→dt=1τ[ σ→1 ( f→+m→ ) −r→ ]+η→ ( t ) where ( 2 ) f→ ( r→ ) =2r→−2xye→x+ ( y2−x2 ) e→yis a polynomial vector field with threefold symmetry and the nonlinear function ( 3 ) σ→1 ( f→ ) =tanh‖f→‖f→‖f→‖ensures that the dynamics is bounded . The term ( 4 ) m→=m→0+l1m→1+l2m→2integrates a default bias towards the default fate 3° , m→0 , and the effect of EGF and Notch signaling , parameterized by two vectors , m→1 and m→2 ( red and green arrows in Figure 1C ) , which are combined linearly according to the levels of EGF and Notch ligands , l1 and l2 , on the cell . Variability in the dynamics is described by the stochastic term η→ ( t ) , parameterized by a coefficient of diffusion D in phase space: ( 5 ) ηi ( t ) ηj ( t' ) =2Dδijδ ( t−t' ) A fixed exponential gradient of EGF is assumed ( 6 ) l1={ γ2 , γ , 1 , γ , γ2 }while the level of Notch ligands a cell exposes to its neighbors is a function of its current state ( 7 ) L2 ( r→ ) =σ2 ( n0+n→1⋅r→ ) with ( 8 ) σ2 ( u ) =1+tanh ( 2u ) 2and varies continuously from 0 to 1 across a line parameterized by n0 and n→1 ( dashed line in Figure 1E ) . The level of Notch ligands received by a cell integrates contributions from its neighbors and from autocrine signaling , e . g . for P6 . p ( 9 ) l2 ( P6 . p ) =L2 ( P5 . p ) +αL2 ( P6 . p ) +L2 ( P7 . p ) where α is the relative strength of autocrine signaling . Modulations in signaling activity are represented as additive or multiplicative changes in the ligand levels . For example , EGF overexpression is described by ( 10 ) l1=λ{ γ2 , γ , 1 , γ , γ2 }where λ>1 denotes the dosage of EGF relative to WT , and ectopic EGF expression in a lin-15 mutant ( assumed uniform ) by ( 11 ) l1={ γ2 , γ , 1 , γ , γ2 }+δlwhere δl denotes the level of ectopic EGF . Simulations proceed as follows . The VPCs are initially equivalent and their state is drawn from the ( Gaussian ) steady state distribution obtained in the model when the term f→ that makes the dynamics multistable is removed and there is no signaling ( an ansatz for dynamics prior to VPC fate specification ) . The full dynamics are then computed for one unit of time , representing the period during which VPCs respond to EGF and Notch signaling , after which the fate of each cell is scored according to its final position . For this purpose , the dynamics is run for one more unit of time in the absence of signaling to allow convergence toward an attractor , then fractional fate assignments are computed according to the distance to the three attractors . With this fractional fate assignment , fate proportions output by the model under fixed realizations of the noise are continuous functions of model parameters , which is numerically convenient for parameter fitting . It may be noted that the form of our equation for VPC dynamics bears a resemblance to standard models for gene expression , comprised of a sigmoidal production term and a linear degradation term . However , the equation is vectorial , and the two coordinates in phase space do not stand for the levels of particular molecular species , providing instead an effective representation of the progression of a cell toward its eventual fate . Model parameters are fit to a set of experimental data , that is , the proportions of the different fates , 1°−3° , adopted by each cell , P4-8 . p , in different conditions ( Table 1 ) . In some backgrounds in which EGF signaling is perturbed , lin-3 mRNA levels have been measured and are used as a proxy for EGF levels . Elsewhere , ligand levels are treated as unknown parameters and fit to the data . Following the principles of Bayesian inference , we compute the posterior probability of a parameter set according to the deviation between fate proportions in the model ( computed from multiple simulation runs ) and in experiments , and to priors that disfavor values that are deemed biologically unreasonable , for example very large pathway sensitivities ( m→1 , m→2 ) or a very short response time ( τ ) ( Corson and Siggia , 2012 ) . Numerically , local maxima of the posterior probability are computed using the Levenberg-Marquardt algorithm . Several runs initialized from different , random parameter sets drawn from the prior distribution identify a unique global maximum . We then sample from the posterior using a Markov chain Monte Carlo algorithm initialized at the global maximum ( Corson and Siggia , 2012 ) . The posterior probability of a parameter set Θ is given by ( 12 ) P ( Θ|D ) ∝P ( Θ ) P ( D|Θ ) where P ( Θ ) is the prior probability of Θ and P ( D|Θ ) is the likelihood of the experimental data D given the outcome of the model with parameters Θ . In Corson and Siggia ( 2012 ) , we used the following approximation for the likelihood of the data , ( 13 ) P ( D|Θ ) ≈e−χ2 ( Θ ) 2where ( 14 ) χ2 ( Θ ) =N∑e , c , f ( pe , c , fexp−pe , c , fnum ) 2 In this equation , pe , c , fexp and pe , c , fnum denote the fraction of animals where cell c adopts fate f in experiment e , in experiments and in the model , respectively . N is the typical number of animals and simulation runs per condition , set to 100 . This inference procedure is intended to yield an approximate confidence interval for parameter values , and we use a uniform value of N rather than actual animal counts , in order to give all phenotypes equal weight in the fit . With the addition of new experiments to the data set ( see below ) , we found that some experiments could be improperly fit . Specifically , for the JU1100 line where P4/8 . p adopt the 1° fate in a small fraction of animals , the parameter ensemble obtained by Monte Carlo sampling included both parameter sets that match the phenotype , and parameter sets such that P4/8 . p never adopt the 1° fate . Indeed , the ‘cost’ imposed by Equation 14 for not fitting a small but non-zero pe , c , fexp is small , which is an artifact of that approximate expression: the likelihood of the data , P ( D|Θ ) , should be very small when this occurs . Here , we thus replaced Equation 14 with ( 15 ) χ2 ( Θ ) =N24∑e , c , f ( pe , c , fexp−pe , c , fnum ) 2Npe , c , fnum+12 With the original data set from ( Corson and Siggia , 2012 ) , Equation 15 yields a parameter distribution ( means and standard deviations ) that is very similar to that obtained using Equation 14 . For the augmented data set used here , we obtain a better fit of the JU1100 phenotype . Our model ( Corson and Siggia , 2012 ) was fit to the vulval patterns observed in various mutants . But to be quantitative , it was essential that we used partially penetrant alleles and fit the observed variable outcomes . However , several of the relevant mutants did not have well-defined EGF levels . These could in principle be fit along with the phenotype , but as such only loosely constrained the model . Reference ( Barkoulas et al . , 2013 ) now has provided an abundance of quantitative data for defined EGF levels . While there are no qualitative discrepancies with our prior model , we can make a much more stringent test of our modeling framework against the new data set if we add one of the newly quantified lines ( JU1107 ) to the training data ( Table 1 ) and then predict all the others . Doing this increases the parameter controlling the range of the EGF gradient ( γ; cf . Equation 6 ) by about 50% . With the resulting flatter gradient , it takes a smaller increase in EGF dosage to obtain a phenotype . Our fit for the EGF level in the JU1100 overexpression line ( Hoyos et al . , 2011 ) ( relative to WT ) is now 4 . 2 ± 0 . 5 instead of 7 . 4 ± 1 . 4 This level was not measured and is a testable prediction of the model . The second new piece of data we add to the training set is for epistasis between low EGF and Notch . We previously showed that model outcomes in this regime are sensitive to the directions of EGF and Notch signaling and can be synergistic ( Corson and Siggia , 2012 ) . The new data provides evidence for such a synergy ( Barkoulas et al . , 2013 ) , and is accommodated by a slight adjustment of the directions , which differ by less than 20° from the default values we took in Corson and Siggia ( 2012 ) ( Table 2 gives all the parameter values and SD for the old and new fit ) . To incorporate downregulation of Notch signaling in 1° fated cells ( Shaye and Greenwald , 2002 ) without introducing new parameters , Notch sensitivity is simply taken to vary inversely with a cell's production of Notch ligands . That is , we replace the term m→ describing the response to signals , cf . Equations 1 and 3 , by ( 16 ) m→=m→0+l1m→1+ ( 1−L2 ( r→ ) ) l2m→2 The sensitivity of the cell to Notch signaling thus varies continuously from 1 to 0 as it crosses the dashed line in Figure 1E . With this modification , the parameters of the model must be adjusted , and they are fit in the same way as for our main model . | At first , embryos are made up of identical cells . Then , as the embryo develops , these cells specialize into different types , such as heart and brain cells . Chemical signals sent and received by the cells are key to forming the right type of cell at the right time and place . The cellular machinery that produces and interprets these signals is exceedingly complex and difficult to understand . In the 1950s , Conrad Waddington presented an alternative way of thinking about how an unspecialized cell progresses to one of many different fates . He suggested visualizing the developing cell as a ball rolling along a hilly landscape . As the ball travels , obstacles in its way guide it along particular paths . Eventually the ball comes to rest in a valley , with each valley in the landscape representing a different cell fate . Although this “landscape model” is an appealing metaphor for how signaling events guide cell specialization , it was not clear whether it could be put to productive use . The egg-laying organ in the worm species Caenorhabditis elegans is called the vulva , and is often studied by researchers who want to learn more about how organs develop . The vulva develops from a small number of identical cells that adopt one of three possible cell fates . Two chemical signals , called epidermal growth factor ( EGF ) and Notch , control this specialization process . Corson and Siggia have now constructed a simple landscape model that can reproduce the normal arrangement of cell types in the vulva . When adjusted to describe the effect of genetic mutations that affect either EGF or Notch , the model could predict the outcome of mutations that affect both signals at once . The twists and turns of cell paths in the landscape could also account for several non-intuitive cell fate outcomes that had been assumed to result from subtle regulation of EGF and Notch signals . Landscape models should be easy to apply to other developing tissues and organs . By providing an intuitive picture of how signals shape cellular decisions , the models could help researchers to learn how to control cell and tissue development . This could lead to new treatments to repair or replace failing organs , making regenerative medicine a reality . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"physics",
"of",
"living",
"systems"
] | 2017 | Gene-free methodology for cell fate dynamics during development |
The signal recognition particle ( SRP ) directs translating ribosome-nascent chain complexes ( RNCs ) that display a signal sequence to protein translocation channels in target membranes . All previous work on the initial step of the targeting reaction , when SRP binds to RNCs , used stalled and non-translating RNCs . This meant that an important dimension of the co-translational process remained unstudied . We apply single-molecule fluorescence measurements to observe directly and in real-time E . coli SRP binding to actively translating RNCs . We show at physiologically relevant SRP concentrations that SRP-RNC association and dissociation rates depend on nascent chain length and the exposure of a functional signal sequence outside the ribosome . Our results resolve a long-standing question: how can a limited , sub-stoichiometric pool of cellular SRP effectively distinguish RNCs displaying a signal sequence from those that are not ? The answer is strikingly simple: as originally proposed , SRP only stably engages translating RNCs exposing a functional signal sequence .
The signal recognition particle ( SRP ) in all three kingdoms of life catalyzes the co-translational targeting of membrane and secretory proteins ( Egea et al . , 2005; Zhang and Shan , 2014 ) . At the beginning of the targeting reaction , SRP binds to a ribosome-nascent chain complex ( RNC ) . If the RNC displays a signal sequence , RNC-bound SRP binds the SRP receptor at the target membrane ( the endoplasmic reticulum membrane in eukaryotes , or the inner membrane in prokaryotes ) . The membrane-localized RNC is then transferred to the translocon , a protein translocation channel through which the nascent chain passes across , or into , the target membrane . Whereas mammalian SRP is composed of a 300-nucleotide RNA and 6 protein subunits , the simpler Escherichia coli SRP is composed of a 114-nucleotide RNA ( 4 . 5S RNA ) homologous to a conserved domain of the eukaryotic SRP RNA and a single protein subunit ( Ffh ) , a homolog of the mammalian SRP54 subunit . The E . coli SRP , which is used in these studies , can efficiently replace mammalian SRP in in vitro targeting reactions , demonstrating that it retains the core targeting functionality ( Bernstein et al . , 1993; Powers and Walter , 1997 ) . Despite wide and careful study , a consistent understanding of the initial step of the targeting reaction , in which SRP binds to translating RNCs , remains elusive . Equilibrium measurements of SRP binding affinities to RNCs stalled with nascent chains up to 35 amino acids in length indicated very tight binding ( ∼1–100 nM binding constants ) ( Bornemann et al . , 2008 ) . An estimate based on ribosome profiling of the ∼2000 most expressed proteins in E . coli indicates that at any given moment ∼10% of RNCs have a nascent chain less than 35 amino acids long ( Oh et al . , 2011 ) . Considering that the SRP concentration in E . coli is ∼400 nM ( 100-fold less than the ribosomal concentration ) ( Jensen and Pedersen , 1994 ) , such tight binding affinities would result in 75–100% of the SRP to be bound to these RNCs that are not exposing a signal sequence , and the majority of which ( ∼95% ) never will ( Bornemann et al . , 2008 ) . SRP binding to these RNCs would thus result in a large unproductive sink on the targeting reaction . Kinetic studies attempted to resolve this issue and concluded that , regardless of nascent chain length , SRP arrives at RNCs very quickly ( arrival rates on the order of 106 M−1 sec−1 ) and that nascent chain length mostly affects dissociation rates ( although different studies have determined a wide range of dissociation rates: ∼10–0 . 01 s−1 for RNCs with no nascent chain and ∼0 . 1 to 2 × 10−4 s−1 for RNCs with an exposed signal sequence ) ( Holtkamp et al . , 2012; Noriega et al . , 2014; Saraogi et al . , 2014 ) . The models that emerged had SRP non-specifically , and quickly arriving to RNCs as soon they begin translating and remaining bound until a nascent chain without a signal sequence becomes long enough to emerge from the ribosomal peptide tunnel and sterically displace SRP . Alternatively , the possibility of an additional factor ( such as the co-translational chaperone trigger factor ) was proposed to bind the RNC and displace SRP ( Holtkamp et al . , 2012; Bornemann et al . , 2014 ) . In either model , a large pool of SRP would be unproductively bound for a significant amount of time until displaced . Prior studies of SRP-RNC binding were performed on RNCs that had been stalled while translating the nascent chain . This approach was technically necessary to create homogenous RNC populations , but lacked the key temporal dimension , provided by active translation by RNCs . Multiple parameters are dynamically changing during active translation: ( i ) ribosome conformations , which cycle through pre- and post-elongation states , ( ii ) nascent chain composition , which changes with each new amino acid added , ( iii ) and length and folding state of the chain outside the ribosomal peptide tunnel . These factors all could affect how SRP interacts with , and subsequently targets , the translating RNCs . Here we developed a single-molecule fluorescence resonance energy transfer ( smFRET ) assay that allowed us to observe SRP binding to actively translating RNCs at physiologically relevant SRP concentrations . We show that both association and dissociation rates of SRP binding are sensitive to active RNC translation , with rapid and stable SRP binding to RNCs only upon exposure of a signal sequence outside the ribosomal peptide tunnel .
We used smFRET to detect SRP binding to translating RNCs . To this end , we labeled the 50S ribosome subunit with the FRET donor dye Cy3B at a unique cysteine on ribosomal protein L29 ( Noriega et al . , 2014 ) and , analogously , SRP with the FRET acceptor dye Cy5 at a unique cysteine in the NG domain of Ffh ( Zhang et al . , 2008; Shen et al . , 2012 ) . According to all structurally characterized SRP-RNC conformations , these dye positions have inter-dye distances within ∼40 and 50 Å ( Halic et al . , 2006; Schaffitzel et al . , 2006 ) , allowing for detectable FRET between SRP and the ribosome given that the Forster radius of the dyes is ∼65 Å ( Uphoff et al . , 2010 ) . To observe SRP binding at relevant concentrations , we performed smFRET experiments using zero-mode waveguides ( ZMWs ) , in which fluorescence measurements are taken from reactions occurring within small metallic apertures ( ∼150 nm in diameter ) that are patterned onto a glass substrate ( Chen et al . , 2014a ) . ZMWs limit background fluorescence from labeled reaction components in solution , which allowed us to measure SRP-RNC binding at 100 nM Cy5-labeled SRP , which is very close to the physiological 400 nM SRP concentration in bacterial cells ( Jensen and Pedersen , 1994 ) . We applied the smFRET assay to observe real-time binding of SRP to RNCs actively translating leader peptidase ( gene name lepB ) mRNA . LepB is a well-characterized in vivo SRP substrate with an N-terminal signal sequence ( de Gier et al . , 1996; Bornemann et al . , 2008 ) . A 3′-truncated lepB mRNA encoding the first 155 amino acids was immobilized via a biotinylated linker on ZMWs . Pre-initiation complexes ( ‘PICs’; composed of 30S ribosomal subunit , formylated methionine-tRNAfMet , and initiation factor 2 in complex with GTP ) were then assembled on the mRNA . Finally , we delivered labeled SRP and 50S subunits , as well as a cocktail of unlabeled elongation factors and charged tRNAs ( Johansson et al . , 2014 ) to the PICs while simultaneously measuring smFRET between SRP and RNCs ( Figure 1A , top panel ) . 10 . 7554/eLife . 04418 . 003Figure 1 . SRP-binding to actively translating RNCs . ( A ) Example smFRET trace of Cy5-labeled SRP , Cy3B-labeled 50S subunits , and unlabeled translation mix delivered at time = 0 to PICs pre-assembled on a truncated lepB mRNA ( encoding the first 155 amino acids ) and immobilized on ZMWs . To reduce non-specific interactions of SRP with the ZMWs , we pre-incubated the ZMWs with BSA , Blocking oligo , and unlabeled SRP , all of which were then thoroughly washed away ( Figure 1—figure supplement 2 ) . The top panel shows a schematic representation of the molecular events throughout the trace . The bottom panel shows the fluorescence intensity of the Cy3B ( green ) and Cy5 ( red ) signal upon 532 nm excitation . ‘AU’ indicates ‘arbitrary units’ . * denotes the initial 50S ribosomal subunit joining . ** denotes photobleaching of the Cy3B dye on the 50S ribosomal subunit . ( B ) Cumulative distributions of SRP first arrival times ( blue ) and 2nd–10th arrival times ( red ) to RNCs from the experiment described in ( A ) . ( n ≥ 141 binding events ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 00310 . 7554/eLife . 04418 . 004Figure 1—source data 1 . SRP-binding to actively translating RNCs . Source data for Figure 1 ( including figure supplements ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 00410 . 7554/eLife . 04418 . 005Figure 1—figure supplement 1 . EFRET validation . Histograms of the average EFRET values of SRP-RNC binding events when Cy5-labeled SRP was delivered to Cy3B-labeled RNCs stalled with a lepB mRNA ( encoding the first 75 amino acids ) and immobilized on ZMWs ( left panel , 100 nM SRP , n = 787 binding events ) or TIRFM slides ( right panel , 20 nM SRP , n = 186 binding events ) . Lines indicate the normal fits of low EFRET events ( blue ) , and high EFRET events ( orange ) . The percentage of total events and mean ( ± one standard deviation ) EFRET value for each are indicated . Please see ‘Materials and methods’ ‘smFRET assay characterization’ for further discussion of the figure . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 00510 . 7554/eLife . 04418 . 006Figure 1—figure supplement 2 . ZMW blocking to reduce non-specific SRP interactions . Example traces of 100 nM Cy5-labeled SRP delivered to ZMWs in the presence of BSA , and Blocking oligo in the absence ( top panel ) or presence of a dark SRP pre-incubation and wash-out ( bottom panel ) . Fluorescence intensity of the Cy5 ( red ) signal upon 532 nm excitation is shown . ‘AU’ indicates ‘arbitrary units’ . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 00610 . 7554/eLife . 04418 . 007Figure 1—figure supplement 3 . Wide variance in SRP arrival and residence times when RNCs are stalled in translation . Comparison of individual RNCs with more than five SRP binding events when 100 nM Cy5-labeled SRP was delivered to Cy3B-labeled RNCs stalled on a lepB mRNA ( encoding the first 55 amino acids ) and immobilized on ZMWs . Each column is an individual RNC with the grey dots indicating the arrival times ( top panel , n = 162 RNCs ) or residence times ( bottom panel , n = 162 RNCs ) of SRP-binding events . Red dots indicate median of binding events to each RNC , red line indicates variance of arrival or residence times . RNCs are arranged along the x-axis by decreasing variance . Data on the far-right labeled ‘total’ reflects all of the binding events analyzed together , with the blue dot and line indicating mean and variance , respectively . Note that the y-axis is in log10 scale . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 007 As shown in Figure 1A , we observed a time-resolved image of the translation and SRP recruitment process . Traces of fluorescence as a function of time show substantial and sustained Cy3B fluorescence increase upon delivery of 50S subunits and SRP , indicating translation initiation as the 50S subunit bound stably to the PICs ( green trace ) . Following initiation , we observed a stable period ( Figure 1A , 10–320 s ) devoid of apparent SRP-RNC binding events . This time window was followed by a period with extensive FRET events indicating SRP-RNC binding ( Figure 1A , 320 s and beyond ) . The EFRET values of the observed smFRET SRP-RNC binding signals were consistent with previous structural and single-molecule characterizations of SRP-RNC complexes ( Figure 1—figure supplement 1 , and ‘Materials and methods’ ‘smFRET assay characterization’ ) . To quantify SRP-RNC binding events across multiple translating ribosomes , we compared the first SRP arrival times ( time between subunit joining and the first SRP-RNC FRET event ) to the second to tenth arrival times ( time between individual SRP-RNC FRET events ) . This analysis showed that on average the first SRP binding events were much delayed ( ∼50-fold ) , compared to subsequent events ( Figure 1B ) . The time-dependent progression from a period of no SRP-RNC binding events to one of multiple SRP-RNC binding events is consistent with the original conception of SRP function , which posed that SRP only effectively binds RNCs after the signal sequence on the nascent chain emerged from the ribosomal peptide tunnel ( Walter et al . , 1981 ) . To confirm that the results we observed corresponded to SRP binding to actively translating RNCs , we varied the elongation rate . Since elongation rate is directly related to EF-G concentration , RNCs will translate the lepB mRNA more slowly at lower EF-G concentration . We therefore repeated the experiment at two different EF-G concentrations: 750 nM and 250 nM . We predicted that as RNCs translate more slowly , it would take longer for the signal sequence to become available and hence make observation of multiple SRP binding events during the recorded time course less likely ( Figure 2A ) . Indeed , we observed fewer SRP binding events per RNC at the lower EF-G concentration: ∼50% of RNCs had more than two SRP binding events in reactions containing 250 nM EF-G , whereas ∼80% of RNCs had more than two SRP binding events in reactions containing 750 nM EF-G ( Figure 2B ) . Moreover , as expected , the initial SRP-RNC binding events occurred later at lower EF-G concentrations: the half-time of first arrival was ∼350 s in 250 nM EF-G reactions and ∼280 s in 750 nM EF-G reactions ( Figure 2C ) . SRP arrival events that followed the initial binding event were similarly slower at the lower EF-G concentrations: half-time of arrival was ∼12 s in 250 nM EF-G reactions and ∼5 s in 750 nM EF-G reactions ( Figure 2D ) . These latter results are explained because it takes slowly elongating RNCs longer to translate nascent chains in which the signal sequence is optimally exposed , allowing SRP to bind most effectively ( Noriega et al . , 2014 ) . 10 . 7554/eLife . 04418 . 008Figure 2 . SRP-binding to RNCs translating at different rates . ( A ) Schematic representation of the effect on the number of SRP-RNC binding events and their first and subsequent arrival times when performing the experiment described in Figure 1 at 250 nM ( blue ) and 750 nM ( red ) EF-G concentrations . ( B ) Comparison of SRP-binding events per RNC distributions for the experiment described in A . Colors as in ( A ) ( n ≥ 621 binding events ) ( C–D ) Cumulative distributions of SRP first arrival times ( C ) and 2nd–10th arrival times ( D ) at 750 nM ( red ) and 250 nM ( blue ) ( n ≥ 131 binding events ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 00810 . 7554/eLife . 04418 . 009Figure 2—source data 1 . SRP-binding to RNCs translating at different rates . Source data for Figure 2 ( including figure supplements ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 009 From previous calibrations of translation rates under single-molecule conditions ( ∼5–10 s per amino acid , at the elongation factor concentrations used ) ( Uemura et al . , 2010 ) , we estimate that the first SRP arrival times occurred after approximately 35–45 amino acids were polymerized . This nascent chain length would correspond to the partial emergence of a signal sequence from the ribosome ( Houben et al . , 2005; Bornemann et al . , 2008 ) . To measure nascent chain length directly , we calibrated the extent of translation by observing labeled tRNA transit events during translation ( Figure 3 ) . To this end , we used an engineered 3′-end truncated lepB mRNA encoding 95 amino acids . The mRNA contained a single phenylalanine codon followed by three clusters of three sequential phenylalanine codons as chain length markers at positions 5 , 25–27 , 55–57 , and 85–87 ( Figure 3—figure supplement 1 , cWT for ‘calibration WT’ ) . After replacing unlabeled phenylalanine tRNA ( tRNAPhe ) in the translation mix with Cy3 . 5-labeled tRNAPhe and monitoring Cy3 . 5 fluorescence , we observed pulses in fluorescence intensity when phenylalanine was incorporated into the nascent chain ( Chen et al . , 2014b; Tsai et al . , 2014 ) . The duration of the pulses represent the transit of tRNAPhe through two rounds of peptide elongation and departure , whereas interpulse durations represent the translation time between the phenylalanine codons . 10 . 7554/eLife . 04418 . 010Figure 3 . SRP binding to translation-calibrated RNCs . ( A ) Representative smFRET trace of Cy5-labeled SRP , Cy3B-labeled 50S subunits , Cy3 . 5-labeled F-tRNA and unlabeled translation mix delivered at time = 0 to PICs pre-assembled on a lepB cWT mRNA ( encoding the first 95 amino acids ) and immobilized on ZMWs ( see text and Figure 3—figure supplement 1 ) . Fluorescence intensity of the Cy3B ( green ) , Cy3 . 5 ( orange ) , and Cy5 ( red ) signal under 532 nm excitation are shown . ‘AU’ indicates ‘arbitrary units’ . * denotes the initial 50S ribosomal subunit joining . ** denotes photobleaching of the Cy3B dye on the 50S ribosomal subunit . ( B ) Schematic showing when , relative to the x-axis shared by panels C–E , the signal sequence is exposed from RNCs . ( C ) Histogram showing how many amino acids have been polymerized when SRP first arrives to RNCs actively translating the lepB cWT mRNA ( blue ) or cMT mRNA ( orange ) . Y-axis shows both total events , and percent of total for RNCs translating cWT mRNA . Note x-axis is shared by ( C–E ) . ( D ) Scatter plot of SRP-RNC binding residence times relative to the number of amino acids polymerized when the event starts ( black dots ) , and average lifetimes of the residence times between the tick-marks ( red squares and dashed red line , with associated error bars that are too small to be seen ) . mRNA translated is lepB cWT . Note that for clarity the y-axis is split at 100 s , as indicated by the dashed grey line . Only traces in which four tRNA pulses were detected were included in the analysis in this panel and panel ( E ) . ( E ) Histogram showing how many RNCs are occupied by SRP relative to the number of amino acids translated when RNCs are actively translating a lepB cWT mRNA ( blue ) or cMT mRNA ( orange—with so few events that , at this y-axis scale , they are not visible ) . Y-axis shows both total events , and percent of total for RNCs translating cWT mRNA . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 01010 . 7554/eLife . 04418 . 011Figure 3—source data 1 . SRP binding to translation-calibrated RNCs . Source data for Figure 3 ( including figure supplements ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 01110 . 7554/eLife . 04418 . 013Figure 3—figure supplement 1 . Translation-calibration lepB mRNA constructs . Amino acid sequences of lepB WT , cWT ( for ‘calibrated WT’ ) , or cMT ( for ‘calibrated mutant signal sequence’ ) mRNA translation products . Blue box marks amino acids in N-terminal signal sequence . Orange boxes show expected F-tRNA pulses in calibration sequences . Red residues have been mutated relative to WT sequence . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 01310 . 7554/eLife . 04418 . 014Figure 3—figure supplement 2 . tRNA pulse characterization . ( Upper left panel ) Histogram of observed tRNA pulses per RNC with SRP-RNC binding events . ( Upper right panel ) Average tRNA pulse lifetimes , derived from single exponential fits to the pulse lifetimes . Error bars indicate 95% confidence of fits ( n ≥ 269 ) . ( Lower left panel ) Cumulative distributions of tRNA inter-pulse times ( as determined from the beginning of one to the beginning to the next ) . ( Lower right panel ) Half-times of tRNA inter-pulse times ( as shown in lower-left panel , normalized to the shortest half-time ( 0–5 amino acids ) . Colored bars indicate observed values and hatched , grey bars indicate expected values based on relative number of amino acids translated . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 01410 . 7554/eLife . 04418 . 015Figure 3—figure supplement 3 . Translation-calibrated SRP arrivals to RNCs . Histograms showing how many amino acids have been translated when SRP arrives for the first ( top panel ) , second ( middle panel ) , or third ( bottom panel ) time to RNCs actively translating a lepB cWT mRNA . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 01510 . 7554/eLife . 04418 . 016Figure 3—figure supplement 4 . SRP binding before and after each tRNA pulse . Heat plot showing the number of SRP-RNC binding events 60 s before or after the start of each tRNA pulse . DOI: http://dx . doi . org/10 . 7554/eLife . 04418 . 016 This experimental set-up consistently yielded clear tRNAPhe pulses superimposed on SRP-RNC FRET binding events ( Figure 3A ) . The tRNAPhe pulses behaved as expected: ( i ) The majority of RNC traces that showed SRP binding displayed three or four tRNAPhe pulses ( Figure 3—figure supplement 2 , upper left panel ) ; ( ii ) the second , third , and fourth tRNA pulses , during which clusters of three phenylalanines were incorporated , showed the same average lifetimes ( ∼30 s ) and lasted three-times longer than the average lifetime of the first pulse ( ∼10 s ) , in which only a single phenylalanine was incorporated ( Figure 3—figure supplement 2 , upper right panel ) ; and ( iii ) the inter-pulse times were proportional to the number of codons translated between them ( Figure 3—figure supplement 2 , lower panels ) . These results indicate that the RNCs under our assay conditions are actively translating and that we can accurately calibrate the reaction to map SRP binding events on nascent chain length . Using this assay , we next determined the distribution of initial SRP-RNC binding events . We observed that the majority of first binding events occurred when RNCs had translated between 40 and 55 amino acids ( 68% of all events ) ( Figure 3C , Figure 3—figure supplement 3 ) , corresponding to a nascent chain length at which the signal sequence emerges from the ribosomal peptide tunnel ( Houben et al . , 2005; Bornemann et al . , 2008 ) . Only 9% of observed first arrivals occurred before the nascent chain was 40 amino acids long . To confirm the dependence of SRP-RNC binding on the presence of a functional signal sequence , we measured SRP binding in a reaction translating lepB with a mutated signal sequence , previously shown to be inactive ( Houben et al . , 2005 ) ( Figure 3—figure supplement 1 , cMT for ‘calibration mutant’ ) . SRP binding events were virtually absent upon translation of the cMT lepB mRNA ( Figure 3C , E ) . We also observed that , past a 50 amino acid nascent chain length , SRP-RNC binding events occurred in quick succession ( at a rate of 1 per ∼1–2 codons translated after 50 amino acid chain length , as opposed to 1 per ∼50 codons before 50 amino acid chain length ) ( Figure 3C , Figure 3—figure supplement 3 ) , consistent with the ∼50-fold increase in SRP association rates upon exposure of a functional signal sequence shown above in Figure 1B . We also observed that SRP-RNC residence times were dependent on translation . Residence times were longest when RNCs were translating nascent chains in the 50–60 amino acid range ( Figure 3C , red squares and dashed line; average lifetimes of 74 . 4 ± 5 . 0 s ) . Shorter or longer nascent chain lengths resulted in ∼two-fold briefer SRP-RNC residence times ( Figure 3C , red squares and dashed line; average lifetimes of 34 . 7 ± 4 . 1 and 37 . 8 ± 3 . 5 s for chains in the 40–50 and 60–70 amino acid range , respectively ) . These results are consistent with SRP binding most effectively ( with the longest residence time ) to nascent chains of an optimal length ( Noriega et al . , 2014 ) . As nascent chains grew from 45 to 60 amino acids , we observed a rapid increase in the RNC occupancy by SRP ( where occupancy refers to the number of RNCs that are bound by SRP , with RNCs grouped by amino acids translated ) ( Figure 3D and Figure 3—figure supplement 4 ) , which depended on a functional signal sequence . When the signal sequence was mutated , occupancy was lost .
In this work , we applied the power of single-molecule approaches to observe directly the dynamics of SRP-RNC interaction on actively translating mRNAs . We show that under close to physiological conditions , SRP-RNC interactions change as the nascent chain grows and a signal sequence becomes exposed . Traditional analyses of binding rates , previously deduced from binding reactions using static , stalled RNCs , demonstrated that SRP-RNC binding kinetics are sensitive to nascent chain lengths but yielded conflicting results when describing SRP occupancy ( Siegel and Walter , 1988; Bornemann et al . , 2008; Holtkamp et al . , 2012; Noriega et al . , 2014 ) . The results presented here establish a new experimental paradigm in which the association of numerous other factors that interact co-translationally with RNCs , such as chaperones and nascent chain modifying enzymes , can be characterized dynamically . Our results suggest that SRP does not engage stably with translating RNCs that do not expose a functional signal sequence . These results are consistent with past biochemical and structural work showing that SRP can directly bind to an exposed signal sequence ( Zopf et al . , 1990; Keenan et al . , 1998; Janda et al . , 2010 ) . However , the conclusions contrast with those of previous work on stalled ribosomes , which suggested that SRP can bind prominently to RNCs with nascent chains as short as 20 amino acids containing a signal sequence still occluded within the ribosome ( Bornemann et al . , 2008; Holtkamp et al . , 2012 ) . We determined that the kinetic parameter most responsible for the observed signal sequence discrimination is a highly variable SRP-RNC association rate . When no signal sequence is exposed on RNCs ( >3 . 5 × 104 M−1 s−1 , for the experiments performed at 750 nM EF-G ) , we observed only negligible SRP occupancy . After translation advances far enough to have a signal sequence exposed on an RNC , the SRP association rates become at least 50-fold faster ( ∼1 . 8 × 106 M−1 s−1 , for the experiments performed at 750 nM EF-G ) . This contrasts with previous work suggesting that stable SRP-RNC complex association rates are insensitive to nascent chain length ( Holtkamp et al . , 2012; Saraogi et al . , 2014 ) . Our system is based on a FRET signal between dyes on the ribosome and SRP , limiting detection to interactions in which the inter-dye distance is less than ∼95 Å ( which would yield an expected EFRET of ∼0 . 1 ) . Additionally , our experiments have a 100 ms temporal resolution , and no binding would be detected , even if there was a FRET signal , if binding events had residence times much shorter than 100 ms . Given that the known SRP-RNC structures predict inter-dye distances of less than 50 Å ( Halic et al . , 2006; Schaffitzel et al . , 2006 ) , and the shortest reported average residence time for an SRP-RNC binding event is ∼70–100 ms ( Holtkamp et al . , 2012 ) , we are confident that we would detect binding events that are conformationally similar to the known SRP-RNC complexes . However , it is possible that there are intermediates in the binding reaction ( including shot-lived , unproductive diffusional collision events ) that we would not detect due to distance and temporal resolution limitations . Such intermediates could be encounter complexes previously observed to form with ∼106 M−1 s−1 association rates , regardless of nascent chain length ( Holtkamp et al . , 2012; Saraogi et al . , 2014 ) . However , if such intermediates formed in our reaction , they would need to be conformationally distinct from known SRP-RNC structures , or they would have been detected in this work . An alternative explanation for the discrepancies of our results to the previous work is that they may arise from differences between actively translating RNCs and stalled RNCs . Our assays show that when SRP binding events were compared among stalled RNCs we observed an enormous range in the variance of their arrival and residence times ( 2 . 4–40 . 9 s and 0 . 4–58 . 7 s median arrival and residence times , respectively; Figure 1—figure supplement 3 , and ‘Materials and methods’ ) . This variability indicates that individual stalled RNCs may exist in numerous different conformational states , many of which are likely to be inactive and perhaps off-pathway , despite each displaying the same length nascent chain . These data argue that quantitative results obtained with purified and stalled RNCs may be less physiologically relevant than results obtained with similarly purified but actively translating RNCs , which , by the very nature of the assay , represent functional states . Studies in yeast and mammalian cell observed some preference of SRP for RNCs carrying a signal sequence before it was exposed from the ribosome ( Berndt et al . , 2009; Mariappan et al . , 2010 ) . These studies used stalled RNCs in crude cell extracts , suggesting that perhaps factors absent in our assays and/or differences between prokaryotic and eukaryotic systems might influence early SRP-RNC interactions and affect signal sequence discrimination . The tools presented here are an important step towards quantitatively testing such a possibility with actively translating ribosomes . More generally , the methods presented here promise to be useful in studies of other RNC-associating factors , including nascent-chain modifying enzymes and co-translational chaperones . Our results resolve a long-standing question in the field: how can a limited , sub-stoichiometric pool of cellular SRP effectively distinguish RNCs that display a signal sequence from those that do not ? The answer appears strikingly simple: as originally proposed ( Walter et al . , 1981 ) and here confirmed using dynamic single-molecule measurements at physiologically relevant concentrations , SRP only engages translating RNCs that expose a functional signal sequence .
All proteins used in this study were derived from E . coli strain MC4100 and expressed in E . coli . The Ffh and L29 expression constructs and purification protocols have been described previously ( Noriega et al . , 2014 ) . The Ffh ( E153C ) single-cysteine mutant was engineered using the QuikChange mutagenesis kit ( Agilent , Santa Clara , CA ) . The DNA templates for the calibration lepB mRNAs ( cWT and cMT ) were ordered as GeneART oligos ( Invitrogen ) and cloned into pCR-Blunt II-TOPO vector according to the Zero Blunt TOPO kit protocol ( Invitrogen , Carlsbad , CA ) . The mRNA transcripts were in vitro transcribed and prepared for single molecule immobilization as described before ( Noriega et al . , 2014 ) . Blocking oligos ( 5′-CGTTTACACGTGGGGTCCCAAGCACGCGGCTACTAGATCACGGCTCAGCT-3′ , and its reverse complement ) were annealed in 50 mM TrisAcetate ( pH 7 . 5 at 25°C ) and 100 mM KCl by heating to 95°C for 1 min and then cooling down to 25°C on the bench . All chemicals , unless otherwise stated , where purchased from Sigma ( St . Louis , MO ) . Ffh ( E153C ) and L29 ( Q38C ) single cysteine variants were labeled with Cy5 and Cy3B , respectively as described previously ( Noriega et al . , 2014 ) . Labeling efficiency was typically >90% for both proteins . Labeled Ffh and L29 were reconstituted into SRP and RNCs as previously described ( Noriega et al . , 2014 ) . We tested binding of SRP to RNCs stalled on a 3′-truncated mRNA encoding the first 75 amino acids of lepB . This RNC construct has been shown to bind SRP robustly under single-molecule conditions ( Noriega et al . , 2014 ) . When the dye-labeled SRP and RNCs were incubated together to allow SRP-RNC complex formation , smFRET was observed with 0 . 2–0 . 32 efficiency ( EFRET ) ( Figure 1—figure supplement 1 , left panel ) . To calibrate these EFRET values , we performed the same experiment using total internal reflection microscopy ( TIRFM ) , which does not quench the Cy5 signal as the aluminum walls of the apertures in the ZMW set-up do ( Chen et al . , 2014a ) . The calibration revealed that the observed signal corresponded to a corrected value of 0 . 33–0 . 5 EFRET ( Figure 1—figure supplement 1 , right panel ) . This is lower than the expected EFRET of ∼0 . 85 predicted from molecular modeling onto cryo-EM structures of SRP bound to RNCs ( Halic et al . , 2006; Schaffitzel et al . , 2006 ) . However , these are relatively low resolution structures ( 9 . 6–16 Å ) , and our previous work has shown that SRP can take on a variety of conformations on the ribosome ( Noriega et al . , 2014 ) , consistent with the observed bimodality in the EFRET distributions ( Figure 1—figure supplement 1 ) . The EFRET signals are too close to each other to be clearly resolved , so we did not pursue conformational distinctions further in this study . We also compared individual RNCs based on the arrival and residence times of SRP binding events . To do this we immobilized PICs on a truncated mRNA encoding the first 55 amino acids of lepB . We then delivered our single molecule translation mix ( as described in the section below ) and allowed RNCs to translate for 20 min at room temperature , which based on our active translation experiments , is enough time to ensure full translation of the mRNA . We then delivered 100 nM of Cy5 labeled SRP in to the stalled RNCs and visualized SRP-RNC binding events . This assay allowed us to determine that stalled RNCs , despite being homogeneously stalled with a 55 amino acid nascent chain , show as much as 1–3 orders of magnitude differences in the median and variance of SRP arrival and residence times ( Figure 1—figure supplement 3 ) . This broad variance among individual RNCs suggests that simply stalling RNCs induces a variety of different RNC conformations that affect SRP-binding , many of which may be physiologically irrelevant because the RNCs are not actively translating . The experiments with stalled RNCs were performed as described before ( Noriega et al . , 2014 ) . The ZMW delivery experiments with translating RNCs were performed as previously described ( Chen et al . , 2014b; Johansson et al . , 2014; Tsai et al . , 2014 ) with the following modifications: the Tris-based polymix translation buffer was supplemented with 5 mg/ml of Ultrapure BSA ( Ambion , Carlsbad , CA ) and 10 µM Blocking oligo . Other standard blockers such as PEG , poly-L-lysine , aprotinin , kappa- , and beta-Casein had no effect on non-specific SRP-ZMW interactions . An additional ZMW chip preparatory step was also added after immobilization of the PICs: 20 µl of wash solution supplemented with 400 nM of unlabeled SRP was added to the ZMW chip for 3 min , removed and then rinsed and washed with 20 µl of wash solution without unlabeled SRP . For all ZMW experiments , the concentration of unlabeled charged tRNAs , EF-Tu , and GTP ternary complexes was 2 . 45 µM . For experiments with labeled tRNAPhe , the concentration of tRNAPhe ternary complexes was 200 nM . The concentration of EF-G was 750 nM unless otherwise stated . None of the SRP-RNC binding kinetics curves presented in the figures ( except for Figure 3C ) fit single or double exponential curves . This is expected given that active translation is a complex multi-step process . We did not attempt to fit these curves to a theoretical equation . Instead , when rate estimates were necessary , we compared values at which the curves reached 50% of the measured effects . The data were analyzed using custom Matlab scripts , similar to those described previously ( Chen et al . , 2014b; Johansson et al . , 2014; Noriega et al . , 2014; Tsai et al . , 2014 ) , and made available at: https://github . com/trnoriega/Matlab-Single-Molecule . | Genes contain the instructions needed to make proteins from smaller building blocks called amino acids . These instructions are first transcribed to produce molecules of messenger RNA , which are then translated by a ribosome . This ‘molecular machine’ translates the instructions in the messenger RNA into the sequence of amino acids needed to make the protein . For some proteins to carry out their role , they need to be delivered to the outside of the cell , or inserted into one of the cell's membranes . As they are being built , these proteins are identified by a so-called ‘signal recognition particle’ , which is often called an SRP for short . The SRP attaches to the new protein when it is still joined to the ribosome , and pulls the protein-ribosome complex to an opening in the target membrane . The new protein chain then enters this opening and either passes through to the other side of the membrane , or ends up embedded within it . To date , most studies that have investigated this process have involved scientists stalling the building of the new protein to see how SRP interacts with inactivated protein-ribosome complexes . Unfortunately , this means that some of the details of what happens during this process have likely been missed . Now , Noriega et al . have addressed this problem by developing a method to watch , in real-time , a single active protein-ribosome complex interacting with individual SRPs . This was achieved by attaching fluorescent molecules to SRP and protein-ribosome complexes purified from the bacterium E . coli . The distance between the two fluorescent molecules was then tracked over time . This revealed that the SRP typically binds to the protein-ribosome complex after 40–55 amino acids have been built into the protein . At this point , a so-called ‘signal sequence’ of amino acids has emerged from the complex and can be recognized by the SRP . Earlier studies had suggested that signal sequences might tell the SRP when to bind , but this had not been demonstrated in experiments using active protein-ribosome complexes . The strategy of using fluorescent molecules to follow single molecules undergoing this process in real-time could now be used by other scientists to re-examine and determine new properties of the protein-ribosome complex in action . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | Real-time observation of signal recognition particle binding to actively translating ribosomes |
Copy number alterations ( CNAs ) play an important role in molding the genomes of breast cancers and have been shown to be clinically useful for prognostic and therapeutic purposes . However , our knowledge of intra-tumoral genetic heterogeneity of this important class of somatic alterations is limited . Here , using single-cell sequencing , we comprehensively map out the facets of copy number alteration heterogeneity in a cohort of breast cancer tumors . Ou/var/www/html/elife/12-05-2020/backup/r analyses reveal: genetic heterogeneity of non-tumor cells ( i . e . stroma ) within the tumor mass; the extent to which copy number heterogeneity impacts breast cancer genomes and the importance of both the genomic location and dosage of sub-clonal events; the pervasive nature of genetic heterogeneity of chromosomal amplifications; and the association of copy number heterogeneity with clinical and biological parameters such as polyploidy and estrogen receptor negative status . Our data highlight the power of single-cell genomics in dissecting , in its many forms , intra-tumoral genetic heterogeneity of CNAs , the magnitude with which CNA heterogeneity affects the genomes of breast cancers , and the potential importance of CNA heterogeneity in phenomena such as therapeutic resistance and disease relapse .
Research into the genetics of breast tumors has yielded comprehensive portraits of the somatic alterations acquired during the evolution of breast cancer genomes . Catalogues of recurrent driver alterations have been identified using an array of technologies ( Teixeira et al . , 2002; Adeyinka et al . , 2003; Chin et al . , 2006; Sjöblom et al . , 2006; Fridlyand et al . , 2006; Banerji et al . , 2012; Cancer Genome Atlas Network , 2012; Shah et al . , 2012; Ciriello et al . , 2015; Nik-Zainal et al . , 2016 ) . This information has been instrumental in furthering our understanding of the basic biology that underlies breast cancer development and has led to the development of diagnostic and prognostic tests and more importantly , the development of efficacious targeted therapies ( Dawson et al . , 2013 ) . While the adoption of therapeutic strategies targeting somatic cancer alterations and the pathways they control has had a profound impact on the treatment of breast cancer ( Alvarez et al . , 2010 ) , disease relapse and therapeutic resistance remain important challenges ( Ma et al . , 2015; Majewski et al . , 2015 ) . Recent research into these phenomena has implicated genetic heterogeneity ( i . e . sub-clonal variation or intra-tumoral heterogeneity ) as a common mechanism to explain recurrence and treatment failure ( Toy et al . , 2013; Carey et al . , 2016; Miller et al . , 2016 ) . However our knowledge of intra-tumoral genetic heterogeneity is limited and thus advancing it is instrumental in combating cancer . CNAs are an important class of somatic mutations that are acquired during the evolution of breast cancer genomes ( Dawson et al . , 2013 ) . Studies exploring the landscape of CNAs have considerably advanced our knowledge of breast cancer biology , with translational efforts leading to advances in the clinic . Most notably , the amplification of the ERBB2 ( HER2 ) oncogene defines a biological subtype of breast cancers and targeting this CNA has led to the development of a multitude of efficacious therapeutic agents that have dramatically increased the survival of HER2+ patients ( Arteaga and Engelman , 2014 ) . Additionally , numerous studies have demonstrated the utility of copy number information in the prognostic stratification of breast tumors ( Russnes et al . , 2010; Curtis et al . , 2012 ) and studies of the genes targeted by this class of somatic mutations have substantially advanced our understanding of breast cancer biology ( Gatza et al . , 2014; Cai et al . , 2016 ) . However , our knowledge of CNA intra-tumoral genetic heterogeneity is limited and given the importance of this class of somatic alterations warrants further investigation . Single-cell DNA sequencing methods have recently emerged as powerful tools for the study of intra-tumoral heterogeneity with recent applications in breast ( Navin et al . , 2011; Eirew et al . , 2015 ) and other tumors ( Francis et al . , 2014; Bolhaqueiro et al . , 2019; Bian et al . , 2018 ) revealing novel cancer genetics and biology . Here , we present a comprehensive analysis of CNAs in breast cancer based on the sequencing of thousands of single-cell genomes across a cohort of breast tumors . Our results offer in-depth views of the magnitude of intra-tumoral CNA heterogeneity present in breast cancer genomes , reveal novel and important observations that have been missed by earlier bulk studies , highlight the unique nature of the data and its ability to retrieve novel knowledge , and provide an important foundation for the future exploration of intra-tumoral genetic heterogeneity of this important class of somatic mutations .
Samples from sixteen patients were selected from a cohort population enrolled in two phase II open-label clinical trials conducted by the Brown University Oncology Group ( BrUOG ) . Two pre-treatment , freshly frozen biopsies per patient were obtained . Two blood samples from normal healthy individuals were also obtained and processed ( discussed below – Methods ) . For the tumor samples , one biopsy was subject to bulk DNA and RNA isolation for transcriptome sequencing and copy number profiling by sparse sequencing . The second biopsy was subjected to single-cell copy number analysis as described previously ( Baslan et al . , 2012; Baslan et al . , 2015 ) . In brief , tissue was processed to obtain a suspension of DAPI stained single-nuclei . Nuclei were subsequently flow sorted by ploidy ( measured by DNA-content ) for single-nucleus deposition , genome amplification and sequencing . For samples with multi-modal ploidy distributions ( i . e . diploid and polyploid populations ) ( Figure 1—figure supplement 1A ) , single-nuclei were sorted and sequenced from each distribution . A mean of 116 single-nuclei per tumor sample were sequenced in multiplex fashion for a total of 2086 single-cell genomes . For each nucleus , we targeted a sequencing depth of roughly 2 million reads , sufficient to call large copy number events ( >1 MB ) as well as focal events such as amplifications and deletions at a resolution of roughly 300kbs . Selected patient samples were representative of various biological and clinical variables such as PAM50 subtype , ploidy , and ER/PR/HER2 status as well as other parameters such as age and tumor size ( Figure 1—figure supplement 1B–D ) . Nuclei processed from the diploid ( 2N ) distribution of sorted samples could either represent cancer cells or normal cells present in the tumor mass ( i . e . normal epithelia , fibroblasts , and/or immune cells ) . When sequencing 2N nuclei of patient samples , we obtain copy number neutral , ‘flat’ genome profiles representative of non-cancer , stromal cells ( Figure 1—figure supplement 2A ) . Intriguingly , we observed recurrent focal homozygous deletions on chromosomes 14 and 7 , in otherwise diploid genomes ( Figure 1A ) . Further inspection localized the deletions to T-cell receptor alpha ( TCRa ) and beta ( TCRb ) loci ( Figure 1—figure supplement 2B ) . These deletions are the result of T-cell maturation and generated via the process of V-D-J recombination . Similarly , we observed focal homozygous deletions at the heavy and light Immunoglobulin loci ( IGH and IGL ) on chromosomes 14 and 22 , respectively , denoting B-cells ( Figure 1B ) . Thus , using single-nuclei copy number sequencing we can detect T- and B-cells present within the tumor mass . To gauge the sensitivity of T- and B-cell detection , we retrieved highly purified CD4+/CD8+ T-cells and CD27+/CD13+ B-cells and performed single-nuclei copy number analysis on approximately 96 nuclei for both samples with the expectation that all single-nuclei should carry their respective deletions . Deletion analysis led to a calculated detection sensitivity of ~80% and 50% for T- and B-cells , respectively ( Figure 1—figure supplement 2C ) . This allowed the quantification of T- and B-cells in the tumor mass in our samples and led to the detection of T-cells ( in varying proportions ) in almost all patient samples , with B-cell infiltration limited to a subset of tumors , ( Figure 1C ) consistent with previous studies ( Ruffell et al . , 2012 ) . Given the importance of T-cells in emerging immunotherapies ( Sharma and Allison , 2015 ) , we performed more detailed analysis on the identified T-cells and queried whether they represent clonally expanded T-cells that have infiltrated the tumor mass , or TCRa unique , non-clonal T-cells . To gauge the specificity in detecting genetically unique T-cells using single-cell sequencing we examined the pattern of breakpoints resulting from TCRa deletions . In the purified CD4+/CD8+ T-cell nuclei ( n = 95 ) , deletion breakpoints were found distributed over 14 genomic bins and in the majority of T-cells were not shared , suggesting unique TCRa recombination events ( Figure 1D and Figure 1—figure supplement 2D ) . Extending this breakpoint analysis to T-cells identified in patient biopsies , we found that within a given tumor none of the T-cells shared identical breakpoints , similarly suggesting that the identified T-cells were non-clonally derived and genetically distinct . This is in contrast to the breakpoints observed in nuclei derived from a T-cell leukemia which were recurrent and positionally identical across different single nuclei ( Figure 1D and Figure 1—figure supplement 2E ) . To quantitatively corroborate these observations , we devised a breakpoint distance metric and applied it to the abovementioned T-cell groupings ( Methods ) . Indeed , we find lower breakpoint pairwise-distance values in the single-nuclei derived from the T-cell leukemia compared to both CD8+/CD4+ T-cells and T-cells found in breast cancer biopsies ( Figure 1—figure supplement 2F ) . All the T-cell leukemia nuclei in this analysis shared identical breakpoints , p-value<10−6 , in contrast to T-cells found in tumor biopsies . In addition to the heterogeneity of the TCRa deletion breakpoints in T-cells , we observed large karyotypic abnormalities affecting some T-cell genomes ( Figure 1—figure supplement 2G ) . These karyotypic abnormalities were represented overwhelmingly by a loss of an entire copy of the X chromosome ( Figure 1E ) consistent with previous karyotyping studies of proliferating lymphocytes ( Nowinski et al . , 1990 ) . These X-loss T-cells , when found in the same patient sample , were genetically distinct at their respective TCRa locus as judged by the deletion breakpoints as well as the above mentioned breakpoint-distance metric ( Figure 1—figure supplement 2E and F ) . Interestingly , T-cells displaying X-loss were found more frequently in estrogen receptor negative ( ER- ) compared to estrogen receptor positive ( ER+ ) tumors ( p-value=0 . 0047 ) ( Figure 1E , insert ) . Given that BRCA1 has been shown experimentally to be associated with the process of X-inactivation ( Ganesan et al . , 2002 ) and that BRCA1-mutant tumors are mostly ER- tumors , this association is intriguing . Additionally , we and others have reported the detection of individual , aneuploid nuclei that carry non-clonal , genetically heterogeneous CNAs termed pseudo-diploids ( Navin et al . , 2011; Demeulemeester et al . , 2016 ) . Initially identified in two triple negative primary tumors and not their matching metastasis ( Navin et al . , 2011 ) , here , we extend their observation to tumors from all four PAM50 subtypes studied ( i . e . LumA , LumB , HER2+ , and Basal ) ( Figure 1c ) . Interestingly , we find that some carry prototypical breast cancer CNAs such as gain of 1q and loss of 16 p ( Figure 1F – right panel ) . The pseudo-diploids observed in our cohort however do not appear to be precursors to the clonal cancer cells constituting the tumor mass as they do not share their respective CNAs ( Figure 1—figure supplement 3 ) . We interpret the observation of these nuclei as a manifestation of the inherent genomic instability present in normal breast epithelial cells of cancer patients . Further analyses , across a larger number of single pseudo-diploid nuclei while also factoring somatic single-nucleotide variant ( SNV ) information , are required to definitively prove this . Thus , using single-cell sequencing we were able to detect immune cells present within the tumor mass and infer their genetic heterogeneity as well as extend the identification of pseudo-diploids to all PAM50 breast cancer subtypes and observe the occurrence of cancer associated CNAs in these cells . We then proceeded to analyze the genetic heterogeneity of large-scale CNAs present in sequenced single cancer nuclei . These events include CNAs larger than 3 MB as well as whole chromosome or arm level events . As a way of illustrating the data , we plot all single-nuclei copy number profiles from a given patient on the same graph . Doing so for tumor P5 ( LumA ) , one can visually distinguish clonal from sub-clonal alterations by observing either one or multiple copy number states at any given genomic position , respectively ( Figure 2A ) . For example , losses of 8 p and 18 p ( recurrent CNAs in breast cancers ) are found at one copy number state and are thus clonal events ( Figure 2A – red arrows ) . Alternatively , gains at 1q and 8q ( also recurrent breast cancer CNAs ) as well as losses on chromosomes 13 and X are found to be sub-clonal ( Figure 2A – blue arrows ) . For P5 , computing the fraction of the genome found to be clonal/sub-clonal using a stringent threshold ( minimum 10% of cells showing sub-clonality – Materials and methods ) , we find P5 to be sub-clonal in over half of its genome ( 56% sub-clonal ) ( Figure 2A - top bar ) . Extending this analysis to all tumors , we find a wide range of fractional sub-clonal values with tumors showing: ( 1 ) significant ( sub-clonal at ~>40% genome – Ex: P5 , P22 , and P59 ) , ( 2 ) moderate ( 5%–20% subclonal-- Ex: P14 , P58 , and P31 ) or ( 3 ) low levels of heterogeneity ( subclonal at <5% of genome – Ex: P16 , P41 , P1 ) ( Figure 2B ) . Representative examples of significant , moderate , and low level CNA heterogeneity samples , as illustrated for P5 , are provided ( Figure 2—figure supplement 1 ) . Importantly , regions found to be sub-clonal in tumor samples included regions: ( 1 ) known to be recurrently altered by CNAs ( ex: 8 p and 13q ) , ( 2 ) harboring genes encoding therapeutic targets such as ESR1 , CYP19A ( Ellis et al . , 2012 ) , CDK6 ( Turner et al . , 2015 ) , and PD-L1 ( Nanda et al . , 2016 ) , ( 3 ) harboring genes known to be recurrently mutated by single nucleotide variants ( SNVs ) such as PIK3R1 , PDGFRA , and RUNX1 ( Cancer Genome Atlas Network , 2012; Nik-Zainal et al . , 2016 ) , and ( 4 ) harboring genes experimentally shown to be involved in metastasis ( Valiente et al . , 2014; Wagenblast et al . , 2015; Ross et al . , 2015; Figure 2C - F and Figure 2—figure supplement 2A ) . DNA-FISH was used to validate a selected subset of these alterations ( Figure 2—figure supplement 2B ) . Importantly , for some of the alterations , we find that the sub-clonal CNAs exist at three or more distinct copy number states in different single cells , for example , 8q gains in P5 ( LumA ) and P22 ( Basal ) ( Figure 2G ) . This may have an effect on the level of expression/dosage of genes encoded on those chromosomes and thus may affect phenotypic heterogeneity . Reasoning that the increase of dosage of genes at 8q and/or 1q might be associated with advanced disease and hence bad prognosis , we devised an analysis approach to measure the relative dosage of these events in bulk copy number datasets and test their association with patient survival in a large , carefully annotated breast cancer dataset; METABRIC ( 20 ) ( Methods ) . Indeed , we find that 1q/8q high tumors in ER+/HER2- patients ( regardless of ploidy status of the tumor genome ) are associated with worst distant relapse free survival ( Figure 2H and Figure 2—figure supplement 2C ) . Further , among the newly discovered breast cancer subgroups ( IntClust , IC ) groups , we find that 1q/8q high tumors are enriched in the IC9 subgroup which is associated with high-risk of late distant relapse ( Rueda et al . , 2019; Figure 2—figure supplement 2D ) . Together , these data show that heterogeneity of large copy number events can: ( 1 ) affect a large proportion of the genome in any given tumor , ( 2 ) exist at multiple levels ( ex: dosage - three or more copy number states in the same tumor ) , and ( 3 ) that this heterogeneity can affect regions of the genome that are important for treatment or disease relapse . Focal amplifications and deletions comprise another important class of CNAs . Prototypical driver amplifications containing ERBB2 , CCND1 , MYC , and CCNE1 , as well as less commonly identified amplifications such as PPM1D and MDM2 , were identified in our patient cohort and in some cases were clonal ( i . e . identified in all single-nuclei sequenced ) ( Figure 3A and Figure 3—figure supplement 1A ) . However , as seen with larger CNAs , many tumors displayed chromosome amplifications genetic heterogeneity . For example in tumor P22 ( basal ) , three focal amplicons encompassing the VEGFA , MYC and CCNE1 loci were found in varying proportions and in certain instances ( VEGFA and CCNE1 ) in a mutually exclusive manner in sequenced single-nuclei ( p-value=0 . 003 – Fisher’s Exact Test ) . A sub-population lacking any of the amplicons , but having lost an additional copy of the RB1 gene via a focal deletion was also identified ( Figure 3B and C and Figure 3 – figure supplement B-D ) . Interestingly , the amplifications segregate with genetically distinct , geographically resolved tumor sub-populations ( Figure 3—figure supplement 1D and E ) . Thus , P22 amplifications are somatically mosaic . Other examples of amplicon mosaicism include P5 ( LumA ) where an amplification targeting GATA6 was found only in a sub-population of cancer cells and P59 ( Basal ) , where amplifications on multiple chromosomes ( chromosomes 3 , 5 , 6 and 18 ) targeting important genes such as LOX ( Cox et al . , 2015 ) and PRDM1 ( Nik-Zainal et al . , 2016 ) where found in the majority of cells but absent in others ( Figure 3—figure supplement 2A ) . In total , 6 out of the 16 ( 37% ) tumors analyzed displayed amplicon mosaicism in the form of presence/absence of one or more amplicons in different subpopulations of cancer cells . Presence or absence of an amplicon in single cells was not the only form of variation we observed in chromosomal amplifications . Another form of variation observed came in the form of the level ( or dosage ) of amplification . An example is P6 ( HER2+ ) where different spatially resolved sub-populations were identified carrying either: ( 1 ) only ERBB2 amplification , ( 2 ) ERBB2 and SKIL amplifications , or ( 3 ) ERBB2 , SKIL , and MYC amplifications ( Figure 3D and E ) . Importantly , for the SKIL and MYC amplicons , different single-nuclei were found to have varying levels/dosage of amplification . Where in some single-nuclei , SKIL and MYC levels reached over 60 copies , in others the level of amplification was less pronounced . For the MYC locus for example certain subpopulations contained MYC amplifications at less than 30 copies ( Figure 3F ) . This was confirmed using DNA-FISH analysis based on fluorescence signal intensity ( Figure 3G ) . This heterogeneity in amplicon level/dosage was also observed for the ERBB2 locus in another one of the six HER2 amplified tumors we analyzed ( Figure 3—figure supplement 2B and C ) . Thus , chromosomal amplicons can exist at different levels within different single cells in a tumor mass . We also observed homozygous deletions affecting known breast cancer genes such as MLLT4 , MAP2K2 , and NCoR1 , among others ( Cancer Genome Atlas Network , 2012; Nik-Zainal et al . , 2016; Figure 3—figure supplement 2D ) . However , we did not observe heterogeneity in this class of CNAs . We cannot rule out genetic heterogeneity of this form of variation since homozygous deletions are generally smaller in size than chromosomal amplifications and at the resolution of our analysis may have gone undetected . Nonetheless , our results illustrate varied forms of genetic heterogeneity in chromosomal amplicons affecting important breast cancer genes in a significant proportion of breast tumors . These observations are particularly important given that driver cancer genes commonly found in amplicons are generally perceived to be good targets for drug development ( Cancer Genome Atlas Network , 2012 ) , have been found to be associated with metastatic breast cancer ( Bertucci et al . , 2019 ) , and because sub-clonality of amplifications is difficult to infer from bulk sequence data ( especially targeted sequencing ) given its quantitative nature , as opposed to qualitative nature of SNVs . Previous studies utilizing multi-region sequencing to investigate somatic SNVs in breast and lung cancer have shown that a significant proportion of the variation that is found between different spatially resolved biopsies can be detected at the sub-clonal level in one of the biopsies with deeper sequencing ( de Bruin et al . , 2014; Zhang et al . , 2014; Yates et al . , 2015 ) . This has not been investigated in a genome-wide , unbiased manner for CNAs . For 9 of the 16 tumors analyzed , we were able to sequence and compare bulk DNA from two spatially resolved biopsies for CNAs . In concordance with previous studies , we find substantial differences in CNAs between biopsies ( Yates et al . , 2015 ) . Interestingly however , we find that much of the variation observed between the two biopsies can also be observed as sub-clonal variation in the single-cell data ( Figure 4A and Figure 4—figure supplement 1 ) . For example , in tumor P22 ( Basal ) , differences on chromosomes 8 , 12 , 15 and X were observed between the two bulk profiles but were also observed sub-clonally at the single-cell level ( Figure 4A and Figure 4—figure supplement 1A and B ) . Of 54 alterations differentially found in the two analyzed biopsies from the nine patients , 35 ( 65% ) were identified in the data from single nuclei . This was observed for broad copy number events as well as focal amplifications , for example CCNE1 and GATA3 ( Figure 4B ) . Importantly , additional heterogeneous CNAs were detected only in the single-cell data and not in the bulk comparisons , for example variation on chromosome 1q and 1 p in P22 as well as alterations on 1q and chromosome four in P5 , with the converse also being true ( Figure 4A and Figure 4—figure supplement 1 ) . Thus , in our cohort , a substantial proportion ( but not all ) of variation in CNAs between spatially resolved biopsies is detected at the single-cell level with higher resolution analysis and some variation is only observed at the single-cell level or via multi-region sequencing . To associate CNA heterogeneity with molecular and clinical parameters we utilized two metrics from the single-cell data , applying stringent thresholds ( Methods ) . The first metric is the fraction of the genome found to be sub-clonal . The second metric is the number of recurrent , sub-clonal breakpoints . We chose four discrete variables for association analysis: PAM50 subtype annotation and ploidy status ( biological parameters ) and HER2 and estrogen receptor status ( clinical parameters ) . Plotting the data in bar graph format with tumors rank ordered according to increasing levels of heterogeneity while applying Wilcoxon rank order testing for associations ( Figure 5A and Figure 5—figure supplement 1A–C ) reveals several points . ( 1 ) Tumor samples of different PAM50 cancer subtypes display variable levels of CNA heterogeneity ( i . e: inter-tumor CNA heterogeneity ) . For example , both P9 and P16 belong to the HER2+ subtype but differ markedly in their CNA heterogeneity profiles . Similarly , both P5 and P1 belong to the LumA subtype but display notable differences in CNA heterogeneity . Thus , for any given subtype , there appears to be a range in terms of CNA heterogeneity in the analyzed samples . We cannot associate or rule out an association between heterogeneity and any particular subtype due to the size of our cohort . Increasing sample sizes as well as analyzing multiple biopsies per sample in future studies will be required to delineate this . ( 2 ) Polyploid tumors are significantly more likely to be heterogeneous on the copy number level and this relationship is unrelated to the total number of clonal alterations found in each tumor . ( 3 ) As a group ER- tumors are significantly more heterogeneous on the copy number level than ER+ tumors . Thus copy number heterogeneity is variable within breast cancer tumors of the same subtype and is associated with tumor polyploidy and ER- disease . There is one exceptional outlier , an ER+/LumA ( P5 ) with extensive copy number and clonal heterogeneity . Interestingly , unlike most ER+ tumors , P5 is mutated for the TP53 gene ( Figure 5B ) , as are the majority of ER- tumors . We also attempted to associate copy number heterogeneity across all samples with continuous variables such as patient age and tumor size and found no significant relationships ( Figure 5—figure supplement 1D ) . However , when restricting the analysis to ER+ tumors ( and excluding the ER+/TP53 mutated case ) we find a relationship where tumor size and CNA heterogeneity are inversely correlated ( p-value=0 . 02793 ) . The largest four tumors ( measuring at more than 10 cm2 ) display the least amount of CNA heterogeneity ( Figure 5C and D ) . This could be explained by the expansion and clonal sweep of a dominant clone in these tumors that is then captured when biopsying a geographically restricted region of a large tumor mass . Lastly , we examined the clonal composition of cancer cells from analyzed tumor samples ( i . e . # of genetically distinct clones in a given sample and their relative frequencies ) and find that tumor samples in general conform to three different clonal classes: ( 1 ) homogenous tumors where only a single clone is observed , found only in ER+ tumors; ( 2 ) heterogeneous tumors where multiple clones exist but with one clone dominant over others ( i . e . dominant >50% ) , a pattern found throughout the cohort subtypes; and ( 3 ) heterogeneous tumors where many clones are found at varying frequencies with none of them being dominant , mainly in ER- tumors ( Figure 5E and F and Figure 5—figure supplement 2 ) .
We provide a comprehensive analysis of breast cancer copy number alterations at the single cell level , and in the process , make many novel observations regarding the genetic heterogeneity of breast tumors . First , we observe an association between ER- tumors and T-cells exhibiting X-chromosome loss . This observation is at present of unclear clinical significance , but nevertheless might be a useful marker for this subtype of disease . Second , we observe pseudo-diploid single cells in the majority of our studied tumors , across different PAM50 subtypes , providing further evidence for the existence of such cells in breast cancer tissue . This , coupled with the observation of cancer-specific alterations ( e . g . 1q gain and 16q loss ) in these cells raises the question of whether these cells are a manifestation of the intrinsic genomic instability of the mammary epithelial cells of breast cancer patients and whether their detection might be useful in early stage risk assessment . Third , we observe profound genetic heterogeneity in CNAs affecting genes and regions with known biological relevance in breast cancer , such as genes associated with metastasis and therapeutic response . Thus , many cancer phenotypes might result from the selection for sub-clonal CNAs . Fourth , we find that CNA heterogeneity is not binary ( i . e . sub-clonal or clonal ) . CNAs can be found at different copy numbers in different populations of cells and that this , in the case of 1q/8q gains is associated with clinical outcome , likely a consequence of dosage dependent biology . Fifth , the observation of mosaicism in recurrent driver amplicons in a significant proportion of tumors has clinical implications given that driver amplicon genes are often targets for drug development ( e . g . , ERBB2 , MET , and EGFR ) and have been proposed to be good targets for drug development in breast cancer ( Cancer Genome Atlas Network , 2012 ) . Further , we find additional layers of heterogeneity in this class of CNAs: the presence or absence of an amplicon and variation in the level/dosage of amplification , which might be the result of extra-chromosomal amplicon instability ( Turner et al . , 2017 ) . This is of importance given that recent studies have provided evidence for a role of variation in gene dosage in therapeutic resistance to targeted therapy ( Xue et al . , 2017 ) . Sixth , the single-cell data show that a significant proportion , but not all , of the heterogeneity in CNAs found differentially between two spatially resolved biopsy samples is also captured at the single-cell level in one of the biopsies . Importantly , some variation is captured only in the single-cell data and not in the bulk comparisons . This provides a rationale for performing both spatial sampling of tumors and deep analysis of biopsy material for precision medicine applications ( Yates et al . , 2015 ) . Seventh , we see somewhat of an inverse correlation between tumor size and CNA heterogeneity in ER+ tumors , many of which present as a single large clone . Thus , for such large ER+ tumors , extensive spatial multi-region sampling might be necessary to capture a true picture of a tumor’s genetic heterogeneity . Eight , the association of CNA heterogeneity with parameters such as polyploidy and ER- status provides an indirect association of heterogeneity with clinical outcome since both parameters have been shown to associate with a worse prognosis ( Carey et al . , 2006; Pinto et al . , 2017 ) . Ninth , we observe that some tumors exhibit a multitude of sub-clones with no single sub-clone being dominant ( largely observed in ER- tumors ) . This raises questions regarding the importance of clonal cooperation in the growth and evolution of breast tumors , an area for which experimental evidence has recently emerged ( Marusyk et al . , 2014 ) . All of the above mentioned findings provide novel observations upon which more focused investigations can be based ( ex: X-loss in ER- tumors and pseudo-diploid cells ) as well as a solid foundation for future , more expansive studies of CNA heterogeneity in breast , as well as other cancers . Importantly , in contrast to bulk sequencing where elegant studies have used deep sequencing information to enable phasing and inference of genetic heterogeneity and clonality ( Shah et al . , 2012; Nik-Zainal et al . , 2012 ) , our study highlights the power inherent in single-cell genomic investigations for the direct observation and quantification of genetic heterogeneity . Most of the novel observations revealed by our study would not have been possible were it not for the single-cell nature of the data . For example , the identification of T-cells with X-chromosome losses and their association with ER- disease , as well as the detection and quantification of pseudo-diploid cells would have been obscured by bulk genomic analysis . Similarly the observation that CNAs can be found at more than two distinct copy number states , which we show is associated with clinical outcome , and the identification of somatic amplicon mosaicism in a substantial proportion of breast cancer biopsies is the result of the single-cell resolution of the data and importantly , has been missed by many prior bulk genomic studies . Thus , the results provide a strong rationale for the continued investment in the development of single-cell genomic technologies ( which have lagged behind single-cell transcriptomic technologies ) and their utilization in the study of tumor genomes ( Baslan and Hicks , 2017 ) to complement bulk sequencing efforts . This will require developing approaches that help in: scaling single-cell genome library construction as well as reducing costs ( for example via the implementation of microfluidics ) ( Laks et al . , 2019; Li et al . , 2020 ) , pairing single cell genome with single cell transcriptome information ( Dey et al . , 2015; Macaulay et al . , 2015 ) , and working with formalin fixed paraffin embedded specimen ( Jin et al . , 2015; Martelotto et al . , 2017 ) , challenges which many groups have taken efforts to address . Ultimately , single-cell genomic investigations will play an important role in advancing our knowledge of breast cancer genetics and heterogeneity and in the process advance our knowledge of the genetics and biology of the disease and help in its clinical management .
Tissue samples were collected from 16 patients enrolled on two , neoadjuvant - phase II , open-label clinical trails conducted by the Brown University Oncology Group ( BrUOG ) : BrUOG 211B and BrUOG 211A . Two fresh , pre-treatment core biopsies were obtained per patient , placed into optimal cutting temperature ( OCT ) and frozen at −80Co for subsequent processing . Patient sample associated metadata ( ex: age , ER/PR/HER2 status , among other variables ) are provided in Figure 1—source data 1 . RNA-sequencing libraries we generated and sequenced as previously described ( Varadan et al . , 2016 ) . For sample subtyping , a nearest centroid classifier was implemented using log2 transformed gene expression data aligned on PAM50 list for all BrUOG samples ( n = 127 ) where RNA-seq data was available . Nuclei isolation from frozen cores was achieved by finely mincing frozen tissue , at room temperature , in 1 . 0 mL of NST-DAPI buffer as previously described ( Baslan et al . , 2012 ) . Prior to sorting , nuclei suspensions were filtered with a 5 mL Falcon round-bottom tube with a cell-trainer cap and kept on wet-ice for a minimum of 30 min . Single-cell sorting was performed using a FACS AriaIIU SORP instrument ( BD Biosciences ) with the ACDU option ( Automated Cell Deposition Unit ) . DAPI signal was detected by a 355 mM UV laser ( 450/50 band-pass filter ) . For tumors with bimodal ploidy distributions ( indicative of polyploidy ) , single-nuclei were sorted from both the diploid and polyploidy distributions for processing . For tumors with only a unimodal diploid distribution , diploid nuclei were sorted and processed . Single-nuclei were deposited into 96-well plates pre-loaded with 9 uL of cell lysis buffer per well as previously described ( Baslan et al . , 2012 ) . All sorted plates were processed on the same day for single-cell genome amplification . Single-nuclei were amplified using WGA4 ( Sigma-Aldrich ) as per manufacturers protocol . WGA DNA was sonicated to + / - 300 bps using the Covaris instrument ( duty cycle −10% , Intensity −4 , cycles/burst – 200 and time 80 s ) . WGA sonicated DNA ( processed in 96 well format , i . e . 96 nuclei per plate ) was end-repaired , A-tailed and ligated to 96 custom in-house developed adaptors ( Iossifov et al . , 2012 ) . Adaptor ligated products were purified using AMPure XP beads ( Beckman Coulter ) . All adaptor ligated single-cell barcoded libraries ( n = 96 ) were pooled , amplified and quantified for sequencing . All pools were sequenced on the HiSeq instrument using either SR76 or SR101 sequencing . Single-nuclei copy number information was inferred from sequencing data as previously described ( Baslan et al . , 2012; Baslan et al . , 2015 ) . In brief , sequence reads were mapped to human reference genome ( reference hg19 ) , sorted , PCR duplicated removed , and subsequently indexed . Uniquely mapped reads were counted in genomic bins using a previously developed algorithm; Varbin ( Navin et al . , 2011 ) . Read counts were then normalized and segmentation performed using circular binary segmentation ( CBS ) . To retrieve absolute ( integer ) copy number information , we utilized an algorithm that assigns ploidy and associated copy number states in single-nuclei data by utilizing a least-squares fitting algorithm used in determining a multiplier value that minimizes the variance from integer copy number states ( Baslan et al . , 2015 ) . All inferred integer copy number values of single-nuclei for all tumors ( at bin sizes of ~600 KB , dividing the genome in 5000 bins/5 k ) are provided in Figure 2—source data 1 . For analysis of fraction ( % ) of genome sub-clonal , we developed a pipeline where for each genomic bin we counted the number of observable copy number states found in all single-nuclei sequenced from a given patient sample . For a bin to be called sub-clonal , we required the observation of a particular copy number state in any genomic bin in at least 10% of sequenced single-nuclei . Summing all sub-clonal bins and dividing by the total number of bins is used to derive the metric: % genome sub-clonal . For this analysis , the genome was divided into five thousand bins ( i . e . bin resolution of ~600 kb ) . Derivation of breakpoint information was performed as described previously ( Alexander et al . , 2018 ) . In brief , bin positions at which changes in segmented copy number states occur were annotated across all sequenced single-nuclei . To account for the inherent uncertainty in breakpoint bin positioning in segmentation algorithms we utilized a window of 3 bins where a breakpoint in any of those bins was treated as equal ( i . e . same breakpoint ) . Histograms of breakpoint distributions were then constructed and any breakpoint found in less than 90% and more than 10% of all sequenced nuclei from a given tumor sample was deemed sub-clonal . All breakpoints found in over 90% of single-nuclei data were annotated as clonal pins/breakpoints . Bulk DNA was purified using phenol-chloroform extraction as previously described ( Baslan et al . , 2015 ) . Bulk DNA was sonicated , end-repaired , A-tailed , and ligated to custom in-house developed adaptors . Libraries were enriched , quantified , and pooled at equimolar concentration for sequencing on the HiSeq instrument . Bulk copy number sequence analysis was performed as described above with the exception that the multiplier , utilized in the least squares-fitting algorithm , was constrained by values derived from flow cytometry data ( i . e . ploidy ) to achieve absolute copy number values . Purified CD8+/CD4+ T-cells and CD19+/CD27+ memory B-cells purified cells were purchased from ALLCELLS ( California , USA ) . Cells were thawed on ice , pelleted , and re-suspended in NST-DAPI buffer for nuclei isolation . Single-nuclei ( n = 96 ) from both sets were sorted from the diploid peak , genome amplified , and processed for multiplex sequencing as described for the cancer biopsies above . The distance between TCR deletions in two nuclei is defined to be the number of bins offset between the left breakpoints plus the number of bins offset between the right breakpoints . In order to normalized for depth and noise , nuclei with at least 1 million mapped reads were each down-sampled to exactly 1 million mapped reads and re-segmented . All T-cell leukemia nuclei share identical breakpoint bin positions . An equal number of T-cell nuclei found in breast cancer tissue with over 1 million uniquely mapped reads included in this analysis were randomly sampled in proportional sets , 1 million times . None of these 1 million random samples had all nuclei sharing identical breakpoints indicating a p-value of <10−6 . FISH analysis was performed on frozen section using home-brew Probe . The probe mix consisted of BAC clones containing the full length target gene and labeled with Red , Green , or Orange-dUTP as indicated in in figure legends . DNA was labeled by nick translation using fluorochrome-conjugated dUTPs from Enzo Life Sciences Inc , supplied by Abbott Molecular Inc Tissue processing , hybridization , post-hybridization washing , and fluorescence detection were performed according to standard procedures established at the MSKCC Molecular Cytogenetics Core Facility . Slides were scanned using a Zeiss Axioplan 2i epifluorescence microscope equipped with CoolCube 1 CCD camera controlled by Isis 5 . 5 . 10 imaging software ( MetaSystems Group Inc , Waltham , MA ) . Prior to hybridization on tissue section , the probe-mix was hybridized on peripheral blood from normal healthy male to ensure locus specificity . Following hybridization , the tissue was scanned through 63X to assess signal pattern and representative regions imaged ( each image was a compressed/merged stack of 12 z-section images taken at 0 . 5 micron intervals under the Red , Green and Orange filter respectively ) . Analysis was performed on captured images . Bilateral breast MRI evaluation with a dedicated breast coil ( without compression ) was done on a 1 . 5 Tesla magnet . Images were collected for 6 min at 1 min intervals for following bolus IV gadolinium administration ( 0 . 1 mmol/kg ) . Images were Axial 3D SPGR fat-suppressed T1-weighted . A board certified radiologist interpreted the images . We used published copy number data from the METABRIC cohort ( Curtis et al . , 2012 ) and associated distant relapse free survival data to assess the association between higher dosages of 1q or 8q and patient outcomes . For 1992 cases ( from the discovery and validation sets in the initial study ) , segmented absolute copy number calls were derived using circular binary segmentation . Specifically , the copy number data were smoothed and analyzed with the R package DNAcopy using default parameters , followed by applying the MergeLevels algorithm to the segmented data . In order to remove the dependence between cellularity and the proportion of alterations , we employed different thresholds for calling alterations and high-levels events according to the cellularity of each sample , as previously reported . The clinical outcome analysis included only cases from the original cohort that possessed a 16q loss and either a 1q gain or an 8q gain ( 446 cases total ) . Copy number status of the three chromosome arms of interest ( 1q , 8q , and 16q ) was determined by calculating a weighted arithmetic mean of the segments comprising the chromosome arm . In order to allow for parallel analysis of diploid and polyploid tumors , the mean copy number for the 16q chromosome arm ( loss of 16q is cytogenetically linked to 1q/8q gain ) was used to normalize the degree of gain present in the 1q and 8q arms . For each case , the maximum of the normalized mean copy number from either 1q or 8q was used for downstream analysis . Cases were divided into those with evidence of high-level Gain on 1q or 8q ( 66 cases , normalized CN >1 . 5 ) and those with low-level Gain ( 380 cases ) . Kaplan-Meier analysis was performed to compare these groups within all breast cancer patients using time to distant relapse as the outcome . Since overexpression of the ER and HER2 receptors and lymph node status are strong independent predictors of clinical outcome , we also stratified our findings by ER/HER2 status and lymph node status . In a Cox proportional hazards models that accounts for other clinical covariates ( age , grade , size , LN status , ER , PR , and HER2 expression ) 1q/8q ratio was not independently significant . All statistical tests were performed using the statistical package R . For the X-loss in T-cells comparison in ER- and ER+ disease a chi-square test was used . For analysis of mutual exclusivity of CCNE1 and VEGFA amplification ( copy number higher than 5 ) in P22 , Fisher’s Exact Test was used with the alterative hypothesis that true odds ratio is less than 1 . For correlations of copy number heterogeneity with biological and clinical variables , Wilcoxon rank order tests were implemented . Spearman rank was used to study the association of tumor size with CNA heterogeneity in ER+ cases . p-values are provided for all statistical tests either in the figures , figure legends , or both . Normal and cancer single-nuclei WGA DNA ( n = 10 for each category ) was subjected to PCR amplification of tp53 exons followed by Sanger sequencing . In brief , 50 ng of WGA DNA for 10 nuclei per above category were amplified in a PCR reaction using 1X Amplitaq 360 master mix with 0 . 2 nM of forward and reverse primers . Resultant PCR products were sequenced using standard Sanger sequencing protocols . A Sanger trace from one representative WGA DNA ( one for each normal and cancer ) is illustrated in Figure 5B . Data generated for this study are available thought Short Read Archive ( SRA ) under BioProject accession number PRJNA555560 . All single-cell raw sequencing data were processed using code provided in detailed in Baslan et al . ( 2012 ) . The R Source code for the calculation of % of genome sub-clonal is included as Source code 1 . The R source code used for the derivation of clonal/sub-clonal pins , as described in Materials and methods section , is available on GitHub at https://github . com/jysonganan/SCclust/blob/master/R/selectpin . R . | Cells in the body remain healthy by tightly preventing and repairing random changes , or mutations , in their genetic material . In cancer cells , however , these mechanisms can break down . When these cells grow and multiply , they can then go on to accumulate many mutations . As a result , cancer cells in the same tumor can each contain a unique combination of genetic changes . This genetic heterogeneity has the potential to affect how cancer responds to treatment , and is increasingly becoming appreciated clinically . For example , if a drug only works against cancer cells carrying a specific mutation , any cells lacking this genetic change will keep growing and cause a relapse . However , it is still difficult to quantify and understand genetic heterogeneity in cancer . Copy number alterations ( or CNAs ) are a class of mutation where large and small sections of genetic material are gained or lost . This can result in cells that have an abnormal number of copies of the genes in these sections . Here , Baslan et al . set out to explore how CNAs might vary between individual cancer cells within the same tumor . To do so , thousands of individual cancer cells were isolated from human breast tumors , and a technique called single-cell genome sequencing used to screen the genetic information of each of them . These experiments confirmed that CNAs did differ – sometimes dramatically – between patients and among cells taken from the same tumor . For example , many of the cells carried extra copies of well-known cancer genes important for treatment , but the exact number of copies varied between cells . This heterogeneity existed for individual genes as well as larger stretches of DNA: this was the case , for instance , for an entire section of chromosome 8 , a region often affected in breast and other tumors . The work by Baslan et al . captures the sheer extent of genetic heterogeneity in cancer and in doing so , highlights the power of single-cell genome sequencing . In the future , a finer understanding of the genetic changes present at the level of an individual cancer cell may help clinicians to manage the disease more effectively . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"genetics",
"and",
"genomics",
"cancer",
"biology"
] | 2020 | Novel insights into breast cancer copy number genetic heterogeneity revealed by single-cell genome sequencing |
Mitotic spindle orientation is crucial for symmetric vs asymmetric cell division and depends on astral microtubules . Here , we show that distinct subpopulations of astral microtubules exist , which have differential functions in regulating spindle orientation and division symmetry . Specifically , in polarized stem cells of developing mouse neocortex , astral microtubules reaching the apical and basal cell cortex , but not those reaching the central cell cortex , are more abundant in symmetrically than asymmetrically dividing cells and reduce spindle orientation variability . This promotes symmetric divisions by maintaining an apico-basal cleavage plane . The greater abundance of apical/basal astrals depends on a higher concentration , at the basal cell cortex , of LGN , a known spindle-cell cortex linker . Furthermore , newly developed specific microtubule perturbations that selectively decrease apical/basal astrals recapitulate the symmetric-to-asymmetric division switch and suffice to increase neurogenesis in vivo . Thus , our study identifies a novel link between cell polarity , astral microtubules , and spindle orientation in morphogenesis .
The fundamental functions of the mitotic spindle include not only the faithful partition of the genome into both daughter cells , but also controlling whether cell fate determinants are distributed symmetrically or asymmetrically to those daughters ( Gonczy , 2008; Gillies and Cabernard , 2011 ) . Cell division symmetry is controlled by orienting the metaphase spindle along a specific plane . Cytokinesis then segregates cell components symmetrically or asymmetrically , depending on their distribution on either side of that plane . Pioneering work in fungi and nematodes has shown spindle orientation to involve mitotic astral microtubules . These astrals dynamically link the spindle poles with the cell cortex ( Pearson and Bloom , 2004; Siller and Doe , 2009 ) . In polarized epithelial cells , the orientation of the mitotic spindle with respect to the apico-basal axis determines the distribution of components located differentially along this axis ( Knoblich , 2008; Gillies and Cabernard , 2011 ) . A classic example is Drosophila neurogenesis , where neuroepithelial cells proliferate by dividing symmetrically , with a cleavage plane parallel to the apico-basal axis . Neuroblasts derived from them delaminate from the apical surface and divide in turn asymmetrically , to self-renew and produce neurogenic progenitors . The mitotic spindle in these asymmetric divisions is re-oriented by 90° , with the cleavage plane now perpendicular to the apico-basal axis . This leads to the asymmetric distribution of polarized fate-determinants to the daughter cells ( Southall et al . , 2008; Sousa-Nunes et al . , 2010 ) . This major spindle re-orientation in Drosophila requires interactions between cell cortical Gαi , a heterotrimeric G protein subunit , and Partner of Inscuteable ( Pins ) , which are in turn linked to the Par polarity complex ( Par3 , Par6 , aPKC ) by Inscuteable ( Knoblich , 2008; Brand and Livesey , 2011 ) . Spindle and cleavage plane orientation has also been implicated in the neurogenesis of vertebrates , including mammals ( reviewed in Lancaster and Knoblich , 2012; Shitamukai and Matsuzaki 2012; see also Das and Storey , 2012; Asami et al . , 2011; Delaunay et al . , 2014 ) . Mammalian neurogenesis , however , shows major differences to Drosophila with regard to spindle orientation in symmetric vs asymmetric divisions of polarized neural stem cells . In the developing neocortex , neuroepithelial cells progressively become radial glia , and both of these highly related subtypes of neural stem cells exhibit a characteristic polarized , apico-basal architecture and undergo apical mitosis , hence the collective term ‘apical progenitors’ ( APs ) ( Kriegstein and Götz , 2003; Götz and Huttner , 2005; Miller and Gauthier , 2007; Corbin et al . , 2008; Martynoga et al . , 2012 ) . Importantly , the switch of APs from symmetric proliferative to asymmetric neurogenic divisions occurs mostly without large and defined re-orientations of the spindle , but with only subtle deviations ( Huttner and Brand , 1997; Haydar et al . , 2003; Kosodo et al . , 2004; Konno et al . , 2008; Shitamukai et al . , 2011 ) . These can nonetheless tilt the division plane enough to no longer bisect , but rather bypass the small apical end-foot , leading to its asymmetric distribution ( Kosodo et al . , 2004 ) . Similarly , subtle spindle deviations can also influence the inheritance of the basal process ( Shitamukai et al . , 2011 , see also Kosodo et al . , 2008 ) . The proper regulation of these symmetry changes is thought to be an important determinant in neocortical neurogenesis ( Götz and Huttner , 2005; Kriegstein and Alvarez-Buylla , 2009; Miyata et al . , 2010 ) , and its perturbation can lead to neurodevelopmental and neurodegenerative disorders ( Feng and Walsh , 2004; Fish et al . , 2006; Yingling et al . , 2008; Gauthier-Fisher et al . , 2009; Godin et al . , 2010; Lizarraga et al . , 2010 ) . However , it is unknown how the subtle spindle orientation changes that occur in the switch of mammalian APs from symmetric proliferative to asymmetric neurogenic divisions are generated and controlled . Proteins evolutionarily conserved between Drosophila and mammals , such as Inscuteable and the Pins homologue LGN , have been shown to be involved in the regulation of AP cleavage plane orientation and AP daughter cell fate ( Konno et al . , 2008; Postiglione et al . , 2011 ) . Despite these findings , it remains to be elucidated how alterations in these or other spindle regulators actually lead to the spindle orientation changes that switch the cell division mode of APs ( Buchman and Tsai , 2007; Lancaster and Knoblich , 2012 ) . Here we show that the spindle orientation that controls symmetric vs asymmetric division of polarized neural stem cells in the developing neocortex is not determined by all astral microtubules , as previously assumed , but by a distinct subpopulation of these microtubules , the apical/basal astrals . Specifically , a decrease in apical/basal astrals causes the subtle spindle orientation deviations that lead to the switch of mammalian APs from symmetric proliferative to asymmetric neurogenic divisions . Moreover , we find that a selective down-regulation of LGN at the basal cell cortex , rather than a general reduction of cell cortical LGN , explains the specific decrease in apical/basal astrals . Our study , which combines live and immunofluorescence microscopy of neocortical tissue from transgenic mice , provides a new model for the control of symmetric vs asymmetric cell division in cells that maintain spindle orientation perpendicular to the apico-basal axis , such as cortical neural stem cells .
During metaphase , the orientation of the mitotic spindle of mammalian APs dynamically deviates from the apical , ventricular surface of the developing cortical wall ( Chenn and McConnell , 1995; Haydar et al . , 2003; Roszko et al . , 2006 ) . Crucially however , whether such deviations are different in proliferating vs neurogenic progenitors is unknown . To ask if the amplitude of spindle deviations is involved in neurogenesis , a transgenic mouse was used where neurogenic divisions are identified by EGFP driven by the promoter of the pan-neurogenic marker Tis21 ( Iacopetti et al . , 1999; Haubensak et al . , 2004 ) and thereby distinguished from proliferative divisions . Specifically , embryonic day ( E ) 14 . 5 mouse dorsolateral telencephalon in organotypic slice culture stained with a vital DNA dye was analysed by dual-colour live 3D imaging . Consistent with previous studies , the metaphase spindles of most progenitors , as reported by the chromosome plates , continuously changed their angles with respect to the apical–basal axis until anaphase , and did so in apparently random fashion . Interestingly , these dynamic deviations increased in amplitude from Tis21::GFP– ( proliferating ) AP divisions to Tis21::GFP+ ( neurogenic ) AP divisions ( Figure 1A , B , D , E; Videos 1 , 2 ) . In neurogenic APs , the total amplitude of deviations was on average twice as high as in proliferating APs ( Figure 1G; AP proliferating 15 . 7 ± 1 . 4° , AP neurogenic 30 . 7 ± 4 . 3° ) . Importantly , the range of metaphase spindle orientations ( Figure 1D , E ) was consistent with the previously reported cleavage plane orientations of fixed ( Kosodo et al . , 2004 ) and live ( Konno et al . , 2008 ) mouse APs , which have been observed to be mostly perpendicular to the apical surface . 10 . 7554/eLife . 02875 . 003Figure 1 . Dynamic spindle orientation variability increases when progenitors become neurogenic . Live tissue imaging of spindle orientation as reported by chromosome plate orientation in organotypic slice culture of coronal sections from E14 . 5 Tis21::GFP mouse dorsolateral telencephalon . Observations focused on metaphase , when all chromosomes congressed to the equatorial plane and the dynamics observed were likely more related to spindle orientation , rather than the prometaphase establishment of a functional spindle and chromosome plate . 0 min is anaphase onset . ( A–C ) Apical progenitor ( AP; A , B ) and basal progenitor ( BP; C ) undergoing either proliferative ( A , Tis21::GFP– ) or neurogenic ( B , C; Tis21::GFP+ ) division . Merge: chromosomes in red , EGFP in green . Chromosome plate orientation was determined by measuring angular deviations from the apico-basal axis ( 0° , corresponding to vertical ) , which runs perpendicular to the apical ventricular surface ( 90° , horizontal white dotted lines in A , B ) . Time-lapse is 3 min . Vertical or oblique dashed lines indicate chromosome plate orientations in metaphase or anaphase onset; maximal deviation angles for each plate were set by the orientation at an early time-point ( yellow ) and a later time-point ( red ) , which are quantified in G . White dashed lines indicate intermediate orientations . Scale bar = 5 μm . ( D–F ) Quantification of chromosome plate orientations from metaphase ( meta . ) to anaphase ( ana . ) . To facilitate tracing , individual tracks are colour-coded according to the range in which most of the track remained ( blue , beyond 30° , cyan 30°–15° , green 15°–0° , yellow 0° to −15° , red −15° to −30° , dark red , beyond −30° ) . ( G ) Mean ± standard error of the mean ( SEM ) of the maximal amplitude of deviations for proliferative ( prolif . ) and neurogenic ( neurog . ) divisions . **p<0 . 001 , ***p<0 . 0001 , Kruskal–Wallis ANOVA ( K–W ) with Dunn's multiple comparison ( DMC ) test; n = 20 progenitors per category , from three independent litters and experiments . See also Videos 1–3 . DOI: http://dx . doi . org/10 . 7554/eLife . 02875 . 00310 . 7554/eLife . 02875 . 004Video 1 . Dynamic spindle orientation variability in a representative proliferating AP . Related to Figure 1A . Live tissue imaging of spindle orientation , as reported by the chromosome plate ( DNA ) orientation , in organotypic slice culture of dorsolateral telencephalon coronal sections , from an E14 . 5 Tis21::GFP mouse . Time-lapse 3 min . Total time elapsed 18 min . DOI: http://dx . doi . org/10 . 7554/eLife . 02875 . 00410 . 7554/eLife . 02875 . 005Video 2 . Dynamic spindle orientation variability in a representative neurogenic AP . Related to Figure 1B . Live tissue imaging of spindle orientation , as reported by the chromosome plate ( DNA ) orientation , in organotypic slice culture of dorsolateral telencephalon coronal sections , from an E14 . 5 Tis21::GFP mouse . Time-lapse 3 min . Total time elapsed 18 min . DOI: http://dx . doi . org/10 . 7554/eLife . 02875 . 005 Spindle orientation variability was also measured in basal progenitors ( BPs ) , which originate from APs but are delaminated from the ventricle and divide basally . Mouse BPs also lack the canonical apical–basal polarity features of APs ( Attardo et al . , 2008 ) and mostly undergo terminal divisions that produce two neurons ( Haubensak et al . , 2004; Miyata et al . , 2004; Noctor et al . , 2004 ) . Compared to proliferating APs , neurogenic BPs showed deviations of the metaphase plate axis from the apical–basal axis of the developing cortical wall that were even greater than those of neurogenic APs ( Figure 1C , F; Video 3 ) . The total amplitude of deviations in BPs increased to 38 . 3 ± 5 . 6° ( Figure 1G ) . Importantly , the range of metaphase spindle orientations ( Figure 1F ) was consistent with the highly variable cleavage plane orientation of mouse BPs reported previously ( Haubensak et al . , 2004; Attardo et al . , 2008 ) . This leads to the conclusion that metaphase spindle orientation variability increases as progenitors become neurogenic . Interestingly , this variability is characteristic of both , neurogenic APs that divide asymmetrically with a cleavage plane largely parallel to the apical–basal tissue axis , and of neurogenic BPs which divide symmetrically with a near-random cleavage plane orientation . 10 . 7554/eLife . 02875 . 006Video 3 . Dynamic spindle orientation variability in a representative neurogenic BP . Related to Figure 1C . Live tissue imaging of spindle orientation , as reported by the chromosome plate ( DNA ) orientation , in organotypic slice culture of dorsolateral telencephalon coronal sections , from an E14 . 5 Tis21::GFP mouse . Time-lapse 3 min . Total time elapsed 18 min . DOI: http://dx . doi . org/10 . 7554/eLife . 02875 . 006 To investigate the causes of increased dynamic spindle deviations in neurogenic progenitors , the mitotic microtubules were analysed ( Figure 2A–F ) . Changes in spindle dynamics could be caused by different numbers of astral microtubules interacting with the cell cortex . Interestingly , the number of astrals reaching the cell cortex was highest in proliferating APs ( Figure 2G; 21 . 65 ± 1 . 07 ) followed by neurogenic APs ( 16 . 70 ± 0 . 98 ) , and lowest in neurogenic BPs ( 11 . 36 ± 0 . 49 ) . 10 . 7554/eLife . 02875 . 007Figure 2 . Fewer astral microtubules when progenitors become neurogenic . ( A ) Cartoons of the mitotic spindle in APs as appearing upon ‘Front’ and ‘Side’ viewing of the same chromosome plate , rotating around its apico-basal axis . Apical- and basal-reaching astrals ( apical/basal astrals ) were defined as those extending beyond the main chromosomal area , delimited by the apical-most ( pink dashed line ) and basal-most ( blue dashed line ) peri/centromeric heterochromatin foci ( brightest points in DNA , see B–F ) , and reaching either the apical or basal region of the cell cortex , which are delimited by these same pink and blue dashed lines , respectively . In the ‘Side’ view , yellow dashed lines delimit the peri/centromeric heterochromatin region laterally , beyond which the astrals defined as central-reaching ( central astrals ) extend . ( B–F ) α-tubulin immunofluorescence ( maximum intensity projections of two 0 . 75 μm optical sections ) of E14 . 5 Tis21::GFP mouse dorsolateral telencephalon showing mitotic microtubules in proliferating APs ( Tis21::GFP– ) vs neurogenic APs ( Tis21::GFP+ ) vs neurogenic BPs ( Tis21::GFP+ ) . Dashed boxes show regions including the basal and apical cell cortex that are shown in the second and third row , respectively , at higher magnification and brightness . DNA staining ( DAPI ) and Tis21::GFP fluorescence are single 0 . 75 μm optical sections . Merge: DNA in blue , microtubules in yellow . Scale bars = 5 μm . ( G–J ) Mean numbers per cell of astrals reaching the cell periphery: all astrals ( G ) , apical/basal astrals ( H ) , central astrals ( I ) , and apical or basal astrals ( J ) . ( G , H , J ) In BPs , astrals are considered as apically , basally , and centrally oriented ( see ‘Materials and methods’ ) . n = 40 progenitors per category and I n = 20 progenitors per category , all from 8 independent litters and experiments . *p<0 . 05 , **p<0 . 001 , ***p<0 . 0001; K–W with DMC post test . Error bars are SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02875 . 007 The next question was whether this decrease was spatially homogeneous around the cell . Considering the highly polarized architecture of APs ( Götz and Huttner , 2005; Mora-Bermúdez et al . , 2013 ) , the progenitor soma was divided into three regions , to count the astrals reaching the cell cortex in each ( Figure 2A and ‘Materials and methods’ ) : the central region , defined by the presence of the centromeres and containing most of the spindle—including some astrals—and an apical and a basal region , both containing only astrals . Surprisingly , the decrease in astrals was specific to the apical and basal regions . The number of astrals reaching the apical or basal region of the cell cortex ( apical/basal astrals ) was lower in neurogenic than in proliferating APs ( Figure 2H; neurogenic 7 . 2 ± 0 . 36 , proliferating 11 . 8 ± 0 . 43 ) . By contrast , the number of astrals reaching the central region of the cell cortex ( central astrals ) did not change significantly ( Figure 2I; neurogenic APs 8 . 8 ± 0 . 7 , proliferating APs 9 . 1 ± 0 . 6 ) . To compare APs and BPs , the apical , basal , and central populations of astrals from APs were compared to the apically , basally , and centrally oriented populations of astrals in BPs ( see ‘Materials and methods’ ) . In contrast to APs , BPs showed a general decrease , with fewer astrals than APs for both populations ( Figure 2H , I , apically/basally oriented 5 . 5 ± 0 . 28 , centrally oriented: 6 . 1 ± 0 . 46 ) . Together with our data on spindle orientation variability ( Figure 1 ) , this suggests that a higher abundance of apical/basal astrals increases the stability of spindle orientation . The numbers of basal and apical astrals were then compared . Surprisingly , in proliferating APs , basal astrals were more abundant than apical ones ( Figure 2J , 6 . 7 ± 0 . 31 vs 5 . 1 ± 0 . 25 ) . By contrast , in neurogenic APs , the numbers of both basal and apical astrals were lower than in proliferating APs , but were not significantly different from each other ( basal 3 . 7 ± 0 . 2 , apical 3 . 4 ± 0 . 2 ) . This shows that the number of basal astrals is initially higher in proliferating APs , but their decrease is such that neurogenic APs have a homogeneously low number of apical and basal astrals . This trend continued in BPs , where the number of apically and basally oriented astrals further decreased homogeneously ( basally oriented 2 . 7 ± 0 . 2 , apically oriented 2 . 6 ± 0 . 2 ) . In mammals , the proportion of neurogenic divisions of APs increases throughout embryonic neurogenesis , which in the mouse goes from around E10 to E17 . Therefore , the total number of apical/basal astrals could decrease as neurogenesis progressed , from E11 . 5 to E14 . 5 to E16 . 5 ( early , mid , and late neurogenesis , respectively ) . Indeed , the number of astrals in APs decreased 2 . 5-fold from early to late neurogenesis , from 15 . 2 ± 0 . 4 at E11 . 5 to 9 . 4 ± 0 . 5 at E14 . 5 , and further to 6 . 2 ± 0 . 3 at E16 . 5 ( Figure 3 ) . This provides further evidence that dividing neurogenic APs have fewer apical/basal astrals than proliferating ones and suggests a role for these astrals in the switch to neurogenesis . 10 . 7554/eLife . 02875 . 008Figure 3 . Fewer apical/basal astral microtubules in progenitors as neurogenesis progresses . A , B ) α-Tubulin immunofluorescence ( maximum intensity projections of two 0 . 75 μm optical sections ) of E11 . 5 and E16 . 5 mouse dorsolateral telencephalon showing mitotic microtubules in APs . Dashed boxes in the first column show regions including the basal and apical cell cortex that are shown at higher magnification and brightness in the second column . DNA stainings ( DAPI ) are single 0 . 75 μm optical sections . Merge: DNA in blue , microtubules in yellow . Scale bars = 5 μm . ( C ) Mean number per cell of apical/basal astrals in APs at E11 . 5 , E14 . 5 ( Figure 2H , mean of both AP categories ) , and E16 . 5 . ***p<0 . 0001 , K–W with DMC post test; for E11 . 5 and E16 . 5 , n = 20 APs , from 4 independent litters and experiments; Error bars are SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02875 . 008 Previous studies have shown LGN to be a key player in spindle orientation . LGN is enriched in the cell cortex , especially around the central part of the cell , where it participates in spindle orientation in non-neural cells in vitro ( Du and Macara , 2004 ) . Also , upon LGN perturbation in neural progenitors in vivo , spindle orientation variability increased , changing progenitor identity ( Morin et al . , 2007; Konno et al . , 2008; Shitamukai et al . , 2011 ) . How LGN regulates spindle orientation in neurogenesis remains unknown . LGN could control spindle orientation in neural progenitors by regulating the number of astrals that anchor the spindle to the cell cortex . Therefore , the higher spindle orientation variability observed upon LGN perturbation could be caused by fewer apical/basal astrals . Indeed , in the E14 . 5 neocortex of LGN knock-out ( KO ) mice ( Konno et al . , 2008 ) ( Figure 4A–D ) , fewer apical/basal astrals were found in APs of homozygous compared to heterozygous mice ( Figure 4E; homozygous 8 . 4 ± 0 . 5 , heterozygous 11 . 2 ± 0 . 7 ) . By contrast , no significant reduction was seen in the central astrals ( Figure 4F; homozygous 9 . 1 ± 0 . 5 , heterozygous 8 . 9 ± 0 . 3 , similar to unperturbed APs; Figure 2I ) . No significant change was observed in BPs either ( Figure 4G; homozygous 5 . 9 ± 0 . 4 , heterozygous 6 . 4 ± 0 . 4 ) , where the number of apically/basally oriented astrals was also similar to unperturbed BPs ( Figure 2H ) . 10 . 7554/eLife . 02875 . 009Figure 4 . Fewer astral microtubules in LGN KO progenitors and upon dominant-negative LGN-C overexpression . ( A–D ) α-Tubulin immunofluorescence ( maximum intensity projections of two 0 . 75 μm optical sections ) of E14 . 5 heterozygous ( LGN +/− ) vs homozygous ( LGN −/− ) E14 . 5 LGN KO mouse dorsolateral telencephalon showing mitotic microtubules in APs and BPs . Dashed boxes show regions including the basal and apical cell cortex that are shown in the second and third row , respectively , at higher magnification and brightness . DNA staining ( DAPI ) and Tis21::GFP fluorescence are single 0 . 75 μm optical sections . Merge: DNA in blue , microtubules in yellow . Scale bars = 5 μm . ( E ) Mean number per AP of astrals reaching the apical and basal cell cortex ( apical/basal astrals ) . ( F ) Mean number per AP of astrals reaching the central cell periphery ( central astrals ) . ( G ) Mean number per BP of astrals that are apically and basally oriented . ***p=0 . 0007 , one-tail t-test , n ( 20 progenitors per category from 3 independent litters and experiments ) . Error bars are SEM . See also Figure 5 . ( H–J ) α-Tubulin immunofluorescence ( maximum intensity projections of two 0 . 75 μm optical sections ) of in utero electroporated E13 . 5 wt mouse dorsolateral telencephalon analysed 24 hr later ( E14 . 5 ) . APs in the electroporated regions were either not electroporated ( H ) , or electroporated with LGN-wt ( 6Myc-wt-LGN , I ) or with dominant-negative LGN ( 6Myc-LGN-C , J ) . Dashed boxes in the first column show regions including the basal and apical cell cortex that are shown at higher magnification and brightness in the second column . DNA staining ( DAPI ) and Myc immunofluorescence are single 0 . 75 μm optical sections . Merge: DNA in blue , microtubules in yellow . Scale bars = 5 μm . ( K ) Mean number per cell of astral microtubules reaching the apical or basal cell cortex ( apical + basal astrals ) of APs in the electroporated region that were either not electroporated , or transfected with 6Myc-LGN-wt or with 6Myc-LGN-C . ***p<0 . 0001 , one-way ANOVA with TMC test; n = 20 progenitors per category , from 3 independent litters and experiments . Error bars are SEM . Scale bars = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02875 . 009 To corroborate and complement these results , in utero electroporation was used to overexpress the dominant-negative C-terminal region of LGN ( LGN-C ) in the dorsolateral telencephalon of wt mice . LGN-C has the GoLoco motifs that bind to Gαi at the cell cortex , and its overexpression saturates the Gαi anchoring sites for microtubules . However , it lacks the TPR motifs in the N-terminal region and cannot interact with microtubules via NuMA and dynein ( Morin et al . , 2007; Shitamukai et al . , 2011 ) . APs overexpressing 6Myc-LGN-wt or 6Myc-LGN-C showed staining mostly around the basolateral cell cortex ( Figure 4I , J ) , similar to previous LGN stainings ( Konno et al . , 2008 ) . Consistent with our LGN-KO mouse results , APs overexpressing LGN-C also showed a significant reduction in apical/basal astrals at E14 . 5 ( Figure 4J , K , 7 . 9 ± 0 . 5 ) , similar to LGN KO homozygous mouse APs ( Figure 4E ) . By contrast , APs in the electroporated area that were either non-electroporated or overexpressing LGN-wt had a similar number of apical/basal astrals ( Figure 4H , I , K , 11 . 0 ± 0 . 6 and 11 . 2 ± 0 . 4 ) as unperturbed ( Figure 2H ) or LGN-KO heterozygous ( Figure 4E ) APs . Together , these results show that LGN perturbations lead to a reduction specifically in apical/basal astrals , likely through reduced anchoring of astrals to the cell cortex . The higher spindle orientation stability in proliferating APs could depend on cortical LGN levels . To test this , the distribution of LGN was analysed in different progenitors . No major differences in overall levels of LGN immunoreactivity were observed , but the subcellular distribution of the signal was different between proliferating and neurogenic E14 . 5 APs ( Figure 5A–C ) . In proliferating APs , cortical LGN increased from the apical to the central region of the cell cortex and remained high in most of the central and basal regions ( Figure 5A , E ) . In neurogenic APs , cortical LGN levels were similar to proliferating APs at the apical and central regions , but markedly lower at the basal region ( Figure 5B , E ) . These data show that neurogenic APs lose LGN enrichment in the basal region of the cell cortex , which could lead to less anchoring of basal astrals . Consistent with this , the strongest decrease in astrals of APs was in the basal region of neurogenic progenitors ( Figure 2J ) . A more modest but significant decrease in astrals in the apical region was also observed , yet the levels of apical LGN remained low in all progenitors . This suggests that additional factors may contribute to spindle orientation stability at the apical side . 10 . 7554/eLife . 02875 . 010Figure 5 . Less cell cortical LGN in neurogenic progenitors . ( A–C ) Double immunofluorescence for LGN and GFP , with DNA staining ( DAPI ) , of E14 . 5 Tis21::GFP mouse dorsolateral telencephalon showing representative examples of a proliferating AP ( Tis21::GFP– ) vs neurogenic AP ( Tis21-GFP+ ) vs neurogenic BP ( Tis21-GFP+ ) . Merge: DNA in cyan , LGN in red . ( D ) Same mitotic AP as in ( I ) , illustrating how LGN cortical immunofluorescence intensity was measured , starting at 0 . 0 in the middle of the apical region and continuing clockwise along the entire cell cortex ( dashed red line ) until 1 . 0 . Apical , central , and basal regions are indicated ( regions 1 , left side; regions 2 , right side ) . ( E ) Mean ± SEM of LGN cortical immunofluorescence intensity along 100 equidistant points of the normalized cell perimeter length; the light grey boxes highlight the central regions; the red box highlights the basal region; n = 15 progenitors per category , from 4 independent litters and experiments . Scale bars = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02875 . 010 Furthermore , compared to APs , cortical LGN levels in BPs were much lower , distributed more homogeneously over the cell and with only minor cortical enrichments ( Figure 5C , E ) . This is consistent with BPs having the lowest overall numbers of apical/basal and central astrals ( Figure 2G–I ) , as well as the highest amplitudes of dynamic spindle deviations ( Figure 1F , G ) . Together , these results strongly support a regulation of astral microtubule abundance by an LGN-dependent anchoring that depends on the specific cortical enrichment of LGN . An increase in asymmetric AP divisions could lead to changes in daughter cell fate , such as more AP daughter cells becoming BPs , and eventually to an increase in neurogenesis . To detect changes in the populations of progenitors and neurons , E13 . 5 BRO culture was used with or without 30 pM nocodazole . Compared to the DMSO control , incubation with 30 pM nocodazole for 24 hr resulted in a decrease in the proportion of APs , as identified by being in the ventricular zone ( VZ ) and having a characteristic transcription factor expression pattern ( Pax6-positive , Tbr2-negative; Figure 10A–C , 62 . 5 ± 1 . 8% vs 53 . 8 ± 0 . 7% ) . Interestingly , a concomitant increase was observed in the proportion of Tbr2-positive newborn BPs in the VZ ( Figure 10A–C , 34 . 3 ± 2 . 1% vs 43 . 1 ± 2 . 4% ) . These results indicate that more AP daughter cells became BPs upon an induced increase in asymmetric divisions . This is consistent with results showing an increase in BPs upon perturbation of other factors that also increase AP spindle orientation variability ( Fish et al . , 2006; Konno et al . , 2008; Postiglione et al . , 2011; Shitamukai et al . , 2011 ) . 10 . 7554/eLife . 02875 . 029Figure 10 . Increase in basal progenitors , basal mitoses , and neurogenesis with minimal nocodazole . E13 . 5 forebrains from wt mice were incubated for 24 hr in BRO culture , either with solvent ( DMSO ) only ( Control ) or with 30 pM nocodazole , followed by immunofluorescence of coronal sections of the dorsolateral telencephalon , acquisition of 1 μm confocal sections ( A , B , D , E , G , H ) and quantification ( C , F , I ) . ( A , B ) Tbr2 ( top ) and Pax6 ( bottom ) double staining . VZ , ventricular zone; SVZ , subventricular zone . ( C ) Left: percentages of nuclei , identified by DAPI staining ( not shown ) , in the VZ that are either resident APs ( Pax6-positive and Tbr2-negative ) or newborn BPs ( Tbr2-positive ) . Control vs nocodazole: APs , *p=0 . 002; BPs , *p=0 . 004 . Right: percentages of nuclei in the SVZ that are positive for Tbr2 or Pax6 . ( D–E ) Staining for the mitosis marker PH3 ( top ) and DNA ( DAPI , bottom ) . ( F ) Number and location of mitotic cells per 100 μm of apical ( ventricular ) surface; All , sum of mitoses in VZ and SVZ . Control vs nocodazole: SVZ , *p=0 . 031 . ( G , H ) Staining for the neuronal marker Tbr1 ( top ) and DNA ( DAPI , bottom ) . Note that the Tbr1 immunoreactivity basal to the cortical plate has been reported before ( Englund et al . , 2005 ) ; it may also reflect tissue stretching during cryosectioning . ( I ) Number of neurons per 100 μm of pial surface . Control vs nocodazole: *p=0 . 015 . ( C , F , I ) One-tail t-tests , n = 3 independent litters and experiments; error bars are SEM . ( A , B , D , E , G , H ) Scale bars = 20 μm . See also Figure 10—figure supplement 1 . ( J ) Conceptual model of mitotic spindle orientation control by local astral microtubule abundance , in proliferating vs neurogenic progenitors . We propose that mitotic astral microtubules that reach the apical or basal cell cortex ( light green rods ) of neural stem and progenitor cells regulate the dynamics of spindle orientation variability . Like guy ropes do for a camping tent , these astrals help anchor the spindle to the cell cortex , most likely through interactions with factors that have specific cortical enrichments , such as LGN ( red; colour intensity indicates local abundance in the three progenitor types shown ) . In APs undergoing proliferative division ( Tis21::GFP– ) , these apical/basal astrals at relatively high abundance minimize deviations in spindle orientation , thereby maintaining the canonical cleavage orientation , perpendicular to the apical ( ventricular ) surface , and thus favouring symmetric proliferative divisions . In neurogenic ( Tis21::GFP+ ) APs , and even more so in BPs , a reduction in the number of these astrals decreases cortical anchoring and increases random deviations in spindle orientation . For APs , this favours asymmetric neurogenic divisions that can generate a BP or a neuron . By contrast , astrals that reach the central cell cortex ( medium green rods ) of neural progenitors may be more involved in the fundamental establishment of a centrally located and functional bipolar spindle that can congress , and then faithfully segregate , chromosomes . ( Dark green rods are kinetochore microtubules; for simplicity , other microtubule populations , such as interpolar microtubules , are omitted . ) DOI: http://dx . doi . org/10 . 7554/eLife . 02875 . 02910 . 7554/eLife . 02875 . 030Figure 10—figure supplement 1 . BRO culture with 30 pM nocodazole increases Tis21-positive VZ cells and basal mitoses , but not apoptosis . E13 . 5 forebrains from wt mice were incubated for 24 hr in BRO culture , either with solvent ( DMSO ) only ( control ) or with 30 pM nocodazole , followed by immunofluorescence of coronal sections of the dorsolateral telencephalon , acquisition of 0 . 9 μm confocal sections ( A , B , D , E , G , H ) and quantification ( C , F , I ) . ( A , B ) Tis21::GFP ( top ) and DNA ( DAPI , bottom ) double staining . ( C ) Percentages of nuclei ( by DAPI ) in the VZ that are Tis21::GFP-positive . Control vs nocodazole: *p=0 . 030 . ( D–E ) Staining for the mitosis marker phosphovimentin ( pVim , top ) and DNA ( bottom ) . ( F ) Number and location of mitotic cells per 100 μm of apical ( ventricular ) surface; All , sum of mitoses in VZ and SVZ . Control vs nocodazole: SVZ , *p=0 . 026; n = 3 brains from two independent litters and experiments . ( G , H ) Immunofluorescence for cleaved caspase 3 ( top ) and DAPI staining ( bottom ) . ( I ) Pooled number of caspase 3-positive cells per 10 , 000 μm2 of dorsolateral cortex tissue; n = 3 brains from three independent litters and experiments . Error bars are SEM . Scale bars = 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02875 . 030 To complement these data , we analysed the expression of the neurogenic marker Tis21::GFP in VZ progenitors ( Figure 10—figure supplement 1A–C ) . Compared to controls , 30 pM nocodazole increased the proportion of Tis21::GFP-positive cells in the VZ ( 58 . 8 ± 5 . 1% vs 68 . 2 ± 2 . 4% ) . These cells may well comprise not only BP daughters but also AP daughters originating from Tis21::GFP-negative AP divisions that had become increasingly asymmetric . We conclude that minimal nocodazole-induced asymmetric AP divisions result in increased Tis21::GFP expression in daughter cells , consistent with these daughters being neurogenic progenitors . Of note , the increase in newborn BPs in the VZ did not yet result in a detectable increase in interphase Tbr2-positive or Pax6-positive BPs in the subventricular zone ( SVZ ) after 24 hr ( Tbr2 , control: 80 . 1 ± 2 . 0% , nocodazole: 82 . 3 ± 5 . 9%; Pax6 , control: 31 . 9 ± 8 . 2% , nocodazole: 28 . 9 ± 12 . 7% ) , despite the higher number of mitoses in the SVZ ( see below ) . This presumably reflects known differences in cell cycle length between progenitor subtypes ( see ‘Discussion’ ) . A prediction from the increase in newborn BPs after minimal nocodazole is that the number of basal mitoses would increase . Confirming this , more mitoses were found in the SVZ with 30 pM nocodazole than in controls ( phosphorylated histone 3 ( PH3 ) immunoreactivity; Figure 10D–F , 1 . 96 ± 0 . 16 vs 1 . 32 ± 0 . 029 ) . This was corroborated with a second , independent mitosis marker ( phosphovimentin ( pVim ) ; Figure 10—figure supplement 1D–F , 2 . 08 ± 0 . 25 vs 1 . 28 ± 0 . 05 ) . Importantly , the total number of mitoses did not change significantly with minimal nocodazole when analysed by either PH3 ( Figure 10F , control: 7 . 98 ± 0 . 57 , nocodazole: 8 . 22 ± 0 . 58 ) or by pVim ( Figure 10—figure supplement 1F , control: 7 . 80 ± 0 . 52 , nocodazole: 8 . 09 ± 0 . 42 ) . In addition , 30 pM nocodazole did not result in more apoptosis ( Figure 10—figure supplement 1G–I and ‘Material and methods’ ) . Also , all mitotic phases proceeded normally and at the same rate with 30 pM nocodazole as they did in controls ( Figure 9A–D ) , and cellular morphology at mitosis was unperturbed ( Figure 6C , D , Figure 9F , G ) . These data show that normal mitotic progression and the cell cycle were unaffected by 30 pM nocodazole . Importantly , the converse manipulation of microtubules by using 25 pM taxol did not affect mitotic progression ( Figure 8A–F ) or morphology ( Figure 6F , G , I , J ) either . The increase in neurogenic progenitors , notably newborn Tbr2-positive BPs , predicted an increase in neuron generation . Counting the number of cells positive for the neuronal marker Tbr1 confirmed this . More neurons were found in the presence of 30 pM nocodazole than in the presence of only DMSO ( Figure 10G–I , 381 . 2 ± 11 . 4 vs 327 . 2 ± 12 . 0 ) . Together , these results show that an acute reduction in the number of apical/basal astrals that leads to an increase in asymmetric AP divisions can increase BPs and neurogenesis .
Our study reports a novel concept regarding the composition and role of astral microtubule populations in the symmetric vs asymmetric division of polarized stem and progenitor cells . Specifically , we use the established model of mouse neocortex APs to show that the transition between symmetric and asymmetric neural stem cell division in cells with minor spindle variations is mediated by a reduction in the number of astral microtubules that reach the cell cortex . Strikingly , this reduction in astrals is not homogeneous , but happens only in the astrals that reach the basal and apical cell cortex . This shows that a distinct subpopulation of astrals , the apical/basal astrals , governs spindle orientation . Our selective reduction in the number of apical/basal astrals , by using minimal nocodazole , mimics the reduction observed in vivo and increases the proportion of asymmetric AP divisions and the pool of BPs , ultimately leading to more neurogenesis . To the best of our knowledge , this reduction is also the earliest change reported thus far in the cytoarchitecture of APs that switch from proliferation to neurogenesis . A homogeneous decrease in astrals , including those reaching the central cell cortex , happens only in BPs , which , accordingly , have the highest spindle orientation variability ( Figure 10J ) . The spatial , temporal , and cell-type specific regulation of astrals is therefore key to maintain the population of proliferating APs for normal brain development . A new model is therefore proposed to explain how spindle and cleavage plane variability increases during mammalian neurogenesis ( Chenn and McConnell , 1995; Haydar et al . , 2003; Kosodo et al . , 2004; Konno et al . , 2008 ) . The subtle mammalian increase in variability is fundamentally different to the large and defined 90° re-orientation in Drosophila neuroblasts ( Knoblich , 2008; Siller and Doe , 2009 ) . We show that , as mammalian APs switch to neurogenesis , the spindle becomes less restricted to an orientation precisely perpendicular to the apico-basal axis ( Sauer , 1935; Smart , 1972a , 1972b; Huttner and Brand , 1997 ) as a result of fewer apical/basal astrals anchoring the spindle to the cell cortex ( Figure 10J ) . The orientation of spindles that become less anchored is then more sensitive to intra- and extracellular forces that can induce tilts , including those resulting from the dense packing of dynamic cells in the developing cortical wall . The increase in BPs upon increasing spindle orientation variability by minimal nocodazole treatment is consistent with previous studies showing more BPs in mice where spindles were tilted by perturbing the LGN binding partner Inscuteable ( Postiglione et al . , 2011 ) , or LGN itself ( Konno et al . , 2008; Shitamukai et al . , 2011 ) . The increase in mitotic BPs observed upon LGN perturbation was associated with an increasing appearance of Pax6-positive and of Tbr2-positive cells in the SVZ ( Konno et al . , 2008; Shitamukai et al . , 2011 ) . Inscuteable overexpression increased the number of mitotic , Tbr2-positive cells in the SVZ ( Postiglione et al . , 2011 ) . The reduction in apical/basal astrals shown here generated more newborn Tbr2-positive BPs in the VZ and mitotic BPs in the SVZ . Thus , these three distinct approaches that tilt the spindle to increasingly deviate from an orientation perpendicular to the apico-basal AP axis yield similar ( albeit not identical ) phenotypes , that is , an increase in delaminated progenitors that exhibit BP features and undergo basal mitosis . The more specific differences could reflect the distinct types of manipulation , i . e . , the genetic ablation of LGN vs Inscuteable overexpression vs pharmacological perturbation of apico/basal astrals . Taken together , our data decisively support the concept that a small change in cleavage plane orientation of APs can be sufficient to cause a fate change in their progeny ( Fish et al . , 2006; Postiglione et al . , 2011 ) . Importantly , the delaminated progenitors increasingly observed upon minimal nocodazole treatment showed Tbr2 immunoreactivity and underwent basal mitosis , hence showing a number of features that are typical of BPs rather than APs . Nevertheless , these newborn Tis21::GFP-positive BPs , generated by just tilting the spindle , may still temporarily exhibit certain AP features , such as an AP-like cell cycle length , which is shorter than that of mature BPs ( Arai et al . , 2011 ) . These newborn BPs may therefore undergo mitosis soon after arrival in the SVZ , which would explain its increased mitoses without a significant increase in Tbr2-positive interphase cells . A shorter than normal cell cycle of the BPs generated by minimal nocodazole would also be consistent with the newborn neurons increasingly observed after 24 hr of this treatment being derived not only directly from the asymmetrically dividing APs , but also indirectly from these BPs . Spindle orientation studies have faced the challenge of establishing whether the phenotypes observed could be solely attributed to changes in orientation , or to other effects on the spindle and other cellular components . This study aimed at specifically perturbing spindle orientation while avoiding other effects . For this , a minimal concentration of the highly specific microtubule polymerization inhibitor nocodazole was identified that was several orders of magnitude lower than previously used in spindle studies . This minimal nocodazole decreased only the number of apical/basal astrals , those detected to change when APs switch from proliferation to neurogenesis , but did not affect other types of microtubules , notably the central astrals . This suggests that these astrals are more sensitive to depolymerisation , perhaps due to being highly dynamic . Moreover , this decrease was titrated to the physiological level observed in unperturbed neurogenic APs , and other effects , notably on progenitor and spindle features , mitotic morphology , and progression , were not detected . Consistent with this , a similarly low concentration of the specific microtubule stabilizer taxol showed the opposite effect , increasing the number of apical/basal astrals . Therefore , treatment with minimal concentrations of compounds acting specifically on microtubules constitutes an alternative approach to genetic manipulations , allowing direct probing of spindle orientation per se and of the symmetry of cell division in tissue . Furthermore , LGN is here shown to regulate the spatial and temporal abundance of astrals that reach the cell cortex , and thereby the transition between proliferation and neurogenesis . Mammalian LGN has been shown to participate in the anchoring of the spindle to the cell cortex in cultured MDCK cells , as part of a complex with NuMA and dynein , by mediating the attachment of astrals to cell cortical Gαi subunits ( Du and Macara , 2004 ) . Consistent with this role , previous studies showed that LGN is important for normal cleavage plane and symmetric AP divisions in the mouse ( Konno et al . , 2008; Shitamukai et al . , 2011 ) and the chick ( Morin et al . , 2007; Peyre et al . , 2011 ) , but how LGN achieved this was unknown . Using an LGN KO mouse ( Konno et al . , 2008 ) , LGN is now shown to be necessary for APs to have a normal number of apical/basal astrals . Importantly , lack of LGN did not affect the number of central astrals , corroborating that loss of apical/basal astrals is sufficient to increase spindle deviations and asymmetric divisions . These data were complemented by overexpressing a dominant-negative version of LGN that interacts with cell cortical Gαi , but cannot interact with NuMA and dynein to anchor microtubules ( Morin et al . , 2007; Shitamukai et al . , 2011 ) . This also reduced apical/basal astral abundance . Conversely , when wt LGN was overexpressed in the presence of the microtubule-stabilizing compound taxol , neurogenic APs could be switched back to a mitosis behaviour characteristic of proliferating APs . Together , these data show that LGN is required for maintaining symmetric AP divisions through cortical anchoring of apical/basal astrals . Our data also show how different cortical localizations of LGN could regulate the abundance of astrals that reach different regions of the cell cortex . When APs switched to neurogenesis , the enrichment of LGN at the basal cell cortex was largely reduced , and it was further reduced all around the cell cortex in BPs . This closely corresponds to the abundance pattern of astrals and also to the levels of spindle orientation variability in these different progenitors . A spatially selective loss of cortical LGN could therefore reduce cortical anchor points and lead to fewer astrals in specific regions of the cell cortex ( Figure 10J ) . The present study showing the crucial role of a specific subset of LGN-dependent astrals in the control of spindle and cleavage plane orientation also builds upon previous studies on another microtubule-interacting protein , the lissencephaly protein Lis1 ( Sapir et al . , 1999; Faulkner et al . , 2000; Tsai et al . , 2005; Yingling et al . , 2008; Bi et al . , 2009 ) . In the neural tube of mice without Lis1 , cleavage planes of APs were more tilted ( Yingling et al . , 2008 ) . Also , cultured embryonic fibroblasts lacking Lis1 showed general defects in interphase and mitotic microtubules , including astrals ( Yingling et al . , 2008 ) . The present findings suggest that the altered cleavage plane orientation in Lis1-deficient APs could be due , at least in part , to a loss of apical/basal astrals . If apical/basal astral microtubules are key to controlling spindle and cleavage orientation , what then is the role of the central astrals unaffected by the proliferation-to-neurogenesis transition of progenitors ( Figure 10J ) and that remained constant upon minimal nocodazole treatment or functional LGN ablation ? An intriguing scenario is that this baseline number of remaining astrals is required for the primary establishment and structural stability of a spindle . If so , our data suggest a way how cells differentially regulate core aspects of spindle function vs those that are cell type-specific , notably spindle orientation .
All observations were performed in the dorsolateral telencephalon of mouse embryos , at a medial position along the rostro-caudal axis . All mice , wt or mutant , were C57BL/6 . Animals were kept pathogen-free at the Biomedical Services Facility ( BMS ) of the MPI-CBG . All experiments were performed according to the German Animal Welfare Legislation . Embryonic day ( E ) 0 . 5 was set at noon on the day of vaginal plug identification . Neurogenic progenitors were identified in knock-in mice heterozygous for EGFP expressed under the control of the Tis21 promoter ( Haubensak et al . , 2004 ) . To examine astral microtubules ( astrals ) in mitotic cells without LGN , LGN knock-out ( KO ) mice ( Konno et al . , 2008 ) were used . A brain culture method was developed that is termed BRO ( Brain Rotation Organotypic ) culture , which includes aspects of previous culture systems ( Nomura and Osumi , 2004; Attardo et al . , 2008; Schenk et al . , 2009 ) and is as follows: Brains of E13 . 5–E14 . 5 embryos were dissected and placed in Tyrode solution ( Sigma T2145 , Germany ) at 37°C , where most of the meninges were surgically removed . Whole forebrains were dissected and transferred to 20 ml glass flasks with 1 ml brain culture medium ( see ‘Live Tissue Imaging’ ) . Flasks were connected to a whole-embryo culture incubator ( RKI Ikemoto , Japan ) and maintained at 37°C with an atmosphere of 60% O2 , 5% CO2 , 35% N2 , and at 30 rpm , allowing tissue rotation and nutrition without mechanical damage . After 1 hr of equilibration , pre-warmed brain culture medium containing 10 nM nocodazole or taxol and 1% DMSO , added from a 1 μm stock solution in DMSO , or brain culture medium containing 1% DMSO only ( control ) , was added to the culture in amounts appropriate to obtain the indicated final concentration . To count astrals , E14 . 5 forebrains were incubated for 2 . 5 hr . To analyse differences in the populations of progenitors , mitotic cells , and neurons , E13 . 5 forebrains were incubated for 24 hr , during which the tissue remained intact without a significant cell death increase ( Figure 10—figure supplement 1 ) . Brains were then fixed in 4% PFA in phosphate buffer for 2 hr at room temperature followed by 4°C overnight . In utero electroporation ( Takahashi et al . , 2002; Fish et al . , 2006 ) was used to introduce the dominant-negative LGN-C or LGN-wt ( Shitamukai et al . , 2011 ) into APs . E13 . 5 pregnant mice were fully anesthetized with isofluorane and received a subcutaneous analgesic injection . The uterus was then exposed by surgically opening the peritoneal cavity , maintaining moisture and 37°C temperature . Embryos were injected intraventricularly with a 0 . 1% solution of fast green ( Sigma ) in sterile PBS , containing 0 . 2–0 . 5 μg/μl of LGN plasmid , plus mCherry or GAP43-EGFP , driven by the CAGGS constitutive promoter ( Niwa et al . , 1991 ) to locate electroporated regions and cells . Electroporations were with six 50 ms pulses of 30 mV at 1 s intervals . Detection of LGN and the fluorescent reporter occupied two of the four available fluorescence channels , with DAPI and tubulin occupying the other two , so these experiments could only be performed with wt and not with Tis21::GFP mice . For consistency , this scheme was followed for the live imaging of organotypic slice cultures ( see ‘Live tissue imaging’ ) . Images were acquired with a Zeiss LSM 510 Duoscan laser-scanning confocal microscope or a Zeiss LSM 780 NLO 2-photon laser-scanning microscope with an automated tuning pulsed Ti:S Chameleon Vision II laser ( Coherent ) , using 63× Plan-Apochromat 1 . 4 N . A . oil or 40× C-Apochromat 1 . 2 N . A . W objectives ( Carl Zeiss , Germany ) . Multiposition and multidimensional imaging was controlled as described ( Rabut and Ellenberg , 2004 ) . Mouse: mAb anti α-tubulin , ( Sigma ) , mAb anti pan-cadherin ( Sigma ) , mAb anti cleaved caspase 3 ( Sigma ) , mAb anti phosphovimentin ( Abcam , San Francisco , CA ) , mAb anti-Myc-A488 or -A555 ( Millipore , Germany ) ; rabbit: pAb anti Pax6 ( Covance ) ; pAb anti Tbr2 ( Abcam ) , pAb anti Tbr1 ( Millipore ) , pAb anti γ-tubulin ( Sigma ) , pAb anti Arl13b ( Proteintech , UK ) , pAb anti phosphorylated histone H3-S10 ( PH3 , Millipore ) , pAb anti LGN ( Konno et al . , 2008 ) ; goat: antibody anti EGFP ( MPI-CBG ) . Secondary antibodies: Alexa Fluor 488 donkey anti goat IgG , Alexa Fluor 555 donkey anti rabbit IgG , Alexa Fluor 647 donkey anti rabbit IgG , and Alexa Fluor 647 donkey anti mouse IgG , all from Invitrogen . Staining: DAPI ( Sigma ) . Images were analysed and prepared with ImageJ ( http://imagej . nih . gov/ij/ ) and AIM software ( Carl Zeiss ) . The brightness and contrast of images were recorded and adjusted linearly . To analyse spindle and cleavage orientation dynamics by live tissue imaging , a measurement was made for each time point of the angle between the main axis of the chromosome plates , typically along the apico-basal axis of the tissue , and the plane of the local apical ventricular surface . At anaphase onset , this chromosome plate angle is also practically identical to the angle of cleavage of the cell soma , and therefore directly reports on the symmetry or asymmetry of the cell division with respect to cellular features . The angles are given as deviations from full orthogonality with the local apical surface , as seen from a coronal perspective ( Figure 1A–F , Figure 10J ) . The accuracy of angular measurements was corroborated along the 3D stack of optical sections . To quantitate the range of variability that spindle orientation can have in different progenitor types , the mean of the total angular amplitude of deviations from metaphase to anaphase onset was calculated ( Figure 1G ) . To quantitate variations in the cell cortical distribution of LGN , a contour of per-pixel fluorescence intensity values was obtained along the cell cortex , in immunofluorescence images of a single central optical section per 3D stack ( Figure 4K ) . The start was at the centre of the apical region and continued clockwise around the cell soma . The cortical intensity values were obtained by subtracting the diffuse cytoplasmic signal from the intensities measured along the one pixel-wide contour . To compare equivalent values between different cells , 100 equidistant cortical intensity values along the normalized cell soma contours were calculated by linear interpolation . The area , fluorescence intensity , and standard deviation measurements of the spindle in Figure 7 ( see also Figure 7—source data 1 ) were performed as described ( Mora-Bermudez et al . , 2007 ) . To distinguish and count populations of astrals according to their spatial distribution , each set of confocal sections of a metaphase cell soma was divided into three regions: an apical , a central , and a basal region . The apical region was defined from the apical surface lining the ventricle to the parallel plane just before the first centromeric region on the metaphase plate ( below pink dashed line in Figure 2A ) . The central region that followed was defined to contain all centromeric regions , as revealed by the punctate DAPI staining of the centromeric and pericentromeric heterochromatin foci ( between pink and blue dashed lines in Figure 2A ) . The basal region extended after the last centromeric region until the basal end of the cell soma ( above blue dashed line in Figure 2A ) . Comprehensive stacks of high-resolution serial confocal sections ( see Microscopy ) were examined from each progenitor in metaphase . All detectable astrals microtubules that emanated from the 2 centrosomes and reached the cell periphery were counted in each of the above-defined regions . These presumably are single microtubules , although bundles of microtubules cannot be excluded . The apico-basal plane of the chromosome plates in APs can tilt with respect to the apical–basal axis of the tissue . However , it can also rotate around that axis , as seen from an ‘en face’ perspective ( Adams , 1996; Haydar et al . , 2003 ) . Therefore , in coronal sections , it is possible to observe spindles in a ‘front’ view , where all or most of the length of the spindle is visible in one section , or in a ‘side’ view , showing only one side of the spindle ( Figure 2A ) . Spindles were therefore classified accordingly in these two , cell-biologically equivalent , perspective categories , a ‘Front’ view , when the main axis of the spindle was mostly parallel to the sectioning plane and fully visible within 1–3 confocal sections , and a ‘Side’ view , when the main axis of the spindle extended mostly along the sectioning axis ( Figure 2A ) . Astrals in the apical and basal regions could be reliably counted in cells with any such perspective . The central region of the spindle showed a dense α-tubulin staining , however , which in a ‘Front’ view made the distinction of many of its astrals very challenging . Therefore , astrals in the central region were counted in cells with a ‘Side’ view ( Figure 2A ) . This may have led to an underestimation of the total number of central astrals , especially short ones oriented along the optical axis , albeit equally in all progenitor types . In most mouse BPs , canonical apical cell polarity markers are lost ( Attardo et al . , 2008 ) and cell cortex regions can no longer be reliably defined as ‘apical’ or ‘basal’ . BPs were included in our analysis for comparative purposes , with the note that astrals that reach the region of the cell cortex facing mostly the ventricle are considered ‘apically oriented’ , those that reached the region of the cortex facing mostly the pia ‘basally oriented’ , and those that reach the region in between the previous regions ‘centrally oriented’ ( Figure 2H , I; Figure 4G ) . Datasets were tabulated and analysed using Excel ( Microsoft , Redmond , WA ) and GraphPad Prism ( La Jolla , CA ) . Statistical tests: for two groups of observations , the Student’s t-test for parametric analysis or the Mann–Whitney U-test for non-parametric analysis . For three or more groups , the one-way Analysis of Variance ( ANOVA ) with Tukey's Multiple Comparison ( TMC ) Test for parametric analysis , or the Kruskal–Wallis ANOVA ( K–W ) with Dunn's Multiple Comparison ( DMC ) Test for non-parametric analysis . Results were interpreted as statistically significant when p<0 . 05 . Nocodazole is a highly specific inhibitor of microtubule polymerization , and 30 pM is 2–3 orders of magnitude below the concentrations typically considered ‘low’ , where the earliest signs of microtubule and cellular perturbations have been reported to appear ( 5–10 nM ) ( De Brabander et al . , 1976; Jordan et al . , 1992; Thery et al . , 2005; Harder et al . , 2009; Kiyomitsu and Cheeseman , 2012 ) . Therefore , 30 pM was very unlikely to have general or nonspecific effects in the neural progenitors , spindles , or microtubules . Indeed , as described in Results , 30 pM nocodazole did not result in overt perturbations in mitotic morphology , mitotic progression , and general cell and tissue features . Furthermore , no significant differences were found between spindles in untreated , control ( DMSO-only ) , and 30 pM nocodazole-treated forebrains regarding spindle size , the relative amount of microtubules or the distribution and structure of the fluorescence signal of the microtubules ( area , mean fluorescence intensity , and mean standard deviation ( SD ) of the per-pixel fluorescence intensity values , respectively; Figure 7B–D; Figure 7—source data 1 ) . A further hypothesis was that , if the spindle , located in the central plane of the cell soma , was somehow affected , the soma diameter at the spindle plane of the cells may be changed in a tissue with high cellular density . This diameter represented also the length of the entire spindle , including some central astrals ( Figure 7E ) . No significant differences were found , however , between APs in untreated , control and 30 pM nocodazole-treated forebrains ( Figure 7F; Figure 7—source data 1 ) . In line with our in-organ observations , cell culture experiments have shown that a much higher nocodazole concentration ( 5 nM ) did not disrupt the assembly of a functional bipolar spindle ( Thery et al . , 2005 ) . The survival of cells in tissue under 30 pM nocodazole was also analysed , by counting the proportion of cleaved caspase 3-immunoreactive cells in cortical wall tissue . The occurrence of positive cells was overall very low , and also similar between control and 30 pM nocodazole ( Control 1 . 49 vs Noc 1 . 59 ) , showing that this concentration did not result in an apoptosis increase ( Figure 10—figure supplement 1G–I ) . Together , these observations strongly suggest that 30 pM nocodazole does not have general or nonspecific effects in the tissue examined . | A stem cell can divide in two ways . Either it can split symmetrically into two identical daughter stem cells , or it can split asymmetrically into a stem cell and a specialist cell . The structure that forms inside the dividing cell to separate pairs of chromosomes—called the mitotic spindle—also partitions the molecules that determine what kind of cell each daughter cell will become . The mitotic spindle is made up of protein microtubules . Astral microtubules connect the spindle to a structure found at the inner face of the cell membrane called the cell cortex . This helps the spindle to orient itself correctly and control the plane of cell division . This is particularly important in cells that are different at their top and bottom , like polarized neural stem cells . To divide symmetrically , these cells need to split vertically from top to bottom . Then , to divide asymmetrically they tilt the cell division plane off-vertical . Classical studies on neuroblasts from the fruit fly Drosophila have shown that a big , 90° reorientation , from vertical to horizontal underlies this change . However , in the primary stem cells of the mammalian brain , subtle off-vertical tilting suffices for asymmetric divisions to occur . This tilting must be finely regulated: if not , neurodevelopmental disorders , such as microcephaly and lissencephaly , may arise . Mora-Bermúdez et al . investigated how mammalian cortical stem cells control such subtle spindle orientation changes by taking images of developing brain tissue from genetically modified mice . These show that not all astral microtubules affect whether the spindle reorients , as was previously thought . Instead , only those connecting the spindle to the cell cortex at the top and bottom of the cell—the apical/basal astrals—are involved . A decrease in the number of apical/basal astrals enables the spindle to undergo small reorientations . Mora-Bermúdez et al . therefore propose a model in which the spindle becomes less strongly anchored when the number of apical/basal astrals is reduced . This makes the spindle easier to tilt , allowing neural stem cells to undergo asymmetric divisions to produce neurons . The decrease in the number of apical/basal astrals appears to be caused by a reduction in the amount of a molecule that is known to help link the microtubules to the cell cortex . This reduction occurs only in the cortex at the top of the cell . Mora-Bermúdez et al . were also able to manipulate this process by adding very low doses of a microtubule inhibitor called nocodazole , which reduced the number of only the apical/basal astrals , increasing the ability of the spindle to reorient . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2014 | Specific polar subpopulations of astral microtubules control spindle orientation and symmetric neural stem cell division |
Transcription activator-like effectors ( TALEs ) are sequence-specific DNA binding proteins found in a range of plant pathogenic bacteria , where they play important roles in host-pathogen interactions . However , it has been unclear how TALEs , after they have been injected into the host cells , activate transcription of host genes required for infection success . Here , we show that the basal transcription factor IIA gamma subunit TFIIAγ5 from rice is a key component for infection by the TALE-carrying bacterium Xanthomonas oryzae pv . oryzae , the causal agent for bacterial blight . Direct interaction of several TALEs with TFIIAγ5 is required for activation of disease susceptibility genes . Conversely , reduced expression of the TFIIAγ5 host gene limits the induction of susceptibility genes and thus decreases bacterial blight symptoms . Suppression or mutation of TFIIAγ5 can also reduce bacterial streak , another devastating disease of rice caused by TALE-carrying X . oryzae pv . oryzicola . These results have important implications for formulating a widely applicable strategy with which to improve resistance of plants to TALE-carrying pathogens .
Transcription activator-like effectors ( TALEs ) are important effectors of plant pathogenic bacteria of the genus Xanthomonas ( Boch et al . , 2009 ) . The bacteria inject TALEs via their Type III secretion system ( T3SS ) into host cells , where they translocate to the nucleus and bind host gene promoters in a sequence-specific manner . The DNA binding domain consists of variable repeats that together account for a predictable DNA recognition code ( Boch et al . , 2009; Moscou and Bogdanove , 2009 ) . This property has been exploited for programmable DNA binding , and has allowed targeted genome editing by combining TALE DNA binding domains with nucleases ( TALENs ) ( Maggio and Gonçalves , 2015 ) . TALE-like proteins are not restricted to the genus Xanthomonas , and have also been found in the plant pathogen Ralstonia solanacearum ( de Lange et al . , 2014 ) , and in the endosymbiont Burkholderia rhizoxinica ( de Lange et al . , 2014; Juillerat et al . , 2014 ) . TALE-like proteins thus may play not only antagonistic roles in host-microbe interactions . Xanthomonas infect many important crops including barley , bean , brassica , cassava , citrus , cotton , mango , pepper , rice , rye , tomato , triticale , and wheat ( Schornack et al . , 2013; Boch et al . , 2014 ) . In rice , Xanthomonas oryzae pv . oryzae ( Xoo ) causes bacterial blight and X . oryzae pv . oryzicola ( Xoc ) causes bacterial streak , both of which are highly devastating diseases . The recessive resistance gene xa5 is widely used to improve rice resistance to Xoo ( Kottapalli et al . , 2007 ) . xa5 is a natural allele of the gene for the transcription factor IIA gamma subunit 5 ( TFIIAγ5 ) , changing a valine to a glutamine ( TFIIAγ5V39E thereafter ) ( Iyer and McCouch , 2004; Sugio et al . , 2007 ) . TFIIA is a basal transcription factor of eukaryotes and it is essential for polymerase II–dependent transcription ( Høiby et al . , 2007 ) . It consists of two subunits , the large subunit TFIIAαβ and the small subunit TFIIAγ ( Li et al . , 1999 ) . Rice TFIIAγ5 has been suggested to be a cofactor that directly enables TALEs to induce host gene expression ( Iyer-Pascuzzi and McCouch , 2007 ) , either as a helper of TALE function ( Boch et al . , 2014 ) , or as a TALE-targeted host gene ( Gu et al . , 2009 ) . The latter scenario is supported by the finding that the TALE PthXo7 directly activates expression of another TFIIAγ encoding gene , TFIIAγ1 ( Sugio et al . , 2007 ) . In this paper , we reveal that TALEs from two Xanthomonas pathogens , Xoo and Xoc directly interact with TFIIAγ5 to activate host susceptibility genes , and that RNAi-mediated suppression or mutation of TFIIAγ5 confers disease resistance . Our results suggest that modifying host TFIIAγ genes by mutation or suppression may provide a widely applicable strategy to improve plant resistance to TALE-carrying pathogens .
To assess whether host TFIIAγ is required for TALE-regulated transcriptional activation of rice susceptibility genes , we first assessed how pair of rice near-isogenic lines , IR24 carrying TFIIAγ5 and IRBB5 carrying mutant TFIIAγ5V39E in the IR24 background , responded to 15 different TALE-carrying Xoo strains ( Yang and White , 2004 ) . IRBB5 always showed fewer disease symptoms than IR24 ( Figure 1—figure supplement 1A ) . Xoo infection did not induce RNA expression of TFIIAγ5 in IR24 or TFIIAγ5V39E in IRBB5 ( Figure 1A ) , which correlates with the absence of predicted DNA binding motifs for known TALEs in the TFIIAγ5 promoter . In contrast , expression of known disease susceptibility genes Os8N3 , TFIIAγ1 , OsTFX1 , and Os11N3 , each of which is targeted by a different TALE ( Römer et al . , 2010; Sugio et al . , 2007; Yang et al . , 2006 ) , was always lower in IRBB5 ( p<0 . 01 ) , although not necessarily completely abolished ( Figure 1A ) . Together , these results point to TFIIAγ5 being a host co-factor for TALE-dependent induction of susceptibility genes . 10 . 7554/eLife . 19605 . 003Figure 1 . Effects of TFIIAγ5 on the expression of disease susceptibility genes Os8N3 , Os11N3 , TFIIAγ1 , or OsTFX1 , after Xoo infection . Plants were inoculated with Xoo strain PXO99 ( harbouring TALEs PthXo1 , PthXo7 , and PthXo6 ) , PXO86 ( harbouring TALE PthXo3 ) or PXO61 ( harbouring TALE AvrXa7 ) at the booting ( panicle development ) stage . It is known that PthXo1 , PthXo7 , and PthXo6 induce Os8N3 , TFIIAγ1 , and OsTFX1 , respectively , and PthXo3 and AvrXa7 all induce Os11N3 . Each bar represents mean ( three replicates ) ± standard deviation . ( A ) Mutation of TFIIAγ5 ( rice line IRBB5 ) . b , significant difference between IR24 and IRBB5 at p<0 . 01 . ( B ) TFIIAγ5-RNAi lines . b , significant difference between wild-type ( WT ) and transgenic plants at p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 00310 . 7554/eLife . 19605 . 004Figure 1—figure supplement 1 . Effects of TFIIAγ5 on rice resistance to Xoo strains known to carry TALEs . Rice plants at the booting ( panicle development ) stage were inoculated with Xoo . ( A ) The near-isogenic lines IR24 and IRBB5 showed different responses to the infection of Xoo . IRBB5 in IR24 background carries a natural mutated TFIIAγ5 , TFIIAγ5V39E . Each bar represents mean ( total 17 to 29 leaves from 5 plants ) ± standard deviation . b , significant difference between IR24 and IRBB5 at p<0 . 01 . ( B ) The enhanced resistance of TFIIAγ5-RNAi plants to strain PXO99 was associated with reduced transcription of TFIIAγ5 but not TFIIAγ1 . WT , wild-type Zhonghua 11 . Each bar represents mean ( 3 replicates for gene expression and total 5 to 10 leaves from one plant for lesion length ) ± standard deviation . b , significant difference between wild-type ( WT ) and transgenic plants at p<0 . 01 . ( C ) The enhanced resistance of TFIIAγ5-RNAi plants co-segregated with reduced TFIIAγ5 transcription in T1 families . Each bar represents mean ( 3 replicates for gene expression and 5 to 10 leaves from one plant for lesion length ) ± standard deviation . b , significant difference between WT and transgenic plants at p<0 . 01 . ( D ) TFIIAγ5-RNAi plants showed enhanced resistance to all the Xoo strains . Each bar represents mean ( total 35 to 40 leaves from 5 plants ) ± standard deviation . b , significant difference between WT and transgenic plants at p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 00410 . 7554/eLife . 19605 . 005Figure 1—figure supplement 2 . Effect of TFIIAγ5 on Xa23-mediated resistance to Xoo strain PXO99 . Each bar represents mean ( total 40 to 45 leaves from 5 plants for lesion length and 3 replicates for gene expression ) ± standard deviation . ( A ) Xa23 Xa23- mediated resistance required the presence of TFIIAγ5/TFIIAγ5 or TFIIAγ5/TFIIAγ5V39E . Rice plants at the booting stage were inoculated with Xoo . b , significant difference between IR24 and other plants at p<0 . 01 . ( B ) PXO99 infection-induced Xa23 expression required the presence of TFIIAγ5/TFIIA5 or TFIIAγ5/TFIIA5V39E . b , significant difference between non-inoculated plants ( ck ) and PXO99-inoculated plants within each rice line at p<0 . 01 . 1d and 2d , 1 day or 2 days after inoculation of PXO99 . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 005 To determine directly the role of TFIIAγ5 in host gene expression , we suppressed its activity by RNA interference ( RNAi ) . Only the expression of TFIIAγ5 , but not of TFIIAγ1 was reduced in T0 transgenic plants , and the reduction in TFIIAγ5 expression correlated with enhanced resistance to Xoo PXO99 in T0 and T1 plants ( Figure 1—figure supplement 1B , C ) . TFIIAγ5-RNAi plants also had enhanced resistance to a diverse collection of 13 additional Xoo strains ( Figure 1—figure supplement 1D ) , and Xoo-induced expression of Os8N3 and OsTFX1 was reduced in TFIIAγ5-RNAi plants ( Figure 1B ) . Suppressing TFIIAγ5 did not obviously influence growth and development of the transgenic plants . TALE DNA-binding motifs have been detected in the promoters of some disease resistance genes , an apparent evolutionary response against TALE-carrying bacteria ( Gu et al . , 2005; Römer et al . , 2010; Wang et al . , 2015 ) . Xoo TALE AvrXa23 activates the Xa23 resistance gene , resulting in resistance to Xoo ( Wang et al . , 2015 ) . To investigate the role of TFIIAγ5 in Xa23 resistance , we crossed rice lines IRBB5 , with a xa23 susceptibility and a TFIIAγ5V39E resistance allele , and CBB23 , with a Xa23 resistance and a TFIIAγ5 susceptibility allele . F2 plants of genotypes Xa23/Xa23 or Xa23/xa23 were completely resistant to PXO99 in the TFIIAγ5/TFIIAγ5 or TFIIAγ5/TFIIAγ5V39E background , but showed the reduced resistance in the TFIIAγ5V39E/TFIIAγ5V39E background ( Figure 1—figure supplement 2A ) . Consistent with the resistance phenotype , Xa23 expression was rapidly induced by PXO99 in Xa23/Xa23 or Xa23/xa23 plants when they also were of genotype TFIIAγ5/TFIIAγ5 or TFIIAγ5/TFIIAγ5V39E ( Figure 1—figure supplement 2B ) . Xa23 induction was completely lost in TFIIAγ5V39E/TFIIAγ5V39E plants . These results suggest that TFIIAγ5 plays dual roles in Xoo−rice interactions: it is required by TALE-containing Xoo to cause disease , but at the same time it can help to protect against disease in the presence of certain resistance genes that have TALE-binding motifs in their promoters . Xoo TALEs typically have an amino-terminal translocation signal ( TS ) , a central repeat region ( RR ) , a transcription factor binding ( TFB ) region , a nuclear localization signal ( NLS ) , and a carboxyl-terminal transcription activation domain ( AD ) ( Figure 2—figure supplement 1 , Figure 2—source data 1 ) ( Yang et al . , 2006; Schreiber et al . , 2015 ) . When fused to the GAL4 DNA binding domain , Xoo TALE PthXo1 on its own could activate reporter gene expression in yeast . This was observed whenever the TS or AD were present , but not with the RR , TFB or NLS ( Figure 2—figure supplement 1A ) . This is similar to what has been reported for Xoo TALE AvrXa10 and X . euvesicatoria TALE AvrBs3 ( Szurek et al . , 2001; Zhu et al . , 1998 ) . We hypothesized that TALEs use TFIIAγ5 directly as a co-factor to induce transcription of susceptibility genes . In yeast two-hybrid ( Y2H ) assays , truncated PthXo1 , RR-TFB-NLS , lacking transcriptional activation ability , interacted strongly with TFIIAγ5 , somewhat less so with the mutant TFIIAγ5V39E , and not at all with the large subunit of TFIIA , TFIIAαβ ( Figure 2—figure supplement 1B , C ) . The interaction with TFIIAγ5 required the TFB ( Figure 2—figure supplement 1D ) . To determine whether this observation of interaction of a TALE TFB with TFIIAγ5 , was general , we isolated the TFB encoding DNA fragments from 14 of the 18 other TALE genes in Xoo pv . PXO99 ( Salzberg et al . , 2008 ) . These TFBs are 134 to 145 amino acids long , with the Tal7b and Tal8b TFBs predicted to be identical ( Figure 2—source data 2 ) . All 14 TFB fragments interacted with TFIIAγ5 , but only two ( Tal7a and Tal8a ) with TFIIAγ5V39E ( Figure 2—figure supplement 1E , F ) . Notably , different from PthXo1 , Tal7a and Tal8a interacted equally well with TFIIAγ5 and TFIIAγ5V39E . The TFBs of Tal7a , Tal8a , and PthXo1 differed by 1 to 20 residues from the other 12 TFBs that interacted only with TFIIAγ5 ( Figure 2—source data 2 ) . We confirmed the interactions observed in the Y2H system by transient expression of Myc- and FLAG-labeled proteins in Nicotiana benthamiana , followed by co-immunoprecipitation ( CoIP ) ( Figure 2A ) . We found the interaction of full-length PthXo1 with TFIIAγ5 or TFIIAγ5V39E , of the TFBs of PthXo1 , PthXo6 , PthXo7 , Tal3a , Tal7a , and Tal9e with TFIIAγ5 , and of the TFBs of PthXo1 and Tal7a with TFIIAγ5V39E ( Figure 2B ) . 10 . 7554/eLife . 19605 . 006Figure 2 . Detection of interactions between rice TFIIAγs and TALEs from Xoo in planta by co-immunoprecipitation . The protein–protein interaction assays were performed in N . benthamiana leaf cells . Proteins before ( input ) and after immunoprecipitation ( IP ) were detected with anti-myc and anti-FLAG antibodies . ( A ) Interaction of the myc-labelled full-length PthXo1 with FLAG-labelled TFIIAγ5 , TFIIAγ5V39E , and mutated rice TFIIAγ1 ( TFIIAγ1S47E ) . ( B ) Interactions of the myc-labelled TFB regions of six TALEs with FLAG-labelled rice TFIIAγs . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 00610 . 7554/eLife . 19605 . 007Figure 2—source data 1 . The defined domains/motifs and sequences of TALE PthXo1 from Xoo strain PXO99 . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 00710 . 7554/eLife . 19605 . 008Figure 2—source data 2 . Amino acid sequence alignment of the TFB regions of TALEs from Xanthomonas oryzae strains , composed of either 134 or 145 amino acids . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 00810 . 7554/eLife . 19605 . 009Figure 2—figure supplement 1 . Interactions between Xoo TALEs and plant TFIIAγs in yeast cells . The interactions were assessed by growth of yeast cells on synthetic defined premixes ( SD ) medium lacking ( - ) leucine ( L ) , tryptophan ( W ) , histidine ( H ) , and adenine ( A ) . V , empty vector as control . TS , translocation signal; RR , repeat region; TFB , transcription factor binding region; NLS , nuclear localization signal; AD , transcription activation domain . ( A ) Examination of transactivation activity of different domains and motifs of TALE PthXo1 . The full-length and truncated PthXo1 were separately fused to the DNA-binding domain of GAL4 , which is a yeast transcription factor , and transformed into yeast . ( B ) Truncated PthXo1 ( RR-TFB-NLS ) interacted with rice TFIIAγ5 and TFIIAγ5V39E ( mutated TFIIAγ5 ) and Arabidopsis AtTFIIAγ analysed by yeast two-hybrid ( Y2H ) assay . ( C ) Truncated PthXo1 did not interact with rice basal transcription factor TFIIAαβ analysed by Y2H assay . ( D ) The TFB region of PthXo1 was required for the interaction with TFIIAγ5 analysed by Y2H assay . ( E ) The TFB regions of TALEs differentially interacted with rice TFIIAγs . The TFB regions of all the 15 TALEs from Xoo strain PXO99 interacted with rice TFIIAγ5 , and the TFB regions of some of the 15 TALEs interacted with the mutated TFIIAγs from rice ( TFIIAγ5V39E and TFIIAγ1S47E ) analysed by Y2H assay . The Tal6b is a putative non-functional TALE . ( F ) The TFB regions of TALEs interacted with rice TFIIAγs with different strength based on the analysis of LacZ activity . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 009 To learn whether the TFB region of TALEs is directly responsible for TALE-induced host gene expression , we generated recombinant Xoo strains in which the TFB of PthXo1 was replaced with different TFBs , chosen based on their differential interaction with TFIIAγ5 and TFIIAγ5V39E: PthXo1 ( TFIIAγ5 > TFIIAγ5V39E ) , Tal7a ( TFIIAγ5 = TFIIAγ5V39E ) , and PthXo7 and AvrXa23 ( TFIIAγ5 but not TFIIAγ5V39E ) ( Figure 2—figure supplement 1B , E , F ) . In addition , we generated a TFB deletion in PthXo1 . The constructs were introduced into Xoo pv . T7174 and KACC10331 , both of which lack PthXo1 . Rice strain IR24 , which carries TFIIAγ5 , is strongly susceptible to T7174 and moderately susceptible to KACC10331 , while IRBB5 , which carries TFIIAγ5V39E , is resistant to both Xoo strains ( Figure 3A , Figure 1—figure supplement 1A , and Figure 3—figure supplement 1A ) . 10 . 7554/eLife . 19605 . 010Figure 3 . Effects of the TFB region of TALE PthXo1 on the virulence of Xoo strains and on the expression of rice susceptibility gene in rice−Xoo interaction . Each bar represents mean ( total 30 to 35 leaves from five plants for lesion length; three replicates for gene expression and bacterial growth rate ) ± standard deviation . ( A ) Virulence of wild-type strain T7174 and recombinant strains carrying PthXo1 and its derivatives in IR24 and IRBB5 . b , significant difference between T7174 and recombinant strains in each rice line at p<0 . 01 . ( B ) Growth of different Xoo strains in rice leaves . b , significant difference between 0 day ( 30 min after infection ) and 12 days after infection of each strain at p<0 . 01 . ( C ) Expression of susceptibility gene Os8N3 after infection of different strains . b , significant difference between non-inoculated ( ck ) and inoculated ( at 48 hr after infection of a strain ) plants in each rice line at p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 01010 . 7554/eLife . 19605 . 011Figure 3—source data 1 . Effects of leucine residues of PthXo1 TFB region on TALE-mediated infection . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 01110 . 7554/eLife . 19605 . 012Figure 3—figure supplement 1 . Effects of the TFB region of TALE PthXo1 on the virulence of Xoo strains and on the expression of rice susceptibility gene in rice−Xoo interaction . Each bar represents mean ( total 30 to 35 leaves from five plants for lesion length; three replicates for gene expression and bacterial growth rate ) ± standard deviation . ( A ) Virulence of wild-type strain KACC10331 and recombinant strains carrying PthXo1 and its derivatives in near-isogenic lines IR24 and IRBB5 . b , significant difference between T7174 and recombinant strains in each rice line at p<0 . 01 . ( B ) Growth of different strains in rice leaves . b , significant difference between 0 day ( 30 minutes after infection ) and 12 days after infection of each strain at p<0 . 01 . ( C ) Expression of susceptibility gene Os8N3 after infection of different strains . b , significant difference between non-inoculated ( ck ) and inoculated ( at 48 hr after infection of a strain ) plants in each rice line at p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 012 As expected , the deletion control PthXo1-ΔTFB did not change the success of infection by T7174 or KACC10331 ( Figure 3A , and Figure 3—figure supplement 1A ) , while the TFBs from PthXo1 and Tal7a , which can interact with both TFIIAγ5 and TFIIAγ5V39E , enhanced infection success in both hosts , IR24 ( TFIIAγ5 ) and IRBB5 ( TFIIAγ5V39E ) . Consistent with Tal7a , but not PthXo1 , interacting equally well with TFIIAγ5 and TFIIAγ5V39E , only the Tal7a TFB caused similar sized lesions in both IR24 and IRBB5 ( Figure 3A , and Figure 3—figure supplement 1A ) . The TFBs of PthXo7 and AvrXa23 , which can interact only with TFIIAγ5 , accordingly increased disease symptoms only on IR24 . Lesion size in these experiments was correlated with titer of bacterial growth ( Figure 3B , and Figure 3—figure supplement 1B ) and expression of Os8N3 ( Figure 3C , and Figure 3—figure supplement 1C ) . The TFB region of the TALEs harbours an imperfect leucine zipper motif , a known protein-protein interaction domain ( Schreiber et al . , 2015 ) . We generated three TFB mutants of PthXo1 , substituting leucine with alanine residues ( Figure 3—source data 1A ) . The muations did , however , not compromise interaction with TFIIAγ5 , nor infection success ( Figure 3—source data 1B , C ) . The other TFIIAγ encoded in the rice genome , TFIIAγ1 , shares 86% sequence identity with TFIIAγ5 ( Figure 4—source data 1 ) , but has a very restricted expression profile , with highest expression in endosperm and stamens ( Figure 4—figure supplement 1 ) . TFIIAγ1 did not interact with full-length or truncated PthXo1 or other Xoo TALE TFBs in yeast or in planta ( Figure 2 , and Figure 2—figure supplement 1B , E , F ) . We produced eight TFIIAγ1 derivatives with TFIIAγ5 substitutions at six positions ( Figure 4—figure supplement 2 ) . Of 15 TFBs tested , those of PthXo1 , Tal3a , Tal7a , Tal8a , Tal9d and Tal9e could interact in yeast with TFIIAγ1S47E , but not with other TFIIAγ1 mutants ( Figure 4—figure supplement 2 , and Figure 2—figure supplement 1E ) . Four of these interactions could be confirmed in planta ( Figure 2B ) . We then generated TFIIAγ1-RNAi plants as well as transgenic plants expressing the TFIIAγ1S47E mutant from TFIIAγ1 regulatory sequence . Both types of plants were morphologically normal . Some T0 TFIIAγ1-RNAi plants showed enhanced resistance to Xoo pv . PXO99 ( Figure 4—figure supplement 3 ) . Increased resistance was associated with reduced TFIIAγ1 expression , whereas TFIIAγ5 expression was unaffected ( Figure 4—figure supplement 3 ) , which was confirmed in two T1 families ( Figure 4A ) . However , these plants did not show enhanced resistance to other 13 Xoo strains ( Figure 4B ) . This is in agreement with previous suggestions that the TFIIAγ1 promoter is a target of the TALE PthXo7 from PXO99 ( Boch et al . , 2009; Sugio et al . , 2007 ) . PthXo7-induced TFIIAγ1 expression is dependent on TFIIAγ5 ( Figure 1 ) . 10 . 7554/eLife . 19605 . 013Figure 4 . Effects of TFIIAγ1 on response to infections by different Xoo strains . Plants were inoculated with Xoo at the booting stage . Each bar represents mean ( three replicates for gene expression and total 35 to 40 leaves from five plants for lesion length ) ± standard deviation . ( A ) Suppressing TFIIAγ1 enhanced rice resistance to strain PXO99 . b , significant difference between wild-type ( WT ) Zhonghua 11 and transgenic plants at p<0 . 01 . ( B ) Suppressing TFIIAγ1 did not change rice response to other strains . b , significant difference between WT and transgenic plants at p<0 . 01 . ( C ) PTFIIAγ1:TFIIAγ1S47E-transgenic plants showed susceptibility to PXO99 and PXO341 compared to IRBB5 . b , significant difference between IRBB5 and transgenic plants at p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 01310 . 7554/eLife . 19605 . 014Figure 4—source data 1 . Amino acid sequence alignment of basal transcription factor IIA gamma subunit ( TFIIAγ ) from different species . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 01410 . 7554/eLife . 19605 . 015Figure 4—source data 2 . Single nucleotide polymorphisms in the TFIIAγ1 coding region of 1419 rice accessions from RiceVarMap ( http://ricevarmap . ncpgr . cn ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 01510 . 7554/eLife . 19605 . 016Figure 4—source data 3 . Single nucleotide polymorphisms in the TFIIAγ5 coding region of 1419 rice accessions from RiceVarMap ( http://ricevarmap . ncpgr . cn ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 01610 . 7554/eLife . 19605 . 017Figure 4—figure supplement 1 . Expression profiles of TFIIAγ5 and TFIIAγ1 in 28 tissues covering the entire life cycle of rice varieties Minghui 63 and Zhenshan 97 . Data were obtained from a microarray database ( http://www . ncbi . nlm . nih . gov ) . E3 , endosperm at 21 days after pollination; E2 , endosperm at 14 days after pollination; E1 , endosperm at 7 days after heading; Spi , spikelet at 3 days after pollination; Sta , stamen at 1 day before flowering; H , hull at 1 day before flowering; P5 , panicle at heading stage; P4 , panicle at 4- to 5-cm young panicle stage; P3 , panicle at pollen–mother cell formation stage; P2 , panicle at pistil and stamen primordium differentiation stage; P1 , panicle at secondary branch primordium stage; Ste2 , stem at heading stage; Ste1 , stem at 5 days before heading; FL2 , flag leaf at 14 days after heading; FL1 , flag leaf at 5 days before heading; L2 , leaf at 4- to 5-cm young panicle stage; L1 , leaf at secondary branch primordium stage; She2 , sheath at 4- to 5-cm young panicle stage; She1 , sheath at secondary branch primordium stage; Sh , shoot ofseedling with two tillers; R , root of seedling with two tillers; L & R , leaf and root at three-leaf stage; Em & Ra , embryo and radicle at 3 days after germination; Pl1 , plumule at 48 hr after emergence under light; Pl2 , plumule at 48 hr after emergence under dark; Ra1 , radicle at 48 hr after emergence under light; Ra2 , radicle at 48 hr after emergence under dark; S , germinating seed at 72 hr of imbibitions . Expression levels ( log2 transformations of average signal values ) are color-coded: yellow and blue denote high and low expression , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 01710 . 7554/eLife . 19605 . 018Figure 4—figure supplement 2 . Interactions between the TFB region of TALE PthXo1 and mutated TFIIAγ1s in yeast cells . The interactions were assessed by growth of yeast cells on synthetic defined premixes ( SD ) medium lacking leucine ( L ) , tryptophan ( W ) , histidine ( H ) , and adenine ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 01810 . 7554/eLife . 19605 . 019Figure 4—figure supplement 3 . Effect of suppressing TFIIAγ1 on rice resistance to Xoo strain PXO99 . Plants were inoculated with PXO99 at the booting stage . RNA was isolated from the flag leaves . WT , wild-type Zhonghua 11 . Each bar represents mean ( 3 replicates for gene expression and 5 to10 leaves from one plant for lesion length ) ± standard deviation . b , significant difference between WT and transgenic plants at p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 01910 . 7554/eLife . 19605 . 020Figure 4—figure supplement 4 . Effect of mutation of TFIIAγ1 on the expression of disease susceptibility gene Os8N3 after Xoo infection . Plants were inoculated with Xoo PXO99 at the booting stage . Each bar represents mean ( 3 replicates ) ± standard deviation . b , significant difference between IRBB5 and transgenic plants at p<0 . 01 . ck , before Xoo inoculation . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 020 In the background of TFIIAγ5V39E , the TFIIAγ1S47E-transgenic plants showed increased susceptibility to Xoo pv . PXO99 and PXO341 ( Figure 4C ) . The increased susceptibility to PXO99 might be due to an interaction between TFIIAγ1S47E and PthXo1 ( Figure 2B ) to induce the susceptibility gene Os8N3 ( Figure 4—figure supplement 4 ) , while the susceptibility to PXO341 may be explained by another TALE ( see the TFBs tested in Figure 2—figure supplement 1E , F ) that can interact with TFIIAγ1S47E . We searched the RiceVarMap database of 1419 rice accessions ( http://ricevarmap . ncpgr . cn; Zhao et al . , 2015 ) for allelic variation at TFIIAγ1 and TFIIAγ5 . There were no non-synonymous single nucleotide polymorphisms ( SNPs ) in TFIIAγ1 ( Figure 4—source data 2 ) . Thirty-three rice accessions shared the same two non-synonymous SNPs diagnostic for the TFIIAγ5V39E allele ( Figure 4—source data 3 ) . Twenty-nine of these belong to the Aus group , which is mainly from South Asia , and the other four accessions belong to the Indica II group , mainly from Southeast Asia ( Xie et al . , 2015 ) ( Figure 4—source data 3 ) . The regional distribution of the TFIIAγ5V39E resistance allele likely reflects the high disease pressure in these regions . To learn whether TALEs of other pathogenic bacteria also exploit TFIIAγ5 to cause disease , we investigated the interaction of TFIIAγ5 with TALEs from Xoc , which causes bacterial streak . Xoc pv . RH3 has at least 11 TALE genes based on DNA blot analysis ( Figure 5—figure supplement 1 ) . All TFBs of RH3 TALEs ( GenBank accession numbers KU163014 to KU163031 ) interacted with TFIIAγ5 in yeast , and two were confirmed in planta ( Figure 5A , and Figure 5—figure supplement 2A ) . Seven randomly chosen TFBs did not interact with TFIIAγ5V39E or TFIIAγ1 , but three interacted with TFIIAγ1S47E in yeast , and at least one in planta ( Figure 5A , and Figure 5—figure supplement 2B ) . Consistent with these results , rice accession IRBB5 ( TFIIAγ5V39E ) was more resistant to infection by different Xoc strains than IR24 ( TFIIAγ5 ) ( Figure 5—figure supplement 2C ) . TFIIAγ5-RNAi plants also showed enhanced resistance to Xoc , whereas suppressing TFIIAγ1 had no effect on resistance to Xoc ( Figure 5B ) . 10 . 7554/eLife . 19605 . 021Figure 5 . Effect of TFIIAγ on rice-Xoc interaction . ( A ) Interactions of myc-labelled TFB regions of TALEs from Xoc RH3 and FLAG-labelled rice TFIIAγs in N . benthamiana leaf cells analysed by CoIP assays . Proteins before ( input ) and after immunoprecipitation ( IP ) were detected with anti-myc and anti-FLAG antibodies . ( B ) TFIIAγ5-RNAi but not TFIIAγ1-RNAi plants showed enhanced resistance to Xoc strains . Each bar represents mean ( total 30 to 35 leaves from five plants ) ± standard deviation . b , significant difference between wild-type and transgenic plants after infection of a strain at p<0 . 01 . ( C ) Mutation of TFIIAγ5 ( rice line IRBB5 ) reduced expression of disease susceptibility gene OsSULTR3;6 after infection . Each bar represents mean ( three replicates ) ± standard deviation . b , significant difference between IR24 and IRBB5 at p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 02110 . 7554/eLife . 19605 . 022Figure 5—figure supplement 1 . Southern hybridization analysis of TALE genes in different Xanthomonas species . DNA was digested with SphI and probed with the 2 . 9-kb SphI fragment of TALE gene pthXo1 from Xoo strain PXO99 . The size positions of DNA markers are indicated at left . Xac , Xanthomonas axonopodis pv . citri; Xcv , Xanthomonas euvesicatoria; Xoc , Xanthomonas oryzae pv . oryzicola; Xcc , Xanthomonas campestris pv . camperstris . The Xcc strain 8004 is TALE-free ( Qian et al . , 2005 , Genome Res . 15:757-767 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 02210 . 7554/eLife . 19605 . 023Figure 5—figure supplement 2 . Analysis of interactions between Xoc TALEs and rice TFIIAγs . The physical interactions between TFB regions of TALEs from Xoc RH3 and TFIIAγs were assessed by growth of yeast cells on synthetic defined premixes ( SD ) medium lacking leucine ( L ) , tryptophan ( W ) , histidine ( H ) , and adenine ( A ) . Among the 18 TFB regions , the sequences of TFB11 , TFB15 , TFB23 , TFB24 , and TFB27 from RH3 were the same as the TFB regions of Tal9b , Tal11a , Tal5b , Tal3c , and Tal12 from sequenced Xoc strain BLS256 , respectively . ( A ) The TFB regions from RH3 interacted with rice TFIIAγ5 analysed using yeast two-hybrid assay . ( B ) A randomly chosen 7 of the 18 TFB regions did not interact with TFIIAγ5V39E ( the mutated TFIIAγ5 ) , but three interacted with mutated rice TFIIAγ1S47E analysed by yeast two-hybrid assay . ( C ) Mutation of TFIIA5 enhanced resistance to Xoc . Plants were inoculated with Xoc at the booting stage . Each bar represents mean ( total 40 to 50 leaves from 5 plants ) ± standard deviation . b , significant difference between IR24 and IRBB5 plants at p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 023 A recent study has shown that a major quantitative trait locus for resistance to Xoc col-localizes with TFIIAγ5 ( Xie et al . , 2014 ) . Two additional studies have revealed that a TALE that occurs in at least 10 sequenced Xoc strains transcriptionally activates the gene for the sulphate transporter OsSULTR3;6 , a major susceptibility gene in rice−Xoc interactions ( Cernadas et al . , 2014; Wilkins et al . , 2015 ) . Xoc-induced expression of OsSULTR3;6 was significantly reduced ( p<0 . 01 ) in IRBB5 relative to IR24 ( Figure 5C ) , suggesting that TALE-containing Xoc also requires TFIIAγ5 to infect rice via TALE-induced expression of host susceptibility genes .
TFIIAγ is indispensable for polymerase II–dependent transcription ( Li et al . , 1999 ) . We have shown here how TALE-carrying Xoo and Xoc exploit rice TFIIAγ5 for activating transcription of downstream susceptibility genes ( Figure 6 ) . TALEs from Xoo and Xoc bind to TFIIAγ5 through their TFB regions , and the binding and binding strength are associated with the induction of susceptibility genes . Thus , TFIIAγ5 functions as a key component for TALE-induced host gene expression . 10 . 7554/eLife . 19605 . 024Figure 6 . A model showing TFIIAγ5 as a key component of rice infection by Xanthomonas bacteria . The bacteria hijack rice basal transcription factor TFIIAγ5 ( IIAγ ) by the transcription factor binding ( TFB ) region of their TALEs to induce host susceptibility ( S ) genes for infection . TS , amino-terminal translocation signal; RR , central repeat region; NLS , nuclear localization signal; AD , carboxyl-terminal transcription activation domain . The IIAγ belongs to the transcription pre-initiation complex . This complex consists of transcription factors IIA , which is composed of IIAβα subunit and IIAγ subunit , IIB , IID , IIE , IIF , and IIH , RNA polymerase II ( Pol II ) , and TATA-binding protein ( TBP ) . The binding of transcription pre-initiation complex to the TATA box of promoter was adopted and modified based on Yudkovsky et al . ( 2000 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19605 . 024 It is striking that the only TFIIAγ5 paralog in rice , TFIIAγ1 , apparently functions as a downstream susceptibility gene for Xoo PXO99 , such that the TALE PthXo7 directly activates TFIIAγ1 transcription ( Sugio et al . , 2007 ) , which differs from the protein-protein interaction of several Xoo TALEs with TFIIAγ5 . The recessive disease resistance allele TFIIAγ5V39E confers markedly reduced TALE-dependent induction of downstream susceptibility genes , apparently without compromising the overall activity of TFIIA . The rice accession IRBB5 carrying TFIIAγ5V39E is indistinguishable from the near-isogenic line IR24 in plant morphology and agronomic performance , including heading date , flag leaf length , number of panicles per plant , panicle length , grains per panicle , 1000-grain weight , yield per plant , seed setting rate , grain length , width , and thickness , with only slightly reduced plant height ( Supplementary file 1 ) . Here , we have shown that not only the specific point mutant TFIIAγ5V39E has increased Xoo resistance , but also that this can also be achieved by RNAi mediated knockdown of TFIIAγ5 . In addition , we have shown that TALEs from other Xanthomonas pathogens , such as Xoc , exploit TFIIAγ5 . Alteration of TFIIAγ5 activity , either through the introduction of the TFIIAγ5V39E allele , or through other reduction-of-function mutations , can provide a general strategy for improving rice resistance to TALE-carrying pathogens . TALE-carrying bacteria cause diseases in many other crops ( Schornack et al . , 2013; Boch et al . , 2014 ) . If these bacteria also exploit the host TFIIAγ for infection , modification of TFIIAγ may provide a road to improving disease resistance in other crops as well .
A pair of near-isogenic lines , IR24 ( TFIIAγ5 ) and IRBB5 ( TFIIAγ5V39E ) , and the variety Zhonghua 11 were used in this study . Plants were grown during the normal rice growing season under natural field conditions in the Experimental Stations of Huazhong Agricultural University , Wuhan , China . Chinese , Japanese , Korean , and Pilipino Xoo strains were used to study rice resistance to bacterial blight disease ( Gao et al . 2010; Li et al . , 2012 ) . Resistance to Xoc was tested using Chinese strains ( Ke et al . , 2014 ) . X . campestris pv . campestris strain was used for Southern blot analysis of TALE genes ( He et al . , 2007 ) . All Xanthomonas strains were grown at 28°C on nutrient agar medium . Antibiotics were used at the following final concentrations as required: ampicillin at 100 μg ml−1 , rifampicin at 75 μg ml−1 , kanamycin at 25 μg ml−1 , and spectinomycin at 50 μg ml−1 when genetic manipulation of bacteria . To construct RNA interference vector , the 3′ untranslated regions of TFIIAγ5 and TFIIAγ1 were amplified with primers listed in Supplementary file 2 and inserted into vector pDS1301 ( Yuan et al . , 2010 ) . Agrobacterium-mediated transformation of rice was performed ( Lin and Zhang , 2005; Ge et al . , 2006 ) . Xoo inoculation described in more detail at Bio-protocol ( Ke et al . , 2017 ) . To evaluate reaction of rice plants to Xoo , plants were inoculated with the Xoo strains by the leaf-clipping method at the booting ( panicle development ) stage ( Chen et al . , 2002 ) . The disease was scored by measuring the lesion length at 14 days after inoculation . Each bacterial inoculation assay was repeated at least twice . The disease of some plants was also evaluated by analysing bacterial growth based on a count of the colony-forming units as described previously ( Sun et al . , 2004 ) . For measuring bacterial growth , one Xoo-infected leaf from each plant was examined as one replicate , and a total of three plants for each sample were analysed . To evaluate Xoc resistance , rice plants were inoculated with Xoc strains by the penetration method using a needleless syringe at the booting stage ( Ke et al . , 2014 ) . Disease was scored by lesion length at 14 days after inoculation . Each bacterial inoculation assay was repeated twice . The 2-cm leaf segments next to the bacterial infection sites in the rice plants were collected for RNA isolation . Quantitative reverse transcription-PCR ( qRT-PCR ) was conducted using gene-specific primers ( Supplementary file 3 ) as described previously ( Qiu et al . , 2007 ) . The expression level of the rice actin gene was used to normalize the measurement of the expression . Each rice sample was a mixture of leaf tissue from at least five plants , with 8 to 10 leaves per plant . For transgenic plants , segments from three to five leaves of the plant were sampled . Each qRT-PCR assay was repeated at least twice , with each repetition having three technical replicates . The TALE PthXo1 was cloned into pHM1 vector to produce pHM1pthXo1 , and transferred into Xoo strains T7174 and KACC10331 following published method ( Yang and White , 2004 ) . The TFB region of PthXo1 was replaced with TFB regions of other TALEs by Gibson assembly ( Gibson et al . , 2009 ) . The recombinant strains were confirmed by PCR amplification of TALE fragments . A standard procedure for Southern hybridization of the bacterial DNA was performed ( Gu et al . , 2009 ) . Genomic DNA from different Xanthomonas strains was digested with SphI , separated by electrophoresis on 1 . 2% agarose gel in TAE buffer , blotted onto a nylon membrane , and hybridized using a 32P-labeled 2 . 9-kb SphI fragment of PthXo1 . The transactivation activity of PthXo1 was analysed in yeast cells as described previously ( Deng et al . , 2012 ) . The open reading frame of pthXo1 was ligated into pGBKT7 vector and fused in frame with the yeast GAL4 DNA binding domain . The recombinant vector was transformed into yeast strain AH109 . The transformed yeast cells were plated on SD/−Trp or SD/−Trp-His medium and cultured for 3 days as described previously ( Yuan et al . , 2010 ) . The interaction between bacterial TALE proteins and host proteins in yeast cells was assayed using MATCHMAKER GAL4 Two-Hybrid System 3 ( Clontech , Mountain View , CA ) according the manufacturer’s instructions . To construct the interaction vectors , full-length and truncated TALEs and the TFB regions of TALEs and plant genes were amplified using the PCR primers listed in Supplementary file 2 . The amplified DNA fragments were first inserted into vector pBluescript ( Agilent Technologies , Santa Clara , CA ) for sequencing confirmation . The confirmed bacterial DNA fragments were then ligated into pGBKT7 vector , and the confirmed plant DNA fragments were then ligated into pGADT7 Rec vector . The recombinant pGBKT7 and pGADT7 plasmids were co-transformed into yeast strain AH109 for yeast two-hybrid assays following the lithium acetate method ( Yuan et al . , 2010 ) . The yeast clones were first scribed on the synthetic defined premixes ( SD ) medium lacking leucine ( L ) and tryptophan ( W ) ( −LW ) . The growth of yeast cells on SD/−LW medium indicated that they carried both pGBKT7 and pGADT7 plasmids . An aliquot ( 10 μl ) of 1:10 diluted stationary phase cultured yeast clones grown on the SD/−LW medium was then scribed on the selective SD medium lacking L , W , histidine ( H ) , and adenine ( A ) ( −LWHA ) . The growth of yeast cells on SD/−LWHA medium indicated that the examined proteins interacted with each other . The interactions of these proteins were also assessed by examination of β-D-galactopyranoside ( X-α-gal ) activity and β-galactosidase ( LacZ ) activity as described previously ( Yuan et al . , 2010 ) . Each yeast two-hybrid assay was repeated at least twice . CoIP assays were performed to study the interaction between TALE proteins and plant proteins in planta . The 9×myc DNA fragment was amplified from pN-TAPa vector ( Rubio et al . , 2005 ) by using myc-specific primers ( Supplementary file 2 ) and inserted into the SmaI- and BamHI-digested pU1301 vector ( Cao et al . , 2007 ) , resulting in a vector that we named pU1301-9myc . The DNA fragments of full-length , truncated , or TFB region of TALEs were ligated into the pU1031-9myc vector . The DNA fragments of plant genes were ligated into the pU1301-3FLAG vector ( Yuan et al . , 2010 ) . The recombinant vectors were introduced into Agrobacterium tumefaciens strain GV3101 . Agrobacterium-mediated transformation was performed by infiltration into N . benthamiana leaves using a needleless syringe ( Yuan et al . , 2010 ) . CoIP assays were conducted using anti-FLAG antibody ( RRID:AB_259529 , Sigma-Aldrich , St . Louis , MO ) and anti-myc antibody ( Tiangen , Beijing , China ) as described previously ( Yuan et al . , 2010 ) . Each CoIP assay was repeated at least twice . Mutations of plant genes and the Xoo TALE genes were made using the GeneTailor Site-Directed Mutagenesis System ( Invitrogen Life Technologies , Carlsbad , CA ) as described previously ( Yuan et al . , 2011 ) . The mutagenic primers are listed in Supplementary file 2 . Differences between samples were analysed for statistical significance by t-test in Microsoft Excel ( Microsoft , Redmond , WA ) . Correlations between gene transcript level and disease level were calculated using CORREL analysis in the Microsoft Office Excel program . | Around the world , bacterial infections reduce the yields of many important crops like rice , tomatoes , peppers and citrus fruits . Xanthomonas is a particularly widespread genus of bacteria; it consists of almost 30 species that cause diseases in more than 400 plant hosts , including bacterial blight and bacterial streak in rice plants . Plants do have an immune system that is able to detect invading microbes and trigger a defensive response against them; however , many disease-causing bacteria have evolved ways to avoid or counteract this response . For example , at least five Xanthomonas species use proteins called transcription activator-like effectors ( or TALEs for short ) to infect their host plants . The bacterial proteins are essentially injected into the plant’s cells where they activate specific plant genes that make the host more susceptible to infection . Like other organisms , plants use proteins called transcription factors to switch genes on or off . However , it was not clear if the TALEs hijack the plant’s transcriptional machinery to activate these “susceptibility genes” or if they activate the genes via some other means . Now , Yuan et al . show that TALE-carrying bacteria do make use of at least one of rice’s own transcription factors to cause bacterial blight and bacterial streak . The transcription factor in question is rice’s version of a general transcription factor , called TFIIAγ , which is essential for gene activation in plants , animals and fungi . Yuan et al . also identify the region of the TALE that binds to the transcription factor , and show that rice plants with lower levels of the transcription factor are protected against bacterial blight and bacterial streak . Uncovering how disease-causing Xanthomonas bacteria use TALEs to infect plants will hopefully help researchers to develop crop plants that are more resistant to these harmful bacteria . Further work is now needed to see if the gene that encodes TFIIAγ in crop plants can be edited to achieve this goal , or whether genes encoding resistant variants of the protein already exist in other plant species . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology"
] | 2016 | A host basal transcription factor is a key component for infection of rice by TALE-carrying bacteria |
Understanding how gene expression programs are controlled requires identifying regulatory relationships between transcription factors and target genes . Gene regulatory networks are typically constructed from gene expression data acquired following genetic perturbation or environmental stimulus . Single-cell RNA sequencing ( scRNAseq ) captures the gene expression state of thousands of individual cells in a single experiment , offering advantages in combinatorial experimental design , large numbers of independent measurements , and accessing the interaction between the cell cycle and environmental responses that is hidden by population-level analysis of gene expression . To leverage these advantages , we developed a method for scRNAseq in budding yeast ( Saccharomyces cerevisiae ) . We pooled diverse transcriptionally barcoded gene deletion mutants in 11 different environmental conditions and determined their expression state by sequencing 38 , 285 individual cells . We benchmarked a framework for learning gene regulatory networks from scRNAseq data that incorporates multitask learning and constructed a global gene regulatory network comprising 12 , 228 interactions .
Elucidating relationships between genes , and the products they encode , remains one of the central challenges in experimental and computational biology . A gene regulatory network ( GRN ) is a directed graph in which regulators of gene expression are connected to target gene nodes by interaction edges . Regulators of gene expression include transcription factors ( TF ) which can act as activators and repressors , RNA binding proteins , and regulatory RNAs . Identifying regulatory relationships between transcriptional regulators and their targets is essential for understanding biological phenomena ranging from cell growth and division to cell differentiation and development ( Davidson , 2012 ) . Reconstruction of GRNs is required to understand how gene expression dysregulation contributes to cancer and complex heritable diseases ( Barabási et al . , 2011; Hu et al . , 2016 ) . Genome-scale methods provide an efficient means of identifying gene regulatory relationships . Efforts of the past two decades have resulted in the development of a variety of experimental and computational methods that leverage advances in technology and machine learning for constructing GRNs . Previously , we developed a method for inferring transcriptional regulatory networks based on regression with regularization that we have called the Inferelator ( Bonneau et al . , 2006; Ciofani et al . , 2012 ) . This method takes as inputs gene expression data and sources of prior information , and outputs regulatory relationships between transcription factors and their target genes that explain the observed gene expression levels . Subsequent work has enhanced this approach by selecting regulators for each gene more effectively ( Madar et al . , 2010 ) , incorporating orthogonal data types that can be used to generate constraints on network structure ( Greenfield et al . , 2013 ) , and explicitly estimating latent biophysical parameters including transcription factor activity ( Arrieta-Ortiz et al . , 2015; Fu et al . , 2011 ) and mRNA decay rates ( Tchourine et al . , 2018 ) . We have successfully applied this approach to construct GRNs from gene expression data acquired from variation across time , conditions , and genotypes in microbes ( Arrieta-Ortiz et al . , 2015; Tchourine et al . , 2018 ) , plants ( Wilkins et al . , 2016 ) , and mammalian cells ( Ciofani et al . , 2012; Miraldi et al . , 2019 ) . Recently , single-cell RNA sequencing ( scRNAseq ) has exploded in popularity with the development of droplet systems for rapid encapsulation and labeling of thousands of cells in parallel . The DROP-seq system ( Macosko et al . , 2015 ) based on bead capture , and the inDrop ( Zilionis et al . , 2017 ) and 10x Genomics ( Zheng et al . , 2017 ) systems based on hydrogel beads , provide a facile means of generating RNA sequencing data for tens of thousands of individual cells . Although scRNAseq has primarily been used for defining cell types and states , this technology holds great potential for efficient construction of GRNs ( Hwang et al . , 2018 ) . By combining genetic perturbation of transcriptional regulators using CRISPR/Cas9 with scRNAseq , mixtures of genetic perturbations can be assayed in a single reaction ( Adamson et al . , 2016; Dixit et al . , 2016; Jaitin et al . , 2016 ) . This approach , known as Perturb-seq , presents a new opportunity for efficiently inferring GRNs from thousands of individual cells in which different regulators have been disrupted . There are considerable advantages in both scalability and detection of intra-sample heterogeneity with Perturb-seq , but quantifying the the effectiveness of CRISPR/Cas9 targeting in individual cells and distinguishing gene expression variability from noise inherent to mRNA undersampling in scRNAseq ( Brennecke et al . , 2013; Grün et al . , 2014 ) present technical challenges . Computational methods to take advantage of scRNAseq data for inferring GRNs are under active development ( Aibar et al . , 2017; Chan et al . , 2017; van Dijk et al . , 2018 ) . However , benchmarking these methods is difficult; in the absence of a known GRN , model performance is often estimated using simulated data ( Chen and Mar , 2018 ) , and issues regarding the appropriate experimental and computational approaches to GRN construction from scRNAseq data remain unresolved . The budding yeast Saccharomyces cerevisiae is ideally suited to constructing GRNs from experimental data and benchmarking computational methods . Decades of work have provided a plethora of transcriptional regulatory data comprising functional and biochemical information ( de Boer and Hughes , 2012; Teixeira et al . , 2018 ) . As a result , yeast is well suited to constructing GRNs using methods that leverage the rich available information and for assessing the performance of those methods by comparison to experimentally validated interactions ( Ma et al . , 2014; Tchourine et al . , 2018 ) . Budding yeast presents several technical challenges for single cell analysis , and as a result scRNAseq methods for budding yeast reported to date ( Gasch et al . , 2017; Nadal-Ribelles et al . , 2019 ) yield far fewer individual cells ( ~102 ) than are now routinely generated for mammalian studies ( >104 ) . The limitations of existing scRNAseq methods for budding yeast cells limits our ability to investigate eukaryotic cell biology as many signaling and regulatory pathways are highly conserved in yeast ( Carmona-Gutierrez et al . , 2010; Gray et al . , 2004 ) , including the Ras/protein kinase A ( PKA ) , AMP Kinase ( AMPK ) and target of rapamycin ( TOR ) pathways ( González and Hall , 2017; Loewith and Hall , 2011 ) . However , recent work has successfully established single cell sequencing in the fission yeast Schizosaccharomyces pombe ( Saint et al . , 2019 ) . In budding yeast , the TOR complex 1 ( TORC1 or mTORC1 in human ) coordinates the transcriptional response to changes in nitrogen sources ( Godard et al . , 2007; Rødkaer and Faergeman , 2014 ) . Controlling this response are four major TF groups , which are regulated by diverse post-transcriptional processes . The Nitrogen Catabolite Repression ( NCR ) pathway , which is regulated principally by TORC1 , consists of the TFs GAT1 , GLN3 , DAL80 , and GZF3 ( Hofman-Bang , 1999 ) , and is responsible for suppressing the utilization of non-preferred nitrogen sources when preferred nitrogen sources are available . Gat1 and Gln3 are localized to the cytoplasm until activation results in relocalization to the nucleus ( Cox et al . , 2000 ) , where they then compete with Dal80 and Gzf3 for DNA binding motifs ( Georis et al . , 2009 ) . The General Amino Acid Control ( GAAC ) pathway consists of the TF GCN4 ( Hinnebusch , 2005 ) , and is responsible for activating the response to amino acid starvation , as detected by increases in uncharged tRNA levels . Gcn4 activity is translationally controlled by ribosomal pausing at upstream open reading frames in the 5’ untranslated region ( Mueller and Hinnebusch , 1986 ) . The retrograde pathway , consisting of the TF heterodimer RTG1 and RTG3 , is responsible for altering expression of metabolic and biosynthetic genes in response to mitochondrial stress ( Jia et al . , 1997; Liao and Butow , 1993 ) or environmental stress ( Ruiz-Roig et al . , 2012 ) . The Rtg1/Rtg3 complex is localized to the cytoplasm until activation , upon which they translocate into the nucleus ( Komeili et al . , 2000 ) . The Ssy1-Ptr3-Ssy5-sensing ( SPS ) pathway ( Ljungdahl , 2009 ) , consists of the TFs STP1 and STP2 , and is responsible for altering transporter expression ( Didion et al . , 1998; Iraqui et al . , 1999 ) in response to changes in extracellular environment . Stp1 and Stp2 are anchored to the plasma membrane until the SPS sensor triggers proteolytic cleavage of their anchoring domain and releases them for nuclear import ( Andréasson and Ljungdahl , 2002 ) . Construction of GRNs based on the transcription factors in these pathways has had mixed success; the high redundancy of the NCR pathway has proven challenging to deconvolute ( Milias-Argeitis et al . , 2016 ) . The GAAC pathway is more straightforward , although separating direct and indirect regulation remains difficult , even with high-quality experimental data ( Mittal et al . , 2017 ) . As a result , a comprehensive GRN for nitrogen metabolism has remained elusive , despite successes in identifying genes that respond to changes in environmental nitrogen ( Airoldi et al . , 2016 ) and identification of post-transcriptional control mechanisms that underlie these changes ( Miller et al . , 2018 ) . Many signalling regulators involved in environmental response interact with cell cycle programs ( Johnston et al . , 1977; Talarek et al . , 2017 ) , including the TOR pathway ( Zinzalla et al . , 2007 ) ; however , how regulation of the mitotic cell cycle and environmentally responsive gene expression is coordinated is unknown . The regulation of nitrogen responsive gene expression in yeast is well-suited to the development of generalizable methods as the degree of TF redundancy , post-transcriptional regulation of TF activity , which precludes straightforward relationships between TF abundance and target expression , and multifactorial impact on gene expression , including intrinsic and extrinsic processes and stimuli , provide a tractable model system for addressing these challenges in higher eukaryotes . Here , we have developed a method for scRNAseq in budding yeast using Chromium droplet-based single cell encapsulation ( 10x Genomics ) . We engineered TF gene deletions by precisely excising the entire TF open reading frame and introducing a unique transcriptional barcode that enables multiplexed analysis of genotypes using scRNAseq . We pooled 72 different strains , corresponding to 12 different genotypes , and determined their gene expression profiles in 11 conditions using scRNAseq analysis of 38 , 000 cells . We show that our method enables identification of cells from complex mixtures of genotypes in asynchronous cultures that correspond to specific mutants , and to specific stages of the cell cycle . Identification of mutants can be used to identify differentially expressed genes between genotypes providing an efficient means of multiplexed gene expression analysis . We used scRNAseq data in yeast to benchmark computational aspects of GRN reconstruction , and show that multi-task learning integrates information across environmental conditions without requiring complex normalization , resulting in improved GRN reconstruction . We find that imputation of missing data does not improve GRN reconstruction and can lead to prediction of spurious interactions . Using scRNAseq data , we constructed a global GRN for budding yeast comprising 12 , 228 regulatory interactions . We discover novel regulatory relationships , including previously unknown connections between regulators of cell cycle gene expression and nitrogen responsive gene expression . Our study provides a generalizable framework for GRN reconstruction from scRNAseq , a rich data set that will enable benchmarking of future computational methods , and establishes the use of droplet-based scRNAseq analysis of multiplexed genotypes in yeast .
The yeast gene knockout collection ( Giaever et al . , 2002 ) facilitates pooled analysis of mutants using unique DNA barcode sequences that identify each gene deletion strain , but these barcodes are only present at the DNA level , precluding their use with scRNAseq . Therefore , we constructed an array of prototrophic , diploid yeast strains with homozygous deletions of TFs that control distinct regulatory modules: 1 ) NCR , 2 ) GAAC , 3 ) SPS-sensing , and 4 ) the retrograde pathway that coordinately control nitrogen-related gene expression in yeast . We engineered eleven different TF knockout genotypes , using six independently constructed biological replicates for each genotype . In addition , we constructed six biological replicates of the wild-type control in which we deleted the neutral HO locus . Genes were deleted using a modified kanMX cassette such that each of the 72 strains contains a unique transcriptional barcode in the 3’ untranslated region ( UTR ) of the G418 resistance gene , that can be recovered by RNA sequencing ( Figure 1 , Figure 1—figure supplement 1A ) . Homozygous diploids were constructed by mating to a strain containing the same TF knockout marked with a nourseothricin drug resistance cassette . On rich media plates , the 72 strains have an approximately wild-type growth; under nutritional stress , some TF knockouts exhibit growth advantages or disadvantages ( Figure 1—figure supplement 1B ) . ScRNAseq in yeast presents several challenges: cells are small ( 40–90 µm3 ) , enclosed in a polysaccharide-rich cell wall , and contain fewer mRNAs per cell ( 40 k-60k ) than higher eukaryotes . We developed and validated a protocol using the droplet-based 10x genomics chromium platform , and it used it to perform scRNAseq of the pool of TF knockouts in eleven growth conditions that provide a range of metabolic challenges ( Table 1 ) . In addition to variable nitrogen sources in minimal media with excess [MM] and limiting [NLIM-NH4 , NLIM-GLN , NLIM-PRO , NLIM-UREA] nitrogen , some conditions result in fermentative metabolism of glucose in rich media [YPD] , and inhibition of the TOR signaling pathway in rich media by the small molecule rapamycin [RAPA] . We also studied conditions that require respiratory metabolism of ethanol in rich [YPEtOH] and minimal media [MMEtOH] , and in rich media after sugars had been fully metabolized to ethanol and cells have undergone the diauxic shift [DIAUXY] . We tested two different starvation conditions , carbon [CSTARVE] and nitrogen starvation [NSTARVE]; however , the latter condition did not pass quality control during single-cell transcriptome library preparation and was discarded . Cells from the eleven different conditions were sequenced and processed using cellranger ( 10x genomics ) and our custom analysis pipeline ( fastqToMat0 ) , yielding a digital expression matrix ( Source code 1 ) in which each cell is annotated with the environmental growth condition and genotype . Genotype-specific barcodes facilitate identification and removal of droplets that have multiple cells ( doublets ) by determining cell IDs that have more than one annotated genotype . Using our pool of 72 strains , we detect and remove 98 . 5% of doublets . PCR artifacts and duplicates are removed using Unique Molecular Identifiers ( UMIs ) ( Kivioja et al . , 2012 ) to quantify gene expression as unique transcript reads ( counts ) . Following sequence processing , quality control , removal of doublets , and assigning metadata , we recovered 83 , 703 , 440 transcript counts from a total of 38 , 225 individual cells . To initially assess the quality of our data , we examined the expression of genes that are characteristic of different metabolic states . Consistent with our expectations , the core fermentative ( anaerobic ) genes PDC1 and ENO2 are expressed in cells in fermenting culture conditions only , and the core respirative ( aerobic ) gene ADH2 is expressed in cells in respiring culture conditions ( Figure 2A ) . The number of cells recovered varies by over an order of magnitude between conditions; stressful conditions of low nitrogen have lower cell yields overall . The yeast stress response is linked to increased resistance to zymolyase digestion ( Nagarajan et al . , 2014 ) , which may be reflected in decreased cell yield during single-cell sequencing . Each of the 72 strains is found in each of the 11 conditions , although the number of each strain and genotype varies by environmental condition ( Figure 2—figure supplement 1A ) , and some strains are disproportionately affected . However , the number of transcripts per cell is generally equivalent between strains even when they differ in representation within libraries ( Figure 2—figure supplement 1B ) . By contrast , we find that total transcript counts per cell are highly linked to environmental growth conditions ( Figure 2—figure supplement 1C ) , which is consistent with decreased total transcriptome pool size in suboptimal conditions ( Athanasiadou et al . , 2019 ) . For cells growing in rich medium ( YPD ) we recover a median of 2250 unique transcripts per cell , from a median of 695 distinct genes , indicating a capture rate of approximately 3–5% of total transcripts from each cell . The strain genotype does not strongly influence transcript counts per cell ( Figure 2—figure supplement 1D ) . There is a high correlation between single-cell expression data and bulk RNA expression data ( spearman correlation 0 . 941 ) for wild-type cells grown in YPD ( Figure 2—figure supplement 2 ) indicating that the effect of technical bias caused by single-cell processing is minimal . We also find good correlation to other published single-cell yeast data sets , and a comparable published bulk RNAseq experiment , . Mapping the digital expression matrix into two-dimensional space with a Uniform Manifold Approximation and Projection [UMAP] results in clear separation of individual cells into groups based on environmental condition ( Figure 2B ) . Cells from different minimal media or nitrogen-limited growth conditions localize near each other , and cells grown in different rich nitrogen sources are clearly separate from each other . Within environmentally-determined grouping there appears to be no strong ordering by genotype ( Figure 2C ) . These clusters are not driven by sequencing depth ( Figure 2—figure supplement 3B ) , although there are some stress conditions which have subpopulations which are downregulated for ribosomal genes and upregulated for induced environmental stress response ( iESR ) genes ( Figure 2—figure supplement 3C–F ) . The increased relative abundance of ribosomal related gene expression in rich media conditions is consistent with previously-observed correlation of ribosomal gene expression and cellular growth rate ( Brauer et al . , 2008 ) . Some measures of variance per-gene differ in different growth conditions ( Figure 2—figure supplement 4 ) . Interactive figures are provided ( http://shiny . bio . nyu . edu/YeastSingleCell2019/ ) facilitating exploration of expression levels for all genes . To identify sources of gene expression differences between cells within environments , we clustered single cells within each environmental condition separately by constructing a Shared Nearest Neighbor graph ( Xu and Su , 2015 ) and clustering using the Louvain method ( Blondel et al . , 2008 ) . Genes with known roles in mitotic cell cycle are highly represented among the most differentially expressed genes between clusters ( Figure 3—figure supplement 1A ) . Overlaying the expression of three of these genes ( PIR1 , DSE2 , and HTB1/HTB2 ) on UMAP plots illustrates cell cycle effects on single cell gene expression ( Figure 3A and Figure 3B ) . PIR1 expression , a marker for early G1 ( Spellman et al . , 1998 ) , is diagnostic of a distinct cluster . DSE2 is expressed only in daughter cells ( Colman-Lerner et al . , 2001 ) , which allows daughter cells in G1 to be distinguished from mother cells in G1 . Cells that have high expression of the histone 2B genes , which are upregulated in S-phase ( Eriksson et al . , 2012 ) , are localized together in the UMAP plots ( Figure 3B ) . For each cluster of cells within a growth condition we plotted the proportion of cells belonging to each TF deletion genotype , and the mean expression of several cell cycle genes ( Figure 3B ) . Some clusters predominantly contain cells from a single TF deletion genotype; for example , cells deleted for GLN3 ( gln3Δ ) form a separate cluster in YPD and RAPA conditions , as do cells deleted for one of the RTG heterodimer components ( rtg1Δ and rtg3Δ ) . However , differences in expression due to genotype do not appear to be a primary source of expression differences within conditions , as most clusters show a uniform distribution of genotypes ( Figure 3B , Figure 3—figure supplement 1B ) . Similarly , we do not find that differences in expression of metabolic genes underlie overall differences in expression ( e . g . HXK2 ) suggesting that the yeast metabolic cycle ( Silverman et al . , 2010; Tu et al . , 2005 ) is not readily identifiable in single cells using scRNAseq . Three of the high-stress growth conditions ( NLIM-GLN , NLIM-PRO , and MMEtOH ) have clusters that are separate from the majority of the cells analyzed in those conditions . We find that these clusters have higher levels of stress response genes and lower levels of ribosomal genes than other cells in these conditions ( Figure 3—figure supplement 2B–C ) These clusters may reflect cells undergoing early entry into quiescence and provide evidence for a heterogeneous response to stressful conditions . To assess our ability to determine differential gene expression between TF knockout strains , we examine the expression of genes known to respond to nitrogen signalling . GAP1 ( General Amino acid Permease ) is a transporter responsible for importing amino acids under conditions of nitrogen limitation; GAP1 expression is regulated by the NCR activators GAT1 and GLN3 , the NCR repressors GZF3 and DAL80 ( Stanbrough and Magasanik , 1995 ) , and potentially GCN4 ( Natarajan et al . , 2001 ) . We identify differing degrees of dysregulation of GAP1 expression when these TFs are deleted ( Figure 4A ) . The effect of deleting TFs varies by condition: GAP1 is not expressed in YPD and its expression increases in nitrogen-limited media and in response to rapamycin . Deletion of GAT1 results in decreased expression in nitrogen limiting media , but deletion of GLN3 does not affect GAP1 expression . By contrast , in the presence of rapamycin deletion of GLN3 results in reduced GAP1 expression . Deletion of GCN4 only impacts GAP1 expression in the presence of urea . MEP2 and GLN1 are also responsive to nitrogen TFs , and are dysregulated when certain TFs are deleted; expression of the glycolytic gene HXK2 decreases when GLN3 , GCN4 , or RTG1/RTG3 are deleted , but only in conditions of nitrogen limitation ( Figure 4—figure supplement 1A ) . These environmentally dependent impacts of genotype on gene expression demonstrate the importance of exploration of variable conditions for studying genotypic effects on expression . A variety of statistical methods have been proposed and benchmarked for testing different expression of scRNAseq data ( Soneson and Robinson , 2018 ) . Our experimental design allows single-cell measurements to be collapsed into a total count ( pseudobulk ) measurement by summing counts across all cells that correspond to each of the six individual replicates of each genotype within a condition . When we analyze this data using standard approaches to RNAseq analysis ( DESeq2 ) we detect several genes with significant ( adjusted p-value<0 . 05 ) differences in expression ( fold change >1 . 5 ) between wild-type and TF deletion strains ( Figure 4B ) that are consistent with known regulatory pathways . There are considerably fewer changes in gene expression as a result of TF deletions compared to the hundreds of genes that change expression between different conditions ( Figure 4—figure supplement 1B ) . However , in cells grown in rich media [YPD] , we found 96 genes that are differentially expressed in TF deletion strains compared to wildtype ( Figure 4C ) , and expression of 160 genes are perturbed in TF deletion strains compared to wildtype when exposed to rapamycin [RAPA] ( Figure 4—figure supplement 1C ) . Many of these differentially expressed genes are annotated as functioning in amino acid metabolism and biosynthesis . Differential gene expression in a TF knockout strain is not sufficient evidence of a direct regulatory relationship as many significant changes in gene expression upon deleting a TF are indirect , and many direct effects may be subtle . Therefore , we constructed a gene regulatory network using the Inferelator , a regression-based network inference method which is based on three main modeling assumptions . First , we assume that Transcription Factor Activity ( TFA ) is a latent biophysical parameter that represents the effect of a TF binding to DNA and modulating its transcription activity ( Arrieta-Ortiz et al . , 2015; Fu et al . , 2011 ) . The TFA values are not directly measured , and instead must be estimated as a relative value based on prior knowledge of a regulatory network of TF and target relationships . This TFA estimation is essential as many TFs are post-transcriptionally regulated , or are expressed at levels that are not reliably detected by scRNAseq ( Filtz et al . , 2014 ) . Second , we assume that expression of a gene can be described as a weighted sum of the activities of TFs ( Bonneau et al . , 2006 ) using an additive model in which activators and repressors increase or decrease the expression of targets linearly . Finally , we assume that each gene is regulated by a small number of TFs , and that regularization of gene expression models is required to enforce this biologically relevant property of target regulation . Saccharomyces cerevisiae , as a preeminent model organism in systems biology , has a well defined set of known interactions that are of considerably higher quality than is available for more complex eukaryotes providing a validated gold standard for testing model performance ( Tchourine et al . , 2018 ) . To evaluate the performance of data processing methods and model parameter selections within the Inferelator on scRNAseq data , we perform ten cross-validations using the existing gold standard network . During cross-validation , we infer a GRN using half of the gold standard target genes as priors , then evaluate performance based on recovery of TF-target gene interactions for gold standard interactions that are left out of the priors . We tested preprocessing and prior selection options by inferring networks using gene expression models that are regularized by best subset regression to minimize Bayesian Information Criterion ( Arrieta-Ortiz et al . , 2015; Greenfield et al . , 2013 ) and quantified performance in predicting TF-target interactions using the area under the precision-recall curve ( AUPR ) . As negative controls , we employed the same procedure after shuffling priors and after simulating scRNAseq data in which all variance is due to sampling noise . The negative control with shuffled priors establishes a random classifier baseline AUPR of 0 . 02; the negative control with simulated data establishes a circular recovery baseline AUPR of 0 . 06 ( Figure 5A ) . Performance of the Inferelator on our scRNAseq data far exceeds these baselines , with a mean AUPR of 0 . 20 . This performance from our single dataset is comparable to that of a GRN constructed from 2577 experimental observations using bulk gene expression data ( Tchourine et al . , 2018 ) . The sparsity of data for each cell acquired using scRNAseq may negatively impact its utility in GRN construction . A commonly used technique to address missing data is data imputation . We tested the impact of several imputation packages on network inference: MAGIC ( van Dijk et al . , 2018 ) , ScImpute ( Li and Li , 2018 ) , and VIPER ( Chen and Zhou , 2018 ) . Whereas these methods can enhance separation of gene expression states in low-dimensionality projections ( Figure 5—figure supplement 1A ) , we find that they are either ineffective or detrimental to network inference ( Figure 5A ) . When the GRN is reconstructed from interactions selected at a precision threshold of 0 . 5 , which takes into account how many interactions are correct according to the gold standard , no imputation method increases the number of recovered interactions compared to unmodified data . Data imputation with MAGIC increases the total number of confidently predicted ( confidence >0 . 95 ) interactions , but recovers fewer interactions that are correct according to the gold standard . Algorithms for network inference perform poorly when making predictions based only on expression data ( Greenfield et al . , 2010 ) . Including prior knowledge of regulatory relationships and network topology improves model selection , and allows approximation of latent variables like TFA . Priors can be generated from regulatory interactions defined using methods such as chromatin immunoprecipitation sequencing ( ChIP-seq ) or analysis of transposase-accessible chromatin ( ATAC-seq ) and TF binding motifs , or from curated databases of interactions derived from literature . The source and processing of prior knowledge has a substantial effect on the size and accuracy of the learned network ( Azizi et al . , 2018; Siahpirani and Roy , 2017 ) . We tested the impact on GRN reconstruction of prior data derived from literature , and from high-throughput experimental assays that encompass interactions between the entire yeast genome and the majority of known TFs ( Figure 5B ) . The best performance is obtained using a curated set of known TF-gene interactions obtained from YEASTRACT ( Teixeira et al . , 2018 ) . Generating priors using motif searching within open chromatin regions determined by ATAC-seq ( Castro et al . , 2019; Miraldi et al . , 2019 ) , and by modeling TF-DNA affinities in promoters ( Ward and Bussemaker , 2008 ) provides a considerable improvement over GRN reconstruction from TF expression without priors , but have lower performance than priors derived from curated data . Numerous methods exist for integrating information across different conditions and experiments that aim to reduce technical variation while retaining biologically meaningful differences ( Hicks et al . , 2018; Leek et al . , 2010 ) . The appropriate approach to integrating scRNAseq data for the purpose of GRN reconstruction remains unknown . We find that when we separate data based on environmental conditions and infer GRNs we obtain unique networks of differing quality ( Figure 5C ) . Learning a single network from all conditions by first combining the data can be compromised by technical variability and imbalance in the number of cells between conditions . Furthermore , normalizing batches to equal transcript depth risks suppressing differences which are true biological variability . An alternative approach is to treat the cells from each environmental condition as separate tasks . Separate tasks can be learned independently , without sharing information between tasks ( implemented as BBSR ( BY TASK ) ) . This entails learning networks from each task , and then combining task-specific networks into a global network . Alternatively , networks can be learned together in a multitask learning ( MTL ) framework ( Lam et al . , 2016 ) , sharing information between tasks while they are learned , which we have implemented as Adaptive Multiple Sparse Regression ( AMuSR ) ( Castro et al . , 2019 ) . We find that , compared to network inference using all data simultaneously [BBSR ( ALL ) ] , treating conditions as separate network inference tasks provides a considerable improvement in performance ( Figure 5Di ) . This is likely due to the retention of environmentally specific interactions that would otherwise be obscured using methods for normalizing data prior to GRN construction . The performance of the information sharing network inference approach [AMuSR ( MTL ) ] and the non-sharing network inference approach [BBSR ( BY TASK ) ] are very similar overall . We find that some individual tasks had modest improvements in model performance with AMuSR and others with BBSR ( Figure 5Dii ) . We constructed a global gene regulatory network using the YEASTRACT priors ( as determined above ) and our multi-task network inference ( AMuSR ) procedure . Eleven GRNs were jointly learned from each of the eleven environmental growth conditions; for each task a confidence score for each regulator-target interaction was calculated . GRNs learned for each condition were combined by rank summing condition-specific confidence scores to create a global confidence score for each potential interaction . All potential interactions are ranked by global confidence score , and a global GRN is constructed from interactions that meet the precision threshold of 0 . 5 , as measured by recovery of known interactions ( Figure 6A , Source code 2 ) . The resulting GRN comprises 6114 new interactions and 6114 interactions present in the priors , resulting in a total of 12 , 228 regulator-target interactions . We find that 5372 interactions from the priors are not recovered ( recall of 0 . 532 ) . The global GRN comprises an identified regulator for approximately half of all known genes ( Figure 6—figure supplement 1A ) . There is a positive correlation between expression level for a gene and the number of regulators for that gene ( Figure 6—figure supplement 1B ) and 90% of the identified interactions are predicted to have activating effects ( Figure 6—figure supplement 1C ) . Many condition-specific networks have uniquely identified interactions ( Figure 6—figure supplement 1D ) , but more than 75% of the final network is composed of TF-gene interactions found in multiple conditions ( Figure 6—figure supplement 1E ) . Of the novel learned interactions ( i . e . those not in the prior data ) , 60% have evidence of a TF-gene regulatory relationship when compared to the YEASTRACT database ( Figure 6—figure supplement 1F ) . 573 learned TF-gene interactions have evidence for physical localization of the TF to the target gene , and 2957 learned TF-gene interactions have evidence of expression changes when the TF is perturbed . Within the nitrogen-regulated TF subnetwork comprising the 11 deleted TFs ( Figure 6B ) we identify 885 regulator-target interactions , of which 447 are novel , and 438 are present in the priors . This subnetwork contains many features consistent with expectations including co-regulation of targets by the NCR TFs . Overall , the global GRN has the largest number of target genes for general TFs ( including ABF1 , RAP1 , CBF1 , and SFP1 ) , but we also define regulatory relationships for a total of 129 of the predicted 207 yeast TFs ( Figure 6C ) . The poorest recovery of prior data is found for TFs that regulate environmental responses not included in our experimental design , such as the stress response TF MSN2 and the mating TF STE12 , highlighting the necessity of exploration of condition space for complete network reconstruction . Regulators and target genes can be mapped to Gene Ontology ( GO ) biological process slim terms , which are broad categorizations that facilitate pathway analysis . Ordering GO slim terms by the number of interactions in the learned GRN , we find that for target genes eight of the top ten GO slim terms are metabolism-related ( Figure 6D i ) ; in contrast , for regulatory TFs , five of the top ten GO slim terms are stress response related ( Figure 6D ii ) . Analysis of single cell expression in asynchronous cultures allows detection of cell cycle regulated relationships . The learned global GRN contains 257 genes that are regulated both by nitrogen TFs and by cell cycle TFs ( Figure 7A ) . Many of these regulatory connections are novel; likely due to the fact that identifying interactions between metabolism and cell cycle are challenging in asynchronous cultures without single-cell techniques . Of these genes , 38 are annotated with the amino acid metabolic biological process GO term and 20 are annotated with the ion or transmembrane transport biological process GO term . Only 11 are annotated with the mitotic cell cycle biological process GO term , indicating that the majority of the interconnection between cell cycle and nitrogen response genes is due to regulation of metabolism-related genes by cell cycle TFs . We estimated the TFA for every TF in each cell , using the learned GRN and the single-cell expression matrix . The TFA of nitrogen responsive TFs is principally linked to growth condition as these TFs vary in activity between conditions ( Figure 7A ) , but are generally similar within condition ( Figure 7—figure supplement 1A ) . As expected , we find that cells grown in rich media ( YPD ) have low TFA for the NCR TFs GLN3 and the GAAC TF GCN4 . The TFA for these TFs increases substantially upon treatment with rapamycin . By contrast , the estimated TFAs of cell cycle TFs varies within condition ( Figure 7B ) ; and are concordant with cell cycle responsive gene expression ( Figure 7—figure supplement 1B–D ) .
Since the inception of single-cell RNA sequencing ( Tang et al . , 2009 ) , technological advances have resulted in the scale of datasets increasing from tens of cells to tens of thousands in a diversity of organisms . However , the number of cells recovered during scRNAseq in budding yeast has been comparatively limited in studies published to date ( Gasch et al . , 2017; Nadal-Ribelles et al . , 2019 ) . We present here the first report of droplet-based scRNAseq in this widely used model eukaryotic cell . Using a diverse library of transcriptionally barcoded gene deletion strains we were able to efficiently analyze the gene expression state of 38 , 255 cells using 11 experiments . In addition to facilitating multiplexed analysis of genotypes , transcriptional barcoding provides a facile means of identifying doublet cells within droplets thereby increasing the accuracy of single cell analysis . Consistent with our understanding of global gene expression variation first characterized in foundational studies of the transcriptome ( DeRisi et al . , 1997; Gasch et al . , 2000 ) , we find that environmental condition is the primary determinant of the gene expression state of individual yeast cells . However , we observe significant heterogeneity in individual cell gene expression within conditions . Much of this variation can be explained by the mitotic cell-cycle . It is important to note that we do not remove or suppress this cell-cycle driven variance . The cell cycle is itself driven by transcriptional regulators , and our goal is to build a network that integrates cell-cycle regulation with regulated responses to the environment . The ability to access the crosstalk between signalling pathways and the cell cycle program is a key advantage to performing single-cell sequencing in asynchronous cultures , which bypasses many of the limitations of synchronized bulk sequencing experiments . It is also important to note that in several stressful growth conditions , we see heterogeneous cellular responses; some cells appear to be proliferative , while other cells have downregulated translational machinery and upregulated stress response genes . This is an interesting outcome by itself , as it is further evidence of bet-hedging strategies ( Levy et al . , 2012 ) , and we expect that the presence of multiple distinct transcriptional states between cells in the same environmental condition is advantageous for network inference . Model performance , as measured by AUPR , can vary considerably when learning networks from any single growth condition ( Figure 5C ) . Cells in rich YPD media do not require many anabolic pathways to be active , and primarily express genes required for the cell-cycle , translation , and glycolysis; in contrast , cells in minimal MMD media must express these pathways plus many anabolic pathways to synthesize nitrogenous bases , cofactors and amino acids . We find that this increased transcriptional diversity results in better overall performance . Nonetheless , the largest performance gain comes from aggregating networks from cells in different conditions ( Figure 5D ) , which demonstrates a general advantage to learning GRNs from heterogeneous data . Deletion of specific transcription factors results in changes in single cell gene expression for some TFs in some conditions . However , genotypic effects are comparatively minor . We believe that this is due to multiple factors including functional redundancy between TFs , physiological adaptation to the genetic perturbation and the conditional specificity of TFs . It is likely that perturbations that are transiently induced , and result in increased TF activity ( McIsaac et al . , 2013 ) may be effective in eliciting detectable responses in gene expression , facilitating causal inference . The use of precise gene deletions does provide several advantages over the use of CRISPR/Cas9-based perturbations as engineered deletions are unambiguous whereas the efficiency of perturbation by CRISPR/Cas9 varies for different guide RNAs . Constructing GRNs from single cell gene expression data is a universal goal in all organisms . A yeast single-cell expression matrix has several beneficial properties for design and testing of gene regulatory network inference models as there exist high quality known interactions and TF binding motifs . The issues of data sparsity and low sampling rates are likely to be problems common in experiments in any organism using scRNAseq . We find that techniques that have been developed for normalization and imputation do not improve performance of the additive linear model-based inference of the Inferelator algorithm ( Figure 5 ) . However , there are significant opportunities for development of smoothing techniques that would enhance network inference , perhaps targeting latent biophysical parameters like transcription factor activity . It seems reasonable to assume that these biophysical parameters should be stable within the local neighborhood of samples , and the activity calculation that we have used is ill-conditioned and potentially unstable . This is of particular concern when working with undersampled single-cell data and we are actively addressing this issue . We find that the application of multitask learning is well suited to GRN reconstruction from scRNAseq data . Jointly learning multiple related tasks improves generalization accuracy , especially in scenarios in which datasets are undersampled ( Caruana , 1998 ) , and has the desirable side benefit of mitigating the need for complex batch-correction techniques that aim to address technical variation between experiments . Removing batch-effect technical noise from data without suppressing interbatch biological variability remains an unsolved problem , and therefore application of multitask learning approaches to network inference from single-cell data is likely to be generally applicable to integrating scRNAseq data from different cell types and conditions . Using our scRNAseq dataset , we reconstructed a global GRN with several novel regulatory relationships . Among the most novel of these interactions are those between cell-cycle associated TFs and targets and nitrogen TFs and target genes . The cell cycle and metabolism are , by necessity , interconnected , and the mechanism of rapamycin in arresting cell cycle through TOR is well-established ( Heitman et al . , 1991 ) . Several studies have identified metabolic cycling patterns which are believed to be driven by the cell cycle ( Burnetti et al . , 2016; Slavov and Botstein , 2011; Tu et al . , 2005 ) . Although regulatory connections between environmental sensors , metabolism , and the cell cycle have been previously reported , a comprehensive regulatory network does not exist , in large part because of the difficulty of experimentally perturbing cell cycle without confounding metabolic changes . Our study provides a valuable first step in identifying specific regulatory connections that were previously inaccessible , and which are necessary to create a complete map of the yeast regulome . Incorporation of additional information into the network inference process , including information about interactions between transcription factors such as functional redundancy and heterodimerization , would likely improve learning of the network . We note that several TFs have few learned targets reflecting the requirement for surveying conditions in which particular TFs are active . For example , STE12 and TEC1 are mating-related TFs that we expect to be entirely inactive in our diploid cells; MSN2 and YAP1 are stress-responsive TFs that respond to specific stimuli that were not tested in our study . Targeted analysis of the GRN with rationally designed genetic perturbations and environmental conditions will maximize the additional information that can be recovered from future experiments . Single-cell sequencing is a transformative method for systems biology . To date , scRNAseq has been widely applied to the problem of defining different cell types . However , the ability to simultaneously study the expression of hundreds of genotypes in different conditions , and sample the expression state of thousands of cells , presents a rich source of information for the purpose of GRN reconstruction . Our study implements this approach in budding yeast , the workhorse of systems biology , and establishes a generalizable framework for GRN reconstruction from scRNAseq data in any organism .
All yeast strains were generated from the prototrophic FY4 ( MATa ) or FY5 ( MATɑ ) background strains . Yeast were transformed using the standard lithium acetate transformation protocol ( Gietz and Schiestl , 2007 ) . E . coli were transformed using the standard chemically competent transformation protocol . Plasmid constructions were confirmed by sanger sequencing . Yeast genotypes , plasmid sequences , and oligonucleotide sequences are provided as Supplementary file 1-supplemental tables 1-3 . Media formulations are provided as Supplementary file 1-supplemental table 4 . All steps below used RNAse-free reagents . To analyze all growth conditions together , the raw single-cell count matrix was normalized using multiBatchNorm from the scater package in R ( McCarthy et al . , 2017 ) . In short , this calculates size factors that are used to scale cells from different environmental condition batches so that each batch is of approximately the same mean UMI count . Cells were then library size normalized within batches and the normalized data was log-transformed with log2 ( x+1 ) to give a transformed and normalized count matrix . | Organisms switch their genes on and off to adapt to changing environments . This takes place thanks to complex networks of regulators that control which genes are actively ‘read’ by the cell to create the RNA molecules that are needed at the time . Piecing together these networks is key to fully understand the inner workings of living organisms , and how to potentially modify or artificially create them . Single-cell RNA sequencing is a powerful new tool that can measure which genes are turned on ( or ‘expressed’ ) in an individual cell . Datasets with millions of gene expression profiles for individual cells now exist for organisms such as mice or humans . Yet , it is difficult to use these data to reconstruct networks of regulators; this is partly because scientists are not sure if the computational methods normally used to build these networks also work for single-cell RNA sequencing data . One way to check if this is the case is to use the methods on single-cell datasets from organisms where the networks of regulators are already known , and check whether the computational tools help to reach the same conclusion . Unfortunately , the regulatory networks in the organisms for which scientists have a lot of single-cell RNA sequencing data are still poorly known . There are living beings in which the networks are well characterised – such as yeast – but it has been difficult to do single-cell sequencing in them at the scale seen in other organisms . Jackson , Castro et al . first adapted a system for single-cell sequencing so that it would work in yeast . This generated a gene expression dataset of over 40 , 000 yeast cells . They then used a computational method ( called the Inferelator ) on these data to construct networks of regulators , and the results showed that the method performed well . This allowed Jackson , Castro et al . to start mapping how different networks connect , for example those that control the response to the environment and cell division . This is one of the benefits of single-cell RNA methods: cell division for example is not a process that can be examined at the level of a population , since the cells may all be at different life stages . In the future , the dataset will also be useful to scientists to benchmark a variety of single cell computational tools . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology"
] | 2020 | Gene regulatory network reconstruction using single-cell RNA sequencing of barcoded genotypes in diverse environments |
The ability to generalize previously learned information to novel situations is fundamental for adaptive behavior . However , too wide or too narrow generalization is linked to neuropsychiatric disorders . Previous research suggests that interactions between the dopaminergic system and the hippocampus may play a role in generalization , but whether and how the degree of generalization can be modulated via these pathways is currently unknown . Here , we addressed this question in humans using pharmacology , functional magnetic resonance imaging , and computational modeling . Blocking dopamine D2-receptors ( D2R ) altered generalization behavior as revealed by an increased kurtosis of the generalization gradient , and a decreased width of model-derived generalization parameters . Moreover , D2R-blockade modulated similarity-based responses in the hippocampus and decreased midbrain-hippocampal connectivity , which in turn correlated with individual differences in generalization . These results suggest that dopaminergic activity in the hippocampus may relate to the degree of generalization and highlight a potential target for treatment .
Generalization enables neural systems to apply stimulus-outcome associations that have been acquired for one particular stimulus to other , related stimuli . A key aspect of generalization is the degree to which learned associations are applied to novel stimuli: the width of generalization . Generalization relieves individuals from having to learn outcome predictions for every single stimulus from scratch before using them to guide behavior . However , generalizing too widely can be maladaptive because it leads to the indiscriminate approach ( or avoidance ) of stimuli that are unlikely to be associated with reward ( or punishment ) . Indeed , aberrant generalization is implicated in several neuropsychiatric diseases including schizophrenia , anxiety disorders , depression , and drug abuse ( Dunsmoor and Paz , 2015; Gotlib and Joormann , 2010; Lissek et al . , 2014b; Lucantonio et al . , 2015; Moustafa et al . , 2010; Shohamy et al . , 2010 ) . Insights into the neurobiological mechanisms controlling the width of generalization are therefore important for understanding adaptive behavior and its disruption in these conditions . Two basic forms of generalization can be distinguished based on what constitutes the relation among stimuli; associative and stimulus generalization . In the case of associative generalization , such as transitive inference and acquired equivalence , the associative relationship among stimuli determines similarity . This relationship can be established for example by sensory preconditioning ( e . g . , train stimuli A-B and B-C , test whether A comes to predict C ) or by a common associate ( e . g . train A-C and B-C , test whether A and B are associated ) . In contrast , in stimulus generalization the relationship among stimuli is based on the similarity along one or more perceptual dimensions ( frequency of sounds , color , line orientation , etc . ) . Most of what we know about stimulus generalization comes from behavioral experiments utilizing intradimensional stimulus discrimination ( Dunsmoor and LaBar , 2013; Ghirlanda and Enquist , 2003; Guttman and Kalish , 1956; Hanson , 1959 ) . In these paradigms , one stimulus ( e . g . one particular line orientation ) is paired with reward ( rewarded conditioned stimulus , CS+ ) , while a second stimulus ( e . g . a slightly different line orientation ) , which differs from the first in only one dimension , is paired with no reward ( unrewarded conditioned stimulus , CS− ) . Generalization is then tested using a range of stimuli that vary along the defining stimulus dimension ( e . g . line orientation ) . Although the test stimuli have never been paired with reward , animals and humans show robust generalization in that they respond to test stimuli that are similar to the CS+ . Several psychological models have been developed based on these experiments ( Ghirlanda and Enquist , 1998; Pearce , 1987; 1994; Shepard , 1987; McLaren et al . , 2012 ) , but the neurobiological mechanisms regulating the width of stimulus generalization have remained unknown . Early research suggested a role for dopamine in mediating generalization by demonstrating that blockade of dopamine receptors during generalization tests alters response gradients in rats and pigeons ( Lyons et al . , 1973a; 1973b; Terrace , 1963 ) . While the effects of dopamine have not been investigated in humans , evidence from neuroimaging suggests that dopaminoceptive and dopaminergic regions such as the striatum and the midbrain are involved in generalization ( Kahnt et al . , 2012; Wimmer et al . , 2012; Wimmer and Shohamy , 2012 ) . However , because standard neuroimaging relies on indirect measurements of neural responses , these studies were unable to inform questions about neurotransmitter-specific activity . A candidate brain region for a prominent role in stimulus generalization is the hippocampus , based on its involvement in learning , memory , and associative generalization ( Dickerson and Delgado , 2015; Eichenbaum , 2000; Frank et al . , 2006; Kumaran , 2012; Norman and O'Reilly , 2003; Squire and Wixted , 2011 ) . Specifically , the hippocampus is thought to contribute to associative generalization by representing higher-order relationships among different stimuli ( Eichenbaum , 1999; Howard et al . , 2005; Kumaran and McClelland , 2012 ) . However , human imaging and animal lesion studies suggest that the hippocampus may also be involved in similarity-based stimulus generalization not requiring inference ( Casasola et al . , 2007; Kahnt et al . , 2012; Lissek et al . , 2014a; Solomon and Moore , 1975 ) . In the current study , we directly examined the role of dopamine and hippocampal processing in regulating the width of stimulus generalization . For this purpose , we combined a visual intradimensional discrimination task ( Kahnt et al . , 2012 ) with pharmacologic dopamine D2-receptor ( D2R ) blockade ( Kahnt et al . , 2015 ) and functional magnetic resonance imaging ( fMRI ) . Our central hypothesis was that dopamine is involved in regulating the width of stimulus generalization via the modulation of similarity-based processing in the hippocampus . In order to test this hypothesis , we fit a computational model of stimulus generalization to the behavioral data , and examined the model parameters regulating the width of generalization . We then used model-based fMRI and functional connectivity analyses to identify the brain processes that potentially mediate the effects of dopamine on generalization . We predicted that D2R blockade during the test session would reduce the computational parameters governing the width of generalization , and thus lead to narrower generalization gradients . Moreover , we hypothesized that this reduction would be mirrored in hippocampal activity as well as in reduced functional coupling between the dopaminergic midbrain and the hippocampus .
The visual intradimensional discrimination task used the orientation of a Gabor patch as reward-relevant dimension ( Figure 1B ) . Specifically , one orientation ( 39° ) served as the CS+ and was paired with the delivery of 20 cents in 50% of trials , whereas a second orientation ( 51° ) served as the CS− and was paired with no reward in all trials ( Figure 1D ) . The association between orientation and reward was counterbalanced across subjects . In order to track the acquisition of stimulus-outcome associations , subjects were asked to make a discriminatory response after the stimulus was displayed , but before the outcome was shown . Subjects had to indicate whether the current stimulus was the rewarded stimulus , the non-rewarded stimulus , or whether they were unsure , by pressing buttons associated with + , - , or x , respectively . Subjects learned the discrimination within the first 50 trials and performed at high levels afterwards ( Figure 2A ) . Performance increased as a function of time ( two-way , group-by-time ANOVA , main effect of time , F ( 39 , 1716 ) = 18 . 98 , P < 0 . 001 ) , but did not differ between groups ( main effect of group , F ( 1 , 44 ) = 0 . 41 , P = 0 . 53; group-by-time interaction , F ( 39 , 1716 ) = 1 . 27 , P = 0 . 13; two-way , group-by-CS type ANOVA , main effect of group , F ( 1 , 44 ) = 0 . 35 , P = 0 . 56; group-by-CS type interaction , F ( 1 , 44 ) = 0 . 39 , P = 0 . 53 , Figure 2B ) . Moreover , learning-related activity in the ventral striatum ( Delgado , 2007 ) did not differ between groups ( see Figure 3 ) , suggesting that both groups were also comparable in terms of neural responses during discrimination training . This demonstrates that , as expected , both groups acquired stimulus-outcome associations and performed at comparable levels during training . Accordingly , effects of D2R blockade during the generalization test session on the next day can be compared independent of potential group differences in discrimination training . 10 . 7554/eLife . 12678 . 004Figure 2 . Behavioral responses during discrimination training and generalization test . ( A ) Percentage of correct responses during discrimination training across time ( bins of 5 trials ) . Subjects learned the stimulus-outcome associations within the first 50 trials , and maintained performance at a high level afterwards . Given that both groups received placebo during the training session , performance was not expected to , and indeed did not , differ . ( B ) Percentage of correct responses for CS+ and CS− is plotted for both groups separately . ( C ) Generalization gradients reflect the probability of a + response as a function of stimulus orientation during the test session . Responses reveal a peak-shift ( stronger responding on the side of the CS+ that is opposite to the CS− ) . Subjects in the amisulpride group ( PA , blue ) , showed a narrower generalization gradient with a higher peak compared to subjects receiving placebo ( PP , black ) . ( D ) Generalization gradients of a similarity-based generalization model with parameters estimated from subjects’ behavioral responses in both groups , separately . The model accurately reproduces the empirical generalization gradients and the differences between the groups . Error bars are SEM for N=25 ( PA ) and N=21 ( PP ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12678 . 00410 . 7554/eLife . 12678 . 005Figure 2—figure supplement 1 . Comparison of Gaussian and exponential similarity functions . ( A ) Behavioral responses , ( B ) P ( + ) responses derived from models with a Gaussian and ( C ) exponential similarity functions . The model with the Gaussian similarity function better captures the behavior and the effect of the dopaminergic manipulation than the model with the exponential similarity function . DOI: http://dx . doi . org/10 . 7554/eLife . 12678 . 00510 . 7554/eLife . 12678 . 006Figure 3 . Prediction error responses in the ventral striatum during discrimination training . ( A ) Regions in the ventral striatum ( VS , left , x = -15 , y = 8 , z = -16 , t = 6 . 97 , P < 0 . 001 , FWE whole brain corrected; right , x = 6 , y = 14 , z = -10 , t = 6 . 22 , P = 0 . 003 , FWE whole brain corrected ) in which activity is correlated with model-derived prediction errors ( PE ) during the training session across both groups . T-map from one-sample t-test ( across the two groups ) is thresholded at P < 0 . 05 , FWE whole brain corrected and overlaid on a T1-weigthed image averaged across subjects . ( B ) Bar plots depict parameter estimates for PE-related activity in the VS . Given that both groups received placebo during the training session , neural PE-responses were not expected to differ , and indeed did not differ between groups ( two-sample t-test , t = 0 . 58 , P = 0 . 56 ) . Error bars are SEM for N = 25 ( PA ) and N = 21 ( PP ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12678 . 006 On the second day , subjects in the PP group received placebo , whereas subjects in the PA group received 400 mg of the D2R antagonist amisulpride ( Rosenzweig et al . , 2002 ) . One hour later , subjects performed a generalization test session in extinction ( without feedback ) . Specifically , on each trial , one of 15 orientations ( 17°–73° , Figure 1D ) was presented , and subjects performed the same discrimination as during training ( Figure 1C; the original CS+ and CS− orientations were not shown during the test ) . Responses described a bell-shaped gradient around the CS+ and revealed a peak shift ( Derenne , 2010; Purtle , 1973; Wisniewski et al . , 2009 ) , such that subjects responded most frequently to an orientation that was never paired with reward ( Figure 2C ) . Specifically , average responding was stronger to orientations on the side of the CS+ that was farther away from the CS− ( paired t-test on responding to stimuli left vs . right of the CS+ , t = 2 . 99 , P = 0 . 006 ) . Such peak shifts are typical for intradimensional discrimination with one CS+ and one CS−; they are thought to result from the summation of excitatory and inhibitory gradients around the CS+ and CS− , respectively , and have been observed across many species and stimuli ( Ghirlanda and Enquist , 2003; Pearce et al . , 2008; Spence , 1937 ) . Direct group comparisons of the individual data points along the generalization gradient did not reveal any significant differences ( two-sample t-tests , all Ps > 0 . 29 ) . However , visual inspection of the gradients suggested that the amisulpride group had a narrower gradient than the placebo group , with enhanced responding at the peak of the curve , reduced responding at both flanks , and enhanced responding at the tail of the curve . These shape features are parsimoniously described by the 4th moment of probability distributions , namely , their kurtosis . Accordingly , a test for differences in the kurtosis of group-specific distributions ( Pearson type VII distribution , see Materials and methods ) revealed a significantly greater kurtosis in the amisulpride group compared to placebo ( PA: 6 . 73 , PP: 3 . 29; permutation test , P = 0 . 043 ) . This finding demonstrates that amisulpride narrowed the width and increased the peak of the behavioral generalization gradient , and suggests that D2R activity alters the neurocomputational processes that mechanistically control generalization behavior . To further address this possibility , and to identify the specific computational parameters that are affected by D2R blockade , we utilized a mathematical model for similarity-based stimulus generalization ( see Materials and methods and Figure 4 ) . The model assumes that the reward prediction of a given stimulus reflects the integrated excitatory and inhibitory associations of that stimulus , plus the excitatory and inhibitory associations of stimuli that are similar to it ( Pearce , 1987 ) . Critically , associations of stimuli that are similar to the currently presented stimulus have a stronger contribution than the associations of dissimilar stimuli . Because the shape of the function determining the similarity between the currently presented stimulus and other stimuli ( i . e . the generalization coefficient ) is of critical importance ( Ghirlanda and Enquist , 2003 ) , we directly compared the most commonly used models , i . e . one with Gaussian ( Kahnt et al . , 2012 ) and the other with exponential similarity functions ( Shepard , 1987 ) . While the exact shape of the similarity functions differs between models , for both models , the extent to which inhibitory and excitatory associations generalize to the current stimulus is controlled by the width of the similarity functions ( si and se ) ( Figure 4B ) . The larger si and se , the stronger the impact of the inhibitory and excitatory associations of dissimilar stimuli on the currently predicted reward , respectively , and thus the more generalization takes place . 10 . 7554/eLife . 12678 . 007Figure 4 . A computational model of similarity-based generalization . ( A ) Schematic of the model . When a stimulus orientation xk is presented , the predicted reward for this orientation Vk is computed by integrating the excitatory and inhibitory associations E and I of all stimuli j that are similar to k , ( including its own associations , j = k ) , weighted by the similarity between stimuli j and k . The similarity between j and k is determined by the excitatory and inhibitory generalization coefficients eSjk and iSjk , respectively , which are assumed to be Gaussian ( or exponential , not shown here ) . The width of the excitatory and inhibitory generalization coefficients , and thus the degree to which excitatory and inhibitory associations generalize from j to k is determined by the parameters se and si . The reward prediction Vk is used to generate approach behavior P ( + ) and to compute a reward prediction error δ , which in turn updates the excitatory and inhibitory associations of k . ( B ) Illustration of the effects of changes in the width of excitatory and inhibitory generalization coefficients on generalization gradients . DOI: http://dx . doi . org/10 . 7554/eLife . 12678 . 007 In order to assess the explanatory power of the Gaussian and exponential similarity functions , we directly compared their fit to the behavioral data . The free parameters of both models ( width of inhibitory generalization coefficient , si; width of excitatory generalization coefficient , se; slope , β; offset , a; learning rate , α ) were estimated for the entire group of subjects by maximizing the likelihood of subjects’ responses during the generalization test given the model ( see Materials and methods ) . Visual inspections suggested that the model with the Gaussian similarity function fitted the behavioral data better than the model with the exponential similarity function ( Figure 2—figure supplement 1 ) . This was confirmed by a formal model comparison using the Akaike information criterion ( AIC ) and Bayesian information criterion ( BIC ) ( Gaussian: AIC = 9586 . 4 , BIC = 9629 . 2; Exponential: AIC = 9691 . 6 , BIC = 9734 . 4 ) . We also compared the fit of the two models by comparing the regression coefficients from a logistic regression of the trial-by-trial responses on the modeled P ( + ) responses . Although both models predicted behavioral responses reliably ( Gaussian: t = 11 . 06 , P < 0 . 001; exponential: t = 11 . 38 , P < 0 . 001 ) , we found significantly higher regression coefficients for the Gaussian model ( paired t-test t = 5 . 34 , P < 0 . 001 ) . Taken together , this demonstrates that in our experiment , a Gaussian similarity function fits behavior better than an exponential similarity function . To determine the effects of dopamine on generalization , in a next step , the free parameters of the Gaussian model were estimated separately for each group ( see Materials and methods ) . As can be seen in Figure 2D , the model ( with group-specific parameters ) accurately reproduced responses in both groups , including the differences in the shape of the generalization gradients . Logistic regression coefficients were significantly different from zero ( PA: t = 7 . 76 , P < 0 . 001; PP: t = 7 . 95 , P < 0 . 001 ) , and did not differ between groups ( t = -0 . 21 , P = 0 . 84 ) , suggesting that responses in both groups were well described by the model . Notably , comparing the model parameters between groups ( Table 1 ) , revealed significant group differences in the width of the excitatory generalization coefficient , with a smaller coefficient in the amisulpride group compared to the placebo group ( permutation test , P = 0 . 035 ) . The width of the inhibitory coefficient was also smaller in the amisulpride group , but this difference was not significant ( P = 0 . 069 ) . Importantly , the learning rate during test did not differ between groups ( P = 0 . 21 ) , demonstrating that additional learning during the test session , which might have been altered by amisulpride , cannot account for the differences in generalization gradients . In order to obtain individual estimates of model parameters , we re-estimated the model using a leave-one-out procedure ( see Materials and methods ) . Comparing the resulting individual estimates between groups confirmed a significant difference in the width of the excitatory ( two-sample t-test , t = -2 . 03 , P = 0 . 024 ) , and to a lesser degree , the inhibitory generalization coefficient ( t = -1 . 59 , P = 0 . 059 , Figure 5 ) . These results suggest that D2R blockade modulates the computational processes that control the width of stimulus generalization , resulting in narrower generalization . 10 . 7554/eLife . 12678 . 008Figure 5 . Estimates of individual differences in model parameters . ( A-–E ) Each bar plot depicts the average difference between model parameters estimated based on the whole group ( N ) and a reduced group when one subject is left out ( N-1 ) . Values are averaged according to the group of the left out subject ( i . e . PA or PP ) . Positive values indicate that removing subjects from the group decreased the parameter estimates in the N-1 group . Amisulpride reduced parameters controlling the width of generalization ( width of inhibitory and excitatory generalization coefficients , A , B ) but had no effect on the slope ( C ) and offset of responding ( D ) , as well as the learning rate ( E ) . P-values are based on two-sample t-tests . Error bars are SEM for N = 25 ( PA ) and N = 21 ( PP ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12678 . 00810 . 7554/eLife . 12678 . 009Table 1 . Group-specific model parameters . DOI: http://dx . doi . org/10 . 7554/eLife . 12678 . 009GroupInhibitory coefficient siExcitatory coefficient seSlopeβOffsetaLearning rateαtestPA group N=2520 . 12117 . 5993 . 0930 . 3700 . 002PP group N=2130 . 58724 . 5843 . 1240 . 3610 . 006Difference PA-PP-10 . 465-6 . 985-0 . 0310 . 009-0 . 004*P-value ( PA-PP ) 0 . 0690 . 0350 . 4950 . 4620 . 214Note: *P-value is based on 10 . 000 permutations . In principle , the observed effects of amisulpride on the width of generalization could have resulted from a drug-induced improvement in perceptual orientation-discrimination . To control for such perceptual effects , subjects performed a challenging orientation discrimination task , once before the drug took effect and once after ( see Materials and methods and Figure 6 ) . Performance ( percentage correct ) on this task did not differ between groups ( two-way , time-by-group ANOVA , main effect of group , F ( 1 , 43 ) = 0 . 99 , P = 0 . 33 ) and there was no group-by-time interaction ( F ( 1 , 43 ) = 0 . 42 , P = 0 . 52 ) . Moreover , while perceptual discrimination performance was reliable across time ( correlation between pre- and post-drug performance , r = 0 . 52 , P < 0 . 001 ) , it was not related to the width of the estimated generalization coefficients ( all Ps > 0 . 38 ) . Taken together , this control analysis demonstrates that the effects of amisulpride on generalization cannot be explained by perceptual improvements in orientation discrimination per se . 10 . 7554/eLife . 12678 . 010Figure 6 . Amisulpride does not enhance perceptual discrimination performance . ( A ) Design and timing of the perceptual orientation discrimination task . Stimulus I was always 135° , whereas stimulus II was tilted by ± 0 . 2° , ± 0 . 4° , ± 0 . 9° , ± 1 . 9° , or ± 4° . Subjects had to indicate whether stimulus II is tilted counter-clockwise or clockwise relative to stimulus I . ( B ) Performance differed as a function of the difference in orientation between stimulus I and stimulus II ( 3-way ANOVA , F ( 4 , 172 ) = 74 . 46 , P < 0 . 001 ) , but not between groups ( F ( 1 , 43 ) = 0 . 01 , P = 0 . 93 ) or time ( pre vs . post drug , F ( 1 , 43 ) = 2 . 73 , P = 0 . 11 ) . Importantly , there was no significant group-by-time interaction ( F ( 1 , 43 ) = 0 . 07 , P = 0 . 80 ) . ( C ) Same data as in B , but collapsed across levels of orientation difference . Again , performance did not differ between groups ( two-way ANOVA , F ( 1 , 43 ) = 0 . 99 , P = 0 . 33 ) , time ( F ( 1 , 43 ) = 0 . 22 , P = 0 . 65 ) , and there was no significant group-by-time interaction ( F ( 1 , 43 ) = 0 . 42 , P = 0 . 52 ) . Error bars are SEM for N = 24 ( PA ) and N = 21 ( PP ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12678 . 010 Having established an effect of D2R blockade on the computational processes that govern stimulus generalization , we next examined the neural circuits that mediate these changes . We first identified brain regions involved in generalization of reward predictions during retrieval . As a proxy of generalized value , we focused on prediction error responses derived from our model , which reflect the extent to which reward predictions have generalized from the original CS+ and CS− to the current stimulus ( please note that because no outcomes were shown , prediction errors are perfectly but negatively correlated with expected value ) . Accordingly , to identify brain regions involved in similarity-based computations during generalization , we searched for regions in which fMRI activity correlated with generalized prediction errors at the time of the expected outcome . Based on previous empirical and modeling work linking hippocampal activity to the representation of relationships between stimuli and their predicted value ( Kumaran et al . , 2009; Kumaran and McClelland , 2012; Lee et al . , 2012; Lissek et al . , 2014a; Wimmer and Shohamy , 2012 ) , we expected fMRI signals in the hippocampus to positively correlate with generalized prediction errors . In line with this hypothesis , across the entire group ( one sample t-test ) we found significant correlations in the bilateral hippocampus ( extending into the parahippocampal gyrus , left , x = -30 , y = -22 , z = -16 , t = 6 . 53 , P = 0 . 001 , FWE whole brain corrected; right , x = 33 , y = -19 , z = -16 , t = 7 . 11 , P < 0 . 001 , FWE whole brain corrected , Figure 7A ) . Similar effects were found in the left amygdala ( x = -24 , y = -4 , z = -19 , t = 6 . 04 , P = 0 . 006 , FWE whole brain corrected ) and the bilateral middle temporal gyrus ( left , x = -48 , y = -73 , z = 14 , t = 6 . 41 , P = 0 . 002 , FWE whole brain corrected; right , x = 28 , y = -64 , z = 14 , t = 5 . 57 , P = 0 . 022 , FWE whole brain corrected ) . In addition , supporting recent work highlighting a role for medial ( mPFC ) and ventromedial PFC ( vmPFC ) in generalization ( Dunsmoor and Paz , 2015; Onat and Buchel , 2015 ) , at an uncorrected threshold of P < 0 . 001 , we found a cluster in the mPFC ( x = -3 , y = 56 , z = 8 , t = 4 . 03 ) extending into the vmPFC . 10 . 7554/eLife . 12678 . 011Figure 7 . Similarity-based prediction error responses in the hippocampus during generalization . ( A ) Across both groups , activity in the hippocampus is significantly correlated with model-derived prediction errors during generalization test . T-map from one-sample t-test is thresholded at P < 0 . 05 , FWE whole brain corrected and overlaid on a T1-weigthed image averaged across subjects . ( B ) Prediction error-related activity in the hippocampus is significantly reduced by D2R blockade ( two-sample t-test , t = -2 . 12 , P = 0 . 02 ) . Error bars are SEM for N = 25 ( PA ) and N = 21 ( PP ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12678 . 011 We next tested whether reduced behavioral generalization observed in the amisulpride group was paralleled by a decrease in generalization-related activity in the hippocampus . In line with this idea , we found significantly reduced activity in the hippocampus in the amisulpride relative to the placebo group ( two-sample t-test , t = -2 . 12 , P = 0 . 02 , Figure 7B ) . To examine whether these effects of dopamine on generalization-related activity are specific to the hippocampus , as a control , we tested for similar group differences in the amygdala , middle temporal cortex and mPFC . No significant group differences were observed in the amygdala ( P = 0 . 43 ) , the middle temporal gyrus ( P = 0 . 31 ) , or the medial PFC ( P = 0 . 98 ) . However , post-hoc analyses directly comparing the effect of the drug in the hippocampus to the drug effect in the other regions ( i . e . group-by-region interactions ) , demonstrated that while the effect of D2R blockade in the hippocampus was significantly stronger than in the mPFC ( P = 0 . 02 ) similar interactions involving the amygdala ( P = 0 . 097 ) and the middle temporal gyrus ( P = 0 . 127 ) did not reach significance . These data suggest specificity of the effects of D2R blockade on similarity-based processing in the hippocampus relative to the mPFC , but not necessarily relative to the amygdala and middle temporal lobe . In a next step , we examined the neural pathways on which DR2 blockade may mediate its effects on generalization . Given the anatomical origin of the dopaminergic projections to the hippocampus ( Swanson , 1982 ) , the relevance of hippocampal D2R for memory functions ( Takahashi et al . , 2008 ) , and the modulation of hippocampal processing reported above , we hypothesized that amisulpride would reduce the functional connectivity between the midbrain and the hippocampus . In line with this prediction , a functional connectivity analysis with the midbrain as a seed region ( Figure 8A , see Materials and methods ) revealed decreased midbrain connectivity in the right hippocampus ( x = 33 , y = -19 , z = -19 , t = 3 . 68 , P = 0 . 02 , FWE small volume corrected , Figure 8B , C ) and the left striatum ( x = -9 , y = 8 , z = -19 , t = 4 . 26 , P = 0 . 001 , FWE small volume corrected ) , in the amisulpride compared to the placebo group . This finding suggests that D2R blockade may modulate the functional connectivity between the midbrain and dopaminergic target regions such as the hippocampus and the striatum . To examine the specificity of these findings , we tested for similar drug-related effects on functional connectivity in the regions involved in similarity-based processing defined above . While connectivity estimates differed significantly between groups in middle temporal gyrus ( P = 0 . 01 ) , no drug effects were observed in the amygdala ( P = 0 . 296 ) and the mPFC ( P = 0 . 17 ) . Accordingly , for all regions except the middle temporal gyrus ( P = 0 . 13 ) , the corresponding drug-by-region interactions were significant ( all Ps < 0 . 05 ) , suggesting that amisulpride-related decreases in midbrain connectivity are relatively specific to the hippocampus . 10 . 7554/eLife . 12678 . 012Figure 8 . Amisulpride modulates midbrain-hippocampal connectivity . ( A ) Anatomical seed region in the midbrain is overlaid on a T2-weigthed image averaged across subjects . ( B ) Functional connectivity between the midbrain and the hippocampus during generalization test is significantly reduced in the amisulpride group ( P < 0 . 05 , FWE small volume corrected ) . For illustration , t-map from two-sample t-test ( PP > PA ) is thresholded at P < 0 . 001 , uncorrected and overlaid on a T1-weigthed image averaged across subjects . ( C ) For illustration , bar plot depicting average midbrain-hippocampus connectivity in both groups . ( D ) Strength of midbrain-hippocampal connectivity is positively correlated with the width of the model-estimated inhibitory generalization coefficient ( r = 0 . 47 , P = 0 . 01 ) in the amisulpride group . Error bars are SEM for N = 25 ( PA ) and N = 21 ( PP ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12678 . 012 Finally , we examined whether the relationship between D2R blockade and the width of generalization is associated with midbrain-hippocampal connectivity . Specifically , we tested the correlation between midbrain-hippocampal connectivity and the estimated width of generalization coefficients in the amisulpride group . This correlation was significant for the inhibitory generalization coefficient ( r = 0 . 47 , P = 0 . 01 , Figure 8D ) , but not for the excitatory coefficient ( r = 0 . 18 , P = 0 . 19 ) . We replicated these findings in our previous data set ( Kahnt et al . , 2012 ) , showing that midbrain-hippocampal connectivity was significantly correlated with the inhibitory generalization coefficient ( r = 0 . 44 , P = 0 . 036 ) , but not with the excitatory coefficient ( r = 0 . 15 , P = 0 . 49 ) . Interestingly , this relationship was not observed for midbrain-striatal connections ( P > 0 . 16 ) , and was significantly stronger for midbrain-hippocampal compared to midbrain-striatal connectivity ( Z = 2 . 31 , P = 0 . 011 ) , suggesting that the relation between dopamine and the width of generalization may arise primarily from midbrain-hippocampal connections . In summary , these findings suggest a link between dopamine-mediated midbrain-hippocampal coupling and the width of stimulus generalization in humans .
The degree to which individuals generalize outcome predictions across similar stimuli is important for adaptive behavior . Here , using D2R pharmacology , fMRI , and computational modeling , we demonstrate that D2R blockade results in narrower behavioral generalization gradients and changes in the computational parameters controlling the width of generalization . Moreover , D2R blockade altered similarity-based processing in the hippocampus and decreased the functional coupling between the midbrain and the hippocampus . This coupling was in turn related to the computational parameter controlling the width of generalization . Previous empirical and modeling work suggests that the hippocampus contributes to generalization by detecting the relationship between items in memory ( Eichenbaum , 1999; Howard et al . , 2005; Kumaran and McClelland , 2012 ) . However , this function was thought to only apply to associative forms of generalization involving higher-order relationships , whereas basic stimulus generalization involving perceptual similarity is suggested to be hippocampus-independent ( Kumaran , 2012 ) . Opposing this view , here we provide evidence that the hippocampus is involved in stimulus generalization . Accordingly , our findings suggest that hippocampal similarity computations are not restricted to detecting higher-order relationships among stimuli as previously thought , but can also exploit the perceptual similarity between stimuli to establish meaningful relationships . While the hippocampus may not be necessary for stimulus generalization per se , our data suggest that it facilitates generalization by allowing a flexible modulation of its width . Specifically , such flexibility could not be achieved if discrimination and generalization were entirely based on static , hippocampus-independent , stimulus-outcome associations . Our results contribute to an ongoing debate regarding the time point at which hippocampal processes support generalization . Two alternative accounts suggest that the relationship among stimuli is established by the hippocampus either during encoding via overlapping neural codes ( Eichenbaum , 1999; Howard et al . , 2005 ) , or at the time of retrieval by means of recurrent similarity computations that are based on separated neural codes ( Kumaran and McClelland , 2012 ) . The former proposal has received empirical support from studies showing that hippocampal activity during encoding is related to transfer performance at test ( Shohamy and Wagner , 2008; Wimmer and Shohamy , 2012 ) . In contrast , in line with the second account , our model assumes that reward predictions are generalized and integrated during test ( Pearce , 1987 ) . It should be noted though that this model is mathematically equivalent to a model in which generalization occurs at encoding ( Kahnt et al . , 2012 ) , and thus , the model alone does not provide evidence for either account . However , only if generalization occurs at retrieval can the width of generalization be modulated after encoding has occurred , and thus , our results are only compatible with a retrieval-based account of generalization . Specifically , because dopamine receptor blockade modulated generalization gradients during test , without affecting additional learning , our pharmacological manipulation provides evidence that stimulus generalization occurs – at least in part – at retrieval . Our data suggest that the functional connectivity between the dopaminergic midbrain and the hippocampus is related to the width of generalization . Specifically , we found that participants with D2R blockade showed decreased midbrain-hippocampal connectivity , which in turn correlated with the width of generalization . While the presence of dopamine receptors in the hippocampus is undisputed , it is worth keeping in mind that although direct dopaminergic innervation from the midbrain to the hippocampus is present ( Gasbarri et al . , 1994 ) and functionally relevant for memory stabilization ( McNamara et al . , 2014 ) , this pathway is not very strong ( Mingote et al . , 2015; Swanson , 1982 ) . Moreover , it is conceivable that some of the hippocampal dopamine is co-released from noradrenergic neurons , in which it may not have been completely metabolized . As such , our data is in line with the idea that reduced activity of hippocampal D2R decreases the extent to which reward associations generalize across stimuli . By extension , enhanced dopaminergic activity in the hippocampus may increase the width of generalization . More specifically , dopamine release in the hippocampus could increase the likelihood that ensemble patterns representing similar stimuli are activated , which would effectively lower the threshold for similarity detection , and , by broadening the range of stimuli for which reward predictions are taken into account , facilitate generalization . Conversely , reduced levels of dopamine would decrease the likelihood of ensemble pattern activation , increase the threshold for similarity detection , and , by enhancing fine-tuned discrimination , reduce generalization . This provides a simple neurobiological mechanism by which dopamine may flexibly adjust generalization during retrieval . Such flexibility is highly adaptive as it allows for different levels of generalization based on the state of the organism or the environment . We speculate that tonic dopamine levels , similar to their enabling effects on movements ( Schultz , 2007 ) , play an enabling role in generalization . For instance , in situations where dopamine transmission is high , such as in novel environments ( Ihalainen et al . , 1999; Lisman and Grace , 2005 ) , elevated levels of dopamine may not only support memory formation ( Shohamy and Adcock , 2010 ) , but also broaden generalization and thereby facilitate exploratory behavior . Of note , although D2R blockade reduced generalization coefficients for both excitatory and inhibitory associations ( albeit less reliably ) , midbrain-hippocampal connectivity was only correlated with the width of the inhibitory coefficient . This raises the possibility that only the effects of D2R blockade on the inhibitory coefficient are mediated via a modulation of midbrain-hippocampal connectivity , whereas the effects on the excitatory coefficient are mediated via a different , yet to be explored , mechanism . The dissociation between inhibitory and excitatory generalization coefficients is in line with the interpretation of previous results suggesting that administration of chlorpromazine , a nonspecific dopamine receptor antagonist , specifically reduces the strength of inhibitory associations in a dose-dependent manner ( Lyons et al . , 1973b; Terrace , 1963 ) . Moreover , it adds to previous findings indicating that generalization involving appetitive and aversive outcomes may involve different mechanisms ( Schechtman et al . , 2010 ) . By revealing the effects of dopamine on generalization , our current results substantially extend those of our previous study ( Kahnt et al . , 2012 ) . Specifically , the current experiment suggests an association between dopamine and the width of generalization in the hippocampus , and dissociates effects of generalization during retrieval vs . encoding , which could not be achieved with the previous design . However , whereas here we find that prediction errors correlate primarily with activity in the hippocampus , the previous study identified prediction error related activity primarily in the ventral striatum . Also , in the previous study functional connectivity between the striatum and the hippocampus was related to the modeled generalization coefficients , whereas our current results suggest a facilitating role of midbrain-hippocampal connections . Several notable differences in the design of both experiments might explain these discrepancies . Most importantly , the previous design consisted of multiple alternating training and testing blocks , all conducted within one session on a single day , whereas in the current study , training and test did not alternate but were conducted in distinct sessions that were separated by a 24 hr delay . Moreover , in the 2012 study , CS-outcome associations were deterministic ( 100% contingency ) , whereas the current experiment involved a 50% reinforcement schedule in order to slow down extinction during test . These differences could have shifted the primary focus of generalization-related processing from the striatum to the hippocampus . In conclusion , here we propose a neurobiological mechanism for the control of stimulus generalization , in which midbrain dopamine changes similarity computations in the hippocampus , resulting in altered generalization gradients . As such , our results demonstrate that the width of stimulus generalization is not hard-wired but flexible , and can change under pharmacological interventions . Accordingly , our results have important clinical implications for a number of neuropsychiatric disorders in which generalization is disrupted . Specifically , aberrant generalization is implicated in depression , anxiety , and schizophrenia ( Buss and Daniell , 1967; Gotlib and Joormann , 2010; Lissek et al . , 2014b; Shohamy et al . , 2010 ) , and our findings indicate that blocking D2R activity may provide a potential treatment of overgeneralization in these disorders .
Subjects were assigned to one of three groups in a double-blind and pseudo-random fashion: amisulpride-placebo ( AP ) , placebo-amisulpride ( PA ) , and placebo-placebo ( PP ) . To avoid confounds related to the effects of dopamine on neural processing during discrimination training , only subjects in the PA and PP group are considered in this manuscript . A total of seven subjects was excluded because they either failed to acquire stimulus-outcome associations during the discrimination training on day one ( performance <60% , 4 subjects ) or because they failed to follow instructions on day two ( 3 subjects ) . Subjects in both groups received placebo on the first day of the experiment , whereas on the second day subjects in the PP and PA group received placebo and 400 mg of the D2R blocker amisulpride , respectively ( Figure 1A ) . All subjects were healthy and had normal or corrected-to-normal vision . Groups did not differ significantly in number ( PA: N = 25 , PP: N=21 , chi-square = 0 . 348 , P = 0 . 56 ) , average age , ( PA: 22 . 72 years ± 2 . 17 SD , PP: 22 . 19 years ± 1 . 83 SD; t = 0 . 88 , P = 0 . 38 ) , and average weight ( PA: 75 . 04 kg ± 7 . 89 SD , PP: 75 . 48 kg ± 10 . 92 SD; t = -0 . 03 , P = 0 . 98 ) . Moreover , subjects were not aware of whether they received placebo or amisulpride on both days as assessed by a post experimental questionnaire ( day I , chi-square = 0 . 49 , P = 0 . 48 , day II: chi-square = 0 . 29 , P = 0 . 60 ) . The study was approved by the Cantonal Ethics Review Board of Zurich , and subjects provided informed consent to participate . On the first day , subjects were briefed about the details of the experiment , signed the consent form , and were administered a pill that was swallowed in front of the experimenter . To minimize and equalize absorption time across subjects , subjects were asked to not eat 6 hr before the experiment . One hour after taking the pill , subjects entered the MRI scanner to perform an intradimensional discrimination task . During the task , subjects learned the association between oriented Gabor patches ( CS+ and CS− ) and reward or no reward ( 0 and 20 cents , respectively ) . In each trial , subjects were presented with an oriented Gabor patch for 600 ms ( Figure 1B ) . Immediately after the stimulus , subjects had to indicate whether the currently displayed stimulus may lead to reward ( + ) or no reward ( - ) , or whether they did not know ( x ) by pressing a button with the index , middle , or ring finger of their right hand , corresponding to the signs ( +/-/x ) on a response mapping screen . The mapping between buttons ( fingers ) and +/-/x was randomized in each trial to dissociate signals related to motor preparation and execution from reward predictions and prediction errors . When subjects pressed a button , the brightness of the signs on the screen slightly decreased to indicate that a response has been made . The screen disappeared after 1500 ms ( maximum decision time ) and was replaced by an outcome screen ( 1000 ms ) indicating the amount of money they received ( 20 or 0 cents ) . When subjects failed to respond within 1500 ms , “‘too slow”’ was presented instead of the outcome . The CS+ was paired with reward and no reward in 50% of the trials , whereas the CS− was always paired with no reward . The outcome was independent of the correctness of the behavioral response , and the association between stimulus orientation ( 39° and 51° ) and reward was counterbalanced across subjects . The training phase consisted of 100 repetitions of CS+ and CS− trials , in pseudorandom order . Trials were separated by a variable interval ranging from 1 . 9 to 9 . 9 s ( 1 . 9 s fix , plus a variable interval drawn from an exponential distribution , truncated at 8 s ) . On the second day of the experiment , subjects performed the generalization test in extinction , one hour after taking the pill containing either placebo ( PP ) or amisulpride ( PA ) . In each trial , subjects saw one of 15 orientations ( 17° , 21° , 25° , 29° , 33° , 37° , 41° , 45° , 49° , 53° , 57° , 61° , 65° , 69° , and 73°; Figure 1D ) for 600 ms ( Figure 1C ) . The original CS+ and CS− were not shown during the test . Each orientation was presented 14 times in pseudorandom order resulting in a total of 210 trials . Directly after presentation of the stimulus , subjects had to make the same discrimination response as during training ( see above ) . Importantly , the test was performed in extinction , i . e . no outcomes were shown for all orientations . This design ensured that subjects made motor responses to all stimuli and thus allowed us to observe reward prediction error responses to all orientations independent of potential confounds attributable to reward feedback , different visual stimulation , and different cognitive or motor demands . Trials were separated by a variable interval ranging from 2 . 9 to 10 . 9 s ( 2 . 9 s fix , plus a variable interval drawn from an exponential distribution , truncated at 8 s ) . In order to control for potential effects of D2R blockade on perceptual performance per se , subjects performed an orientation discrimination task . In each trial , two oriented Gabor patches were presented for 200 ms each , separated by a blank screen of 300 ms . The first stimulus had an orientation of 135° ( i . e . orthogonal to the stimuli used in the main experiment ) , whereas the second stimulus was tilted -4° , -1 . 9° , -0 . 9° , -0 . 4° , -0 . 2° , 0 . 2° , 0 . 4° , 0 . 9° , 1 . 9° , or 4° relative to the first stimulus . Thus , the task consisted of 10 trial types , reflecting 5 levels of difficulty ( i . e . absolute difference between first and second orientation ) . Subjects had to indicate as fast and accurately as possible whether the second orientation was tilted counterclockwise or clockwise relative to the first orientation by pressing a button . No feedback was provided to minimize feedback-based perceptual learning ( Kahnt et al . , 2011 ) . Each trial type was repeated 7 times , resulting in a total of 70 trials . Discrimination performance was computed by averaging accuracy across trials and difficulty levels . The task was administered twice on each day: once immediately after administration of the pill to obtain a baseline measure in the absence of drug effects , and once after the scanning session ( 2 hr after drug/placebo administration ) to measure discrimination performance under the influence of the drug ( amisulpride plasma levels have a first peak after ~1 hr ( Rosenzweig et al . , 2002 ) ) . Due to technical problems , discrimination accuracy data from one subject were not saved . In order to compare the overall shape of the behavioral generalization gradients , we estimated the 4th moment of the distributions underlying the behavioral gradients ( because the behavioral gradients were bounded ( 17-–73 degrees ) and not centered on 45 degrees , direct numerical estimation of the kurtosis was not possible ) . For this , we fitted a Pearson type VII distribution to the behavioral generalization gradient of each group and numerically computed the kurtosis of the group-specific distributions according to: kurt ( X ) =E[ ( X−μ ) 4]σ4 The kurtosis was then compared between groups , and statistical inference on the observed group difference was performed using a permutation test . We designed a similarity-based generalization model ( Figure 4A ) , based on previous computational approaches to stimulus generalization ( Ghirlanda and Enquist , 1998; Kahnt et al . , 2012; Pearce , 1987 ) . The model assumes that each orientation k holds inhibitory and excitatory associations , Ik and Ek that change with learning . In a given trial t , the net associative strength V ( or predicted reward ) of the currently presented stimulus k equals the aggregated excitatory and inhibitory associative strengths of all stimuli j that that are generalized to stimulus k ( including j = k ) : Vt=∑j Et , j·eSjk−It , j·iSjk The degree to which associations generalize from stimulus j to k is determined by the inhibitory and excitatory generalization coefficients iSjk and eSjk ( Figure 4B ) , respectively , which vary continuously between 0 and 1 ( for j = k ) . These coefficients can take the form of Gaussians or exponential functions , and their widths ( i . e . the width of generalization ) are controlled by the parameters si and se , respectively: iSjk=exp− ( xj−xk ) 22·si2 and eSjk=exp− ( xj−xk ) 22·se2 Where xj and xk are the orientations ( in degrees ) of stimuli j and k , respectively . The corresponding exponential similarity functions are given by: iSjk=exp−|xj-xk|2·si2 and eSjk=exp−|xj-xk|2·se2 The excitatory and inhibitory strengths , E and I , are updated on every trial . Specifically , when the outcome R ( 1 or 0 for reward and no reward , respectively ) is experienced , a prediction error δ is generated according to: δt=R−Vt Because the prediction error is based on generalized reward predictions , it directly reflects the extent to which excitatory and inhibitory associations generalize to the currently presented stimulus . The prediction error is used to update the inhibitory and excitatory associative strengths according to: Et+1 , k=Et , k+α·δt if δt>0 and It+1 , k=It , k−α·δt if δt<0 where α is the learning rate . To account for different learning rates during discrimination training with feedback and generalization test without feedback , separate learning rates were allowed during training and test ( αtrain , and αtest ) . The probability of making an approach response on a given trial P ( + ) t is given by the net associative strength Vt , passed through a biasing sigmoid function ( softmax ) , which is controlled by its slope β and offset a . P ( + ) t=11+exp−β· ( Vt−a ) The free parameters of the model ( si , se , β , a , αtrain , αtest ) were estimated by maximizing the log likelihood estimate ( LLE ) of subjects’ responses during test given the model LLE=∑tlogP ( +|θ ) t , * where P ( +|θ ) t* is the probability of the model for making the same response as the subject in trial t . In order to compare models with Gaussian and exponential similarity functions , we estimated the free parameters of both models by combining the LLE from subjects in both groups , and compared the aggregate LLE from the best fitting parameter sets using AIC and BIC . In order to generate reliable parameter estimates for each group separately , the same fitting procedure was performed for each group ( PA and PP ) separately by combining LLE across all subjects within a group when evaluating the model . This yielded two sets of estimated model parameters ( Table 1 ) . In order to reduce the number of free parameters , and because both groups received placebo during training , the training learning rate ( αtrain ) was estimated based on data from all subjects , and then fixed to that value when estimating group-wise model parameters . We tested the statistical significance of the observed differences in the group-wise model parameters using a permutation test . Specifically , we randomly assigned all subjects into one of the two groups , estimated a set of parameters for each group , and computed the difference between the estimated model parameters between groups . This was repeated 10 , 000 times , resulting in a distribution of group-wise parameter differences that should be expected by chance ( i . e . if group assignments were random ) . This random distribution was then used to generate P-values for the empirically observed group-wise parameter differences . To obtain an individual difference measure of model parameters , and as an alternative way to perform inference on the model parameters , we used a leave-one-out estimation procedure . For this , we first fitted the model using the data from all but one subject ( N-1 ) . The resulting model parameters were then subtracted from the parameters obtained when using all subjects ( N ) . This difference ( parameter based on N-1 subjects – parameter based on N subjects ) is proportional to the relative contribution of the left-out subject to the entire group . For instance , if the left-out subject has a large “‘true”’ parameter , leaving this subject out when estimating the model will reduce the parameter obtained in the reduced sample ( N-1 ) relative to the parameter obtained from the entire sample ( N ) . In other words , this procedure generates estimates that reflect the individual differences in model parameters . Functional imaging was performed on a Philips Achieva 3 T whole-body scanner equipped with an eight-channel head coil . During the training and test sessions , 675 and 711 T2*-weighted whole-brain EPI images with 37 transversal slices were acquired with a repetition time ( TR ) of 2000 ms . Imaging parameters were as follows: slice thickness , 3 mm; in-plane resolution , 2 . 75 x 2 . 75 mm; echo time ( TE ) , 30 ms; flip angle , 90° . For anatomical reference and identification of the dopaminergic midbrain , T1- and T2-weighted high-resolution ( 1 x 1 x 1 mm ) anatomical images were acquired using the following imaging parameters . T1-weigthed: matrix size , 256 x 256; field of view , 256; 181 slices; flip angle , 8°; TR = 8 . 2 ms; TE = 3 . 8 ms . T2-weighted: matrix size , 256 x 256; field of view , 256; 181 slices; flip angle , 8°; TR = 2500 ms; TE = 248 ms . Preprocessing of functional images was performed using SPM12 and consisted of slice-time correction , realignment , coregistration of anatomical ( T1-weighted ) and functional images , spatial normalization to the standard template of the Montreal Neurological Institute ( MNI ) by estimating normalization parameters based on the T1-weighted image , and spatial smoothing using a Gaussian kernel of 8 mm FWHM . To identify brain regions in which activity correlates with prediction errors derived from the similarity-based generalization model , we used a general linear model ( GLM ) with parametric modulators ( Buchel et al . , 1998 ) that included the following regressors: ( 1 ) onset of actual ( training data ) or expected ( test data ) time of outcome ( offset of the response mapping screen ) , ( 2 ) a parametric modulator of stimulus orientation ( z-standardized ) , and ( 3 ) a parametric modulator of prediction errors derived from the model with group-wise parameters ( z-standardized ) . All regressors were convolved with a canonical hemodynamic response function ( HRF ) and together with the head movement parameters from the realignment procedure regressed against the BOLD signal in each voxel . Independent GLMs were estimated for the training and test session . Voxel-wise one-sample t-tests were applied to the resulting parameter estimates of the prediction error regressor . To test for global effects of amisulpride on blood flow , and thereby BOLD response , we tested whether cue-evoked activity in visual cortex differed between groups . For this , we set up a GLM including one regressor for the onset of the visual cue ( HRF convolved ) and the six head movement parameters . Cue-related activity in an anatomical mask of the calcarine sulcus ( AAL ) did not differ between groups ( t = -0 . 80 , P = 0 . 42 ) , suggesting that amisulpride did not unspecifically affect the BOLD response . This is in line with previous studies that found no differences in visually evoked activity between amisulpride and placebo ( Jocham et al . , 2011 ) . We examined dopamine-related differences in the functional connectivity between the midbrain and the hippocampus by using a variant of the psycho-physiological interaction ( PPI ) model ( McLaren et al . , 2012 ) . For each subject , the average time course was extracted from voxels in a 50% probabilistic mask of the substantia nigra ( Murty et al . , 2014 ) , and together with the six head movement parameters regressed against the time course in each voxel . The parameter estimate of the midbrain-seed regressor reflects the correlation between activity in the midbrain and activity in every other voxel in the brain . To identify regions where connectivity differed depending on D2R blockade , the connectivity maps were compared between groups using a two-sample t test . To identify significant voxels in the fMRI analysis ( prediction error during test or training ) , we used one-sample t-tests and a threshold of P < 0 . 05 , FWE whole brain corrected , in combination with a cluster extent threshold of k>10 . To test for group differences , parameter estimates were extracted from significant voxels in the striatum ( training ) and hippocampus ( test ) , and compared between groups using two-sample t-tests at a statistical threshold of P < 0 . 05 . Significant group differences in functional connectivity were identified using a threshold of P < 0 . 05 , FWE small volume corrected for a functional region of interest in the hippocampus that was identified in the independent prediction error contrast during test in the entire group of subjects ( P < 0 . 05 , FWE-corrected ) . A priori comparisons using t-tests and correlations with directed hypotheses are tested one-tailed . Although amisulpride is generally known to not alter RT ( Jocham et al . , 2011; Rosenzweig et al . , 2002 ) , we found a trend-level group difference in RT ( t = 1 . 83 , P = 0 . 074 ) , with slower responding in amisulpride ( mean RT ± SEM = 816 ± 16 . 6 ) compared to placebo ( mean RT ± SEM = 770 ± 18 . 5 ) . Importantly , all group comparisons remained significant when RT was included as a covariate in the statistical models , suggesting that non-significant differences in RT did not confound the imaging results . | In the 1920s , two psychologists taught a young child known as ‘Little Albert’ to fear a white rat . They did so by striking a metal bar with a hammer whenever the rat was present . After experiencing the rat and noise together on multiple occasions , Little Albert eventually began to cry whenever the rat appeared . However , he also showed a similar response to a number of other white furry objects , including a rabbit and even a fur coat . By applying knowledge about a familiar object to other similar stimuli , humans and other animals avoid having to learn about each and every stimulus from scratch . However , this stimulus generalization is only effective if it occurs to the correct degree: under- or over-generalization ( as shown by Little Albert ) can lead to behaviors that are less than optimal . Nerve cells use molecules called neurotransmitters to communicate with each other . For example , one nerve cell might release a neurotransmitter called dopamine , which can be detected on the surface of another nerve cell by a protein called the dopamine D2 receptor . Research suggests that dopamine is involved in stimulus generalization in rat and pigeons , but the effects of dopamine have not been studied in humans . Kahnt and Tobler have now explored how the degree of stimulus generalization is determined in human adults . For the experiments , volunteers viewed images that contained sets of parallel lines , while lying inside a “functional MRI” brain scanner . The volunteers learned to associate lines with a particular orientation ( e . g . 39° from vertical ) with receiving a reward , and lines with another specific orientation with the absence of a reward . The next day , half the volunteers were given a drug that blocks the dopamine D2 receptor , while the other half received a placebo . All volunteers were then asked to classify lines of different orientations as either “rewarded” or “non-rewarded” . Both groups classified lines that were aligned similarly to the previously rewarded orientation as also rewarded . However , those given the D2 receptor blocker classified a narrower range of line orientations as rewarded than those who received the placebo . In other words , blocking the D2 receptors reduced stimulus generalization . It also reduced activity in a region of the brain called the hippocampus , and the extent to which this activity was connected to activity in another area called the midbrain . Taken together , these results suggest that activation of dopamine D2 receptors in the hippocampus may determine the extent to which we generalize between stimuli . Given that over- and under-generalization is a feature of psychiatric disorders such as depression and anxiety , manipulating D2 receptor activity could have therapeutic benefits in these patients . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Dopamine regulates stimulus generalization in the human hippocampus |
Transcriptional memory allows certain genes to respond to previously experienced signals more robustly . However , whether and how the key proinflammatory cytokine TNF-α mediates transcriptional memory are poorly understood . Using HEK293F cells as a model system , we report that sustained TNF-α stimulation induces transcriptional memory dependent on TET enzymes . The hypomethylated status of transcriptional regulatory regions can be inherited , facilitating NF-κB binding and more robust subsequent activation . A high initial methylation level and CpG density around κB sites are correlated with the functional potential of transcriptional memory modules . Interestingly , the CALCB gene , encoding the proven migraine therapeutic target CGRP , exhibits the best transcriptional memory . A neighboring primate-specific endogenous retrovirus stimulates more rapid , more strong , and at least 100-fold more sensitive CALCB induction in subsequent TNF-α stimulation . Our study reveals that TNF-α-mediated transcriptional memory is governed by active DNA demethylation and greatly sensitizes memory genes to much lower doses of inflammatory cues .
The epigenetic system has two important characteristic features: signal-induced plasticity and relatively stable mitotic inheritance . Signal-induced epigenetic plasticity allows changes in gene expression profiles and cell fate switching , whereas mitotic inheritance ensures the maintenance of gene expression profiles and cell fate . In theory , the combinatory effect of signal-induced plasticity and mitotic inheritance should convey epigenetic transcriptional memory , which allows certain genes to respond to previously experienced signals more robustly with faster kinetics and greater magnitude . The first examples of transcriptional memory date back to almost half a century ago ( Bergink et al . , 1973 ) . Current understanding of transcriptional memory events and underlying mechanisms has been nicely reviewed ( D'Urso and Brickner , 2014 ) . In Saccharomyces cerevisiae , the INO1 and GAL genes are the best studied examples of transcriptional memory genes . These genes can remember past inositol or galactose induction and exhibit accelerated induction kinetics during a second induction , and this transcriptional memory can last for a few cell divisions . These transcriptional memory events depend on a number of mechanisms , including gene looping , nuclear pore targeting , histone H2A . Z deposition , and H3K4 methylation ( Bheda et al . , 2020; D'Urso et al . , 2016; Kundu et al . , 2007; Light et al . , 2010; Tan-Wong et al . , 2009 ) . In mammals , interferon-induced transcriptional memory ( Gialitakis et al . , 2010; Kamada et al . , 2018; Light et al . , 2013 ) and inflammation-induced Aim2 gene transcriptional memory ( Naik et al . , 2017 ) also rely on chromatin structure and histone modifications . On the other hand , DNA demethylation has been observed to associate with transcriptional memory of the Tat gene during glucocorticoid induction ( Thomassin et al . , 2001 ) and the IL2 gene during T cell activation ( Murayama et al . , 2006 ) , but whether DNA demethylation plays a causal role in transcriptional memory regulation remains unclear . In summary , transcriptional memory enhances the future response to a previously experienced stimulus , providing a mechanism for cells to adapt to environmental changes . Several critical and intriguing questions regarding transcriptional memory need to be explored . Why do only a small fraction of genes induced by the same stimuli exhibit transcriptional memory ? What differentiates genes with or without transcriptional memory effects ? What are the characteristics of epigenetic modules governing transcriptional memory ? The proinflammatory cytokine TNF-α plays a vital role in the pathogenesis of chronic inflammatory diseases ( Reimold , 2003; Schett et al . , 2013 ) . TNF-α activates the transcription factor NF-κB , which is essential for inflammatory responses ( Duh et al . , 1989; Lowenthal et al . , 1989; Osborn et al . , 1989; Taniguchi and Karin , 2018 ) . Canonical NF-κB contains p65 and p50 and binds to genomic regions termed κB sites to activate target genes ( Hayden and Ghosh , 2008; Sen and Baltimore , 1986; Zhang et al . , 2017 ) . We previously reported that short-term TNF-α treatment activates the methylated IL32 gene in the absence of demethylation and that long-term TNF-α treatment induces DNA demethylation of the IL32 promoter and CpG island , which depends on p65 and TET enzymes ( Zhao et al . , 2019 ) . However , whether any NF-κB target gene exhibits transcriptional memory in response to TNF-α induction is unknown . By combining reporter assays and genome-wide analysis , we found that the proinflammatory cytokine TNF-α can induce transcriptional memory effects in several target genes . These genes are associated with p65 peaks induced by TNF-α treatment . Interestingly , these regions are heavily methylated in cells naïve to TNF-α signaling and tend to become demethylated in a p65 and TET enzyme-dependent manner during sustained TNF-α treatment . Although p65 can associate with its binding motif located within heavily methylated regions , the transcriptional activation capacity of p65 is potentiated by demethylation . This explains why a small fraction of p65 target genes display inflammatory transcriptional memory because NF-κB binding sites embedded within heavily methylated regions are more likely to serve as epigenetic memory modules . Our study reveals that the transcriptional memory established by sustained TNF-α stimulation is dependent on TET enzymes and significantly enhances the sensitivity to inflammatory signals .
We previously generated HEK293F-derived cells containing an integrated EGFP reporter gene under the control of a fully methylated CMV ( cytomegalovirus ) promoter ( Figure 1A ) , with which we performed a number of screens to identify factors and compounds regulating DNA methylation or methylation-mediated gene silencing ( Dong et al . , 2018; Du et al . , 2019; Li et al . , 2018a; Li et al . , 2018b ) . Because the CMV promoter contains several κB sites ( Figure 1A ) , we took advantage of the reporter cells and examined whether TNF-α treatment could activate the reporter gene and even induce transcriptional memory . We first treated the reporter cells naïve to TNF-α with 50 ng/mL TNF-α ( this concentration applies to all of the following experiments , unless indicated otherwise ) for various lengths of time ( Figure 1B ) , and the percentages of EGFP-positive cells were 0 . 96% ( 0 day ) , 43 . 7% ( 2 day ) , 56 . 3% ( 4 days ) , 69 . 0% ( 8 days ) , and 79 . 5% ( 12 days ) ( Figure 1C ) . Obviously , TNF-α treatment gradually potentiated the expression capacity of the reporter . Next , we allowed the above cells to rest for 10 days in TNF-α-free media and then measured EGFP expression . Cells that had experienced longer TNF-α treatment tended to exhibit higher baseline expression ( Figure 1—figure supplement 1A ) . Then , we applied a second TNF-α induction for 12 hr to test the transcriptional memory response ( Figure 1B ) . Apparently , cells that had experienced longer initial TNF-α treatment exhibited much more robust EGFP expression , and EGFP-positive cells were 13 . 8% ( 0 day pretreatment ) , 18 . 6% ( 2 days pretreatment ) , 25 . 6% ( 4 days pretreatment ) , 40 . 1% ( 8 days pretreatment ) , and 56 . 5% ( 12 days pretreatment ) of total cells ( Figure 1D ) . These data indicate that not all initial TNF-α treatments were equal in inducing inflammatory transcriptional memory , and sufficient time was needed during the initial induction phase to consolidate transcriptional memory . To directly compare the induction kinetics at the mRNA level , we performed RT-qPCR for the EGFP gene at various time points using cells naïve to TNF-α and cells that had experienced short-term ( 12 hr ) or long-term ( 12 days ) prior TNF-α treatment . Interestingly , we observed a clear transcriptional memory with faster and stronger induction only in cells that had experienced long-term prior TNF-α treatment ( Figure 1E ) . Together , the above experiments identified an example of inflammatory transcriptional memory and indicated that a consolidation phase during the initial treatment was required to establish such memory . We then exposed naïve cells to TNF-α treatment for various lengths of time and measured the CMV promoter methylation levels , giving the following results: 90 . 6% ( 0 day ) , 86 . 1% ( 2 days ) , 76 . 7% ( 4 days ) , 56 . 7% ( 8 days ) , and 41 . 7% ( 12 days ) ( Figure 1F ) . Importantly , the CMV promoter methylation levels of cells experiencing different durations of TNF-α treatment were still maintained after 10 days of TNF-α withdrawal ( Figure 1—figure supplement 1B ) . The above results indicate that gradual CMV promoter demethylation induced by TNF-α treatment correlates with increasingly stronger EGFP expression and the consolidation of transcriptional memory . However , whether demethylation is required for memory consolidation needs to be determined . We noticed that in cells with 12 days of TNF-α treatment , the CMV promoter methylation level remained 41 . 7% ( Figure 1F ) , and 20 . 5% of the cells remained EGFP-negative ( Figure 1C ) . We suspected that not all cells underwent similar levels of demethylation even after 12 days . Therefore , we sorted the 10% of cells with the highest EGFP signal ( Figure 1—figure supplement 1C ) and measured their CMV promoter methylation level , which was 6 . 0% ( Figure 1G ) , indicating that brighter cells had experienced more demethylation . To examine whether this is an active demethylation process , we treated TET1 knockout ( KO ) , TET2 KO , TET3 KO , and TET triple KO ( TET TKO ) cells ( Zhao et al . , 2019 ) with TNF-α for 12 days; sorted the 10% of cells with the highest EGFP signal; and measured their CMV promoter methylation levels . Obviously , TNF-α-induced CMV promoter demethylation mainly depended on TET2 , with some contribution of TET3 and little participation of TET1 ( Figure 1G ) . Importantly , TET2 KO , TET3 KO , and TET TKO cells failed to reach the highest level of EGFP expression after 12 days of TNF-α treatment ( Figure 1H ) , although their baseline expression ( 0 day ) and expression after 1 day of TNF-α treatment were quite comparable to those of the wild-type cells ( Figure 1—figure supplement 1D+E ) . Importantly , TET2 KO and TET TKO cells also failed to exhibit a wild-type level of transcriptional memory during a second round of TNF-α treatment ( Figure 1I , Figure 1—figure supplement 1F ) . These data demonstrate that TET enzyme–mediated DNA demethylation is required for the optimal activation and inflammatory transcriptional memory of the CMV reporter . From our experiences with the CMV reporter , we reasoned that the inflammatory memory genes likely possess the following features: 1 . they should be activated by the inflammatory signal TNF-α; 2 . they may display higher baseline expression in the absence of TNF-α signal after the initial induction; and 3 . most importantly , they should respond to subsequent TNF-α treatment better . With the above features in mind , to identify endogenous genes with inflammatory transcriptional memory , we performed RNA-seq experiments with the following samples: no induction , 12 hr TNF-α induction , 12-day TNF-α induction , 10-day recovery from a 12-day TNF-α induction , and second induction ( 12 days of TNF-α treatment followed by 10 days without TNF-α , then treated with TNF-α for 12 hr ) . We defined TNF-α-responsive genes as those that displayed greater than twofold upregulation upon 12 hr of TNF-α treatment with statistical significance ( p<0 . 01 ) ( Figure 2A ) . Among them , five endogenous genes together with the EGFP reporter were defined as inflammatory transcriptional memory genes , as they were expressed at least 1 . 3-fold higher in the second TNF-α induction than in the first TNF-α induction and exhibited an FPKM ( fragments per kilobase of transcript per million mapped fragments ) value greater than 5 ( Figure 2B ) . The EGFP reporter and five endogenous memory genes can be categorized into two groups: EGFP , CALCB , and PTGES displayed excellent memory effects , whereas the memory effects of CCL2 , TNF , and LHX2 were relatively moderate ( Figure 2B , Figure 2—figure supplement 1A ) . Indeed , EGFP , CALCB , and PTGES also displayed apparently elevated baseline expression in the absence of TNF-α signal after the initial induction ( Figure 2C ) . IL32 was upregulated in cells with 12-day TNF-α pretreatment ( Figure 2C , Figure 2—figure supplement 1A; Zhao et al . , 2019 ) , but IL32 did not exhibit a stronger second induction ( Figure 2B , Figure 2—figure supplement 1A ) . Among the five endogenous genes displaying inflammatory transcriptional memory , CALCB was of particular interest for two reasons . First , CALCB clearly exhibited the best memory effect ( Figure 2B , Figure 2—figure supplement 1B ) . Second , CALCB encodes β-CGRP . β-CGRP , and α-CGRP ( encoded by CALCA ) are two isoforms of CGRP ( calcitonin gene-related peptide ) . They are 37-amino acid–secreted neuropeptides and differ by three amino acids in humans ( Russell et al . , 2014; Russo , 2015 ) . The two CGRP isoforms share similar biological activities and are potent vasodilators ( Brain et al . , 1985; Russell et al . , 2014 ) . The overexpression of CGRP is an important cause of migraine ( Edvinsson et al . , 2018; Ho et al . , 2010; Pellesi et al . , 2017; Russell et al . , 2014; Russo , 2015; Tepper , 2018 ) , and several monoclonal antibodies ( eptinezumab , fremanezumab , galcanezumab , and erenumab ) against CGRP or CGRP receptors have been approved by the FDA as therapeutics for migraine patients . For a gene whose elevated expression is key to its disease-causing effect ( Ashina et al . , 2000; Fusayasu et al . , 2007; Goadsby et al . , 1990 ) , it is certainly important to know under which circumstances this gene can reach a very high level of expression . We noticed that CALCB gene expression reached very high levels in some conditions we analyzed , with an average FPKM value of 699 after 12 days of TNF-α treatment and 346 after 12 hr of TNF-α treatment in cells with 12 days of prior TNF-α exposure ( Figure 2D ) . We then measured CALCB induction kinetics in cells with or without prior TNF-α exposure , and clearly , CALCB exhibited robust induction with faster kinetics and stronger expression during subsequent induction ( Figure 2E ) . Importantly , the transcriptional memory effect of CALCB was maintained for at least 30 days in the absence of TNF-α ( Figure 2F ) . Therefore , we selected CALCB as a paradigm to further study the mechanisms governing inflammatory transcriptional memory . Importantly , the transcriptomes of the naïve cells and memory consolidated cells were extremely similar ( r = 0 . 999 ) , and the same was true for the transcriptomes of the first and second inductions ( r = 0 . 999 ) ( Figure 2—figure supplement 1C ) , indicating that the general physiology of the cells was not altered during memory consolidation , except for genes primed for memory response . The deletion of the RELA gene ( Zhao et al . , 2019 ) , which encodes p65 , severely impaired TNF-α-induced CALCB activation ( Figure 3A ) and the inflammatory transcriptional memory of CALCB ( Figure 3B ) . Since CALCA is barely transcribed ( Figure 3—figure supplement 1A ) , we were able to measure β-CGRP secreted in the media by performing sandwich ELISA using antibodies against CGRP . Likewise , the induction and memory effect of secreted β-CGRP at the protein level was abrogated in RELA KO cells ( Figure 3C ) . On the other hand , CALCB expression increased gradually during induction ( Figure 3D ) , and cells that had experienced longer initial TNF-α exposure displayed higher expression with subsequent induction ( Figure 3E ) . These features were highly similar to the behavior of the methylated CMV reporter ( Figure 1 ) and prompted us to ask whether CALCB transcriptional memory is also mediated by DNA demethylation . Indeed , TET2 KO and TET TKO cells displayed great defects in reaching the induction peak after 12 days of TNF-α treatment ( Figure 3F ) and failed to maintain robust transcriptional memory during the second round of TNF-α treatment ( Figure 3G ) . These results clearly indicate that DNA demethylation mediated by TET enzymes is critical for the inflammatory transcriptional memory of CALCB . Another gene with excellent inflammatory transcriptional memory is PTGES ( Figure 2B ) , which encodes prostaglandin E synthase that synthesizes a key inflammatory mediator prostaglandin E2 ( Gomez et al . , 2013; Jakobsson et al . , 1999 ) . Similar to CALCB , PTGES also reached higher expression during long-term TNF-α induction ( Figure 3—figure supplement 1B ) , and RELA KO disrupted its TNF-α-induced activation and transcriptional memory ( Figure 3—figure supplement 1C ) . To confirm the roles of TET enzymes in regulating other inflammatory transcriptional memory genes , we performed RNA-seq experiments with TET TKO cells under four conditions: no induction , 9 hr of TNF-α induction , 10 days of recovery from a 12-day TNF-α induction , and second induction ( 12 days of TNF-α treatment followed by 10 days without TNF-α , then induced with TNF-α for 9 hr ) and compared the results with wild-type cells . The elevated expression of all inflammatory transcriptional memory genes during the second TNF-α treatment was abrogated in TET TKO cells ( Figure 3H ) . Therefore , we conclude that inflammatory transcriptional memory is controlled by DNA demethylation mediated by TET enzymes . The CALCB promoter was unmethylated even in untreated cells ( Figure 3—figure supplement 1D ) . This prompted us to ask whether CALCB has a distal TNF-α-responsive element that governs its inflammatory transcriptional memory . To identify putative TNF-α-responsive elements , we performed ChIP-seq experiments to profile p65 and H3K27ac occupancy in the following samples: no induction , 12 hr of TNF-α induction , 10-day recovery from a 12-day TNF-α induction , and second induction ( 12 days of initial TNF-α induction followed by 10 days of recovery and then a second TNF-α induction for 12 hr ) . We identified 405 TNF-α-induced p65 peaks that were co-occupied by H3K27ac ( Figure 4A ) , the best known chromatin indicator of active transcriptional regulatory elements ( Creyghton et al . , 2010; Heintzman et al . , 2009; Zhang et al . , 2020 ) , and we defined them as TNF-α-responsive elements . The five endogenous inflammatory transcriptional memory genes were associated with 10 TNF-α-responsive elements ( Figure 4B , Figure 4—figure supplement 1A ) . For CALCB , two TNF-α-responsive elements were observed at the upstream distal regions ( −10 kb and −3 kb ) of CALCB , which we considered putative enhancers of CALCB , and they displayed TNF-α-induced p65 and H3K27ac occupancy ( Figure 4B ) . Most importantly , p65 and H3K27ac exhibited further elevation of TNF-α-responsive elements associated with CALCB and PTGES ( Figure 4B , Figure 4—figure supplement 1A ) , which were the two endogenous genes displaying the best inflammatory transcriptional memory ( Figure 2B ) . Interestingly , the two TNF-α-induced enhancers associated with the CALCB gene resided in a 7 kb endogenous retrovirus ( ERV ) spanning from −10 kb to −3 kb upstream of the transcription start site of CALCB ( Figure 4B ) . Since these two enhancers are located at the left and right end of an ERV , and each enhancer corresponds to a MER11B element , we termed them MER11B-left and MER11B-right . Both MER11B elements contain a κB site flanked by 7 CpG sites within regions spanning approximately 200 bp ( 209 bp for MER11B-left and 166 bp for MER11B-right; Figure 4—figure supplement 1B ) . We measured the DNA methylation levels of these two regions , and both regions were highly methylated in cells naïve to TNF-α signaling and became gradually demethylated during TNF-α treatment ( Figure 4C ) . The demethylated states were well maintained after culturing for 10 days in the absence of TNF-α ( Figure 4—figure supplement 1C ) and could even last for 30 days ( Figure 4—figure supplement 1D ) , indicating that long-term TNF-α exposure , in certain cases , may permanently alter the future signal response . Similar to CMV promoter demethylation , TNF-α-induced CALCB enhancer demethylation depended on p65 and the TET enzymes , with a major contribution from TET2 ( Figure 4D ) . To monitor DNA demethylation at p65 peaks associated with inflammatory transcriptional memory and all other p65 peaks , we performed enzymatic methyl-seq for genome-wide DNA methylation analysis with wild-type , RELA KO and TET TKO cells under no induction or 12 days of TNF-α induction . Indeed , DNA demethylation occurred at p65 peaks ( Figure 4E ) , and the demethylation effect was stronger in regions adjacent to the peak centers ( Figure 4F ) . The deletion of RELA impaired the above demethylation events , and the deletion of TET2 reduced the extent of demethylation ( Figure 4G ) . To differentiate the contributions of p65 binding and p65-dependent transcriptional activation in TNF-α-induced demethylation , we determined p65 binding peaks in TNF-α-treated cells , selected those that were heavily methylated in naïve cells , and then categorized them into three groups based on their eRNA expression levels . Interestingly , regardless of whether eRNA could be detected from these p65 binding peaks , demethylation occurred in a similar way ( Figure 4H ) . Therefore , we conclude that p65 binding per se , but not active transcription , is the main cause of TNF-α-induced demethylation , which is in line with an early observation that DNA-binding proteins can induce DNA demethylation without gene activation ( Stadler et al . , 2011 ) . We identified putative TNF-α-responsive enhancers for CALCB based on the occupancy of p65 and H3K27ac and established a correlation between NF-κB-induced demethylation at these enhancers and the transcriptional memory of CALCB . To provide direct evidence for the functional significance of these enhancers , we generated KO cells by deleting MER11B-left ( Figure 5—figure supplement 1A ) , MER11B-right ( Figure 5—figure supplement 1B ) , both MER11B elements ( Figure 5—figure supplement 2A ) , and the entire ERV ( Figure 5—figure supplement 2B ) . The activation of CALCB was partially impaired in MER11B-left and MER11B-right KO cells ( Figure 5A ) , suggesting that both MER11B elements were involved in CALCB activation mediated by TNF-α . On the other hand , cells lacking both MER11B elements or the entire ERV completely failed to respond to TNF-α signaling ( Figure 5A and B ) . Cells lacking the entire ERV did not display any inflammatory transcriptional memory of CALCB ( Figure 5C ) or secreted β-CGRP ( Figure 5D ) . Notably , although the CALCB gene is conserved between rodents and humans , this regulatory ERV only exists in some primates ( Figure 5E ) , suggesting that the inflammatory transcriptional memory of CALCB emerged during evolution after an ERV was inserted in the neighborhood and that this ERV was coopted as the epigenetic memory module . Although we identified only five endogenous inflammatory transcriptional memory genes , it is logical to hypothesize that what we discovered represents a general mechanism and that other genes may exhibit inflammatory transcriptional memory in other cell types and/or signal contexts . Therefore , it is highly important to determine the characteristic features of the inflammatory transcriptional memory modules associated with these five genes . The p65 peaks were linked to the nearest genes that are located within a 100 kb distance , and in total , 10 TNF-α-induced p65 peaks were associated with these genes . These peaks displayed several consistent features: they were highly methylated in cells naïve to TNF-α and exhibited obvious demethylation during memory consolidation , their demethylation was impaired upon the loss of RELA and TET2 ( Figure 6A ) , and they exhibited greater H3K27ac enrichment during the second TNF-α induction ( Figure 6B ) . These results imply that although TNF-α-induced p65 chromatin association can occur without DNA demethylation , the p65 chromatin association is favored in regions with less methylation . Indeed , when we grouped p65 peaks according to their initial DNA methylation levels and then plotted p65 occupancy in cells treated with the first dose of TNF-α for 12 hr , it was obvious that p65 could bind methylated regions but favored unmethylated regions ( Figure 6C ) . Thus , one important mechanism in consolidating inflammatory transcriptional memory is to change initially disfavored methylated p65-binding regions into favored unmethylated regions , to facilitate p65 binding and to achieve more effective transcriptional activation when a subsequent stimulus arrives . Active DNA demethylation mediated by the TET enzymes was required for consolidating transcriptional memory ( Figure 3H ) , but not all genes with TNF-α-induced demethylation at their regulatory regions exhibited transcriptional memory . For example , the IL32 promoter and CpG islands were demethylated after 12 days of TNF-α treatment ( Zhao et al . , 2019 ) , but IL32 did not exhibit inflammatory transcriptional memory ( Figure 2B ) . Why ? To address this critical question , we carefully compared p65 peaks in the IL32 promoter and other transcriptional memory genes , and we noticed that the adjacent region of the κB site in the IL32 promoter had fewer CpGs than the other p65 peaks associated with genes displaying a memory effect ( Figure 6D ) , whereas the three genes ( EGFP , CALCB , and PTGES ) with the best memory effects ( Figure 2B ) contained more CpGs in their associated p65 peaks ( Figure 6D ) . To test the effect of the number of CpGs , we chose highly methylated p65 peaks ( methylation level ≥50% ) , grouped them according to the CpG number within 250 bp flanking the κB sites , and then plotted p65 occupancy in cells treated with the first dose of TNF-α for 12 hr . Indeed , p65 occupancy was higher in regions with fewer CpGs even though the methylation levels were equally high ( Figure 6E ) . The above results suggest that the total DNA methylation ( combining CpG number and methylation level ) in the adjacent regions of κB sites is critical for their potential to serve as epigenetic modules for inflammatory transcriptional memory . Indeed , when we grouped p65 peaks according to their total methylation level and plotted p65 occupancy in cells treated with the first dose of TNF-α for 12 hr , regions with higher total methylation were clearly more disfavored for p65 binding ( Figure 6F ) . In retrospect , this is easy to understand because the efficiency of methylation-mediated repression depends not only on the methylation level but also on the regional methylated CpG density ( Curradi et al . , 2002; Hsieh , 1994 ) . We systematically analyzed the occupancy of p65 between the two TNF-α induction times ( Figure 6—figure supplement 1 ) and observed that occupancy-increased peaks were prone to high initial methylation and became demethylated after 12 days of TNF-α treatment ( Figure 6G ) . We also compared p65 occupancy , ATAC-seq signal , and H3K27ac signal according to various levels of total methylated CpGs , and the reduction in the number of methylated CpGs promoted p65 binding , chromatin opening , and H3K27ac ( Figure 6H–J ) . The above results suggest that initially methylated p65 peaks with the largest amount of total demethylation may function as potential inflammatory transcriptional memory modules . To functionally assess this hypothesis , we chose 515 methylated p65 peaks ( methylation level ≥50% ) out of 1143 p65 peaks and divided them into two groups: the highly demethylated p65 peaks comprising the top 25% of total demethylation during 12-day TNF-α treatment ( 129 peaks associated with 106 genes , including all five memory genes ) and the other methylated p65 peaks ( 386 peaks associated with 295 genes; Figure 6K ) . Then , we analyzed these peaks and their associated genes . First , upon second induction , the genes of the highly demethylated group displayed higher expression of their own enhancer RNAs ( Figure 6L ) . This suggests that these p65 peak regions possess the potential ability to serve as inflammatory epigenetic memory modules . Moreover , neighboring genes associated with the highly demethylated p65 peaks also exhibited higher expression during the second induction ( Figure 6M ) . In addition , the impact on neighboring genes was distance-sensitive , and genes with transcriptional start sites ( TSSs ) within 10 kb of the highly demethylated p65 peaks exhibited stronger transcriptional memory ( Figure 6N ) . Taken together , p65 peaks with high initial methylation and CpG density are more likely to serve as inflammatory transcriptional memory modules . Transcriptional memory has been shown to allow cells to respond to previously exposed stimuli more rapidly and more strongly ( Bergink et al . , 1973; D'Urso and Brickner , 2014; D'Urso et al . , 2016; Gialitakis et al . , 2010; Kamada et al . , 2018; Light et al . , 2010; Light et al . , 2013; Murayama et al . , 2006; Naik et al . , 2017; Tan-Wong et al . , 2009; Thomassin et al . , 2001 ) . In theory , transcriptional memory may also allow cells to respond to previously exposed stimuli in a much more sensitive manner , which has rarely been studied . We have already demonstrated that CALCB responded to subsequent TNF-α stimuli with faster kinetics and greater magnitude ( Figure 2E ) , but we have not tested whether CALCB can respond to a much weaker TNF-α signal after memory consolidation . Given the substantial demethylation at MER11B-left and MER11B-right after memory consolidation ( Figure 4C ) , which removes the epigenetic barrier for transcriptional activation and promotes p65 association ( Figure 6C ) , we speculated that MER11B-left and MER11B-right demethylation might allow the CALCB gene to respond to TNF-α signaling in a much more sensitive way . Therefore , we treated naïve cells and cells that had experienced a 12-day initial TNF-α treatment with various concentrations of TNF-α ( 0 . 4 ng/mL , 2 ng/mL , 10 ng/mL , and 50 ng/mL ) for 12 hr . Both cells displayed dose-dependent activation ( Figure 7A ) . Amazingly , cells that had experienced prior TNF-α treatment could achieve an approximately twofold better effect with a 125-fold lower TNF-α concentration ( 0 . 4 ng/mL ) than naïve cells treated with 50 ng/mL TNF-α ( Figure 7A ) . To the best of our knowledge , this is the first transcriptional memory effect reported to increase signal sensitivity by more than 100-fold . To confirm that cells could be sensitized by prior TNF-α treatment ( 50 ng/mL for 12 days ) , we performed RNA-seq experiments using the above cells with 0 . 4 ng/mL or 50 ng/mL TNF-α for a second induction of 12 hr and compared the results with those of naïve cells treated with 0 . 4 ng/mL or 50 ng/mL TNF-α for 12 hr . Indeed , CALCB and EGFP genes displayed even higher expression in cells with consolidated memory , despite having a 125-fold lower TNF-α signal ( Figure 7B ) . In addition , PEGES reached a 60% expression level in memory consolidated cells treated with a 125-fold lower TNF-α signal ( Figure 7B ) , indicating highly improved signal sensitivity .
Transcriptional memory has been studied for several decades , but the number of transcriptional memory genes remains relatively small . More examples need to be identified to reveal the significance of transcriptional memory . Two key difficulties exist in the identification of genes with transcriptional memory . First , they consist of only a small fraction of targeting genes responding to a particular signal of interest . Second , the duration of initial signal treatment can impact the process of memory consolidation . This study offers a pipeline to systematically identify genes with transcriptional memory by combining a defined signaling cue , a time course of the initial stimulus for memory consolidation , a dedicated transcriptome comparative analysis , and a functional validation with key regulator mutated cells . With this pipeline , we believe that many new transcriptional memory genes in response to various signaling cues will be quickly identified , which will certainly boost our understanding of the functional roles of transcriptional memory in many systems , such as the potential roles of transcriptional memory in immune cell memory . It is well known that transcription factor associations promote local demethylation ( Brandeis et al . , 1994; Costa et al . , 2013; de la Rica et al . , 2013; Dubois-Chevalier et al . , 2014; Fujiki et al . , 2013; Kirillov et al . , 1996; Macleod et al . , 1994; Perera et al . , 2015; Rampal et al . , 2014; Sérandour et al . , 2012; Silke et al . , 1995; Tsai et al . , 2014; Wang et al . , 2015; Xiong et al . , 2016 ) , and many transcription factors respond to various signaling cues . Our finding of inflammatory signal-induced DNA demethylation and its role in transcriptional memory consolidation is unlikely to be unique to inflammation signaling . We expect this to be an applicable principle for many other types of long-term transcriptional memory . In theory , long-term transcriptional memory should be governed by epigenetic mechanisms that can be maintained for a long time in quiescent cells and can be faithfully inherited for many cell divisions in proliferating cells . DNA methylation is likely the best candidate because it meets both criteria . Importantly , CpG methylation tends to display a bimodal distribution ( all or none ) within a region , and this kind of digital behavior makes it an ideal candidate to store digital information at regulatory regions with a simple status code such as ‘naïve/methylated/locked’ and ‘memory consolidated/unmethylated/unlocked’ ( Figure 7C ) . Whether this bimodal switch mediated by signal-induced demethylation also governs other types of transcriptional memory events is a highly interesting topic for future investigation and is relatively straightforward to test . The methyl-CpG binding proteins are important for DNA methylation-mediated transcriptional repression ( Bird and Wolffe , 1999 ) . Moreover , the number and density of methylated cytosines are crucial for the efficiency of gene silencing ( Curradi et al . , 2002; Hsieh , 1994 ) . In this study , high CpG density and initial methylation level around the κB sites were found to positively correlate with the functional potential of these regions to serve as inflammatory transcriptional memory modules ( Figure 6H–J ) , likely because DNA methylation is more effective in establishing a repressive environment in these regions , and sustained TNF-α induction is required to demethylate these regions to consolidate transcriptional memory . DNA methylation has a long-established role in gene silencing , but it is also known that gene activation can occur without demethylation ( Amedeo et al . , 2000; Dong et al . , 2018; Li et al . , 2018a ) . Some TFs contain CpG site ( s ) within their binding motifs and are sensitive to CpG methylation . Some TFs can bind and even favor binding motifs containing methylated CpG site ( s ) . Therefore , TFs are classified as methylation-sensitive and methylation-insensitive TFs , respectively ( Tate and Bird , 1993 ) . Some other TFs , including NF-κB , are also methylation-insensitive when their binding motifs do not contain any CpGs . Indeed , p65 can activate methylated target genes without prior demethylation ( Zhao et al . , 2019; Figures 3G and 4D ) . Therefore , the question is why transcriptional memory genes governed by a methylation-insensitive TF ( NF-κB in this case ) exhibit great expression differences in response to the same signal stimulus before and after DNA demethylation ( Figures 2E and 7A ) . We believe that two explanations likely exist , and they may function independently or in combination to regulate transcriptional memory . First , as we presented , p65 exhibits marked preference toward the target sequence within unmethylated regions ( Figure 6C ) , and demethylation lowers the transcriptional induction threshold and leads to higher gene induction with a 100-fold lower stimulus ( Figure 7A ) . Second , methylation-insensitive , signal-dependent TFs may function together with methylation-sensitive TFs to regulate transcriptional memory events . In this context , DNA demethylation induced by the signal-dependent , methylation-insensitive TFs may facilitate the binding of other methylation-sensitive TFs and activate gene transcription in a coordinated manner . The terms methylation-sensitive and methylation-insensitive TFs were originally defined by their interactions with naked DNA . We would like to emphasize that in the context of chromatin , even traditional methylation-insensitive TFs that contain no CpGs in their binding motifs greatly favor their binding sites embedded in unmethylated regions ( Figure 6C ) . κB sites embedded in regions with higher CpG density and initial methylation levels tended to display functional potential to serve as inflammatory transcriptional modules ( Figure 6H–J ) ; however , only a small fraction of these modules could actually drive the transcriptional memory events of adjacent genes ( Figures 2B and 6N ) . One possible explanation is that not all κB sites bound by p65 can drive the expression of neighboring genes . This again emphasizes the point that binding is not functioning , and transcriptional factor binding can only stimulate gene expression when enhancer-promoter connection is functionally wired . Migraine is a recurrent unilateral headache disorder ( Pellesi et al . , 2017 ) , and it is one of the five leading causes of YLDs ( years lived with disability ) in the world , contributing to more than 30 million YLDs ( GBD 2016 Disease and Injury Incidence and Prevalence Collaborators , 2017 ) . CGRP overexpression is a major cause of migraine ( Edvinsson et al . , 2018; Ho et al . , 2010; Pellesi et al . , 2017; Russell et al . , 2014; Russo , 2015; Tepper , 2018 ) . Neurogenic neuroinflammation is proposed to be involved in the increase in migraine frequency , which leads to chronic migraine ( Edvinsson et al . , 2019; Malhotra , 2016; Ramachandran , 2018 ) . In this study , CALCB expression reached an FPKM of 699 after 12 days of TNF-α treatment ( Figure 2D ) , and in memory consolidated cells , CALCB expression reached an FPKM of 225 after 12 hr with 125-fold lower TNF-α treatment ( Figure 7B ) . These discoveries support a role for neuroinflammation in the pathogenesis of migraine . The proinflammatory signal TNF-α induced gradual demethylation of CALCB enhancers within an ERV ( Figure 4C ) , and memory consolidated cells encountering a subsequent TNF-α stimulus exhibited a robust elevation of CALCB expression ( Figure 2E ) and more than 100-fold sensitivity to TNF-α induction ( Figure 7A and B ) . These features imply that inflammatory signals and memory consolidation likely play important pathogenic roles in migraine patients , especially in patients suffering from chronic migraine . Although these mechanisms remain highly speculative at this moment , this is a potentially interesting direction for clinical scientists studying migraine . Additionally , although the CALCB gene is conserved between humans and rodents , we would like to point out that the ERV associated with CALCB , which regulates its inflammatory response , is primate-specific ( Figure 5E ) . This suggests that it may be important to use nonhuman primate models to study whether the inflammation and inflammatory transcriptional memory of CALCB promote migraine pathogenesis .
All the HEK293F-derived cells are maintained in Dulbecco’s Modified Eagle Medium ( Gibco C11995500BT ) with 10% fetal bovine serum ( Biological Industries 04-010-1ACS 500 mL ) and 1× penicillin streptomycin solution ( Sangon Biotech , China E607011-0100 ) using a cell culture incubator at 37°C and 5% CO2 . The HEK293F cells stably inserted with a DNA methylation silenced CMV reporter ( Figure 1A ) and its TET1 KO , TET2 KO , TET3 KO , TET TKO , and RELA KO cells are previously established and described ( Li et al . , 2018b; Zhao et al . , 2019 ) . The MER11B-left KO , MER11B-right KO , MER11B-left+right KO , and ERV KO cells generated in this study are also derived from the HEK293F cells stably inserted with the DNA methylation silenced CMV reporter . The cell lines used in this study are free from mycoplasma contamination and authenticated: The HEK293F cells stably inserted with a DNA methylation silenced CMV reporter is established in Dr . Bing Zhu’s lab and widely used in our previous studies ( Dong et al . , 2018; Du et al . , 2019; Li et al . , 2018a; Li et al . , 2018b; Zhao et al . , 2019 ) ; the flow cytometry results ( Figure 1C ) can confirm the insertion of DNA methylation-silenced CMV reporter . The concentration of recombinant human TNF-α ( Peprotech , USA 300-01A ) for the treatment is 50 ng/mL unless otherwise stated . The schematic of TNF-α-mediated transcriptional memory experiment is shown in Figure 1B . All the cultured HEK293F-derived cells are passaged every 2 days . For the long-term TNF-α induction , the proinflammatory cytokine TNF-α is added into the new culture medium immediately after cell passage . For the CGRP release experiments , 3 × 105 cells were seeded per well in 6-well plates with 2 mL medium added and cultured for 2 days . Then the cell culture supernatants were collected to determine the levels of CGRP secretion . The Calcitonin Gene-Related Peptide ( human ) ELISA Kit ( Cayman Chemical , USA 589101 ) and EnSight Multimode Plate Reader ( PerkinElmer , USA ) were used to measure CGRP concentration in the cell culture supernatants following the manufacturer’s instructions . Cells are cultured and treated as described . Total RNA was extracted from cultured cells with TRIzol Reagent ( Invitrogen , USA 15596026 ) following the manufacturer’s instructions; 500 ng RNA was reverse-transcribed using HiScript II Q RT SuperMix for qPCR ( +gDNA wiper ) ( Vazyme , China R223-01 ) according to the manufacturer’s instructions . The synthesized cDNA was analyzed by quantitative real-time PCR using KAPA SYBR FAST Universal qPCR kit ( Kapa Biosystems KK4601 ) and run on 7500 Fast Real-Time PCR system ( Applied Biosystems , USA ) with validated qPCR primers ( Supplementary file 1 ) . The gene expression changes were normalized to GAPDH transcript as an internal standard . Briefly , cells were fixed with 1% formaldehyde solution ( Sigma , USA F1635 ) for 10 min at room temperature . The crosslinking reaction was stopped with glycine ( 0 . 125 M ) and nuclei were prepared in cell lysis buffer ( 20 mM Tris-HCl pH 8 . 0 , 85 mM KCl , 0 . 5% NP-40 ) . Resuspend cell nuclei in nuclei lysis buffer ( 50 mM Tris-HCl pH 8 . 0 , 10 mM EDTA-NaOH pH 8 . 0 , 1% SDS , 1× complete EDTA-free protease inhibitor cocktail [Roche , Switzerland 04693132001] ) . Chromatin was sonicated using Covaris M220 Focused-ultrasonicator to an average length of approximately 250 bp . Cell debris was cleared by centrifugation and the supernatant was diluted 10-fold with dilution buffer ( 1% Triton X-100 , 20 mM Tris-HCl pH 8 . 0 , 200 mM NaCl , 2 mM EDTA-NaOH pH 8 . 0 ) . Samples were incubated with anti-NF-κB p65 ( C-20 ) rabbit polyclonal antibody ( Santa Cruz Biotechnology , USA sc-372 ) overnight at 4°C . Antibody–chromatin complexes were pulled down with Protein A Dynabeads ( Invitrogen 10002D ) and were washed once with low salt wash buffer ( 0 . 1% SDS , 1% Triton X-100 , 2 mM EDTA-NaOH pH 8 . 0 , 20 mM Tris-HCl pH 8 . 0 , 150 mM NaCl ) , twice with high salt wash buffer ( 0 . 1% SDS , 1% Triton X-100 , 2 mM EDTA-NaOH pH 8 . 0 , 20 mM Tris-HCl pH 8 . 0 , 500 mM NaCl ) , once with LiCl wash buffer ( 0 . 25 M LiCl , 1% NP-40 , 1% deoxycholate , 1 mM EDTA-NaOH pH 8 . 0 , 10 mM Tris-HCl pH 8 . 0 ) , and twice with TE buffer ( 10 mM Tris-HCl pH 8 . 0 , 1 mM EDTA-NaOH pH 8 . 0 ) . Incubate the beads with elution buffer ( 100 mM NaHCO3 , 1% SDS ) to elute the complexes . Following cross-link reversal , RNase A and proteinase K treatment , phenol–chloroform extraction , and isopropanol precipitation , ChIP DNA samples were used to prepare libraries with Kapa hyper prep kit ( Kapa Biosystems KK8504 ) and NEBNext multiplex oligos for Illumina ( index primers set 1 ) ( NEB , USA E7335 ) according to the manufacturer’s protocol . Libraries were sequenced on an Illumina NovaSeq 6000 using 150 bp paired-end mode . The ATAC-seq libraries are prepared as previously described ( Corces et al . , 2017 ) with minor modifications . A total of 45 , 000 cells were collected and resuspended in 50 μL cold ATAC-Resuspension Buffer ( 10 mM Tris-HCl pH 7 . 4 , 10 mM NaCl and 3 mM MgCl2 in sterile water ) containing 0 . 1% NP40 , 0 . 1% Tween-20 , and 0 . 01% Digitonin . Then the cells were incubated on ice for 3 min . The nuclei were washed with 1 mL cold ATAC-Resuspension Buffer containing 0 . 1% Tween-20 and centrifuged at 500 g and 4°C for 10 min . The cell pellets were resuspended in 50 μL transposition mix ( 10 μL 5 × TTBL , 5 μL TTE Mix V50 , 16 . 5 μL PBS , 0 . 5 μL 1% digitonin , 0 . 5 μL 10% Tween-20 , 17 . 5 μL ddH2O ) with TruePrep DNA Library Prep Kit V2 for Illumina ( Vazyme , China TD501 ) and incubated at 37°C for 30 min with shaking in a thermomixer . The reactions were cleaned up with DNA Clean and Concentrator-5 ( Zymo Research D4014 ) . The libraries were constructed through PCR amplification with TruePrep Index Kit V2 for Illumina ( Vazyme , China TD202 ) to barcode samples . The ATAC-seq Libraries were sequenced on an Illumina NovaSeq 6000 ( Berry Genomics Co . , Ltd . , China ) using 150 bp paired-end mode . ChIP-seq data for p65 and H3K27ac were filtered by trim_galore and aligned to human genome ( hg38 ) using Bowtie2 software ( v2 . 3 . 5 . 1 ) . Potential PCR duplicates reads were removed and uniquely mapped reads were used for the subsequent analysis . Peaks were called using MACS2 ( v2 . 1 . 2 ) . p65 peaks in 12 hr TNF-α-treated naïve cells ( first induction ) and memory cells ( second induction ) were merged using Bedtools ( v2 . 29 . 2 ) as consensus peak set , and peaks overlapping hg38 blacklist were removed . ATAC-seq data were processed in a similar way to ChIP-seq data . Aligned read pairs of ATAC-seq were deduplicated and used to generate genome coverage files in bigwig format with the normalization of CPM ( counts per million ) . ChIP-seq and ATAC-seq profiles at regions of interest were drawn using ‘plotProfile’ tool in deepTools ( v3 . 4 . 2 ) . Gene track profiles were visualized using IGV ( v2 . 7 . 2 ) . Gene features of human genome hg38 were extracted from annotations in UCSC genome browser . Repeat elements were extracted from ‘rmsk’ table of UCSC genome browser and grouped by five classes . Nearest genes for each regulatory element were found using ‘bedtools closest’ tool according to the gene transcription start site annotation , and genes that were far than 50 kb were discarded . By using CRISPR-Cas9 system , gRNA sequences are designed to target the primate-specific ERV , MER11B-left element , MER11B-right element adjacent to CALCB for deletion ( Figure 5—figure supplements 1 and 2 ) . The gRNA sequences ( Supplementary file 1 ) were cloned into lentiCRISPR v2 vectors ( Addgene 52961 ) ( Sanjana et al . , 2014 ) . HEK293F-derived cells are transfected with plasmids by lipofectamine 3000 transfection kit ( Invitrogen L3000-015 ) according to the manufacturer’s instructions . One and a half days after transfection , cells were placed under puromycin selection for 2 days . After one round of cell passage , isolation of clonal cells is achieved by serial dilutions in 96-well plates . Genotyping PCR with designed primers ( Supplementary file 1 ) and Sanger sequencing are used for verification of individual clones to obtain the desired genome edited cell line . To perform the CMV promoter , MER11B elements ( Supplementary file 2 ) loci-specific DNA methylation analysis , the purified genomic DNA was bisulfite converted with the EpiTect Bisulfite Kit ( Qiagen , Germany 59104 ) following the manufacturer’s instructions . The bisulfite-converted DNA was then amplified with Jumpstart REDTaq Readymix ( Sigma P0982 ) using locus-specific PCR primers ( Supplementary file 1 ) with nested PCR . The purified PCR products were cloned using the pEASY-T5 Zero Cloning Kit ( TransGen Biotech , China CT501 ) and transformed into competent DH5α cells . Positive individual bacterial clones were selected and the inserted PCR products were sequenced by Sanger sequencing and analyzed by BiQ Analyzer ( Bock et al . , 2005 ) . In the results of bisulfite sequencing , the individual sequenced clones are represented by the horizontal lines . To assess the GFP fluorescent intensity in the HEK293F-derived cells stably inserted with a GFP reporter gene , a single-cell suspension was prepared , sorted , and analyzed on BD FACSAria III ( BD Biosciences ) or BD FACSCalibur ( BD Biosciences ) . The flow cytometry data was analyzed with FlowJo 7 . 6 . 1 . The NEBNext Enzymatic Methyl-seq Conversion Module ( NEB , USA E7125L ) was used for measuring genome-wide DNA methylation level . Purified genomic DNA containing 0 . 05% CpG-methylated pUC19 control DNA and 1% unmethylated lambda DNA was sheared to a mean length of 450 bp with Covaris M220 Focused-ultrasonicator . End repair , A-tailing , methylated adaptor ( Supplementary file 1 ) ligation , and size selection were performed on 125 ng fragmented genomic DNA with Kapa hyper prep kit ( Kapa Biosystems KK8504 ) according to the manufacturer’s instructions . The methylated adaptor ligated DNA fragments were then treated with oxidation of 5mC and 5hmC , clean-up of TET2 converted DNA , denaturation of DNA , deamination of cytosines , and clean-up of deaminated DNA using the NEBNext Enzymatic Methyl-seq Conversion Module ( NEB , USA E7125L ) . The deaminated DNA was amplified with KAPA HiFi Hotstart Uracil+ ReadyMix PCR kit using primers from NEBNext multiplex oligos for Illumina ( index primers set 1; NEB E7335; Kapa Biosystems KK2801 ) according to the manufacturer’s protocol . PCR cycling condition was as follows: 98°C for 45 s , followed by seven cycles of 98°C for 15 s , 60°C for 30 s , 72°C for 30 s , and final extension 72°C for 1 min , and then hold at 4°C . Libraries were sequenced on an Illumina NovaSeq 6000 using 150 bp paired-end mode . Enzymatic Methyl-seq reads in PE150 were first assessed by FASTQC software and then trimmed by trim_galore to remove adapters and low-quality bases . The filtered reads were then mapped to human genome ( hg38 ) using the Bismark software ( v0 . 22 . 3 ) . Duplicate reads were discarded , and methylation information called by Bismark were then processed by ‘DSS’ ( v2 . 34 . 0 ) ( Park and Wu , 2016 ) , a package in Bioconductor ( v3 . 11 ) , including smoothing , fetching average methylation of regions . Differential methylation around p65 peaks were calculated by CpG sites , based on the methylation information in ‘Bismark’ extracted data ( cov ≥ 5 ) , and plotted by the relative distance to κB motif ( Figure 4F ) . All RNA-seq experiments were performed with two biological replicates for each sample by poly-A selection . RNA-seq reads were assessed by FastQC ( v0 . 11 . 9 ) software and trimmed by trim_galore ( v0 . 6 . 4 ) to remove adapters and low-quality bases . Based on the annotation of GENCODE Human Release 24 ( GRCh38 ) , the filtered reads were then mapped to human genome ( hg38 ) using the STAR aligner ( v2 . 7 . 3a ) . FPKM for genes were quantified using Cuffdiff ( v2 . 2 . 1 ) . The data of all protein coding genes were kept for the subsequent gene-based analysis . eRNA reads were quantified by counting RNA-seq reads that were mapped to p65-bound regulatory regions , normalized by sequencing depth and compared between samples . Statistical analyses were performed using Student's t-test in the R software or GraphPad Prism , two-tailed or one-tailed as indicated in the figure legends separately . Boxplots and barplots were drawn using R software . | Genes are the instruction manuals of life and contain the information needed to build the building blocks that keep cells alive . To read these instructions , cells use specific signals that activate genes . The process , known as gene expression , is tightly controlled and for the most part , fairly stable . But gene expression can be modified in various ways . Epigenetics is a broad term for describing reversible changes made to genes to switch them on and off . Sometimes , certain genes even develop a kind of ‘transcriptional memory’ where over time , their expression is enhanced and speeds up with repeated activation signals . But this may also have harmful effects . For example , the signalling molecule called tumour necrosis factor α ( TNF-α ) is an essential part of the immune system . But it is also implicated in chronic inflammatory diseases , such as rheumatoid arthritis . In these conditions , cell signalling pathways triggering inflammation are overactive . One possibility is that TNF-α could be inducing the transcriptional memory of certain genes , amplifying their expression . But little is known about which fraction of genes exhibits transcriptional memory , and what differentiates memory genes from genes with stable expression . Here , Zhao et al . treated cells grown in the laboratory with TNF-α to investigate its role in transcriptional memory and find out what epigenetic features might govern the process . The experiments showed that mimicking a sustained inflammation by stimulating TNF-α , triggered a transcriptional memory in some genes , and enabled them to respond to much lower levels of TNF-α on subsequent exposure . Zhao et al . also discovered that genes tagged with methyl groups are more likely to show transcriptional memory when stimulated by TNF-α . However , they also found that these groups must be removed to consolidate any transcriptional memory . This work shows how TNF-α influences can alter the expression of certain genes . It also suggests that transcriptional memory , stimulated by TNF-α , may be a possible mechanism underlying chronic inflammatory conditions . This could help future research in identifying more genes with transcriptional memory . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2020 | Sustained TNF-α stimulation leads to transcriptional memory that greatly enhances signal sensitivity and robustness |
Tool use has allowed humans to become one of the most successful species . However , tool-assisted foraging has also pushed many of our prey species to extinction or endangerment , a technology-driven process thought to be uniquely human . Here , we demonstrate that tool-assisted foraging on shellfish by long-tailed macaques ( Macaca fascicularis ) in Khao Sam Roi Yot National Park , Thailand , reduces prey size and prey abundance , with more pronounced effects where the macaque population size is larger . We compared availability , sizes and maturation stages of shellfish between two adjacent islands inhabited by different-sized macaque populations and demonstrate potential effects on the prey reproductive biology . We provide evidence that once technological macaques reach a large enough group size , they enter a feedback loop – driving shellfish prey size down with attendant changes in the tool sizes used by the monkeys . If this pattern continues , prey populations could be reduced to a point where tool-assisted foraging is no longer beneficial to the macaques , which in return may lessen or extinguish the remarkable foraging technology employed by these primates .
Humans are currently contributing to one of the most dramatic extinction events in global history ( Barnosky et al . , 2011 ) . In coastal areas , a large component of this pressure comes from our use of increasingly efficient technologies for harvesting food ( Jackson et al . , 2001 ) as well as significantly denser human populations ( Small and Nicholls , 2003 ) . For shellfish , however , archaeological evidence demonstrates that over-harvesting is not a new phenomenon , with intertidal species repeatedly depleted in various parts of the world ( Mannino and Thomas , 2002; Cortés-Sánchez et al . , 2011 ) . Observations and examinations of past and present human populations that exploit shellfish have revealed that over-harvesting and focused collection on the larger individuals , shortens the life histories of the shellfish that are preyed on , resulting in significant reduction in shellfish sizes ( Blackburn et al . , 2004; Erlandson et al . , 2008; Fenberg and Roy , 2008; Langejans et al . , 2012; Morrison and Hunt , 2007; Spennemann , 1987; Branch and Odendaal , 2003; Parkington , 2008 ) . Shellfish exploitation has also been linked to the cognitive , social and technological changes that led to the emergence of modern human behavior ( Marean et al . , 2007; Will et al . , 2016 ) . Recently , island-dwelling wild Burmese long-tailed macaques ( Macaca fascicularis aurea ) were also found to regularly exploit shellfish with the aid of stone tools ( Malaivijitnond et al . , 2007 ) . The macaques use these tools to break open oysters , gastropods , and other intertidal prey ( Gumert and Malaivijitnond , 2012 ) , during intensive foraging episodes that can result in dozens of shellfish being eaten by a single animal using a single tool ( Haslam et al . , 2016a2016 ) . Protein and nutrients obtained in this manner would otherwise be inaccessible to the macaques , providing significant benefits under closed island conditions . Stone tool behavior is currently known from islands off the west coast of Thailand and Myanmar , as well as two islands on the east coast of Thailand ( Tan and , 2017; Gumert and Malaivijitnond , 2013 ) , where Burmese long-tailed macaques have hybridized with common long-tailed macaques ( M . f . fascicularis ) ( Bunlungsup et al . , 2016 ) . The study of island populations is particularly valuable for detecting the effects of predators on local prey species , because of the relatively closed nature of island ecological systems ( Blackburn et al . , 2004; Swadling , 2010 ) . The macaques therefore offer the opportunity to test the effect of non-human stone technology on resource biology and sustainability . Here , we studied a population of long-tailed macaques in Khao Sam Roi Yot National Park , on the east coast of Thailand , that are hybrids of common and Burmese long-tailed macaques . Two groups of macaques live on two neighboring islands – Koram and NomSao – which are separated by less than 400 m , and therefore the overall environmental conditions under which shellfish grow on each island are similar . Koram , however , is densely populated and has a total of at least 80 macaques , while NomSao is sparsely populated by only 9 individuals . On both islands , the macaques are provisioned with food and water , cared for during times of resource scarcity , and thus sustained by human intervention . Despite this , the macaques forage daily for shellfish along the shorelines , like their counterparts on the west coast of Thailand . To investigate the potential effects of macaque stone technology on the sustainability of marine prey , we compared tool use behavior and shellfish properties between the islands , in addition to environmental variables that may affect shellfish size . Wild macaques have been reported to match their tool sizes to prey size ( Gumert and Malaivijitnond , 2013 ) , although this effect has not yet been reported for our study area . We therefore recorded the weight of tools used for respective prey species and compare tool selection patterns between the two foraging groups . To control for environmental factors potentially influencing tool selection , we further compared available shorelines , the local stone availability and stone sizes between both islands . To determine whether macaque tool-assisted predation follows the same trajectory of resource depletion seen in anthropogenic contexts , we compared shellfish densities and sizes between the two islands . We concentrated on the prey most commonly targeted by the macaques: rock oysters ( Saccostrea cucullata ) , tropical periwinkle ( Planaxis sulcatus ) , bifasciated cerith ( Clypeomorus bifasciatus ) , and tooth-lipped snail ( Monodonta labio ) . This approach allowed us to assess whether macaques follow a size-selective harvesting of prey species . The harvesting of large individuals has been reported to leave prey populations with fewer individuals able to reproduce , which additionally hinders population stability and recovery ( Morrison and Allen , 2017 ) . Depending on the life cycle of prey species , size selective harvesting in wild macaques might therefore affect the likelihood of resource depletion . Additionally , species with high reproductive rates , especially at an early age , are reported to be relatively resistant to overharvesting . Prey with high localized aggregation are more vulnerable to foraging , as this clustering enables higher predator foraging efficiency per time ( Morrison and Allen , 2017 ) . We therefore used published life history data ( Angell , 1986; Rohde , 1981; Moore , 1937; Ohgaki , 1997 ) to more accurately judge the impact of tool-assisted foraging might have on marine prey populations . Over-harvesting marine prey has been reported to alter the life history of shellfish as an evolutionary response to increased selection pressure ( Fenberg and Roy , 2008 ) . We therefore compared sexual maturation stages of different shellfish sizes between the islands to assess potential changes in prey life history where predation pressure is high . As a final measure of possible environmental influence , we assessed the rate of resource depletion through tool-assisted foraging by recording the number of prey items consumed during daily foraging events . The estimated rate of depletion could then be judged against the calculated shellfish population on each island . Based on analogy from the human record , we hypothesized that macaque stone technology will exert significant pressure on local shellfish sustainability on the island with the largest number of predators , Koram . We tested this hypothesis through the following predictions . If stone tool size selection by macaques at Khao Sam Roi Yot reflects that seen previously elsewhere , then we expect that stone tools on the two islands would match their respective shellfish prey size , with larger tools selected for larger prey . Additionally , if macaques follow patterns seen in human shellfish foraging ( Mannino and Thomas , 2002 ) , we expect that the macaques would preferentially target larger prey , which would lead to smaller prey available on the heavily populated Koram Island than on NomSao Island . If resource depletion occurred on these islands because of pressure from tool-using macaques , we further expect that shellfish on Koram would be less available than on NomSao . Finally , if the macaques have had an evolutionarily significant effect on prey biology , we would expect to find that shellfish on Koram , which are exposed to increased predation pressure , mature at a younger age ( and smaller size ) than shellfish on NomSao island . If we find our predictions to be true , this may be an indication that macaque stone technology could lead to a depletion of prey populations , mirroring the effects of human predation seen in the archaeological and ethnographic records ( Mannino and Thomas , 2002; Cortés-Sánchez et al . , 2011 ) .
During the observation period for this study , 26 long-tailed macaques on Koram Island and 4 macaques on NomSao Island regularly used stone tools to forage for shellfish . Since the NomSao group contained only adult males , we limited our comparative analyses with Koram Island to tools used by males only ( Koram N = 14 , NomSao N = 4 ) . The tool behavior of macaques was different between the two islands . We found that the two macaques groups select different sized stone tools for shellfish foraging . Tools selected by macaques to open marine gastropods were significantly smaller on Koram Island than on NomSao Island ( Figure 1A , Figure 1—source data 1; LM: N = 67 , E = −0 . 585 , SE = 0 . 113 , F ( 1 , 67 ) = 26 . 645 , p<0 . 001 ) . We found the same result for the stones tools used to crack open oysters; tool used on Koram were smaller than on NomSao ( Figure 1B , Figure 1—source data 1; N = 186 , LMM: E = −1 . 729 , SE = 0 . 562 , χ2 = 6 . 040 , df = 1 , p=0 . 009 ) . Permutation tests revealed the same p values for both models ( LM ( weight of snail tool ) p<0 . 001; LMM ( weight of oyster tool ) p=0 . 009 ) . To investigate potential drivers for the observed difference in tool selection between the two islands , we compared environmental factors that may influence macaque tool selection . Shellfish foraging occurred mainly along the northwest coasts on both islands ( Figure 2 ) . The length of accessible shoreline suitable for shellfish foraging , however , differed between the two islands . On Koram Island , three independent rocky areas , separated by sandy patches , were suitable for shellfish foraging along a total shoreline distance of 1551 m . On average , this resulted in 55 . 4 m accessible shoreline per tool-using macaque ( N = 26 ) . On NomSao Island two rocky foraging zones covered a total of 653 m averaging 163 . 3 m suitable shoreline per tool user ( N = 4 ) . The length of available coastline for foraging is therefore almost three times larger on NomSao island per tool-using macaque . The availability of stones suitable to use as tools did not differ between the islands ( Figure 3A , Figure 3—source data 1; two sample t-test: N = 558 , t ( 19 ) = 1 . 403 , p=0 . 177 ) . Stones available on NomSao , however , were significantly smaller than on Koram ( Figure 3B , Figure 3—source data 1; two sample t-test: t ( 440 ) = −2 . 023 , p=0 . 044 ) . ( Note that the oyster processing tools on NomSao were of a similar size to randomly available stones ( Figures 1B and 3B ) . The availability of marine gastropods between the two islands differed significantly for two species , with a higher number of periwinkles and tooth-lipped snails on NomSao ( Figure 4 , Figure 4—source data 1; t-tests: P . sulcatus: N = 749 individuals , t ( 18 ) = 2 . 885 , p=0 . 010 , M . labio:: N = 72 individuals , t ( 13 ) = 2 . 912 , p=0 . 012 ) . For bifasciated cerith , we found no difference between the two islands ( C . bifasciatus: N = 72 individuals , t ( 16 ) = −1 . 090 , p=0 . 292 ) . Oyster beds were abundant along the lengths of rocky shores of both islands and therefore were not considered a potential limiting foraging factor . However , we found that rock oysters were significantly larger on NomSao than on Koram island ( Figure 5A , Figure 5—source data 1; two sample t-test: N = 1018 individuals , t ( 563 ) = 9 . 873 , p<0 . 001 ) . We found a similar result for two of the investigated marine gastropod species , the periwinkle and the cerith . Both prey species were significantly larger on NomSao ( Figure 5B , Figure 5—source data 2; two sample t-tests: P . sulcatus: N = 223 individuals , t ( 150 ) = 19 . 929 , p<0 . 001; C . bifasciatus: N = 218 individuals , t ( 206 ) = 9 . 762 , p<0 . 001 ) . We did not find enough tooth-lipped snails on Koram ( N = 4 , versus N = 68 on NomSao ) to do a size comparison between the two islands ( see Figure 6B for size differences ) . Sizes of snails between both islands correlated with the maturation stage of these species . Across all collected snails on both islands , more mature specimens had larger shells ( Source data 1 , comparisons full with null model: LM: N = 77 individuals , F ( 4 , 45 ) = 61 . 130 , p<0 . 001 ) . Further , the snail size for a given maturation stage was not significantly different between Koram and NomSao Islands ( LM: E = 0 . 319 SE = 0 . 261 F ( 1 , 48 ) = 1 . 494 P=0 . 228 ) . We assessed macaque predation pressure using observations of prey consumption on Koram Island . On average , one tool-using macaque on Koram Island consumes 46 . 5 shellfish items per day ( 36 rock oysters , 1 . 6 tropical periwinkle and 8 . 9 other species , Source data 2 ) . In total , the studied monkey group on Koram Island ( N = 26 ) consumes approximately 441 , 000 prey items per year , of which almost 61 , 000 are periwinkles ( foraging items consumed per individual multiplied by the number of tool users per island multiplied by 4 hr foraging time per day for 365 days per year ) . Focusing on periwinkles , the four tool-using macaques on NomSao Island would in theory consume some 9344 periwinkles per year . Shellfish foraging areas were estimated to be 4653 m2 on Koram Island and 1959 m2 on NomSao Island . Availability per square meter gives estimates for the current total number of periwinkles as 60 , 163 on Koram and 79 , 477 on NomSao Island . Extrapolating from the consumption data , and without prey population replenishment , then within a year the macaque group foraging on Koram hypothetically would consume more than the estimated current number of periwinkles on the island . On NomSao Island , however , each year the inhabitants hypothetically eat just over a tenth of the current periwinkle population on their island . To assess the vulnerability of macaque prey species to foraging pressure , we merged predictions on prey resilience from published agent based models with available life history data on the two main prey species ( oysters and periwinkles ) . The results are summarized in Table 1 , and show that both species are easily harvested ( they are clustered ) but they also have planktonic dispersal , reducing reliance on local populations for replenishment . In addition , when comparing published reproductive sizes of periwinkles located in our point transects we found that 62% ( N = 123 ) of individuals located in Koram plots were smaller than reported reproductive sizes ( Moore , 1937 ) . On NomSao Island on the other hand , none of the periwinkles in the plots ( N = 100 ) were smaller than the reported reproductive size . Koram is therefore missing around a third of the expected larger and more mature periwinkles . This suggests that macaques preferably harvest larger prey individuals .
Macaque stone tool selection for shellfish processing differed between two closely adjacent islands in Khao Sam Roi Yot National Park , Thailand . On Koram Island , macaques selected significantly smaller stone tools than macaques on NomSao Island , despite targeting the same prey species . This pattern was not explained by available stone material on the islands , with smaller stones on average found on NomSao . The most likely factor influencing tool selection patterns were differences in prey sizes between the two islands . On Koram , the sizes of multiple prey species were significantly smaller than on NomSao , and selected tool sizes correlated positively with targeted prey size on both islands . The fact that macaques on Koram selected smaller tools to use on the shellfish , despite stones there being larger on average than on NomSao , suggests that the macaques actively select task-specific stone sizes . Our result supports a previous finding from a wild macaque population on the west coast of Thailand , where macaques selection of tool size is associated with the size of the prey ( Gumert and Malaivijitnond , 2013 ) . In addition to size differences , multiple prey species on both islands were less available on Koram than on NomSao Island . Feeding pressure , as measured by rates of prey consumption , was estimated to be significantly higher on Koram Isalnd , where , unlike NomSao Island there was a dense population of tool-using predators foraging for shellfish . However , shellfish of similar size were of similar maturation stage on both Koram and NomSao Islands . This outcome suggests that there have been no life history changes in the Koram shellfish , adapting them to higher predation levels on that island . This indicates a rather recent change in feeding intensity , meaning that the difference in prey size between the two islands likely results from a large number of macaques preferentially harvesting large prey on Koram Island , and not a long-term evolutionary response to predation pressure . We do not know how long macaques have been on Koram Island , nor how long they have been provisioned , and range at such high population levels . Based on local reports though , we can infer that macaques have been on these islands for at least 30 years . It is therefore possible that there has been no opportunity for a long-term evolutionary relationship between this population of macaques and their prey . All others tool using macaque populations are in the Andaman Sea ( Gumert and Malaivijitnond , 2013; Carpenter , 1887 ) , so exactly how these macaques arrived in the Thai Gulf , with tool technology , remains a mystery . Assessment of the life history of the main prey species , periwinkles and rock oysters , revealed that the both species appear in large clusters and are therefore susceptible to overharvesting . However , the number of mature individuals and their reproduction rate on the islands turned out to not be useful proxies for estimating population replenishment rates , as both species undergo planktonic reproductive stages . New prey therefore arrives on the ocean currents , with both Koram and NomSao Islands receiving the same input of new individuals , at the same rate . On both islands , shellfish foraging sites ( where the oyster beds are located ) face north-west and the currents transporting planktonic larvae and therefore supply of new prey affect each island similarly . For periwinkles , the continual arrival of new individuals offers an explanation for why the prey population on Koram has not been entirely consumed , despite high foraging rates , as population replacement is not dependent on local mature individuals . Other environmental factors are rather unlikely to have caused a reduction in shellfish size . Sources of shellfish harvesting other than macaques cannot be entirely excluded , but are minimal . For example , shorebirds ( Scolopacidae ) have been occasionally observed foraging on Koram and NomSao Islands , but these birds primarily eat crustaceans ( Moreira , 2008 ) , and are solitary foragers with negligible effect on prey numbers . Both islands are also visited regularly by tourists and locals . On NomSao , we have never observed locals to harvest shellfish , as the island is considered holy and removing anything from there is unacceptable . On Koram Island , seafood harvesting was only observed once during a full year of eight hour daily focal follows . The target prey of this gatherer was exclusively pen shells ( Atrina sp . ) , which are not targeted by the monkeys . Tourists have not been observed harvesting molluscs on either island . Bringing together the observed macaque behavior and environmental conditions , we interpret the variation seen in macaque stone tool selection and shellfish characteristics on Koram and NomSao Islands as the result of a feedback loop driven by the level of predation pressure . The inverse correlation between the number of tool-using macaques , prey size and availability , suggests that predation pressure may be the primary cause of shellfish body size and population on these islands . Particularly , when a population of macaques becomes densely populated , the effect on their prey population becomes obvious . If our conclusions are correct , we interpret the current size distribution of shellfish on Koram as an indication that the macaques are in the process of reducing prey numbers and size which ultimately may lead to unsustainable resource exploitation . This situation could result from tool use being a relatively recent phenomenon on the island , occurring too fast to allow for evolutionary responses from prey species . Alternatively , group sizes on Koram may have only recently reached densely populated levels , likely through increased provisioning , causing an exertion of greater pressure on the shellfish population . These alternative explanations might be testable through archaeological excavation ( Haslam et al . , 2016b ) , for example we might expect to find larger shellfish remains and tool evidence on Koram Island as we go further back in time , with the size approaching that seen today on NomSao Island . This approach would also assist comparisons with human archaeological records of shellfish over-exploitation . Macaques have been reported to impact the local flora and fauna in areas where they are provisioned and densely populated ( Gumert , 2011 ) . However , comparative data from other primates that forage on intertidal resources ( Malaivijitnond et al . , 2007; Hall , 2009; Fernandes , 1991 ) are lacking . Away from the coastlines , there is evidence that non-human primates can hunt prey at unsustainable levels , for example wild chimpanzees ( Pan troglodytes schweinfurthii ) at Ngogo in Uganda hunt red colobus monkeys ( Procolobus rufomitratus ) at a rate that may lead to local extinction of the latter ( Teelen , 2008 ) . Similar to the effects posited for human coastal exploitation , our results suggest that tool-assisted shellfish consumption by a densely populated non-human primate species might lead to unsustainable foraging . Over-harvesting could ultimately lead to the loss of technological knowledge in these macaques . With the decline of prey species , the benefit from using stone tools would decrease , leading to less tool use or its cessation . To definitively answer the question of whether continued shellfish depletion might lead to such a loss at Koram Island , we will need to monitor future developments . However it unfolds , these macaques provide an interesting case for potential feedback systems between a technological predators and its prey . The isolated nature of island-dwelling macaques therefore makes them a useful model for assessing how simple technologies may impact prey population dynamics , life histories , and phenotypes .
Koram ( N12°14'32" , E100°0'34" ) and Nom Sao ( N12°13'51" , E100°0'17" ) Islands are located in Khao Sam Roi Yot National Park ( KSRY ) , Prachuap Khiri Khan province , Thailand . Both islands are arid and contain no freshwater bodies , with limestone karst interiors that are covered with dwarf evergreens and deciduous scrub flora ( Figure 1 ) . Koram Island is located about 1 km offshore from Sam Roi Yot beach on the Thai eastern mainland . It is approximately 0 . 45 km2 , with 3 . 5 km of coastline comprising of limestone cliff shore on the side of the island facing the open gulf , and rocky shores and sandy beaches on the side of the coast facing the mainland . NomSao Island is 0 . 37 km southwest of Koram , with an area of 0 . 10 km2 , including 1 . 32 km of coastline . The shoreline is not completely accessible for the monkeys as steep overhanging cliffs make some parts unsuitable for foraging . We measured the accessible shoreline on which the monkeys were able to forage by taking GPS points at the furthest edge of suitable foraging zones , and we calculated the total length of suitable shoreline using Google Earth . The macaque groups on Koram and NomSao Islands are habituated to the presence of humans as a result of provisioning by locals and tourists with food and water . However , they also forage naturally using percussive stone technology . The macaques process predominantly sessile rock oysters ( Saccostrea cucullata ) , and the main gastropod species processed are the tropical periwinkle ( Planaxis sulcatus ) . They also harvest bifasciated cerith ( Clypeomorus bifasciatus ) , and tooth-lipped snail ( Monodonta labio ) . Other less commonly processed food species are conches , various bivalves and crabs , as well as coconuts that occasionally wash ashore , and the dried seeds of provisioned mangoes . At the end of our first data collection period in December 2014 , the studied Koram macaque group contained 64 individuals ( about 178 ind/km2 ) . Twenty-five ( 12 ♂; 13 ♀ ) of 36 adult and adolescent macaques ( i . e . > 4 years ) were tool users , or 69 . 4% of the group , and 3 of the 14 ( 21 . 4% ) juveniles ( i . e . 1–4 years ) were observed using tools . A smaller group of about 10 individuals is usually excluded from the coast by the larger group and less easily observed , but they also forage on shellfish with tools when they are able to access the shores . On NomSao , there are at least nine adult males that have been observed and individually identified ( approximately 90 ind/km2 ) with at least four being tool users . There are no females or juveniles on NomSao , and it is likely that the males there were disperced from Koram . Macaques live in multi-male multi-female groups where males emigrate from the natal group when they are mature ( so-called male dispersal ) , which is believed to reduce the inbreeding depression in the populations . We collected data over two field seasons , from May to December 2014 , and from September to October 2015 . During the first season , we collected tools from the Koram group directly after observing individuals using them as either oyster or snail foraging tools . In the second field session , we focused on behavioral observations and tool collection on the NomSao group , as well as the sampling of prey size , and food and stone availability on both islands . We collected used stone tools from Koram and NomSao Islands . Since the NomSao group contained only adult males , we limited our analyses on Koram to tools used by adult males , only to keep tool dimensions comparable between sites , as males are known to use larger tools ( Gumert and Malaivijitnond , 2013 ) . Tools on Koram were collected only after we observed male monkeys using and discarding them , to prevent female choices appearing in our data set . On NomSao , we additionally collected abandoned tools from recent cracking sites as only adult males could have used them on this island . Stones at recent cracking sites were identified as tools if they were conspicuously placed on a boulder , accompanied by food debris on or adjacent to the stone , and/or showed use-wear marks from being struck . For each tool collected , we recorded whether oysters or snails were processed ( identified through prey remains on the tool ) , and weighed the stone on a portable digital scale . We measured the length , width and depth by photographing the stone in line with a ruler . Where possible , we also noted the identity of the tool user . Our final data set included 93 oyster tools from 14 individuals on Koram and 93 oyster tools from 4 individuals on NomSao as well as 45 snail tools from 9 individuals on Koram and 22 snail tools from 4 individuals on NomSao . We used point transects to sample the size and availability of stones between Koram and NomSao Islands . We applied 14 point transects at 7 locations spaced 100 m apart along the shores of each island . At each location , we demarcated two 20 × 20 cm plots , one located in the lower littoral zone along the water line where snails were abundant , and one located higher up in the littoral zone amongst the oyster beds . We counted , weighed , and measured the lengths and widths of all stones found within each point transects . We excluded stones that were 20% smaller than the smallest tool used by the macaques on Koram . The upper limit of potential stone tools was set as the size of the plot ( max = 20 cm2 ) . However , no stones found in the plots reached this size . To compare the ecological conditions between the two islands , we collected data on prey size and tool availability . To compare the sizes of oysters exploited by the macaques , we traversed the oyster beds of both islands , and measured the length and width of recently cracked oysters . From these measurements , we calculated the size of 345 oysters on NomSao and 673 oysters on Koram . Recently-cracked oysters were identified by the presence of flesh remains on the opened valve , or the white shiny inner surface of the opened valve ( older oysters are rapidly discoloured by sea water ) . We did not carry out plot sampling for oyster availability as oyster beds were distributed continuously along the lengths of rocky shores of both islands and therefore were not a limiting foraging factor . To compare snail availability and size between Koram and NomSao , we applied 14 systematic point transects on each island . Snails were more abundant in the lower littoral zones and thus exposed during the low tide , so we conducted sampling during the low tide hours of each day . During these low tide hours , when the lower littoral zone was exposed , we demarcated 1 × 1 m plots at each point transect along the shore in a line , 100 m apart from one another . As a measure of snail availability , we counted all snails of the three most commonly processed snail species ( Planaxis sulcatus , Clypeomorus bifasciatus , and Monodonta labio ) in each plot . On Koram Island , we found a total of 236 ( P . sulcatus = 181; C . bifasciatus = 50; M . Labio = 4 ) snails and on NomSao Island a total of 658 ( P . sulcatus = 568; C . bifasciatus = 22; M . Labio = 68 ) snails in all point transects combined . To obtain data on snail size , we used calipers to measure the maximum shell length and width of 100 randomly selected individual snails of each species and on each island , and we weighed each snail on a portable digital scale ( 1–1000 g ) . To further investigate whether the underlying reason for observed size differences between the three main marine snail prey species stems from overharvesting large individuals , or is rather a result of an evolutionary change in snail biology , we collected a representative sample of four different size categories of each species on each island ( N = 77 ) . We assessed the maturation stage of each size category by investigating the developmental stage of the reproductive organs . The samples were stored in 70% ethanol and their sexual organs were inspected under a microscope after breaking their outer shell . Maturation was classified into four developmental stages: Immature ( no traces of reproductive organs ) ; Subadult ( testis/ovaries occupy less than 50% of the upper whorl and are therefore not yet able to reproduce ) ; Mature ( testis/ovaries occupy more than 50% but less than 80% of the upper whorl , these samples might be able to reproduce ) ; and Completely Mature ( testis/ovary occupy more than 80% of the upper whorl ) . We aimed at collecting four different snail sizes ( size equivalent for each island ) , however , since Monodonta labio was very rare on Koram we were not able to find any large specimens of that species and therefore excluded it from the comparative analysis . To investigate the sustainability of the marine prey consumption , we combined published information on the life history of our main prey species with the outcome of an agent-based model ( Morrison and Allen , 2017 ) , concentrating on the main prey species ( P . sulcatus , C . bifasciatus , and M . labio ) . These species have a reproductive system which includes a planktonic stage of the larvae that can disperse great distances by water currents before settlement . Observational data were collected on the proportions of foraging time that the macaques engaged in tool-aided foraging , and on their rates of tool use . During focal sampling , all individuals were sampled in random order on continuous rotation , for a five minute duration each time . During foraging , we recorded the type of food item processed , noting whether the subject was cracking a sessile oyster or unattached gastropod , and recording the species whenever possible . We counted the number of prey items consumed per focal observation time of 5 min and conservatively estimated an average foraging time of 4 h ( during low tide ) per individual to extrapolate the number of shellfish items eaten per day ( and per year ) on Koram Island . We used the amount of daily prey consumption on Koram Island to estimate the total foraging pressure that one tool-using macaque can place on the prey population . We used this same value to estimate foraging pressure for the neighbouring NomSao population , for which long-term focal observations were not available ( Tan and Luncz , personal communication ) . We then used our data on snail abundance per island to extrapolate the potential time needed to deplete the existing prey population . For that calculation we multiplied the length of suitable coastal foraging areas per island with an average width of foraging grounds of 3 m . We used the resulting area and the number of snails found per surveyed square meter to estimate the total number of snails available on each island . To compare oyster size between the two islands , we bootstrapped the measured oyster sizes for each island 1000 times and compared the confidence intervals at the level of 95% to each other . To test for differences in snail size between the islands , we conducted the same procedure for two snail species ( C . bifasciatus ( NNamSao = 101 , NKoram = 119 ) ; P . sulcatus ( NNamSao = 100 , NKoram = 123 ) ) . To compare the availability of the different snail species between islands for each species individually , we bootstrapped the number of snails we found on each transect 1000 times and compared the confidence intervals at the level of 95% between the islands . To compare stone availability and stone size separated for island and location we bootstrapped i ) the number , and ii ) the weight of the stones we found per transect 1000 times and compared the confidence intervals at the level of 95% between the islands . For each test , we additionally applied two sample t-tests . All bootstraps and t-tests were implemented in R ( version 3 . 2 . 3 ) . | Tools have helped us to become one of the most successful species on Earth . However , our use of tools for hunting and foraging has also caused many prey species to become endangered , or even extinct . In some cases , it has also led to evolutionary changes in prey species . For example , over-harvesting of shellfish in coastal areas has driven the shellfish to become smaller in size . Recently , long-tailed macaques living on islands off the coast of Thailand and Myanmar were also found to use stone tools to forage on shellfish . The macaques use these tools to break open oysters , snails and other prey on the seashore . Studying these monkeys offers the opportunity to test how a non-human primate using stone-based technology affects the sustainability of their prey species . Luncz et al . investigated how foraging with stone tools by long-tailed macaques living in Khao Sam Roi Yot National Park in Thailand affects local shellfish populations . This revealed that macaques using stone tools alter prey populations in a similar way to human technologies . Specifically , tool use by the macaques significantly reduced the numbers and size of the prey , especially on islands that were home to larger populations of monkeys . In return , the macaques responded by using smaller and smaller stone tools . This “feedback loop” could lead to the stone tools becoming less useful to the macaques to the point where they stop using them . An important next step is to learn whether continued foraging of shellfish might actually lead to the macaques losing the knowledge on how to use stone tools . Luncz et al . propose that since stone tools first emerged , the size of the tools and the prey species they target may have been gradually decreasing . Future archaeological investigations will clarify if this is indeed the case . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology"
] | 2017 | Resource depletion through primate stone technology |
Ultradian ( ∼4 hr ) rhythms in locomotor activity that do not depend on the master circadian pacemaker in the suprachiasmatic nucleus have been observed across mammalian species , however , the underlying mechanisms driving these rhythms are unknown . We show that disruption of the dopamine transporter gene lengthens the period of ultradian locomotor rhythms in mice . Period lengthening also results from chemogenetic activation of midbrain dopamine neurons and psychostimulant treatment , while the antipsychotic haloperidol has the opposite effect . We further reveal that striatal dopamine levels fluctuate in synchrony with ultradian activity cycles and that dopaminergic tone strongly predicts ultradian period . Our data indicate that an arousal regulating , dopaminergic ultradian oscillator ( DUO ) operates in the mammalian brain , which normally cycles in harmony with the circadian clock , but can desynchronize when dopamine tone is elevated , thereby producing aberrant patterns of arousal which are strikingly similar to perturbed sleep-wake cycles comorbid with psychopathology .
Ultradian rhythms with periods ranging from one to several hours have been linked to various aspects of mammalian physiology . Usually superimposed on the 24-hr diurnal or circadian rhythm , ultradian oscillations have been observed in the context of locomotion , sleep , feeding , body temperature , serum hormones , and brain monoamines in species ranging from fruit flies to humans ( Tannenbaum and Martin , 1976; Ibuka et al . , 1977; Rusak , 1977; Daan and Slopsema , 1978; Honma and Hiroshige , 1978; Dowse et al . , 1987; Rivkees , 2003; van Oort et al . , 2007; Dowse et al . , 2010; Seki and Tanimura , 2014 ) . These physiological cycles most frequently exhibit periods of 2–6 hr , adopting harmonics of the 24-hr daily light–dark cycle or the endogenous circadian rhythm , when external timing cues are absent . However , the generation of such ultradian rhythms does not depend on a functional circadian system nor a light:dark cycle . Ultradian locomotor oscillations persist in rodents housed in constant darkness even upon ablation of the suprachiasmatic nucleus ( SCN ) , the site of the master circadian pacemaker ( Ibuka et al . , 1977; Rusak , 1977 ) , or genetic disruption of the circadian clock ( Vitaterna et al . , 1994; Bunger et al . , 2000 ) ( Figure 1A , B ) . These ultradian activity cycles may not simply be driven by metabolic demand since the 2- to 3-hr rhythm in foraging activity observed in the common vole persists even in the absence of food ( Gerkema and van der Leest , 1991 ) . Studies in this species further indicate that one adaptive value of ultradian activity rhythms may lie in the facilitation of social synchrony , which is suggested to reduce predator risk in this species ( Gerkema and Verhulst , 1990 ) . However , despite their prevalence and hypothesized biological significance , ultradian locomotor rhythms have received little research attention , likely owing to their frequently masked expression and unstable periodicity in contrast with circadian activity rhythms ( Ruis et al . , 1987; Schibler , 2008 ) ( Figure 1C , D ) . 10 . 7554/eLife . 05105 . 003Figure 1 . Inter- and intra-animal variability of ultradian activity rhythms across time in circadian incompetent mice . ( A ) Representative double-plotted actograms of running wheel activity in SCN-lesioned ( SCNx ) and Bmal1−/− mice in DD . ( B ) Dot plot of locomotor period length in DD based on Lomb-Scargle periodogram analysis of seven consecutive days of activity recording ( N = 65 for Bmal1−/− and N = 48 for SCNx; t111 = 0 . 2785 , p = 0 . 78 , unpaired t-test ) . ( C ) Intra-animal period variability expressed as mean ± SD for each animal , ranked according to mean period length derived from continuous wavelet transforms ( CWT ) for the 1–12-hr frequency range ( same animals and timespans for calculation as in B ) . ( D ) CWT-heatmaps showing decibel scaled and normalized amplitude of oscillations according to frequency and time with black traces indicating the ridge of local amplitude maxima . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 003 In addition to ultradian oscillations , methamphetamine-induced rhythms of locomotor activity occur in the absence of a functional circadian clock . When provided in the drinking water , methamphetamines ( Meth ) induce a daily activity bout in addition to the expected circadian rhythm ( Honma et al . , 1986 ) . This Meth-dependent component—which typically adopts a period in the circadian range—is not abolished upon SCN lesion nor by genetic disruption of circadian clock function ( Honma et al . , 1987; Mohawk et al . , 2009 ) . It was thus concluded that a methamphetamine-sensitive circadian oscillator ( MASCO ) outside the SCN exists which is capable of driving daily cycles of locomotor activity ( Tataroglu et al . , 2006 ) . Despite the longstanding recognition of ultradian and Meth-dependent rhythms , the underlying cellular and molecular identity of the oscillator ( s ) driving them is unknown . Here , we provide evidence for a highly tunable dopaminergic ultradian oscillator ( DUO ) which is continuously operative in the mammalian brain and which , together with the circadian clock , orchestrates the daily pattern of arousal . Our data suggest that dopamine acts as both the principal oscillator output as well as an integral component of the DUO , determining oscillator period . Our findings further indicate that the previously described MASCO represents a long-period manifestation of the DUO resulting from elevated dopamine tone . Importantly , our data support an intriguing proposition: that DUO , rather than circadian clock , dysregulation critically contributes to the sleep-wake abnormalities associated with psychopathology .
To gain insights into the mechanistic basis of ultradian locomotor rhythm generation , we considered that locomotor activity is associated with an awakened state ( Welsh et al . , 1988 ) and consequently , the ultradian locomotor rhythms observed in mice that lack circadian clock function ( Figure 1 ) could be interpreted as rhythms of heightened wakefulness or arousal . In mammals , a key role in arousal promotion has been attributed to distinct monoaminergic neuronal populations located in the upper brainstem and midbrain ( Brown et al . , 2012 ) . While altering extracellular levels of the arousal-associated monoamines serotonin , norepinephrine , or histamine by genetic manipulation has only limited effects on locomotion ( Thomas and Palmiter , 1997; Bengel et al . , 1998; Xu et al . , 2000; Parmentier et al . , 2002; Zhao et al . , 2006 ) , depleting the brain of dopamine ( DA ) profoundly abrogates locomotor activity ( Zhou and Palmiter , 1995 ) . Moreover , increasing extracellular DA levels induces hyperlocomotion ( Giros et al . , 1996 ) and lengthens the time spent awake ( Wisor et al . , 2001 ) . We therefore speculated that altering DA tone may affect ultradian rhythm generation . To test this , we examined running wheel activity in mice carrying a disruption in the Slc6a3 gene , which encodes the dopamine transporter ( DAT ) . Slc6a3−/− mice exhibit hyperdopaminergia due to the lack of DAT-mediated DA reuptake into dopaminergic neurons , leading to a hyperactivity phenotype ( Giros et al . , 1996; Gainetdinov et al . , 1998 ) . As the presence of the circadian clock and/or a light:dark cycle frequently masks ultradian activity rhythms ( Schibler , 2008 ) , we assessed the locomotor behavioral consequences of DAT elimination in the absence of the master SCN circadian pacemaker . To do so , we electrolytically lesioned the SCN of Slc6a3−/− mice and their wildtype littermates and monitored their running wheel behavior in constant darkness ( DD ) . While control mice ( SCNx-Slc6a3+/+ ) exhibited ultradian activity rhythms with the expected ∼4-hr period , SCNx-Slc6a3−/− mice showed rhythms whose periods were three times longer ( Figure 2A , B ) . Analysis of mice that were deficient for both DAT and the essential clock component BMAL1 ( Bmal1−/− , Slc6a3−/− ) corroborated this finding . Bmal1−/− , Slc6a3−/− mice exhibited ∼12–14-hr rhythms in locomotor activity , largely phenocopying the SCNx-Slc6a3−/− mice , while their Bmal1−/− , Slc6a3+/+ littermates showed ∼4-hr periods as expected for isodopaminergic mice lacking circadian clock function ( Figure 2C , D ) . Together , these results suggest that DAT removal markedly increases ultradian cycle length . Alternatively , the ∼12-hr rhythms observed in SCNx-Slc6a3−/− and Bmal1−/− , Slc6a3−/− animals may originate from an independent oscillator , one that is activated by DAT elimination , while the short period ultradian oscillator that operates in DAT intact , SCNx or Bmal1−/− animals is disengaged or otherwise obscured . 10 . 7554/eLife . 05105 . 004Figure 2 . Dopamine transporter knockout alters periodicity of ultradian locomotor rhythms in circadian incompetent mice under constant darkness . ( A ) Representative , double-plotted actograms demonstrating marked lengthening of ultradian locomotor periods in SCNx-Slc6a3−/− mice as compared to SCNx-Slc6a3+/+ littermates . Tau ( τ ) indicates individual period . ( B ) Period length averages of ultradian activity ( N = 7; F1 , 6 = 253 . 8 , ***p < 0 . 0001 , ANOVA ) from Lomb-Scargle periodogram analysis of 7-day time-spans . ( C ) Representative , double-plotted actograms demonstrating markedly increased ultradian period lengths in Bmal1−/− , Slc6a3−/− mice as compared to Bmal1−/− , Slc6a3+/+ mice . ( D ) Period length averages of ultradian activity ( N = 4; F1 , 3 = 194 . 2 , ***p < 0 . 001 , ANOVA ) from Lomb-Scargle periodogram analysis of 7-day time- spans . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 004 In order to corroborate that DAT removal indeed lengthens the period of the ultradian activity cycles , we took into account that DAT-mediated DA uptake can be reversed by the selective action of the psychostimulant methamphetamine ( Meth ) ( Howell and Kimmel , 2008 ) . Because Meth leads to increased extracellular DA concentrations as in the case of Slc6a3 gene disruption , we speculated that Meth treatment would result in a , possibly gradual , period lengthening of the ultradian locomotor rhythms . Indeed when we treated Bmal1−/− animals with increasing concentrations of Meth via drinking water in DD , we observed a gradual lengthening of the initial ∼4-hr locomotor oscillations and this was accompanied by a corresponding increase in activity bout length ( Figure 3A , B ) . Of note , the period increase in response to elevating Meth concentrations did not halt in the circadian range , rather , oscillations continued to lengthen with individual animals reaching periods of 100 hr or more ( Figure 3C ) . Gradual period lengthening of ultradian locomotor rhythms was also observed in Bmal1−/− animals exposed to amphetamine ( Figure 3—figure supplement 1A , B ) , a drug similarly targeting DAT ( Howell and Kimmel , 2008 ) , but with lower efficacy than Meth ( Goodwin et al . , 2009 ) . These results argue that DAT targeting psychostimulants affect an endogenous ultradian rhythm generator by increasing period length . However , due to the mode of delivery ( drinking water ) and the rhythmic Meth uptake that may in turn result from it , it is conceivable that the Meth-dependent , long-period oscillations are ‘driven’ by rhythmic drug intake rather than being generated endogenously . To address this possibility , we subcutaneously implanted Bmal1−/− mice with osmotic minipumps that continuously infused Meth over a period of 2 weeks . Running wheel analysis in DD of Meth-infused Bmal1−/− animals demonstrated a significant ultradian period lengthening upon drug infusion ( Figure 3—figure supplement 1C , D ) suggesting that rhythmic uptake is not required for Meth to exert its period lengthening effect , in line with a previous study performed in rats ( Honma et al . , 1987 ) . The relatively limited change in period observed in this experimental paradigm could be due to the short , 2-week infusion timespan , which is perhaps insufficient to robustly lengthen periods beyond 12 hr ( Figure 3—figure supplement 1D ) . Together , these findings support the notion that the long-period oscillations observed in Meth-treated animals , and likewise in Slc6a3−/− mice , are due to period expansion of an endogenous ultradian rhythm of arousal . 10 . 7554/eLife . 05105 . 005Figure 3 . Pharmacological tuning of DAT activity by incrementally increasing methamphetamine concentrations lengthens ultradian locomotor period into the infradian range . ( A ) Representative actograms of Meth-treated Bmal1−/− mice in DD . Treatment intervals are highlighted with corresponding concentrations indicated in ( B ) . ( B ) Mean periods from the last 7 days at a given Meth concentration . Repeated measures ANOVA revealed a significant main effect of concentration/time ( F5 , 40 = 34 . 30 , p < 0 . 001 ) and significant period lengthening between consecutive concentrations ( mean ± SEM; N = 9; **p ≤ 0 . 005 , planned comparisons; 30 day , 30 day exposure to 100 mg/l ) . ( C ) Modulo 110-hr actogram of a Bmal1−/− animal after extended exposure to Meth revealing an ultra-long activity rhythm . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 00510 . 7554/eLife . 05105 . 006Figure 3—figure supplement 1 . Amphetamine treatment lengthens ultradian locomotor rhythms in Bmal1−/− mice . ( A ) Representative actogram of Bmal1−/− mice in DD provided with d-amphetamine ( 100 mg/l ) in their drinking water . ( B ) Average activity rhythm periods calculated for the indicated days post treatment-onset ( mean ± SEM , N = 10; F4 , 9 = 26 . 51 , p < 0 . 0001 , repeated measures ANOVA; *p ≤ 0 . 001 , planned comparison ) . ( C ) Representative running wheel activity actogram ( plotted modulo 13 . 6 hr ) of Bmal1−/− mice in DD , which received chronic subcutaneous infusions of Meth ( 0 . 6 mg/day ) for 14 days . Red arrow indicates time of pump implantation . ( D ) Average activity rhythm periods during the last 7 days prior to pump implantation ( 0 mg/day ) and the last 7 days of treatment ( mean ± SEM , N = 6; F1 , 5 = 3 . 525 , *p < 0 . 01 , ANOVA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 006 We next aimed to confirm these findings in intact mice . While activity rhythms in the ultradian range are easily discernable in voles and hamsters ( Gerkema et al . , 1990; van der Veen et al . , 2006; Prendergast et al . , 2012 ) , the presence of such rhythms in mice or rats is much less obvious and this has been attributed to masking by the circadian clock ( Schibler , 2008 ) . Indeed , running wheel activity data from intact mice often do not provide strong indications of ultradian activity rhythms . However , when we recorded ambulatory behavior using telemetry , we frequently detected three , evenly spaced activity bouts during the active ( night ) phase ( Figure 4A ) . The observed ∼4 hr peak-to-peak spacing is in line with a previous study on several circadian competent mouse strains where ultradian rhythms with similar period length were detected ( Dowse et al . , 2010 ) . The absence of a clearly discernable ∼4-hr rhythmic component during the light portion of the daily LD cycle ( Figure 4A ) is likely due to masking by the SCN and/or light . Unlike WT mice , Slc6a3−/− mice never showed a triple-peak activity pattern at night ( Figure 4B ) , and spectral density analysis in the ultradian ( 2–8 hr ) range ( Figure 4C ) underscores that rhythm generation is significantly altered in these animals while total daily activity is not ( Figure 4D ) . When we provided C57BL/6 mice carrying a telemetry implant with Meth in their drinking water , we observed a gradual lengthening of the interval between the night-time activity peaks ( Figure 4F , G ) , resulting in the transformation of the triple peak ( Figure 4G , white trace and white triangles ) into a dual or single night-time activity peak pattern at the highest ( 100 mg/l ) Meth concentration ( Figure 4G , black trace and black triangles ) . This was corroborated by spectral density analysis revealing a marked reduction in ultradian frequencies upon Meth treatment ( Figure 4E ) . Together , these results indicate that blocking DA reuptake also lengthens ultradian activity bout intervals in circadian competent mice , arguing that an ultradian oscillator is not ‘activated’ by circadian disruption , but rather continuously operative alongside the circadian timer . As with Bmal1−/− mice , prolonged Meth-treatment in intact WT mice can lead to profound period lengthening of ultradian rhythms , often into the circabidian ( 48 hr ) range ( for examples see Figures 5E , 7F , 9B ) . 10 . 7554/eLife . 05105 . 007Figure 4 . Ultradian activity in SCN-intact Slc6a3−/− mice and their wildtype littermates . ( A and B ) Ambulatory activity recorded by telemetry implants across multiple days ( left ) and averaged over 24 hr ( right ) . Traces represent 2-hr recursive smoothing ( black ) of the underlying raw DATa ( dark grey; SEM envelope , light grey ) . Areas in yellow indicate lights on . ( C ) Amplitude spectral density in the ultradian range ( 2–8 hr ) is significantly different between Slc6a3−/− mice and their wildtype littermates revealing a loss of the ultradian component ( mean ± SEM; N = 10; F1 , 18 = 26 . 40 , **p < 0 . 0001 , ANOVA ) . ( D ) Although the temporal pattern of locomotion differs between genotypes there is no significant difference in daily activity averaged over multiple days ( mean ± SEM; N = 10 , F1 , 18 = 0 . 1793 , ANOVA ) . ( E–G ) Addition of Meth to the drinking water of C57BL/6 mice lengthens the night-time activity bouts in a concentration-dependent manner . Averaged daily locomotor activity of individual mice at different Meth concentrations ( G ) derived from the time-span indicated by colored bars next to the representative actogram ( F ) and subjected to Butterworth filtering . The three night-time activity peaks before treatment ( white triangles ) , transform to 2 peaks after exposure to the highest concentration ( black triangles ) . The night-time bout lengthening is also reflected in the reduction of ultradian amplitude spectral density in the 1 to 5-hr range ( E , mean ± SEM , N = 7; F2 , 20 = 8 . 08 , p < 0 . 005 , repeated measures ANOVA; *p < 0 . 05 , **p < 0 . 01 , post-hoc Bonferroni ) . White asterisks ( in E ) indicate cage changes . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 007 Given that both DAT disruption and Meth treatment elevate extracellular DA and concurrently lengthen ultradian rhythm period , we speculated that manipulations aimed at lowering DA tone should conversely lead to period shortening . We therefore provided Bmal1−/− mice in DD with the antipsychotic drug haloperidol ( Hal ) in their drinking water . Hal selectively blocks the DA receptor D2 ( DRD2 ) ( Strange , 2008 ) , which is expressed on postsynaptic target sites of DA neuronal projections but also presynaptically as an autoreceptor ( Schmitz et al . , 2003 ) . Importantly , when given chronically as in our case , Hal has been reported to electrically silence midbrain DA neurons ( White and Wang , 1983 ) and to markedly lower extracellular DA levels in the striatum/nucleus accumbens regions in rats ( Lane and Blaha , 1987; Ichikawa and Meltzer , 1991 ) . As predicted , Bmal1−/− mice responded with successive locomotor period shortening to increasing concentrations of Hal ( Figure 5A , B ) . We also observed Hal-mediated shortening of the long-period behavioral rhythms in Bmal1−/− and wild-type mice treated concurrently with Meth ( Figure 5C–F ) . A period shortening effect of Hal could also be discerned from the core body temperature fluctuations of Slc6a3−/− mice , which increased in frequency ( Figure 5G ) . This response , which mirrored the locomotor behavioral response ( not shown ) , was concentration dependent and spectral density analysis confirmed a successive increase of the ultradian component upon Hal exposure ( Figure 5H ) . Together , these data strongly support the hypothesis that both the ultradian rhythms observed in wildtype and Bmal1−/− mice and the long-period rhythms previously attributed to the MASCO are driven by the exact same oscillatory mechanism . 10 . 7554/eLife . 05105 . 008Figure 5 . Haloperidol shortens circadian-clock-independent locomotor rhythms . Representative actograms with Hal treatment periods indicated by colored bars ( left ) ; bar graphs indicate corresponding locomotor period ( right ) . ( A ) Increasing concentrations of Hal provided in the drinking water gradually shortens the endogenous ultradian activity rhythms of Bmal1−/− mice in DD ( B , mean ± SEM , N = 12; F2 , 10 = 14 . 36 , p < 0 . 0001 , repeated measures ANOVA; *p < 0 . 05 and **p < 0 . 01 , planned comparison ANOVA ) . ( C and E ) Hal shortens the infradian rhythms in Meth-treated Bmal1−/− mice ( D , mean ± SEM , N = 9; F1 , 8 = 2 . 357 , *p < 0 . 05 ANOVA ) and WT mice ( F , mean ± SEM , N = 9; F1 , 8 = 3 . 525 , **p < 0 . 01 , ANOVA ) in DD . ( G ) Hal treatment increases the frequency of temperature fluctuation in Slc6a3−/− mice under LD measured by telemetric implants . ( H ) Changes of amplitude spectral density in the ultradian range ( 2–8 hr ) in response to increasing Hal concentrations ( mean ± SEM; N = 8; F2 , 7 = 14 . 74 , repeated measures ANOVA , *p < 0 . 05 , ***p < 0 . 0005 , post-hoc Bonferroni ) . For all experiments , periods are calculated based on the running wheel activity during the last 7 days of treatment at the indicated Meth/Hal concentration . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 008 As altering extracellular DA is a common denominator of all three manipulations—DAT elimination , ( meth ) amphetamine , and haloperidol treatment—these findings collectively suggest that DA tone determines ultradian period . Given that DA is known to mediate arousal/wakefulness ( Brown et al . , 2012 ) , it appears plausible that DA also serves as principal oscillator output , driving rhythms in arousal . We thus hypothesized that extracellular DA must oscillate in synchrony with the observed activity rhythms . To test this hypothesis , Bmal1−/− mice were unilaterally implanted with a microdialysis probe positioned along the ventro-dorsal extent of the right striatum ( Figure 6G ) , a site heavily innervated by DA neurons . Mice were kept under constant dim red light ( <5lux ) and ultradian activity rhythms were monitored using infra-red beam breaking . The animals showed the expected short period rhythms in locomotor activity throughout the experiment ( Figure 6A ) , and analysis of dialysates revealed that extracellular DA levels fluctuated concordantly with the activity cycles ( Figure 6A and Figure 6—figure supplement 1 , compare solid red and black traces ) . Importantly , we generated a model oscillation using the cosinor method from locomotor rhythms recorded for 20 hr after dialysate sampling with no investigator present and compared it to the DA fluctuations observed . The goodness of fit between the cosinor and the DA profiles indicated by the sum of squared errors ( SSE = 0 . 050 ± 0 . 027 , mean ± SEM , N = 7 ) suggests that the dialysate sampling procedure itself did not perturb the generation of the endogenous ultradian activity cycles . To determine the statistical significance of the observed agreement between the cosinor and the DA trace , we randomly permutated the individual time points of the DA trace and determined the percentage of permutated traces with an equal or better fit to the cosinor in comparison to the observed DA fluctuation . On average , only 3 . 2 ± 1 . 7% of 100 , 000 permutated DA profiles correlated as well or better than the experimentally observed DA measurements ( Figure 6B ) , confirming that extracellular DA fluctuates in synchrony with the ultradian activity cycles . Of particular note , linear regression revealed a highly significant correlation between mean DA concentration in the dialysate and ultradian locomotor period in the Bmal1−/− animals we tested ( Figure 6C ) , which again supports a role of ( extracellular ) DA as a period determinant of the ultradian activity cycle . 10 . 7554/eLife . 05105 . 012Figure 6 . DA fluctuations correlate with ultradian locomotor behavior in Bmal1−/− mice . ( A ) Two representative examples of in vivo striatal microdialysis in Bmal1−/− mice . Upper: locomotor activity as measured by beam breaks . Lower: DA dialysate concentration measured at 20-min intervals ( red trace ) plotted alongside the corresponding locomotor activity ( solid black trace ) . Cosinor ( dotted line ) was computed from the 20 hr of locomotor activity following dialysate sampling . ( B ) False discovery rate of the fit between the DA profiles and corresponding cosinors ( mean ± SEM , N = 7 ) . ( C ) Linear regression analysis of period length vs mean DA concentration , dots representing individual animals . ( D ) Tissue punches of CPu , NAcc , and midbrain were analyzed for DA and DOPAC content in animals sacrificed during their active ( ■ ) vs their non-active ( □ ) phase ( mean ± SEM; active phase , N = 7 , inactive phase , N = 6; *p < 0 . 05 , **p < 0 . 01 , ∼p = 0 . 066 , ANOVA ) . ( E ) Linear regression for period vs DA content in animals sacrificed during the active phase revealed significant correlations for the CPu and midbrain . ( F ) Representative actograms displaying activity rhythms of Meth-treated Bmal1−/− mice used for tissue collection . Triangles indicate collection time points ( in ° ) , with locomotor activity bout onset set to 180° . ( G ) Illustrations of coronal mouse brain sections ( Paxinos and Franklin , 2001 ) indicating position of the active membrane ( black bar ) and tissue punch placements ( circles , colors correspond to labels in D and E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 01210 . 7554/eLife . 05105 . 013Figure 6—figure supplement 1 . Rhythms of extracellular DA in the striatum of freely-moving Bmal1−/− mice . ( A ) Individual records of striatal in vivo microdialysis showing DA dialysate concentrations ( red ) , horizontal activity in 20-min bins ( black ) and the cosinor trace ( dotted ) . τ indicates locomotor rhythm period computed from the 20 hr of horizontal activity recorded after dialysate collection . ( B ) Histogram plots showing the distribution of the summed square of errors values for the fit between the cosinor and the DA profile after random permutation of the latter ( 100 , 000 operations ) . False discovery rate is calculated by determining the percentage of permuted profiles which fit the cosinor as well or better than the experimentally determined DA profile ( highlighted in red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 013 As our findings suggest that the observed >24-hr activity rhythms are due to period lengthening of the ultradian oscillations , we similarly expected DA levels to fluctuate in synchrony with the activity cycles in Meth-treated animals . To test this , we measured DA content in tissue punches of Meth-treated Bmal1−/− mice from the dorsal ( caudate putamen , CPu ) and ventral ( nucleus accumbens , NAcc ) striatum as well as the ventral midbrain , which includes both the substantia nigra ( SN ) and the ventral tegmental area ( VTA ) ( Figure 6G ) . For punch collection , animals were sacrificed during their active and inactive phases , respectively ( Figure 6F ) . We detected significantly higher levels of DA as well as its immediate metabolite , 3 , 4-dihydroxyphenylacetic acid ( DOPAC ) , in tissue extracts of the CPu during the animal's active phase ( Figure 6D ) . There was also a significant increase in DOPAC levels and a trend towards elevated DA ( p = 0 . 066 ) during the active period in the NAcc , while extracts from the midbrain region , which contains the cell bodies of the striatal dopaminergic afferents , exhibited no detectable change ( Figure 6D ) . We also found a significant correlation between locomotor period and DA content in CPu and midbrain punches collected during the active phase ( Figure 6E ) , again consistent with a role for DA as a period determinant of ultradian oscillations . As Slc6a3−/− mice lack the triple peak night-time activity pattern that is characteristic of wild-type mice , we subjected Slc6a3−/− mice to long-term running wheel activity monitoring to examine their locomotor behavior in more detail . Under LD conditions , we observed more fragmented daily activity in Slc6a3−/− than their wildtype counterpars ( Figure 7A , D ) . Upon release into constant darkness ( DD ) , wild-type animals exhibited a daily rhythm with a period slightly shorter than 24 hr as expected for mice ( Bunger et al . , 2000 ) ( Figure 7A ) . While this principal locomotor component was also observed in Slc6a3−/− mice , they repeatedly exhibited an additional activity component with periods longer than 24 hr ( Figure 7D–F ) . This component , which was more evident after the first few weeks in DD , extended from the main activity component and lasted for several cycles before disappearing late in the subjective day ( Figure 7D ) . Overall , there was no significant difference in the total amount of daily activity even though the temporal pattern of locomotion in these mice is severely altered ( Figure 7C ) . The observed dual component activity pattern shows striking similarities to the activity patterns of Meth-treated C57BL/6 mice ( Figure 7G–I ) , suggesting that the second rhythmic component is generated by the dopaminergic oscillator . 10 . 7554/eLife . 05105 . 009Figure 7 . Slc6a3−/− mice show a second rhythmic locomotor activity component . Representative actograms displaying daily running wheel activity of Slc6a3+/+ ( A ) , Slc6a3−/− mice ( D ) and C57BL/6 mice on Meth ( G ) . ( B , E and H ) Lomb-Scargle periodograms generated from the time-span indicated by red bars , in ( A , C , G ) , respectively . ( C ) There is no significant difference between genotypes in daily activity averaged over the time-span of analysis ( mean ± SEM; N = 6 , F1 , 10 = 1 . 848 , ANOVA ) ( F and I ) Average periods of the highest 2 periodogram peaks for Slc6a3−/− mice ( E , mean ± SEM , N = 6 ) and C57BL/6 mice on Meth ( H , mean ± SEM , N = 9 , Meth-treatment started on day 1 of the recordings ) . Areas in yellow indicate lights-on . Green line in the periodograms indicates the confidence threshold for rhythmicity ( α = 0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 00910 . 7554/eLife . 05105 . 010Figure 7—figure supplement 1 . The SCN of Slc6a3−/− mice is unperturbed . ( A ) In situ hybridization indicates normal expression of Per1 in the SCN of Slc6a3−/− mice with strong staining of the SCN in the subjective day ( DAY ) and barely detectable probe signal in the subjective night ( NIGHT ) , a pattern similar to wildtype SCN . ( B ) Subjective day and night were determined based on running wheel records using the onset of the circadian alpha band . Red triangles indicate time of sacrifice . 3V , third ventricle; OC , optic chiasm . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 010 However , such locomotor activity pattern , with two components oscillating at different frequencies , has also been observed in rats housed under a 22-hr LD cycle and have been attributed to a dissociation of molecular rhythms between the dorsal and ventral SCN ( de la Iglesia et al . , 2004 ) . To rule out that the SCN is similarly affected in Slc6a3−/− mice , we sacrificed mice at the peak and trough of the main activity component when the second rhythm was maximally antiphasic ( Figure 7—figure supplement 1A ) . In situ hybridization directed against Period1 transcripts did not indicate rhythmic desynchrony between or within SCN hemispheres . The riboprobe showed strong , homogenous staining throughout the SCN in the early subjective day , whereas little to no signal was detectable in the early subjective night , suggesting that SCN pacemaker function is unperturbed in Slc6a3−/− mice ( Figure 7—figure supplement 1B ) . Together , these findings thus argue that the second rhythmic component of Slc6a3−/− mice is indeed a product of the above described dopaminergic oscillator . Since striatal and midbrain DA levels vary concordantly with activity rhythms in Bmal1−/− mice , and DA neurons of the SN/VTA are by and large the exclusive source of striatal and midbrain DA , these neurons are a likely site of ultradian rhythm regulation , if not generation . To corroborate the importance of these neurons in the oscillator process , we employed a chemogenetic approach based upon DREADD ( designer receptors exclusively activated by designer drugs ) technology ( Krashes et al . , 2011 ) . We stereotaxically delivered the adeno-associated virus AAV-DIO-hM3Dq-mCherry ( Krashes et al . , 2011 ) into the SN/VTA region of mice that carried an Slc6a3 promoter driven Cre transgene ( DAT-Cre ) ( Figure 8A ) ( Turiault et al . , 2007 ) and that additionally were either BMAL1-deficient ( DAT-Cre , Bmal1−/− ) or received an SCN lesion ( DAT-Cre , SCNx ) . Upon Cre-mediated recombination , virally transfected cells express the stimulatory DREADD hM3Dq , which has been demonstrated to increase neuronal firing frequency upon binding of the compound clozapine-N-oxide ( CNO ) , which is otherwise physiologically inert ( Alexander et al . , 2009 ) . As augmenting firing frequency in DA neurons enhances DA release ( Sulzer , 2011 ) , and CNO stimulation of successfully transduced DAT-Cre mice evokes vigorous firing of dopamine neurons ( Wang et al . , 2013 ) , CNO is expected to increase DA release . Consistently , i . p . -injection of CNO elevated extracellular DA in the striatum of those mice and led to a prolonged increase in locomotor activity ( Figure 8C , D ) , which was absent in vehicle injected animals ( Figure 8—figure supplement 1D ) . Responsive mice showed mCherry fusion tag expression selectively in tyrosine hydroxylase ( TH ) positive cells ( Figure 8B , Figure 8—figure supplement 1A ) . Cell counting revealed that , 92 . 9 ± 5 . 1% and 87 . 8 ± 6 . 0% ( mean ± SEM , N = 6 for both groups ) of the midbrain TH+ neurons also expressed the mCherry fusion tag in responsive DAT-Cre , SCNX and DAT-Cre , Bmal−/− mice , respectively , confirming effective Cre-mediated hM3Dq expression in midbrain DA neurons . Running wheel activity monitoring of virus-injected mice showed the expected short period locomotor oscillations , however , upon switching to CNO-containing drinking water ( red arrows ) , both , AAV-hM3Dq transduced DAT-Cre , SCNx and DAT-Cre , Bmal1−/− mice responded with locomotor period lengthening ( Figure 8E–H ) , an effect that was reversed when mice were returned to pure water ( Figure 8G , blue arrows ) . As expected , a period lengthening was not observable in AAV-hM3Dq injected DAT-Cre0/0 , Bmal1−/− mice upon CNO exposure ( Figure 8—figure supplement 1B , C ) . These results further corroborate that elevating DA tone , in this case by selective excitation of DA neurons , lengthens ultradian locomotor period and that the oscillator driving ultradian rhythmicity comprises DA neurons of the VTA/SN region . 10 . 7554/eLife . 05105 . 014Figure 8 . Chemogenetic activation of midbrain DA neurons lengthens ultradian locomotor period . ( A ) AAV-DIO-hM3Dq-mCherry was stereotaxically and bilaterally delivered into the VTA/SN region of DAT-Cre transgenic mice as indicated . ( B ) Representative immuno-fluorescent image of the ventral midbrain from a virus-injected and behaviorally responsive mouse showing extensive co-expression of the mCherry fusion-tag in TH-positive cells of the midbrain . ( C ) Locomotor response to CNO ( red arrow , 1 mg/kg body weight i . p . ) of a representative , AAV-hM3Dq transduced DAT-Cre mouse undergoing microdialysis . ( D ) 20min- binned locomotor activity and extracellular striatal DA content ( 20 min ) of AAV-hM3Dq transduced mice 1 hr prior ( Baseline ) and 2 hr after CNO injection . Mice were implanted with a striatal microdialysis probe and DA content was measured as in Figure 6A ( mean ± SEM; N = 3 , *p < 0 . 05 , **p < 0 . 01 , paired t-test ) . ( E–H ) Representative actograms of AAV-hM3Dq transduced DAT-Cre , SCNx ( E ) and DAT-Cre , Bmal1−/− mice ( G ) . Switch to CNO-containing drinking water ( 7 . 5 mg/l ) is indicated by red arrow . Blue arrow marks return to regular water ( in E ) . Animal activity is plotted modulo according to the period measured during the last 7 days of CNO treatment . Periodogram analysis reveals CNO-dependent period lengthening in both DAT-Cre , SCNx ( F , mean ± SEM; N = 6 , F1 , 5 = 3 . 68 , *p < 0 . 05 , ANOVA ) and DAT-Cre , Bmal1−/− ( H , mean ± SEM; N = 11; F1 , 10 = 20 . 48 , **p < 0 . 0001 , ANOVA ) mice . VTA , ventral tegmental area; SNc , substantia nigra pars compacta . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 01410 . 7554/eLife . 05105 . 015Figure 8—figure supplement 1 . AAV-h3MDq targeting of DA neurons . ( A ) Midbrain expression pattern of mCherry and tyrosine hydroxylase ( TH ) in DAT-Cre mice upon bilateral injection with AAV-DIO-h3MDq-mCherry . Shown are immunofluorescent images of representative sections of CNO-responsive mice that were stained with dsRed ( red ) and TH ( green ) antibodies demonstrating comprehensive h3MDq-mCherry expression in TH neurons . Corresponding illustrations indicate sites of viral expression ( red ) . ( B ) AAV-h3MDq injected DAT-Cre0/0 , Bmal1−/− mice which lack the DAT-Cre transgene do not exhibit a change in ultradian activity when exposed to CNO . Shown are representative actograms of running wheel activity in DD . Red arrow indicates switch to CNO ( 7 . 5 mg/l ) containing drinking water . ( C ) Periodogram analysis indicates no significant difference in locomotor period between pre and post-treatment ( mean ± SEM; N = 7 , F1 , 6 = 0 . 401 , ANOVA ) . ( D ) Acute locomotor response ( running wheel activity ) of AAV-h3MDq injected DAT-Cre mice to midday ( ZT 6 ) i . p . injections of either vehicle ( 0 . 5% DMSO in 0 . 9% Saline ) or CNO ( 1 mg/kg b . w . in vehicle ) . VTA , ventral tegmental area; SNc , substantia nigra pars compacta; Cli , caudal linear nucleus of the raphe; RMC , red nucleus , magnocellular part . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 015
Collectively , our results provide strong evidence that a dopaminergic ultradian oscillator ( DUO ) driving rhythms of behavioral arousal is continuously operative in the mammalian brain . We propose that under normal conditions , this DUO cycles in harmony with the circadian SCN pacemaker and that the rhythmic information of both the SCN and the DUO are integrated at a common downstream site to create the daily pattern in locomotor activity ( Figure 9A ) . However , elevation of DA tone can lead to DUO period lengthening , which either results in DUO free-run ( for example , Figure 7D , G ) or reinstatement of oscillator synchrony albeit at a different harmonic ( Figure 9B , e . g . , 48 hr ) . The DUO appears as a highly tunable oscillator , able to adopt period lengths from a few hours to multiple days . This is in stark contrast to the circadian timer which cannot adopt periods that are more than a few hours off from 24 hr when its limits of entrainment are tested experimentally ( Aschoff and Pohl , 1978 ) . 10 . 7554/eLife . 05105 . 011Figure 9 . Proposed model of circadian and ultradian oscillator output integration to govern daily locomotor behavior . ( A ) Light input entrains the circadian ( SCN ) and ( indirectly ) the ultradian ( DUO ) oscillators creating a stable phase relationship . Their rhythmic outputs , upon integration at a common downstream effector , generate the daily pattern of locomotor activity . ( B ) Under conditions of high DA tone ( e . g . , DAT elimination or Meth-treatment ) , DUO period lengthens , leading to a second , separate activity rhythm . This rhythm either free-runs ( see Figure 7D , G ) or phase locks with the SCN pacemaker by adopting a subharmonic , that is , 48-hr period , as frequently observed upon Meth treatment . Representative output plots show average activity of individual mice computed from 8 days of ambulatory activity ( A ) or 14 days of running wheel activity under Meth treatment ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05105 . 011 Given that DA is known to stimulate locomotor activity ( Zhou and Palmiter , 1995 ) , our observation of a cyclical rise in extracellular striatal DA , which is in synchrony with ultradian activity rhythms , argues that DA acts as an output of the DUO . Our finding that manipulations affecting extracellular DA levels alter oscillator period and that extracellular DA tone shows a remarkably high correlation with activity cycle length , strongly suggests that DA is a period determinant and therefore must be an integral component of the oscillatory mechanism itself . As all period-altering manipulations directly impinge upon DA neuronal physiology , this suggests that either ( i ) DA neurons are the site of ultradian rhythm generation or ( ii ) they are a key cog in the oscillatory mechanism . The chemogenetic activation experiments indicate that the relevant DA neuronal population is located in the midbrain as selective activation of DA neurons in this region had a period lengthening effect on the ultradian activity . Future experiments will be aimed at delineating the precise DA population ( s ) required for the ultradian rhythm generation process and how rhythmic synchrony between neurons is maintained , if the DUO is indeed composed of a population of cellular oscillators . Our data indicate that the circadian and ultradian locomotor rhythms are normally harmonized ( for instance , Figure 9A ) , suggesting that the circadian pacemaker and the DUO interact . Of note , extracellular DA was reported to fluctuate diurnally in the rodent striatum ( Hood et al . , 2010 ) , which was abrogated in Slc6a3−/− mice ( Gallardo et al . , 2014 ) . In these studies , only group averaged DA profiles were presented and DA was solely measured in circadian intact mice and rats , thus it is conceivable that any ultradian component ( in WT ) or infradian component ( in Slc6a3−/− ) escaped detection . Critically , the observation that the DA levels , on average , followed a diurnal pattern with a night-time peak , together with the observation that ultradian locomotor period in female hamsters is longer in the dark vs the daily light phase ( Prendergast et al . , 2012 ) , is consistent with the LD cycle and/or the circadian pacemaker affecting DUO rhythmicity by altering ( extracellular ) DA . Note however , that dopamine content in whole brain extracts from circadian-intact rats kept under LD showed a strong ultradian variation with no obvious evidence for a diurnal rhythmic component ( Scheving et al . , 1968 ) . Also , microdialysis did not reveal a day:night difference in extracellular DA levels in the striatal NAcc region when measured in DD ( Chung et al . , 2014 ) . This same study reported elevated DA levels and hyperactivity in mice lacking clock gene Rev-erbα , suggesting that the circadian clock , possibly intrinsic to DA neurons themselves , has a role in DA regulation . Indeed , knockdown of the core circadian clock component Clock in DA neurons of the murine VTA , increases electrical firing rate in VTA neurons and enhances locomotor response to novel objects ( Mukherjee et al . , 2010 ) . Interestingly , we did not observe any systematic differences in ultradian locomotor period between SCNx and Bmal1−/− mice in constant darkness conditions , regardless of DAT status ( Figures 1 , 2 ) , suggesting that extra-SCN circadian clocks have no role in DUO-mediated ultradian locomotor rhythm generation . Our data also provide evidence that the postulated MASCO ( Tataroglu et al . , 2006 ) represents a long-period manifestation of the DUO as a result of methamphetamine's action on the dopamine transporter , blocking dopamine reuptake thereby increasing extracellular DA levels . Interestingly , the dopamine system has been also implicated with another behavioral timing system: the food-entrainable oscillator ( FEO ) . This circadian independent oscillator ( Pitts et al . , 2003; Pendergast et al . , 2009; Storch and Weitz , 2009 ) drives food-anticipatory locomotor activity ( FAA ) that emerges when food access is restricted to a few hours each day ( Mistlberger , 2011 ) . D1 ( DRD1 ) and D2 ( DRD2 ) receptor antagonists attenuate FAA additively ( Liu et al . , 2012 ) while pharmacological DRD2 , but not DRD1 , activation altered the phase of FAA ( Smit et al . , 2013 ) . Most recently , using knockout mice , it was revealed that DRD1 but not DRD2 is necessary for the appropriate expression of FAA and that rescuing dopamine signaling selectively within the dorsal striatum was sufficient to restore FAA in dopamine-deficient mice ( Gallardo et al . , 2014 ) . Given the links to the dopaminergic system , it will be of interest to investigate whether the DUO has a role in the temporal regulation of FAA . While these findings argue that food cues engage the dopamine system to alter daily locomotor activity patterns , it is clear that dopamine signaling also affects food intake as mice lacking dopamine are lethargic and do not actively consume food ( Zhou and Palmiter , 1995 ) . Considering that the DUO is a universal driver of ultradian behavioral rhythms in mammals , the finding that ultradian bouts of running-wheel and feeding activity are co-expressed in the common vole ( van der Veen et al . , 2006 ) suggests that dopamine can synchronously drive food seeking and general activity , which is in line with the view that dopamine acts as a general promoter of motivated arousal . Slc6a3−/− mice have been proposed as a model for schizophrenia ( Gainetdinov et al . , 2001 ) and the DA hypothesis of schizophrenia states that DA elevation is causal to the behavioral symptoms of this psychiatric condition . Intriguingly , circabidian ( 48 hr ) or free-running rhythms in locomotor activity reminiscent of the behavioral patterns we detected in Slc6a3−/− or Meth-treated mice ( Figures 4F , 7C , F , 9B ) have been observed in schizophrenic subjects ( Wirz-Justice et al . , 2001; Wulff et al . , 2012 ) , suggesting that DUO dysregulation underlies the rest-activity aberrations associated with schizophrenia . Furthermore , actigraphy recordings in schizophrenic patients also revealed that Hal treatment reduces circadian/diurnal locomotor amplitude and leads to the emergence of ultradian activity bouts ( Wirz-Justice et al . , 1997 , 2009 ) , effects we likewise observed in Slc6a3−/− mice in response to Hal . Transient DUO period lengthening might equally account for the rest-activity pattern abnormalities observed during manic episodes in bipolar disorder , which has been also associated with altered DA tone ( Berk et al . , 2007 ) . Bipolar subjects have been reported to show rapid , 48-hr cycling between mania and depression ( Gann et al . , 1993; Wilk and Hegerl , 2010 ) , one to multiple 48-hr sleep–wake cycles when switching from depression to mania ( Wehr et al . , 1982 ) , or a long-period ‘free-running’ rhythm of wakefulness ( Wehr et al . , 1998 ) . Thus schizophrenic and bipolar subjects both appear to exhibit rest-activity cycle aberrations strikingly similar to those observed in Slc6a3−/− or Meth-treated mice , suggesting that DUO dysregulation is a common indicator for these psychopathologies and perhaps even a common disease cause . A switch to sleep–wake cycles with a period much longer than 24 hr have also been observed in subjects that were studied in temporal isolation ( Aschoff , 1965 ) . Because other physiological parameters such as urine secretion and core body temperature showed phase-aligned , circadian fluctuations with a period much closer to 24 hr , the subjects were considered internally desynchronized . Notably , affected subjects tend to exhibit high scores of neuroticism ( Wever , 1979 ) . It is therefore conceivable that the observed internal desynchronization is also due to a dysregulation of the DUO .
Bmal1−/− ( Bunger et al . , 2000 ) and DAT-Cre ( Turiault et al . , 2007 ) mice were on a C57BL/6J genetic background , while Slc6a3−/− mice ( Giros et al . , 1996 ) were maintained on a mixed C57BL/6JxDBA/2J background ( Morice et al . , 2004 ) . Animals were housed under LD 12 hr:12 hr unless otherwise stated . Slc6a3−/− mice were found to exhibit high attrition rates ( ∼80% ) in experiments that involve long-term locomotor activity monitoring . To increase survival , Slc6a3−/− mice were first group-housed for 1 week after transfer into the light-tight , ventilated cabinets used for activity monitoring . This was followed by 1 week of individual housing in running wheel cages with the wheels locked before commencing baseline activity recordings . During this adaptation phase and throughout the subsequent experimental period , both Slc6a3−/− mice and their wild-type littermates were provided with ad libitum chocolate flavored chow ( Supreme Mini-Treats , BioServ , Flemington , NJ ) as well as cotton nestlets and ample shredded corrugated card stock . This regimen significantly reduced attrition rates to less than 20% on average . ‘Material and methods’ were performed in accordance with the Canadian Council on Animal Care guidelines and approved by the McGill University Animal Care Committee . Running wheels: Animals were individually housed in light-controlled cabinets and activity was recorded continuously ( ClockLab , Actimetrics , Wilmette , IL ) . Actograms , displaying binned running wheel revolutions per 6 min ( 0 . 1 hr ) , and the associated Lomb-Scargle periodograms , displaying amplitude , were generated using ClockLab software . Telemetry: Animals were individually housed in standard cages placed atop energizer/receiver units ( ER-4000 , Starr Life Science Corp . , Oakmont , PA ) . 1 week prior to data collection , electromagnetic induction powered telemetry probes ( G2 E-mitter , Starr Life Science Corp . ) were implanted intraperitoneally . Locomotion , measured in counts per minute , and core body temperature ( BT , in °C ) , was collected in 6-min bins ( 0 . 1 hr ) using Vitalview software ( Starr Life Science Corp . ) . BT data was exported into Clocklab to generate actogram-style data displays with tick mark height corresponding to temperatures from 34–38°C; for waveform generation , averaging , and 2 hr recursive smoothing , locomotor activity and BT data was exported to Excel ( Microsoft , Redmond , WA ) . Matlab ( Mathworks , Natick , MA ) was used for ribbon plot generation , and low pass Butterworth filtering ( 1 hr ) . Stock solutions of ( + ) -Methamphetamine hydrochloride ( 100 mg/l , Sigma-Aldrich , St . Louis , MO ) , D-Amphetamine hemisulfate ( 100 mg/l , Sigma-Alrdich ) , Haloperidol ( 11 . 25 mg/ml , Sigma-Aldrich ) and Clozapine-N-Oxide ( 15 mg/l , National Institute of Mental Health , Bethesda , MD ) were prepared using distilled water . Methamphetamine solutions were adjusted to pH 7 using sodium hydroxide and haloperidol was dissolved by stirring at 40°C . Mice were continuously infused with methamphetamine ( 0 . 6 mg/day in 0 . 9% saline ) for 14 days using subcutaneously implanted osmotic minipumps ( Alzet Model 1002 , Durect Corp . , Cupertino CA ) . Pumps were fitted with a 65-mm polyvinyl chloride catheter ( 0 . 69-mm inner diameter , Plastics One , Roanoke , VA ) and backfilled with saline to allow for an approximately 4 day post-surgery recovery prior to drug exposure . In situ hybridization was performed as previously described ( Kraves and Weitz , 2006 ) . Briefly , following decapitation , brains were removed and quickly frozen at −80°C . Serial coronal hypothalamic brain slices ( 25 µm ) were collected using a cryostat and stored at −80°C until hybridization . Sections were hybridized overnight at 60°C to a digoxigenin-labeled riboprobe targeting the coding regions of mouse Per1 ( nucleotides 579–1478 of the Per1 mRNA , Genbank , NM_011065 . 4 ) . Electrolytic lesions of the suprachiasmatic nucleus were performed as described ( Storch et al . , 2007 ) . Briefly , an electrode ( RNE-300X , Rhodes Medical Instruments , Woodland Hills , CA ) was lowered through a hole drilled in the skull at the mid-sagittal sinus according to stereotaxic coordinates ( AP −0 . 25 mm , DV −6 . 00 mm from Bregma ) and a constant current ( 2 mA , 10 s; D . C . Constant Lesion Maker , Grass Instruments , Quincy , MA ) was applied . Only mice with behavioral circadian arrhythmia ( in the 20–28-hr range assessed by Lomb-Scargle periodogram analysis ) and subsequent post-mortem histological verification using DAPI staining mounting medium ( Vectashield , Vector Labs ) were included for analysis . This represented approximately 50% of lesioned mice when averaged across all studies . 1 week prior to sampling , mice were stereotaxically implanted with a guide cannula ( C312G/spc 2 . 5 mm below pedestal , Plastics One ) , targeting the striatum ( AP +1 . 10 mm , DV +5 . 5 mm , ML +0 . 9 mm ) . 1 day prior to dialysis , mice were transferred to a beam break monitor ( VersaMax , Omnitech Electronics , Columbus , OH ) and tethered using a dummy probe assembly under constant , dim , red light ( <5lux ) . On the day of sampling , in-house recording probes ( Lupinsky et al . , 2010 ) , were connected to a central swivel ( Model 72-0000 , Harvard Apparatus , Holliston , MA ) . Artificial cerebrospinal fluid was delivered via syringe pump ( Model 403 , CMA Microdialysis , Kista , Sweden ) at a flow rate of 1 . 0 μl/min . This procedure was similarly followed to test the effects of CNO in AAV-transduced DAT-Cre mice where intraperitoneal injections of 1 mg/kg CNO were given after approximately 1 hr of baseline sampling . Samples were collected every 20 min for 4 hr and immediately admixed with 1 μl of perchloric acid . DA and DOPAC content in the dialysate were determined by high-performance liquid chromatography with electrochemical detection ( HPLC-EC ) as described ( Domenger et al . , 2012 ) . Chromatographic peak analysis was conducted using CoulArray software ( ESA Inc . , Chelmsford , MA ) . After data collection , brains were removed , sectioned , and stained with hematoxylin to verify probe placement . Locomotor activity data , recorded as the number of beam breaks per minute , were exported into Matlab ( Mathworks ) for waveform generation , 20-min binning , cosinor modeling , and false discovery rate analysis . After decapitation , brains were quickly removed and frozen ( −80°C ) . 320 μm coronal brain slices were obtained by cryosectioning and then microdissected ( 1- or 2-mm diameter sample corers , Fine Science Tools Inc . , Foster City , CA ) to obtain tissue of the CPu ( 2 mm ) , NAcc ( 1 mm ) , and SN/VTA midbrain ( 2 mm ) regions ( see Figure 6G for punch location ) . DA/DOPAC was quantified as previously described ( Domenger et al . , 2012 ) . Briefly , individual punches from each region were homogenized in 45 μl perchloric acid ( 0 . 25 M ) to which 15 μl of a 100 ng/ml solution of 3 , 4-dihydroxybenzylamine was added , which served as the internal standard . Concentrations were determined by HPLC-EC . After perchloric acid extraction , the protein containing pellets were reconstituted in 0 . 1 N sodium hydroxide for protein quantification ( Pierce BCA Kit , ThermoFisher Scientific , Waltham , MA ) . Mice were anaesthetized with isofluorane and placed in a stereotaxic aparatus ( David Kopf Instruments ) . Recombinant AAV8-DIO-h3MDq-mCherry ( Krashes et al . , 2011 ) ( titer = 3 × 1012 genomes copies per ml , UNC Vector Core Services , Chapel Hill , NC ) was bilaterally injected into to the VTA/SN area ( coordinates from bregma: AP: −3 . 44 mm , DV: −4 . 40 mm , L: ±0 . 48 mm ) ( Tsai et al . , 2009 ) through a cannula ( 33 gauge , Plastics One ) at a flow rate of 0 . 05 μl/min for 10 min ( 0 . 5 μl total volume per side ) using a syringe pump ( Harvard Apparatus ) . Mice were subsequently maintained in individual housing for at least 2 weeks prior to CNO treatment . Immunostaining was performed as previously described ( Chu et al . , 2013 ) using cryoprotected ( 30% Sucrose ) , free-floating coronal sections from fixed brains collected after intra-cardiac perfusion ( 4% paraformaldehyde ) and cut at 40 microns using a cryostat ( Leica , Solms , Germany ) . For fluorescent labeling , sections were incubated overnight with primary antibodies for mCherry ( rabbit anti-RFP , 1:1000 , Rockland , Limerick , PA ) to enhance detection of the mCherry expression and antibodies for tyrosine hydroxylase ( mouse anti-TH , 1:1000 , EMD Millipore , Etibicoke , Canada ) . This was followed by 2 hr incubation with secondary antibodies ( anti-rabbit Alexa Fluor 567 and anti-mouse Alexa Fluor 647 , 1:250 , Life Technologies , Carlsbad , CA ) after which sections were mounted on superfrost slides ( VWR , Radnor , PA ) , coverslipped with Vectashield mounting medium ( Vector Labs , Burlington , Canada ) and imaged by fluorescence microscopy ( AxioObserver Z1 , Zeiss , Jena , Germany ) . Quantification of co-expression was performed on a single , medial , midbrain section from each behaviorally responsive mouse using the cell counter plugin of ImageJ ( National Institute of Health , Bethesda , MD ) in order to determine the percentage of all TH-positive cells that also expressed the hM3Dq--mCherry fusion protein ( N = 6 per genotype ) . One-way and repeated measures ANOVA , t-tests , and linear regression analyses were performed using Prism 5 ( GraphPad , La Jolla , CA ) . Planned comparisons for the repeated measures were carried out to determine significant period changes between subsequent measurements . Bonferroni correction was used for post-hoc testing of individual group differences . To evaluate the probabilistic significance of the least-square fit between the DA trace and the cosinor model that was computed based on the animal's locomotor activity oscillations , a false discovery rate approach was employed ( Storch et al . , 2007 ) . For cosinor computation , we first determined the locomotor period of Bmal1−/− mice by Lomb-Scargle periodogram analysis of 20 hr of locomotor recordings following dialysate sampling . The determined period was then used as an input parameter for a least-squares cosinor analysis ( Nelson et al . , 1979 ) using a custom script and function written for Matlab ( https://github . com/storchlab/Cosinor-FDR . git ) . This procedure generated a best-fit ( co ) sine wave modeling the activity time-series . This model wave was then used to assess the concordance between the extracellular DA fluctuations and locomotor oscillations . To calculate FDR , we randomly permutated the temporal order of the measured DA concentrations 100 , 000 times , and assessed its fit to the cosinor . The permutation procedure will degrade the fit of any truly rhythmic signal but will have little to no effect on noisy or flat profiles . The probability of false discovery was calculated as the percentage of randomly permuted traces that show an equal or better fit to the cosinor than the measured DA fluctuation . Low pass filtering ( 1-hr cut-off ) of raw data was conducted using a Butterworth zero-phase filter ( Butterworth , 1930 ) in Matlab . This allowed for better visualization of ultradian frequencies in the >1-hr range , in a manner similar to recursive smoothing but without phase distortion . Ultradian rhythms show variability in amplitude and period . We thus used Lomb-Scargle periodogram analysis ( Clocklab ) to estimate ultradian period length as this method is relatively tolerant to noisy data and data gaps , which can result from intermittent ultradian rhythm expression ( Press and Rybicki , 1989 ) . Unless otherwise stated , period length was determined by identifying the highest peak above the significance threshold ( α = 0 . 01 ) in the Lomb-Scargle periodogram . Subsequent plotting of an actogram at the determined modulus was conducted in each case in order to visually confirm rhythmicity and rule out the false identification of harmonic frequencies or side-lobes due to leakage ( Van Dongen et al . , 1999 ) . To visualize and quantify the dynamics of period and amplitude in circadian incompetent mice , continuous wavelet transforms using the Morlet wavelet ( Goupillaud et al . , 1984 ) were performed for the 1–12-hr range using custom algorithms written in Matlab ( https://github . com/storchlab/CWT . git ) . Locomotor activity data of seven consecutive days per individual SCN-lesioned or Bmal1−/− mouse was used . For each spectrogram , a ridge identifying the peak amplitude of the spectrum was generated as described ( Leise , 2013 ) and the associated values were used to calculate the average and standard deviation of the dominant ultradian frequency for each animal . To determine the prevalence of oscillations in a given period range we calculated the area under the curve of all significantly rhythmic periodicities ( α = 0 . 01 ) in the Lomb-Scargle periodogram , that is , the spectral density . For normalization , the obtained spectral density was divided by the total significant spectral density in the 0–40-hr range and expressed as percentage . | The sleep-wake cycle of mammals is controlled by a ‘circadian clock’ within the brain , which is synchronized to the day–night cycle . However , other aspects of mammalian physiology including alertness and activity levels , as well as appetite and body temperature—fluctuate in cycles that repeat every few hours . These cycles are known as ultradian rhythms , and they may offer survival benefits by enabling potentially risky behaviors , such as foraging , to be coordinated between members of a group . Despite their widespread nature and the fact that they appear to be conserved in evolution , virtually nothing is known about the molecular basis of ultradian rhythms . Blum et al . have now identified a second internal clock within the brain , which they name ‘the DUO’ , and shown that this clock normally works in concert with the circadian clock to regulate daily patterns of activity and alertness . Experiments in mice revealed that the DUO uses the brain chemical dopamine to generate bursts of activity roughly every four hours . Moreover , it continues to work when the circadian clock has been destroyed . Measurements of dopamine in freely moving mice showed that levels of the chemical fluctuate in synchrony with the animals' activity levels . Moreover , drugs that flood the brain with dopamine , such as methamphetamine , disrupt the 4-hour cycle by lengthening the period between bursts of activity , whereas drugs that block dopamine receptors have the opposite effect . As well as revealing a mechanism by which the brain coordinates processes that repeat several times per day , the identification of the DUO could also provide insights into the biological basis of psychiatric disorders . Conditions such as schizophrenia and bipolar disorder are often accompanied by disturbances in patterns of activity and rest . While these have previously been attributed to the disruption of circadian rhythms , there is little direct evidence for this , which raises the possibility that these changes might instead reflect the disruption of ultradian rhythms . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | A highly tunable dopaminergic oscillator generates ultradian rhythms of behavioral arousal |
The widely accepted model of G1 cell cycle progression proposes that cyclin D:Cdk4/6 inactivates the Rb tumor suppressor during early G1 phase by progressive multi-phosphorylation , termed hypo-phosphorylation , to release E2F transcription factors . However , this model remains unproven biochemically and the biologically active form ( s ) of Rb remains unknown . In this study , we find that Rb is exclusively mono-phosphorylated in early G1 phase by cyclin D:Cdk4/6 . Mono-phosphorylated Rb is composed of 14 independent isoforms that are all targeted by the E1a oncoprotein , but show preferential E2F binding patterns . At the late G1 Restriction Point , cyclin E:Cdk2 inactivates Rb by quantum hyper-phosphorylation . Cells undergoing a DNA damage response activate cyclin D:Cdk4/6 to generate mono-phosphorylated Rb that regulates global transcription , whereas cells undergoing differentiation utilize un-phosphorylated Rb . These observations fundamentally change our understanding of G1 cell cycle progression and show that mono-phosphorylated Rb , generated by cyclin D:Cdk4/6 , is the only Rb isoform in early G1 phase .
The retinoblastoma tumor suppressor protein ( Rb ) functions to regulate multiple critical cellular activities , including the late G1 restriction point , the DNA damage response checkpoint , cell cycle exit , and differentiation ( Burkhart and Sage , 2008; Paternot et al . , 2010; Henley and Dick , 2012; Johnson and Skotheim , 2013 ) . However , the Rb gene is infrequently mutated or deleted , instead upstream pathways that regulate Rb by phosphorylation on Cdk sites are altered in the majority of human cancers , including deletion and mutation of the p16 tumor suppressor and upregulation and mutation of cyclin D1 , D2 , D3 , Cdk4 and Cdk6 genes ( Sherr and McCormick , 2002; Knudsen and Knudsen , 2006; Burkhart and Sage , 2008; Henley and Dick , 2012; Choi and Anders , 2013 ) . Rb contains 16 putative Cdk phosphorylation sites that are spread throughout the protein , and all but one ( S567 ) lie outside of Rb's structured A'/B' and A/B-box or ‘pocket’ protein–protein binding domains ( Figure 1A ) . Rb is thought to exist in three generalized isoforms: ( 1 ) un-phosphorylated Rb; ( 2 ) hypo-phosphorylated Rb , also referred to as ‘under’ phosphorylated Rb or ‘partially’ phosphorylated Rb; and ( 3 ) inactive hyper-phosphorylated Rb , present in late G1 , S , G2 and M phases that is readily identifiable by SDS-PAGE as a slower migrating species ( Burkhart and Sage , 2008; Paternot et al . , 2010; Henley and Dick , 2012 ) . Surprisingly , given the scientific scrutiny of Rb over the last 25 years , the biochemical identification of the biologically active isoform ( s ) of Rb required for early G1 phase regulation , DNA damage checkpoint control , cell cycle exit , and differentiation remains unknown . To dissect Rb function and regulation , many early reports utilized supra-physiologic overexpression studies using various cyclins ( A , B , D , E ) and Cdks ( −1 , −2 , −4 , −6 ) that resulted in Rb inactivation by hyper-phosphorylation associated with an accelerated S-phase entry , and induction of E2F-dependent target genes ( Hinds et al . , 1992; Ewen et al . , 1993; Resnitzky et al . , 1994; Lundberg and Weinberg , 1998 ) . Likewise , supra-physiologic overexpression studies using Rb constructs where many , but not all , of the putative Cdk Ser/Thr consensus sites were mutated to Ala residues resulted in repressed E2F-dependent transcription and cell cycle arrest , as did overexpression of Cdk inhibitors , p16 , p21 , and p27 ( Sherr , 1994; Knudsen and Wang , 1997; Leng et al . , 1997; Sherr and McCormick , 2002; Knudsen and Knudsen , 2006; Burkhart and Sage , 2008; Paternot et al . , 2010; Henley and Dick , 2012; Choi and Anders , 2013 ) . Collectively , over the last 20 years , these studies have led to a widely accepted model of G1 cell cycle progression that proposes cyclin D:Cdk4/6 inactivates Rb during early G1 phase by progressive multi-phosphorylation , termed ‘hypo-phosphorylation’ , resulting in release of E2F transcription factors that induce expression of cyclin E , resulting in activation of cyclin E:Cdk2 complexes that complete Rb inactivation by hyper-phosphorylation in late G1 phase . The key tenet of this model is the progressive multi-phosphorylating , hypo-phosphorylation of Rb by cyclin D:Cdk4/6 complexes; however , the putative hypo-phosphorylated Rb and un-phosphorylated Rb co-migrate on 1D SDS-PAGE and cannot be separated ( Ezhevsky et al . , 2001 ) . Moreover , there is no biochemical data defining the extent or timing of phosphorylation that constitutes hypo-phosphorylated Rb . Consequently , it remains entirely unknown if hypo-phosphorylated Rb contains one , two , three , five , seven or more phosphates and at what phosphate number does the putative hypo-phosphorylated Rb become inactive to release E2F transcription factors . Thus , the critical core tenet of the G1 model that cyclin D:Cdk4/6 inactivates Rb by progressive multi-phosphorylation or hypo-phosphorylation remains unproven biochemically . We have previously found in kinetic analyses from highly synchronized normal cells and p16-deficient cancer cells that cyclin D:Cdk4/6 is constitutively active throughout early G1 phase at the same time when Rb is repressing E2F target genes ( Ezhevsky et al . , 1997 , 2001; Haberichter et al . , 2007 ) . In fact , we only observed induction of E2F target genes upon the activation of cyclin E:Cdk2 complexes and the appearance of hyper-phosphorylated Rb . These observations questioned the core biological consequences of cyclin D:Cdk4/6 progressive hypo-phosphorylation of Rb during early G1 phase . Here , for the first time , we separated all Rb isoforms by two-dimensional isoelectric focusing ( 2D IEF ) and find that Rb is exclusively mono-phosphorylated in early G1 phase in both normal and p16-deficient tumor cells . We found no experimental evidence to support the notion of progressive multi-phosphorylating hypo-phosphorylation of Rb . Using Cdk4/6-specific inhibitors and triple cyclin D-deleted MEFs , we determined that cyclin D:Cdk4/6 is the Rb mono-phosphorylating kinase that generates 14 independent mono-phosphorylated Rb isoforms in early G1 phase . At the late G1 Restriction Point , activation of cyclin E:Cdk2 complexes perform a quantum hyper-phosphorylating inactivation of all mono-phosphorylated Rb isoforms . Cells undergoing a DNA damage response activate cyclin D:Cdk4/6 complexes to generate active mono-phosphorylated Rb that regulates global transcription , whereas cells exiting the cell cycle use un-phosphorylated Rb . Together , our observations demonstrate that mono-phosphorylated Rb , generated by cyclin D:Cdk4/6 complexes , is the functionally active Rb isoform present in early G1 phase .
Rb contains 15 putative Cdk phosphorylation sites located on loops between or after structured A'/B' and A/B pocket domains ( Burke et al . , 2012; Lamber et al . , 2013 ) ( Figure 1A ) . Rb is thought to exist in three generalized biochemical states: un-phosphorylated Rb; progressive hypo-phosphorylated Rb ( also termed ‘under’ or ‘partially’ phosphorylated Rb ) ; and inactive hyper-phosphorylated Rb ( Lee et al . , 1987; DeCaprio et al . , 1989; Ludlow et al . , 1989; Mittnacht et al . , 1994; Ezhevsky et al . , 2001 ) . Although early G1 phase Rb hypo-phosphorylation was first reported 25 years ago ( Ludlow et al . , 1989 ) , the actual number , kinetics and location of phosphates on hypo-phosphorylated Rb remains entirely unknown . Analysis of synchronized primary human foreskin fibroblasts ( HFFs ) arrested in early G1 phase by contact inhibition in the presence of serum and released by replating at low density progressed through early G1 phase with constitutively active cyclin D:Cdk4/6 complexes and no evidence for increased Rb phosphorylation or transcriptional induction of Cdc6 , a key E2F target gene repressed by active Rb ( Morris and Dyson , 2001 ) ( Figure 1B ) . In contrast , Cdc6 was induced ten-fold upon cyclin E:Cdk2 activation and Rb hyper-phosphorylation , which migrates more slowly on 1D SDS-PAGE . However , the putative hypo-phosphorylated Rb isoforms co-migrated as a single fastest migrating species during all of the early G1 phase time points ( Figure 1B ) . Thus , there is either no evidence for progressive hypo-phosphorylation of Rb and/or 1D SDS-PAGE is not capable of separating all Rb phospho-isoforms . 10 . 7554/eLife . 02872 . 003Figure 1 . Rb is exclusively mono-phosphorylated in early G1 phase . ( A ) Schematic diagram of human Rb Cdk phosphorylation sites , A'/B' and A/B pocket domains . ( B ) Kinetic analysis of contact inhibited early G1 phase arrested ( +FBS ) and released primary Human Fibroblasts ( HFFs ) by 1D SDS-PAGE Rb immunoblot , anti-Cdk4/6 , and anti-Cdk2 immunoprecipitation-kinase assay , and qRT-PCR of cyclin E and cdc6 mRNA normalized to B-2-microglobulin levels . ( C ) Schematic diagram of two-dimensional isoelectric focusing ( 2D IEF ) . Immunoprecipitated Rb is loaded at origin on acidic end of IEF strip and separated first by pI . IEF strip is then soaked in SDS , run in second dimension into SDS-PAGE and immunoblotted for Rb . ( D ) 2D IEF Rb-HA immunoblot of Rb construct standards expressed in cycling cells and containing 0 ( ΔCdk ) , 1x , 2x , 3x , 6x , 9x or 15x Cdk phosphorylation sites . ( E ) Top panels: 2D IEF Rb immunoblot of primary HFFs serum deprived G0 arrested ( −FBS ) , contact inhibited early G1 phase arrested ( +FBS ) , or asynchronously cycling . Bottom panels: 2D IEF Rb immunoblot of serum deprived G0 arrested ( −FBS ) HFFs mixed with ΔCdk Rb standard , contact inhibited early G1 phase arrested ( +FBS ) HFFs mixed with single Cdk site Rb standard and contact inhibited treated with λ phosphatase . ( F ) 2D IEF Rb immunoblot from cycling human tumor cell lines expressing wild-type Rb and deregulated cyclin D:Cdk4/6 due to p16 deletion , HCT116 colon carcinoma , H1299 lung adenocarcinoma , U2OS osteosarcoma , HL60 promyelocytic leukemia . ( G ) 2D IEF Rb immunoblot from serum deprived G0 arrested ( −FBS ) and released ( +FBS ) primary HFFs from 0 to 16 hr ( H ) 2D IEF Rb immunoblot from contact inhibited early G1 phase arrested ( +FBS ) and released HFFs from 0 to 10 hr ( I ) 2D IEF Rb immunoblot from contact inhibited early G1 phase arrested ( +FBS ) and released U2OS from 0 to 10 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 00310 . 7554/eLife . 02872 . 004Figure 1—figure supplement 1 . Rb is exclusively mono-phosphorylated in early G1 phase . ( A ) Primary HFFs were synchronized in G0 by serum deprivation ( −FBS ) and released ( +FBS ) , and analyzed by Rb immunoblot , and Cdk4/6 and Cdk2 immunoprecipitation-kinase assays . ( B ) p16-deleted U2OS tumor cells were arrested in early G1 phase by contact inhibition ( +FBS ) and released ( +FBS ) , and analyzed by Rb immunoblot , and Cdk4/6 immunoprecipitation-kinase assay . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 004 Phosphates are highly acidic modifications that significantly change the isoelectric point ( pI ) of a protein . Unlike 1D SDS-PAGE , two-dimensional isoelectric focusing ( 2D IEF ) can separate specific phospho-isoforms of a protein based on total phosphate numbers regardless of position within the protein or nature of the modified residue ( Figure 1C ) . Therefore , we utilized 2D IEF to ascertain the extent and kinetics of the putative progressive multi-phosphorylated hypo-phosphorylation of Rb during early G1 phase . First , we calibrated the 2D IEF by generating a non-phosphorylatable Rb construct ( ΔCdk Rb ) standard where 15 of the 16 putative Cdk sites were converted to Ala residues ( Figure 1A ) , plus we added an N-terminal HA tag . We left S567 unaltered because it is buried in the central core of Rb's A-box and solvent inaccessible ( Lee et al . , 1998 ) . The isoelectric point of un-phosphorylated Rb is 8 . 1 , and 2D IEF of the ΔCdk Rb construct expressed in asynchronous cycling cells focused as a single basic species with a pI ∼8 ( Figure 1D ) , confirming that Rb is only phosphorylated on Cdk sites and that S567 is not phosphorylated in vivo . Starting with ΔCdk Rb , we generated Rb phospho-isoform standards by restoring one ( Rb1xCdk ) , two ( Rb2xCdk ) , three ( Rb3xCdk ) , six ( Rb6xCdk ) , nine ( Rb9xCdk ) , or fifteen ( Rb15xCdk ) Cdk sites on Rb . 2D IEF of single Cdk site Rb1xCdk construct expressed in cycling cells focused as a single phosphorylated Rb species with a more acidic pI ∼7 . 0 , that we termed mono-phosphorylated Rb ( Figure 1D ) . The double Cdk site Rb2xCdk construct focused as two spots: mono-phosphorylated Rb and di-phosphorylated Rb . Surprisingly , 2D IEF of the Rb3xCdk construct focused as mono-phosphorylated and tri-phosphorylated Rb , with no di-phosphorylated Rb ( Figure 1D ) . Likewise , 2D IEF of Rb constructs containing six ( Rb6xCdk ) , nine ( Rb9xCdk ) and fifteen ( Rb15xCdk ) Cdk sites resulted in the appearance of mono-phosphorylated Rb , plus either a six , nine , or >12 phosphate ( pI < 4 ) Rb species , respectively , and the absence of any intermediate Rb phospho-isoforms ( Figure 1D ) . Thus , unlike 1D SDS-PAGE , 2D IEF quantitatively separated all Rb isoforms from un-phosphorylated Rb to mono-phosphorylated Rb and all multi-phosphorylated Rb isoforms up to hyper-phosphorylated Rb . We next analyzed endogenous , wild-type Rb from primary HFFs by 2D IEF . Consistent with no 32P-labeling of Rb from G0 arrested cells ( DeCaprio et al . , 1989; Ezhevsky et al . , 2001 ) , Rb from serum deprived G0 arrested primary HFFs focused as a single , basic isoform with a pI ∼8 ( Figure 1E ) . 2D IEF of Rb from G0-arrested primary HFFs mixed with non-phosphorylatable ΔCdk Rb standard , co-focused as a single un-phosphorylated Rb species . In contrast , Rb from contact inhibited early G1 phase arrested ( +FBS ) HFFs focused as a single mono-phosphorylated species with a pI ∼7 . 0 ( Figure 1E ) . 2D IEF of Rb from contact inhibited HFF Rb mixed with the single Rb1xCdk construct standard confirmed that contact inhibited cells contain only mono-phosphorylated Rb . Lambda phosphatase treatment of contact inhibited mono-phosphorylated Rb collapsed into un-phosphorylated Rb ( Figure 1E ) . 2D IEF of Rb from asynchronously cycling primary HFFs focused as two isoforms: mono-phosphorylated Rb plus a very acidic hyper-phosphorylated Rb isoform ( pI < 4 ) ( Figure 1E ) . The majority of human tumors expressing wild-type Rb contain oncogenic mutations that upregulate cyclin D:Cdk4/6 kinase activity ( Sherr and McCormick , 2002; Choi and Anders , 2013 ) . We examined four disparate human tumor cell lines that express wild-type Rb , and are deleted for the p16 tumor suppressor gene , a specific inhibitor of Cdk4/6 . Surprisingly , 2D IEF of Rb from cycling populations of all four tumor cell lines , HCT116 colon carcinoma , H1299 lung adenocarcinoma , U2OS osteosarcoma , and HL60 promyelocytic leukemia , showed the presence of only mono-phosphorylated Rb and hyper-phosphorylated Rb , with no evidence of multi-phosphorylated , hypo-phosphorylated Rb isoforms , even though cyclin D:Cdk4/6 was deregulated in all four tumor cell types ( Figure 1F ) . Together , these results present several significant insights into the biochemical properties of Rb phosphorylation in vivo . First , G0-arrested cells contain un-phosphorylated Rb . Second , Rb from both cycling normal and p16-deleted tumor cells is only mono-phosphorylated or hyper-phosphorylated in vivo . Third , these observations point to phosphorylation of Rb by two entirely independent cyclin:Cdk activities: ( 1 ) a Rb mono-phosphorylating Cdk activity that places one , and only one , phosphate on Rb; and ( 2 ) a Rb hyper-phosphorylating Cdk activity that places >12 phosphates on Rb . The current widely accepted model of G1 cell cycle progression proposes that Rb becomes progressively more hypo-phosphorylated by cyclin D:Cdk4/6 complexes as cells advance through early G1 phase . To test this notion , we performed kinetic analyses on HFF cells arrested in G0 by serum deprivation and restimulated by serum addition ( +FBS ) to enter early G1 phase ( Figure 1G ) . G0 cells contained only un-phosphorylated Rb , but by 1 hr post-stimulation , a small amount of mono-phosphorylated was detected , concurrent with activation of cyclin D:Cdk4/6 complexes ( Figure 1—figure supplement 1A ) . By 3 hr , only mono-phosphorylated Rb was present . Surprisingly , Rb remained exclusively mono-phosphorylated throughout the entire early G1 phase time points at 3 , 4 , 5 , 6 , 8 , 10 , 12 . and 14 hr with no higher order phosphorylation species detected ( Figure 1G ) , even though cyclin D:Cdk4/6 complexes were constitutively active . At 16 hr post-release , we detected a quantum switch-like shift to hyper-phosphorylated Rb concomitant with activation of cyclin E:Cdk2 complexes ( Figure 1—figure supplement 1A ) . We next performed a kinetic analysis on primary HFFs arrested in early G1 phase by contact inhibition ( +FBS ) and released by replating at low density . Rb remained exclusively mono-phosphorylated throughout the entire early G1 phase time points from 0 to 7 hr with no higher order Rb phosphorylation species detected ( Figure 1H ) . Cyclin D:Cdk4/6 kinase activity was constitutively active in contact arrested early G1 phase cells and throughout all of early G1 phase with no detectable cyclin E:Cdk2 kinase activity ( Figure 1B ) . At 10 hr post-release , we detected a strong shift to hyper-phosphorylated Rb ( >12 phosphates ) , concomitant with activation of cyclin E:Cdk2 complexes and transcriptional induction of cdc6 , an E2F target gene ( Figure 1B ) . Again , we surprisingly did not detect any evidence for the progressive multi-phosphorylation or hypo-phosphorylation of Rb in early G1 phase or increases in E2F target genes , even though cyclin D:Cdk4/6 complexes were constitutively active . We next examined synchronized p16-deleted human U2OS osteosarcoma tumor cells ( Figure 1I ) . Contact inhibited early G1-arrested U2OS cells ( +FBS ) contained only mono-phosphorylated Rb , with no higher order Rb phosphorylated species , and active cyclin D:Cdk4/6 complexes ( Figure 1—figure supplement 1B ) . Rb remained exclusively mono-phosphorylated throughout all of the early G1 phase time points at 1 , 2 , 3 , 4 , 5 , 6 , 7 , and 8 hr in the presence of constitutively active cyclin D:Cdk4/6 complexes with no evidence of progressive hypo-phosphorylation . We first detected hyper-phosphorylated Rb at 10 hr post-release ( Figure 1I ) . Together , these observations demonstrated that both primary and tumor cells exclusively generate mono-phosphorylated Rb during all of early G1 phase before being converted in a quantum step to hyper-phosphorylated Rb at the late G1 phase Restriction Point . Collectively , we performed hundreds of 2D IEFs on Rb from 11 cell types and found no biochemical evidence to support the notion of progressive multi-phosphorylation or hypo-phosphorylation of Rb in early G1 phase . In all of our 2D IEFs of Rb , we detected one , and only one , phosphate on Rb during early G1 phase . However , Mittnacht et al . ( 1994 ) reported that tryptic phospho-peptide mapping ( where the 32P-labeled protein is cleaved into small peptides by trypsin digestion and then separated by charge and hydrophobicity ) of total Rb isolated from early G1 phase cells ( labeled as hypo-phosphorylated Rb ) retained the vast majority of the same phospho-peptide spots that hyper-phosphorylated Rb contained . In light of our new observations showing only mono-phosphorylated Rb present in early G1 phase , the Mittnacht et al . ( 1994 ) study suggested the potential for the presence of many mono-phosphorylated Rb isoforms that when summed together would result in the observed phospho-peptide pattern . To ascertain how many of Rb's 15 Cdk sites ( Figure 1A ) are mono-phosphorylated by cyclin D:Cdk4/6 in early G1 phase , we used a series of phospho-specific Rb antibodies to immunoblot Rb from HFF cells arrested in G0 by serum deprivation ( un-phosphorylated Rb ) and early G1 phase arrested by contact inhibition ( mono-phosphorylated Rb ) ( Figure 2A ) . While none of the antibodies recognized un-phosphorylated Rb from G0 arrested HFFs , all of the phospho-specific Rb antibodies recognized mono-phosphorylated Rb isoforms from early G1 phase arrested cells , including S249/S252 , T373 , S608 , S612 , S795 , S807/S811 , T821 and T826 ( Figure 2A ) . Similar mono-phosphorylated Rb isoform results were obtained from contact inhibited early G1 phase arrested p16-deficient U2OS cells ( Figure 2A ) . We also note that all of the phospho-specific antibodies recognized hyper-phosphorylated Rb from S phase arrested cells ( Figure 2—figure supplement 1 ) . Together , these observations suggested the presence of at least 8 individual mono-phosphorylated Rb isoforms . 10 . 7554/eLife . 02872 . 005Figure 2 . Mono-phosphorylated Rb exists as fourteen individual isoforms . ( A ) Phospho-specific Rb immunoblot of un-phosphorylated Rb ( G0 , −FBS ) and mono-phosphorylated Rb ( contact inhibited early G1 , +FBS ) from HFF and U2OS cells . ( B ) T826 and S608 phospho-specific Rb immunoprecipitation of mono-phosphorylated Rb ( contact inhibited early G1 , +FBS ) and hyper-phosphorylated Rb ( S phase ) from HFF cells , followed by phospho-specific Rb immunoblot analysis , as indicated . Note the absence of other phosphates at these locations on mono-phosphorylated Rb , but present on hyper-phosphorylated Rb . ( C ) 2D IEF Rb-HA immunoblot of single Cdk site Rb-HA constructs expressed in cycling cells . Numbering indicates single Cdk site location on Rb . ( D ) Immunoblot of Rb-HA single Cdk site constructs and control ΔCdk Rb construct from co-immunoprecipitated and co-expressed E1a , E2F1 , E2F2 , E2F3 or E2F4 ( Myc tagged ) as indicated . Numbering indicates single Cdk site location on Rb . Note that all 14 avidly bind to at least two E2F family members and there are no completely inactive mono-phosphorylated Rb isoforms . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 00510 . 7554/eLife . 02872 . 006Figure 2—figure supplement 1 . Mono-Phosphorylated Rb exists as fourteen individual isoforms . U2OS S phase arrested cells ( hydroxyurea ) were immunoblotted with indicated phospho-specific Rb antibodies or pan Rb antibody . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 006 To independently confirm the presence of mono-phosphorylated Rb isoforms , we immunoprecipitated mono-phosphorylated Rb from contact inhibited early G1 phase arrested HFFs with either the T826 or S608 phospho-specific Rb antibodies , and then immunoblotted with five phospho-specific Rb antibodies ( Figure 2B ) . Consistent with the 2D IEF data , the phospho-specific immunoprecipitation of T826 mono-phosphorylated Rb was only recognized by immunoblot with the T826 phospho-specific antibody and not by T373 , S608 , S612 , or S795 phospho-specific antibodies . Likewise , immunoprecipitation of S608 mono-phosphorylated Rb was only recognized by the S608 phospho-specific antibody and not by the other phospho-specific antibodies . In contrast , immunoprecipitated hyper-phosphorylated Rb from S phase arrested HFF cells with either the T826 or S608 phospho-specific antibodies was recognized by multiple phospho-specific antibodies ( Figure 2B ) , supporting the presence of multiple phosphates on individual hyper-phosphorylated Rb molecules . To identify the extent of mono-phosphorylated Rb isoforms , we generated all 15 individual single Cdk site Rb constructs . 2D IEFs on each of the single Cdk site Rb constructs expressed in cells determined that 14 of the single Cdk site Rb constructs were mono-phosphorylated in vivo ( Figure 2C ) . T5 , which is not evolutionary conserved below primates , was not phosphorylated . Together , these observations demonstrate the presence of 14 independent mono-phosphorylated Rb isoforms in early G1 phase and explain the large number of Rb tryptic phospho-peptide spots observed by Mittnacht et al . ( 1994 ) . Moreover , because early G1 phase cells exclusively contain mono-phosphorylated Rb , by definition , some , most or all of the 14 mono-phosphorylated Rb isoforms must be biologically active . Rb has been shown to bind to four members of the E2F family of transcription factors ( E2F1-4 ) and over 100 additional cellular proteins ( Morris and Dyson , 2001 ) . We hypothesized that the generation of 14 mono-phosphorylated Rb isoforms may serve as a post-translational mechanism to diversify Rb from a single un-phosphorylated protein in G0 into 14 independently functionalized mono-phosphorylated Rb isoforms that each bind specific cellular targets during early G1 phase . To test this hypothesis , we independently co-transfected each single Cdk site mono-phosphorylated Rb-HA construct and control un-phosphorylatable ΔCdk Rb-HA into cells co-expressing the E1a oncoprotein or E2F-1 , E2F-2 , E2F-3 , E2F-4 transcription factors ( Myc tagged ) , then individually immunoprecipitated E1a and each E2F , and immunoblotted for the associated Rb mono-phosphorylated isoforms ( Figure 2D ) . Given it's role in driving adenovirus infected quiescent G0 cells ( containing un-phosphorylated Rb ) into G1 phase and then S phase , it was not too surprising that the E1a oncoprotein bound equally well to un-phosphorylated Rb and all 14 mono-phosphorylated Rb isoforms . This observation also showed that all single Cdk site Rb constructs were correctly folded in vivo . Surprisingly , we found that none of the mono-phosphorylated Rb isoforms were completely inactive for binding to E2Fs and that each E2F showed a preferential binding specificity for individual mono-phosphorylated Rb isoforms ( Figure 2D ) . While E2F2 and E2F3 showed subtle biases for specific mono-phosphorylated Rb isoforms , E2F1 and E2F4 showed the largest enhanced or decreased binding specificities to each Rb mono-phosphorylated isoform . E2F1 showed enhanced binding to Rb when it was mono-phosphorylated at positions S230 , S249 , T356 and S612 and decreased avidity when Rb was mono-phosphorylated at positions S608 , S795 and T826 ( Figure 2D ) . E2F1 bound all of the other mono-phosphorylated Rb isoforms with comparable avidity as control un-phosphorylated ΔCdk Rb . E2F4 showed enhanced binding when Rb was mono-phosphorylated at positions S230 , S788 and S811 , and decreased binding when it was mono-phosphorylated at positions S249 , T373 , S780 and T826 ( Figure 2D ) . While phosphorylation of T373 has recently been singled out as an inactivating phosphorylation on a fragment of Rb ( Burke et al . , 2012 ) , in our hands E1a , E2F1 , E2F2 , and E2F3 all bound T373 mono-phosphorylated Rb when the full-length protein was expressed in cells . Together , these observations demonstrated the presence of 14 independent mono-phosphorylated Rb isoforms that are present in early G1 phase and showed that each has differential binding preferences to E2F family members . These results parallel other signaling proteins where phosphorylation of specific sites enhance or decrease binding to cellular targets . Based on the constitutive cyclin D:Cdk4/6 activity in early G1 phase when Rb was exclusively mono-phosphorylated ( Figure 1B ) , cyclin D:Cdk4/6 became a prime candidate for the Rb mono-phosphorylating kinase . To dissect the role of cyclin D:Cdk4/6 to phosphorylate Rb , we used triple knockout ( TKO ) cyclin D genetic deletion in mouse embryonic fibroblasts ( MEFs ) ( Choi et al . , 2012 ) . MEFs containing a deleted cyclin D2 gene and homozygous LoxP cyclin D1f/f and D3f/f genes were treated with Adenovirus Cre recombinase to generate TKO cyclin D− MEFs . TKO cyclin D− MEFs continuously cycled and contained hyper-phosphorylated Rb by 1D SDS-PAGE ( Figure 3A , Figure 3—figure supplement 1 ) . 2D IEF of Rb from asynchronously cycling parental D1+/D3+ MEFs showed the presence of both mono-phosphorylated Rb and hyper-phosphorylated Rb isoforms ( Figure 3B ) . In contrast , 2D IEF of Rb from cycling TKO cyclin D− MEFs contained un-phosphorylated Rb and hyper-phosphorylated Rb , with no mono-phosphorylated Rb detected ( Figure 3B ) . Moreover , contact inhibited early G1 phase arrested ( +FBS ) parental D1+/D3+ MEFs contained only mono-phosphorylated Rb , whereas contact inhibited ( +FBS ) TKO cyclin D− MEFs contained only un-phosphorylated Rb ( Figure 3B ) . Retroviral expression of cyclin D1 in contact inhibited TKO cyclin D− MEFs resulted in the appearance of mono-phosphorylated Rb ( Figure 3B ) . 10 . 7554/eLife . 02872 . 007Figure 3 . Cyclin D:Cdk/6 is the Rb mono-phosphorylation kinase . ( A ) Immunoblot of cyclin D1 , D3 , Rb in cycling parental D1+/D3+ MEFs and triple knockout ( TKO ) cyclin D− MEFs . ( B ) 2D IEF Rb immunoblot from cycling parental D1+/D3+ MEFs and TKO cyclin D− MEFs , and contact inhibited early G1 phase arrested ( +FBS ) parental D1+/D3+ MEFs and TKO cyclin D− MEFs plus/minus retroviral cyclin D1 expression . ( C ) 2D IEF Rb immunoblot from serum deprived G0 arrested ( −FBS ) and released ( +FBS ) HFFs plus control DMSO , Cdk4 inhibitor ( PD0332991 ) or retroviral p16 expression . ( D ) 2D IEF Rb immunoblot from serum deprived G0 arrested ( −FBS ) and released ( +FBS ) p16-deleted U2OS tumor cells or released ( +FBS ) plus TET-induced p16 expression . ( E , F ) 1D SDS-PAGE ( E ) and 2D IEF ( F ) Rb immunoblot of G2/M phase nocodazole arrested ( Noc . ) and released U2OS cells plus DMSO ( con ) or Cdk4 inhibitor ( PD0332991 ) . ( G , H ) 1D SDS-PAGE ( G ) and 2D IEF ( H ) Rb immunoblot of G2/M phase nocodazole arrested ( Noc . ) and released HeLa cells plus DMSO ( con ) or Cdk4 inhibitor ( PD0332991 ) . ( I ) Late G1 phase primary HFFs were treated with Cdk2 inhibitor Roscovitine [15 μM] or control ( DMSO ) and analyzed by 1D SDS-PAGE Rb immunoblot , anti-Cdk4/6 and anti-Cdk2 immunoprecipitation-kinase assays . ( J ) 2D IEF Rb immunoblot from late G1 phase HFFs treated with dose curve of Cdk2 inhibitor Roscovitine or control ( DMSO ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 00710 . 7554/eLife . 02872 . 008Figure 3—figure supplement 1 . Cyclin D:Cdk/6 is the Rb mono-phosphorylation kinase . Growth curve analysis of parental D1+/D3+ MEFs and triple knockout ( TKO ) cyclin D− MEFs plated at low density ( 10 , 000 ) in 10% FBS and counted each day over 7 days . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 008 To further test the notion that cyclin D:Cdk4/6 is the Rb mono-phosphorylating kinase , we treated serum deprived G0 arrested ( −FBS ) and restimulated ( +FBS ) HFFs to enter early G1 phase with a selective Cdk4/6 inhibitor ( PD0332991 ) ( Fry et al . , 2004 ) . Consistent with TKO cyclin D− MEFS , treatment of HFFs with the Cdk4/6 inhibitor at restimulation ( +FBS ) resulted in the presence of un-phosphorylated Rb , whereas control DMSO-treated cells contained mono-phosphorylated Rb ( Figure 3C ) . Likewise , specific inhibition of cyclin D:Cdk4/6 complexes by retroviral expression of the Cdk4/6 inhibitor p16 resulted in the presence of un-phosphorylated Rb ( Figure 3C ) . Furthermore , induction of p16 in serum deprived G0 arrested ( −FBS ) and restimulated ( +FBS ) U2OS cells ( Jiang et al . , 1998 ) also resulted in the presence of un-phosphorylated Rb , whereas control cells ( repressed p16 ) contained mono-phosphorylated Rb ( Figure 3D ) . Collectively , these results from three independent approaches ( genetic TKO cyclin D− MEFs , expression of p16 , treatment with Cdk4/6 inhibitor ) in three different cell types ( MEFs , HFFs , U2OS ) confirmed that cyclin D:Cdk4/6 was the Rb mono-phosphorylating kinase in vivo . Rb remains hyper-phosphorylated in M phase and is dephosphorylated by activation of the PP1 phosphatase as cells exit mitosis into the next early G1 phase ( Ludlow et al . , 1993 ) . To ascertain if phosphatase activity dephosphorylates Rb to mono-phosphorylated or un-phosphorylated isoforms , we arrested U2OS cells at G2/M phase by addition of nocodazole , a microtubule depolymerizer , and detected only hyper-phosphorylated Rb ( Figure 3E , F ) . After nocodazole washout and release of U2OS cells into early G1 phase for 4 hr , all of the hyper-phosphorylated Rb was converted to mono-phosphorylated Rb . However , release of U2OS cells from the nocodazole arrest in the presence of the Cdk4/6 inhibitor ( PD0332991 ) resulted in the exclusive appearance of un-phosphorylated Rb ( Figure 3E , F ) . Likewise , nocodazole block and release of HeLa cells in the presence of the Cdk4/6 inhibitor ( PD0332991 ) resulted in the presence of only un-phosphorylated Rb ( Figure 3G , H ) . These results determined that as cells exit mitosis , Rb is fully dephosphorylated to un-phosphorylated Rb , and then rapidly mono-phosphorylated by cyclin D:Cdk4/6 complexes in early G1 phase . Previous studies from our lab and many others have shown that Rb becomes inactivated by hyper-phosphorylation at the late G1 Restriction point and remains hyper-phosphorylated throughout late G1 phase , S phase , G2 phase , and M phases ( DeCaprio et al . , 1989; Mittnacht et al . , 1994; Ezhevsky et al . , 2001 ) . Hyper-phosphorylated Rb first appears concomitant with activation of cyclin E:Cdk2 complexes ( Figure 1B ) . To ascertain the role of cyclin E:Cdk2 as the initial Rb hyper-phosphorylating kinase , we analyzed Rb from contact inhibited and released HFFs that were allowed to enter late G1 phase at 12 hr . We found that the vast majority of Rb was hyper-phosphorylated , with active Cdk2 and active Cdk4/6 complexes ( Figure 3I ) . Selective inhibition of Cdk2 by 15 μM roscovitine , a Cdk2 ATP competitive inhibitor , resulted in no Cdk2 activity , continued Cdk4/6 activity and the presence of mono-phosphorylated Rb ( Figure 3I , J ) . Titration of roscovitine from 2 μM to 15 μM resulted in a dose-dependent appearance of intermediate phosphorylated Rb isoforms ( Figure 3J ) , suggesting that unlike cyclin D:Cdk4/6 , cyclin E:Cdk2 is a processive Rb kinase . These observations are entirely consistent with our previous reports that cyclin E:Cdk2 complexes are the initial Rb hyper-phosphorylating kinase at the late G1 Restriction Point ( Ezhevsky et al . , 1997 , 2001; Haberichter et al . , 2007 ) . Although Rb regulates many processes in early G1 phase ( Burkhart and Sage , 2008 ) , to ascertain if mono-phosphorylated Rb was functionally active , we focused on Rb's regulation of a DNA damage response cell cycle arrest ( Harrington et al . , 1998; Brugarolas et al . , 1999; Knudsen et al . , 2000; Avni et al . , 2003 ) . Treatment of cycling MEFs with a sub-lethal dose ( 100 ng/ml ) of doxorubicin , a DNA damage-inducing topoisomerase II inhibitor , resulted in a G1 phase cell cycle arrest with constitutive cyclin D:Cdk4/6 activity , mono-phosphorylated Rb , and loss of cyclin E/A:Cdk2 activity ( Figure 4A , B ) . Surprisingly , doxorubicin treatment of serum-deprived G0 arrested MEFs that contained un-phosphorylated Rb and no cyclin D:Cdk4/6 activity , resulted in induction of cyclin D1 and Cdk6 , activation of cyclin D:Cdk4/6 complexes and Rb mono-phosphorylation ( Figure 4C , D ) . Although several studies have suggested that Rb is phosphorylated during a DNA damage response by non-Cdk kinases , including Chk1/2 and Aurora B ( Inoue et al . , 2007; Nair et al . , 2009 ) , inhibition of Cdk4/6 activity by retroviral p16 expression in doxorubicin-treated MEFs resulted in the exclusive presence of un-phosphorylated Rb ( Figure 4B , D ) , thereby excluding the involvement of other kinases . Together , these observations demonstrated that in response to DNA damage , cells select for mono-phosphorylated Rb by activating cyclin D:Cdk4/6 complexes . 10 . 7554/eLife . 02872 . 009Figure 4 . DNA damage induces cyclin D:Cdk4/6 activity and mono-phosphorylated Rb . ( A ) Asynchronously cycling MEFs ( +FBS ) treated with Doxorubicin ( +Doxo , 100 ng/ml ) were analyzed by 1D SDS-PAGE Rb immunoblot , anti-Cdk4/6 and anti-Cdk2 immunoprecipitation-kinase assays . ( B ) 2D IEF Rb immunoblot from asynchronously cycling MEFs ( +FBS ) treated with control or Doxorubicin ( +Doxo ) or Doxorubicin plus retroviral p16 ( +Doxo/+p16 ) expression . ( C ) Serum-deprived G0 arrested MEFs ( −FBS ) treated with Doxorubicin ( +Doxo/−FBS ) were immunoblot analyzed for cyclin D1 , Cdk4 , Cdk6 , and Cdk4/6 immunoprecipitation-kinase activity . Note activation of cyclin D:Cdk4/6 in absence of serum growth factors . ( D ) 2D IEF Rb immunoblot from serum-deprived G0 arrested MEFs ( −FBS ) treated with control or Doxorubicin ( +Doxo ) or Doxorubicin plus retroviral p16 ( +Doxo/+p16 ) expression . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 009 We developed a genetic system to test if un-phosphorylated Rb and/or mono-phosphorylated Rb were functionally active to mediate the DNA damage cell cycle arrest checkpoint . Parental homozygous LoxP Rbf/f MEFs ( Marino et al . , 2000 ) were treated with TAT-Cre protein ( Wadia et al . , 2004 ) to delete Rb ( Rb−/− ) , followed by infection with carefully engineered and titered wild type ( WT ) Rb-HA or non-phosphorylatable ΔCdk Rb-HA retroviruses ( Figure 5A ) . This approach resulted in an acute loss of endogenous Rb protein combined with simultaneous replacement by ectopic Rb at physiologic levels ( Figure 5A ) . Treatment of Rb-deleted MEFs expressing ectopic WT Rb-HA with doxorubicin resulted in the presence of mono-phosphorylated Rb , whereas Rb-deleted MEFs expressing ectopic ΔCdk Rb-HA contained un-phosphorylated Rb ( Figure 5B ) . 10 . 7554/eLife . 02872 . 010Figure 5 . Mono-phosphorylated Rb is active during a DNA damage response . ( A ) Rb immunoblot of contact inhibited parental MEFs , conditionally deleted Rb−/− MEFs , and retrovirally-expressed wild type ( WT ) Rb-HA or non-phosphorylatable ΔCdk Rb-HA in deleted Rb−/− MEFs . ( B ) 2D IEF Rb immunoblot of cycling and doxorubicin ( +Dox ) ( 100 ng/ml ) treated WT Rb-HA MEFs , and ΔCdk Rb-HA MEFs in deleted Rb−/− background . ( C ) Microarray heat map of mRNA levels from early G1 phase Rb−/− MEFs , WT Rb-HA MEFs , and ΔCdk Rb-HA MEFs compared to parental MEFs treated with Doxorubicin ( +Doxo ) for 3 hr . Only genes increased >1 . 6-fold are shown . ( D ) Focused mRNA expression analysis of E2F target genes and p21 control gene from ( C ) . ( E–G ) qRT-PCR mRNA analysis of endogenous E2F-dependent target genes , cdc6 ( E ) and DHFR ( F ) , and a non-E2F control gene , p21 ( G ) , from parental MEFs ( con ) , Rb−/− MEFs , WT Rb-HA MEFs , and ΔCdk Rb-HA MEFs in deleted Rb−/− background treated with doxorubicin ( +Doxo ) . Mean values were normalized to β2-microglobulin levels , reported as fold change from parental MEFs ( con ) . Error bars indicate SEM from three independent experiments . ( H and I ) Quantification of percent tetraploid ( >4n DNA ) nuclei in parental MEFs ( con ) , Rb−/− MEFs , WT Rb-HA MEFs , and ΔCdk Rb-HA MEFs in deleted Rb−/− background four days after treatment with Doxorubicin ( +Doxo ) ( H ) or 2 days post-treatment with ionizing radiation ( 20 grays ) ( I ) . Error bars indicate SEM from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 01010 . 7554/eLife . 02872 . 011Figure 5—figure supplement 1 . Mono-phosphorylated Rb is active during a DNA damage response . ( A ) Total RNA isolated from contact inhibited and released Rb−/− MEFs , WT Rb-HA MEFs and ΔCdk Rb-HA MEFs in deleted Rb−/− background compared to the control parental MEFs treated with 100 ng/ml doxorubicin for 3 hr was used to probe whole-genome microarrays . Blue line indicates 1 . 6-fold increase/decrease filter . Red and green dots indicate genes with increased or decreased expression >1 . 6-fold , respectively . ( B ) Microarray heat map of mRNA levels from early G1 phase Rb−/− MEFs , WT Rb-HA MEFs and ΔCdk Rb-HA MEFs in deleted Rb−/− background compared to parental MEFs treated with Doxorubicin ( +Doxo ) for 3 hr restricted to 308 genes that were decreased >1 . 6-fold . ( C ) Total number of genes altered from ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 011 We next analyzed the ability of WT Rb-HA and ΔCdk Rb-HA to regulate global transcription during an acute DNA damage response . Parental MEFs expressing endogenous Rb , Rb-deleted MEFS , and Rb-deleted MEFS expressing either WT Rb-HA or ΔCdk Rb-HA were contact arrested in early G1 phase ( +FBS ) and released for 4 hr , then treated with a sub-lethal dose of doxorubicin ( 100 ng/ml ) for 3 hr and analyzed for whole genome transcriptional changes ( Figure 5C ) . Rb is a transcriptional repressor and Rb-deleted MEFs showed a >1 . 6x increase in expression of 173 genes , primarily involved in DNA replication ( 24% ) , cell cycle control ( 20% ) , and regulation of transcription ( 18% ) ( Table 1 ) . Expression of physiologic levels of mono-phosphorylated WT Rb-HA in Rb-deleted cells restored repression of many of these genes , especially E2F target genes ( Figure 5C , D; Figure 5—figure supplement 1 ) . However , expression of un-phosphorylated ΔCdk Rb-HA at physiologic levels failed to repress genes and gave a pattern of global transcriptional deregulation similar to Rb-deleted MEFs , suggesting that un-phosphorylated Rb was functionally inactive during a DNA damage response checkpoint during early G1 phase . Further qRT-PCR analysis of two strong E2F target genes , DHFR and cdc6 , showed similar levels of deregulation in both Rb-deleted and un-phosphorylated ΔCdk Rb-HA MEFs , whereas WT Rb-HA repressed both of these genes ( Figure 5E–G ) . A non-Rb regulated gene , p21 , showed no difference between all three genotypes . Rb-deleted ( Rb−/− ) MEFs also failed to prevent the appearance of tetraploid cells several days after treatment with either doxorubicin or ionizing radiation ( 20 Grays ) ( Figure 5H , I ) . Consistent with the inability to regulate transcriptional control , un-phosphorylated ΔCdk Rb-HA MEFs showed similar high levels of tetraploid cells in response to DNA damage as Rb-deleted ( Rb−/− ) MEFs . In contrast , WT Rb-HA expression rescued the tetraploid phenotype to levels near parental MEFs expressing endogenous wild type Rb ( Figure 5H , I ) . These observations demonstrated that cells undergoing a DNA damage response activate cyclin D:Cdk4/6 to generate biologically active , mono-phosphorylated Rb , whereas un-phosphorylated was functionally inactive for regulating E2F transcription and preventing the appearance of tetraploid cells . 10 . 7554/eLife . 02872 . 012Table 1 . Mono-phosphorylated Rb is active during a DNA damage responseDOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 012InducedRepressed DNA replication24% Localization18% Cell cycle20% Post-translational protein modification10% Regulation of transcription18% Biosynthetic process9% Organelle organization and biogenesis12% Organelle organization and biogenesis8% Response to stress12% Intracellular signaling cascade8% DNA repair8% Proteolysis6% DNA packaging7% Catabolic process6% Macromolecular complex assembly6% Cell cycle6% Cellular component assembly6% Nervous system development6% Chromatin assembly6% Response to external stimulus4% Response to wounding3% Dephosphorylation2%Gene ontology of mRNA differences showing a >1 . 6-fold increase/decrease level by microarray analysis between parental MEFs and Rb−/− MEFs treated with 100 ng/ml doxorubicin for 3 hr after release from contact inhibition . The above results demonstrated that un-phosphorylated Rb was non-functional during a DNA damage response . However , serum-deprived G0 arrested HFFs and U2OS cells contained un-phosphorylated Rb ( Figure 1E , G; Figure 3D ) , as do quiescent G0 peripheral blood lymphocytes ( PBLs ) ( Ezhevsky et al . , 2001 ) . Moreover , other studies have documented a role for Rb in cell cycle exit and differentiation , including proper myogenic development as myoblasts exit the cell cycle into G0 ( Gu et al . , 1993; Zacksenhaus et al . , 1996; Chen and Wang , 2000; Sage et al . , 2003; Blais et al . , 2007 ) . Together , these observations suggested a potential functional role for un-phosphorylated Rb during cell cycle exit and differentiation . To evaluate the phosphorylation status of Rb during differentiation , we used the well-established C2C12 myoblast to myotube differentiation system ( Blais et al . , 2005 ) . Asynchronous cycling C2C12 myoblasts grown in high mitogen media ( FBS ) contained both mono-phosphorylated Rb and hyper-phosphorylated Rb with no detectable un-phosphorylated Rb ( Figure 6A , B ) . However , addition of low mitogen , differentiation media induced expression of myotube specific myogenin and resulted in the exclusive presence of un-phosphorylated Rb at day 2 , concomitant with loss of cyclin D:Cdk4/6 kinase activity and expression of p18 , a Cdk4/6-specific inhibitor ( Halevy et al . , 1995; Franklin and Xiong , 1996; Wang and Walsh , 1996; Zhang et al . , 1999 ) ( Figure 6B–D ) . We also observed the exclusive appearance of un-phosphorylated Rb when human HL60 promyelocytic cells were induced to undergo differentiation by addition of retinoic acid ( Figure 6—figure supplement 1A , B ) . 10 . 7554/eLife . 02872 . 013Figure 6 . Un-phosphorylated Rb promotes cell cycle exit and differentiation . ( A ) Rb immunoblot of cycling C2C12 myoblasts ( Asynch ) , and after 2 days in differentiation medium ( Diff ) . ( B ) 2D IEF Rb immunoblot from cycling C2C12 myoblasts ( Asynch ) and after 2 days in differentiation medium ( Diff ) . ( C ) Cdk4/6 immunoprecipitation-kinase assay of cycling C2C12 myoblasts ( 0 ) and at 1 and 2 days post-addition of differentiation medium ( Diff ) . Negative ( neg ) control , irrelevant antibody . ( D ) Myogenin immunoblot of cycling C2C12 myoblasts ( Asynch ) , and after 2 days in differentiation medium ( Diff ) . ( E ) Immunoblot of endogenous Rb , and retroviral expressed wild type ( WT ) Rb-HA or ΔCdk Rb-HA in C2C12 myoblasts co-infected with short hairpin ( sh ) RNAi retroviruses targeting endogenous Rb ( 3' UTR ) or scrambled ( Scr ) control . ( F and G ) Proliferation analysis ( number of nuclei ) of C2C12 myoblasts co-infected with short hairpin ( sh ) RNAi retroviruses targeting endogenous Rb ( 3' UTR ) or scrambled ( Scr ) control , and wild type ( WT ) Rb-HA or ΔCdk Rb-HA retroviruses . Cells were stained with Hoechst 33342 DNA dye , visualized by microscopy ( F ) and quantified by flow cytometry ( G ) ( number of nuclei x 103 ) at 2 days post-addition of differentiation medium . Error bars indicate SEM . ( H ) qRT-PCR mRNA analysis of Mcm3 DNA replication factor in C2C12 myoblasts treated as above ( G ) for 2 days in differentiation media . Mean values were normalized to β2-microglobulin levels and reported as fold change from C2C12 myotubes expressing endogenous Rb ( ScrshRNA ) . Error bars indicate SEM from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 01310 . 7554/eLife . 02872 . 014Figure 6—figure supplement 1 . Un-phosphorylated Rb promotes cell cycle exit and differentiation . ( A and B ) 1D SDS-PAGE ( A ) and 2D IEF ( B ) Rb immunoblot of human HL60 promyelocytic cells treated with retinoic acid to undergo cell cycle exit and differentiation . ( C ) qRT-PCR mRNA analysis of Mcm5 DNA replication factor in C2C12 myoblasts treated for 2 days in differentiation media . Mean values were normalized to β2-microglobulin levels and reported as fold change from C2C12 myotubes expressing endogenous Rb ( ScrshRNA ) . Error bars indicate SEM from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 014 To directly test for a role of un-phosphorylated Rb in regulating cell cycle exit and differentiation in myoblasts , we devised a strategy similar to the MEF DNA damage response approach used above . Endogenous Rb from C2C12 myoblasts was knocked down by shRb or control shScramble ( Scr ) RNAi retroviruses , followed by retroviral physiologic expression of WT Rb-HA or ΔCdk Rb-HA ( Figure 6E ) . As per the method of Blais et al . ( 2007 ) , the efficiency of cell cycle exit was evaluated 2 days post-addition of low mitogen differentiation media by counting the number of nuclei ( Figure 6F–H ) . Reduction of Rb levels by shRNAi resulted in an increased proliferation index ( number of nuclei ) , and increased expression of Mcm3 and Mcm5 ( markers of DNA replication ) ( Figure 6F–H , Figure 6—figure supplement 1C ) . However , physiologic expression of WT Rb-HA compensated for the reduction in endogenous Rb by bringing the proliferation index back to the shScrambled RNAi control . Strikingly , physiologic expression of ΔCdk Rb-HA resulted in a dramatic reduction of proliferation as cells prematurely exited the cell cycle and decreased Mcm3 and Mcm5 levels ( Figure 6F–H; Figure 6—figure supplement 1C ) . Thus , in contrast to the DNA damage response where un-phosphorylated Rb was non-functional , these observations demonstrated that un-phosphorylated Rb was functionally active to drive cell cycle exit and differentiation of myoblasts into myotubes .
For the last 20 years , the key tenet of the prevailing model of G1 cell cycle progression proposed that cyclin D:Cdk4/6 complexes inactivated Rb by progressive multi-phosphorylation , termed hypo-phosphorylation , resulting in the gradual release of E2F transcription factors that drive cells into late G1 phase . This notion was reinforced when tumor cells expressing wild-type Rb were found to have genetic and epigenetic alterations of the p16 tumor suppressor gene or oncogenic expression of cyclin D1 , D2 , D3 , Cdk4 , and Cdk6 genes , which became known as the ‘p16-cyclin D-Rb’ pathway ( Sherr , 1994; Sherr and McCormick , 2002; Burkhart and Sage , 2008; Paternot et al . , 2010; Henley and Dick , 2012; Choi and Anders , 2013 ) . Further reinforcing this notion were experiments utilizing supra-physiologic overexpression of D-type cyclins and Cdk4/Cdk6 that inactivated Rb and drove cells into S phase . Likewise , experiments overexpressing p16 , a Cdk4/6 inhibitor , arrested cells in a 2n DNA content and were interpreted to confirm the notion that cyclin D:Cdk4/6 complexes inactivated Rb . Moreover , the large number of tryptic phospho-peptides of ‘hypo-phosphorylated’ Rb from early G1 cells reported by Mittnacht et al . ( 1994 ) was unknowingly misinterpreted and further reinforced the notion that cyclin D:Cdk4/6 complexes inactivate Rb by progressive multi-phosphorylating , hypo-phosphorylation . Although there was a complete absence of rigorous biochemical evidence as to the extent of phosphate numbers or kinetics on what was loosely termed hypo-phosphorylated Rb , the notion that cyclin D:Cdk4/6 inactivated Rb by progressive multi-phosphorylating , hypo-phosphorylation was solidified in the 1990s as the model of G1 cell cycle progression . In sharp contrast to the prevailing model , we had previously performed kinetic analyses from highly synchronized normal cells and p16-deficient tumor cells , and found that cyclin D:Cdk4/6 was constitutively active throughout all of the early G1 phase at the same time points that Rb was actively binding E2Fs and repressing E2F target gene expression ( Ezhevsky et al . , 1997 , 2001; Haberichter et al . , 2007 ) . Consequently , we suspected that cyclin D:Cdk4/6 complexes were not inactivating Rb , but may , in fact , be activating Rb by phosphorylation . The key to understanding the relationship between Rb and cyclin D:Cdk4/6 complexes was developing the ability to quantitatively separate all Rb isoforms by 2D IEF combined with the generation of Rb phosphorylation standards . During the course of this study , we performed hundreds of 2D IEFs on Rb from 11 independent normal and tumorigenic cell types under conditions of asynchronous cycling cells , cells arrested in G0 , early G1 , late G1 or G2/M phases , and cells that were arrested/released and followed kinetically . Under all of these conditions and cell types , we found that Rb was exclusively mono-phosphorylated throughout all of early G1 phase in both normal and tumor cells , and hyper-phosphorylated in late G1 , S , G2 , and M phases . In fact , we found no biochemical evidence to support the prevailing G1 cell cycle model that cyclin D:Cdk4/6 progressively hypo-phosphorylates Rb . Moreover , using three independent approaches to dissect cyclin D:Cdk4/6 function on Rb , namely: triple cyclin D genetic deletion , addition of a Cdk4/6-specific chemical inhibitor , and p16 expression , we determined that cyclin D:Cdk4/6 is the Rb mono-phosphorylating kinase ( Figure 7 ) . Given that mono-phosphorylated Rb is the only isoform of Rb present in early G1 phase , by definition , some , most or all of the mono-phosphorylated Rb isoforms must be biologically active . Consequently , it was not too surprising that mono-phosphorylated Rb was the active Rb isoform mediating a DNA damage response cell cycle arrest and regulating global transcription . However , we note the unanticipated observation that exposure of quiescent G0 ( −FBS ) primary cells , containing no cyclin D:Cdk4/6 activity , to DNA damaging agents induced and activated cyclin D:Cdk4/6 complexes to mono-phosphorylate Rb in the absence of serum growth factors . This observation is consistent with a role for cyclin D:Cdk4/6 complexes in the DNA damage response checkpoint ( Choi et al . , 2012 ) , but also raises a cautionary concern for a potential increase of genomic DNA damage in the normal cells of patients being simultaneously treated with a cyclin D:Cdk4/6 inhibitor and a DNA damaging chemotherapy or ionizing radiation . 10 . 7554/eLife . 02872 . 015Figure 7 . Revised working model of G1 cell cycle progression . Un-phosphorylated Rb regulates G0 cell cycle exit and differentiation . Growth factor signaling and DNA damage stimulate activation of cyclin D:Cdk4/6 complexes that diversify Rb into 14 mono-phosphorylated isoforms that independently bind specific cellular factors to regulate early G1 phase functions and the DNA damage response . Cyclin D:Cdk4/6 mono-phosphorylation of Rb inactivates un-phosphorylated Rb G0 functions and thereby prevents cells from exiting the cell cycle . Activation of cyclin E:Cdk2 complexes inactivates all 14 mono-phosphorylated Rb isoforms by hyper-phosphorylation ( >12x phosphates ) at the late G1 Restriction Point . Cyclin A:Cdk2 and cyclin B:Cdk1 maintain Rb in an inactive hyper-phosphorylated state during S , G2 and M phases . As cells complete cytokinesis , hyper-phosphorylated Rb is de-phosphorylated by phosphatases and rapidly mono-phosphorylated by cyclin D:Cdk4/6 complexes . We speculate that an unknown metabolic sensor is upstream of cyclin E:Cdk2 activation . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 015 The Rb gene is transcribed and spliced into a single 4 . 7 kb mRNA composed of 27 coding exons that does not undergo appreciable alternative splicing and encodes a single 105 kDa protein ( Burkhart and Sage , 2008 ) . Although the complete structure of Rb has yet to be solved , several groups have recently solved the core structure as an intra-strand pseudo-dimer of dimers with the Cdk sites present on loops between structured regions or on the C-terminus abutting the B-box ( Burke et al . , 2012; Lamber et al . , 2013 ) ( Figure 1A ) . Based on the binding preferences of individual Rb mono-phosphorylated isoforms to specific E2F transcription factors , our data suggest that the generation of 14 mono-phosphorylated Rb isoforms may serve as a mechanism to post-translationally functionally diversify Rb in early G1 phase from a single un-phosphorylated isoform in G0 . In addition to E2F transcription factors , Rb has also been shown to bind to over 100 additional cellular proteins ( Morris and Dyson , 2001 ) , leaving open the potential for differential binding preferences of mono-phosphorylated Rb to specific cellular targets . Excluding hyper-phosphorylated Rb , we were surprised at the complete absence of any multi-phosphorylated Rb isoforms . This raised the mechanistic question of how does cyclin D:Cdk4/6 complexes place one , and only one , phosphate on Rb while leaving the remaining 13 Cdk sites un-phosphorylated ? While this will require extensive structural analyses beyond the scope of our study , we speculate that the substrate recognition of Rb by cyclin D's N-terminal LxCxE motif weak binding to Rb's pocket domain ( Dowdy et al . , 1993 ) vs cyclins E and A avid binding to four C-terminal R/KxL substrate motifs ( Adams et al . , 1999 ) outside of the pocket likely serves as the defining mechanism between Rb mono-phosphorylation and Rb hyper-phosphorylation ( Figure 1A ) . In this hypothesis , cyclin E/A:Cdk2's strong binding to the C-terminal tail of Rb would allow access to all 14 Cdk sites on Rb even when transcription factors and chromatin remodeling factors were bound to Rb's pocket and N-terminal binding sites . This also allows for a simultaneous switch-like inactivation of all 14 mono-phosphorylated Rb isoforms by one processive hyper-phosphorylation mechanism . We believe that the utilization of the same phosphorylation sites for the activation and inactivation of a protein , albeit with a >12-fold increased phosphate stoichiometry , is an unprecedented mechanism in the literature , but one that also likely applies to Rb-related genes p107 and p130 . Our study addresses several critical problems arising from numerous biochemical analyses of Rb phosphorylation going back more than 20 years . The results presented here fundamentally change the understanding of G1 cell cycle regulation to show that cyclin D:Cdk4/6 activates Rb for binding cellular targets during early G1 phase by generating 14 independent mono-phosphorylated Rb isoforms . Given that the majority of human tumors contain wild-type Rb , but select for deregulated cyclin D:Cdk4/6 activity ( Sherr and McCormick , 2002; Burkhart and Sage , 2008; Knudsen and Knudsen , 2006; Choi and Anders , 2013 ) , we hypothesize that the oncogenic activation of cyclin D:Cdk4/6 results in Rb mono-phosphorylation to drive quiescent G0 cells into a more metabolically active , but early G1 phase arrested phenotype ( Figure 8 ) . By constitutively mono-phosphorylating Rb , the nascent neoplastic cell avoids cell cycle exit and differentiation mediated by un-phosphorylated Rb , and also maintains a high level of metabolism . This notion is entirely consistent with the observed subtle and highly tolerated cancer predisposing mutations of p16 deletion and cyclin D overexpression in mouse models that avoid activation of oncogene-induced apoptosis ( Burkhart and Sage , 2008 ) . The net effect is a subtle , but irreversible , oncogenic step forward . While our study determined the role of cyclin D:Cdk4/6 in mono-phosphorylating Rb , it leaves wide open the question of what the rate-limiting switch-like mechanism is to activate cyclin E:Cdk2 , the first domino in Rb inactivation . Cyclin D:Cdk4/6 activity combined with other signal transduction pathway mutations contributes to increased cellular metabolism that we speculate is monitored by an unknown metabolic sensor . Once the metabolic threshold has been exceeded , the sensor activates cyclin E:Cdk2 resulting in Rb inactivation by hyper-phosphorylation , induction of E2F target gene transcription and progression across the Restriction Point into late G1 phase ( Haberichter et al . , 2007 ) . We are currently investigating the mechanics of this putative mechanism and the identity of the metabolic sensor . 10 . 7554/eLife . 02872 . 016Figure 8 . Deregulated cyclin D:Cdk4/6 in cancer mono-phosphorylates Rb to prevent cell cycle exit . Deregulation of cyclin D:Cdk4/6 activity in cells occurs by a variety of mechanisms , including: p16 deletion , cyclin D1 , D2 and D3 amplification or overexpression , and mutation or overexpression of Cdk4 or Cdk6 . Cyclin D:Cdk4/6 mono-phosphorylation of Rb simultaneously inactivates Rb's G0 functions and activates Rb's early G1 phase functions thereby driving cells from a low metabolism G0 quiescence into a high metabolism early G1 arrested state that also prevents subsequent cell cycle exit or differentiation . Similar to p53 mutations and Bcl2 overexpression , deregulated cyclin D:Cdk4/6 activity is a well tolerated priming oncogenic mutation that avoids activation of oncogene-induced apoptosis . The net effect is a subtle , but irreversible , oncogenic step forward . We predict that additional oncogenic and metabolic pathways ultimately converge on and activate cyclin E:Cdk2 complexes to inactivate Rb by hyper-phosphorylation at the Restriction Point and drive cells into late G1 phase . DOI: http://dx . doi . org/10 . 7554/eLife . 02872 . 016
Cells were G0 arrested by serum deprivation for 5 days , followed by addition of 10% FBS . Cells were plated at high density in 10% FBS to contact inhibit arrest in early G1 phase for 48 hr , followed by replating at low density in 10% FBS . DNA damage was induced by addition of 100 ng/ml doxorubicin ( Sigma , St . Louis , MO ) or exposure to 20 Grays of ionizing radiation . MEFs were prepared from Rbf/f mice ( Marino et al . , 2000 ) and cyclin D1f/f/D2−/−/D3 f/f mice ( Choi et al . , 2012 ) . Rb and cyclin D inactivation was performed by addition of TAT-Cre protein ( Wadia et al . , 2004 ) . U2OS-p16 cells ( Jiang et al . , 1998 ) were maintained in 1 μg/ml tetracycline to repress p16 expression . C2C12 myoblasts were differentiated into myotubes by incubating with DMEM plus 2% horse serum for 2 days . Human RbΔCDK−HA and murine RbΔCDK−HA were generated by changing all 15 Ser/Thr Cdk acceptor sites to Ala , with Ser567 and S561 , respectively , left unaltered , with a HA tag placed on the N-terminus and C-terminus , respectively , and expressed from pCMV . Human Rb single Cdk sites were generated by individually adding back each single Cdk site to RbΔCDK−HA . Rb2xCdk retained T373 , S811; Rb3xCdk retained T373 , S612 , S811; Rb6xCdk retained the N-terminal Cdk sites; Rb9xCdk retained spacer and C-terminal Cdk sites . Murine RbWT−HA and RbΔCdk−HA MSCV retroviruses were generated from transfected HEK 293 cells and stored at −80°C . E2F constructs were expressed from pCMV and contained a C-terminal Myc tag . The Rb shRNA vector was generated by inserting a 3′ UTR region of the endogenous Rb mRNA ( GCTTTGAACTGAAGACTAT ) into pSM2c-scramble ( Stegmeier et al . , 2005 ) . 2D-IEF was performed as described ( Ezhevsky et al . , 2001 ) by immunoprecipitating Rb and eluting in 7 M urea/2 M thiourea/2% CHAPS ( pH 8 . 4 ) , then loading onto the acidic end of a 3–10 immobiline strips ( GE Healthcare ) with the current ramped up from 200 V for 2 hr , 500 V for 1 hr , 800 V for 1 hr , 1000 V for 0 . 5 hr , 1200 V for 0 . 5 hr , 1400 V for 0 . 5 hr , 1600 V for 0 . 5 hr , 1800 V for 2 . 5 hr , and 2000 V for 2 . 5 hr . Second dimension was performed by soaking IEF strip in 2% SDS/6 M urea/75 mM Tris ( pH 8 . 8 ) , 29% ( wt/vol ) glycerol , and placing the strip on top of a 6% SDS-PAGE containing a single large well to accommodate the IEF strip with Mw marker side wells . Co-immunoprecipitations were performed as described ( Ezhevsky et al . , 1997 ) using anti-Rb ( G3-245 , BD Pharmingen , San Jose , CA ) , anti-HA ( 3F10 , Roche , Basel , Switzerland ) , anti-Myc ( 9E10 , Developmental Studies Hybridoma Bank , Iowa City , IA ) , or anti-E1a ( M73 ) . Immunoblotting was performed as described ( Ezhevsky et al . , 1997 ) using anti-Rb ( G3-245 , BD Pharmingen ) , anti-HA ( 3F10 , Roche ) , anti-actin ( C4 , Abcam ) , anti-E1a ( 13-S5; Santa Cruz ) , and anti-Myc ( 9E10 , Developmental Studies Hybridoma Bank ) antibodies . Rb immunoblots were performed using 6% SDS-PAGE for separation or 10% SDS-PAGE for quantification . All immunoblots were quantified utilizing ChemiDoc XRS ( Bio-Rad , Hercules , CA ) sub-saturating linear signals . Rb phospho-specific antibodies: T356-PO4 ( AB4780 , Abcam , Cambridge , England ) , S608-PO4 ( 2181 , Cell Signaling , Danvers , MA ) , S612-PO4 ( OPA1-03891 , Thermo Scientific , Waltham , MA ) , S780-PO4 ( 3590 , Cell Signaling ) , S807-PO4/S811-PO4 ( 9308 , Cell Signaling ) , T821-PO4 ( AB4787 , Abcam ) , T826-PO4 ( AB4779 , Abcam ) , T821-PO4/T826-PO4 ( sc-16669 , Santa Cruz ) , T373 ( AB52975 , Abcam ) , S249-PO4/T252-PO4 ( sc-16671 , Santa Cruz ) . Immunoprecipitation-kinase assays were performed as described ( Ezhevsky et al . , 1997 ) using anti-CDK4 ( C22 ) , anti-CDK6 ( C21 ) , and anti-Cdk2 ( M2 ) polyclonal antibodies ( Santa Cruz ) . qRT-PCR was performed as described ( Eguchi et al . , 2009 ) using 6-FAM labeled TaqMan probes ( Dhfr , 00515663; Cdc6 , 00488573; Ccna2 , 01282245; p21 , 00432448; β2M , 00437762; Mcm3 , 00801867; Mcm5 , 00484839; Life Technologies , Grand Island , NY ) . Mean values of triplicate samples were normalized to beta-2-microglobulin . Whole-genome microarray analysis was performed as described ( Eguchi et al . , 2009 ) using MouseWG-6 v2 . 0 BeadChips ( Illumina , San Diego , CA ) at Biogem core ( UCSD ) . Heat maps were created with Cluster 3 . 0 and Java TreeView 1 . 1 . 3 and gene ontology classifications were based on DAVID Bioinformatics Resources ( Dennis et al . , 2003; Huang da et al . , 2009 ) . Full microarray data set has been submitted to GEO ( GSE56453 ) http://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? acc=GSE56453 . | Cells go through a tightly controlled , multi-step procedure before they divide . This cell division program—the cell cycle—is necessary for preventing unrestrained cellular growth , which may lead to cancer . Proteins called cyclins control the progression through each of the phases of the cell cycle , with different cyclins working during different phases . During the G1 phase of the cell cycle , cells grow in size and produce the proteins that are required to copy DNA . Once a cell passes a checkpoint called the 'restriction point' at the end of the G1 phase , it is committed to dividing . It is therefore particularly important to keep events during G1 phase in check . The Retinoblastoma tumor suppresor protein ( Rb ) is a key player in regulating the G1 phase . Rb sequesters transcription factors that are essential for the cell cycle to progress . Previously , it was thought that a complex called cyclin D added more and more phosphates to the Rb protein during the G1 phase . This process predicted a slow release of transcription factors , which attach to DNA and start the process of DNA replication . While many studies have presented data that is consistent with this model , direct biochemical evidence of these events is lacking . Narasimha , Kaulich , Shapiro et al . now present biochemical analyses of Rb proteins that show—completely unexpectedly—that the cyclin D complex adds just one phosphate group to Rb during the G1 phase , although this group can be added to one of fourteen different sites . The resulting 'mono-phosphorylated' Rb varieties can each sequester different transcription factors and stop them working . At the restriction point , many more phosphate groups are then rapidly added , and the Rb protein is inactivated by a different cyclin . This cyclin—called Cyclin E—then drives cells into the next phase of the cell cycle . Establishing how cyclin E is activated is a priority for future research . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2014 | Cyclin D activates the Rb tumor suppressor by mono-phosphorylation |
Supramolecular signaling assemblies are of interest for their unique signaling properties . A µm scale signaling assembly , the central supramolecular signaling cluster ( cSMAC ) , forms at the center of the interface of T cells activated by antigen-presenting cells . We have determined that it is composed of multiple complexes of a supramolecular volume of up to 0 . 5 µm3 and associated with extensive membrane undulations . To determine cSMAC function , we have systematically manipulated the localization of three adaptor proteins , LAT , SLP-76 , and Grb2 . cSMAC localization varied between the adaptors and was diminished upon blockade of the costimulatory receptor CD28 and deficiency of the signal amplifying kinase Itk . Reconstitution of cSMAC localization restored IL-2 secretion which is a key T cell effector function as dependent on reconstitution dynamics . Our data suggest that the cSMAC enhances early signaling by facilitating signaling interactions and attenuates signaling thereafter through sequestration of a more limited set of signaling intermediates .
T cell activation is governed by spatiotemporal organization of signal transduction across scales . At the nanoscale , receptors form clusters of dozens of molecules that can coalesce into microclusters and are commonly associated with active forms of signaling intermediates ( Boyle et al . , 2011; Hu et al . , 2016; Lillemeier et al . , 2006; Schamel et al . , 2005; Sherman et al . , 2011; Varma et al . , 2006; Yokosuka et al . , 2005 ) . This association suggests that such receptor clusters mediate efficient T cell signaling . Larger , µm scale assemblies were first described at the center and periphery of T cells activated by antigen-presenting cells ( APC ) for the TCR , PKCθ and LFA-1 , talin , respectively , as central and peripheral supramolecular activation clusters ( cSMAC and pSMAC ) ( Grakoui et al . , 1999; Monks et al . , 1998; Monks et al . , 1997 ) . µm scale of assemblies , in particular in the form of supramolecular protein complexes , provides unique biophysical and signaling properties ( Banani et al . , 2016; Li et al . , 2012; Shin and Brangwynne , 2017 ) . Supramolecular protein complexes play critical roles in viral sensing ( Cai et al . , 2014 ) , inflammation ( Franklin et al . , 2014 ) , embryonic development ( Brangwynne et al . , 2009 ) , protein folding in cancer ( Rodina et al . , 2016 ) , nuclear ubiquitinoylation ( Marzahn et al . , 2016 ) , and chromatin compaction ( Larson et al . , 2017 ) . Such complexes are readily observed by fluorescence microscopy , held together by a network of multivalent protein interactions and often have distinct phase properties ( Banani et al . , 2016; Li et al . , 2012; Shin and Brangwynne , 2017 ) . The cSMAC has many properties of such supramolecular protein complexes: It contains various multivalent signaling intermediates ( Balagopalan et al . , 2015 ) , prominently LAT ( linker of activation of T cells ) , components of this complex including LAT and PKCθ exchange with the remainder of the cell to a moderate extent and slowly ( Roybal et al . , 2015 ) , and components of this complex can be assembled into supramolecular structures in vitro ( Su et al . , 2016 ) . Therefore , understanding biophysical properties of the cSMAC and how it regulates T cell activation is of substantial importance . cSMAC function is controversial despite decades of work . Limitations in investigating the cSMAC are that its properties are largely unresolved and that its composition and/or assembly have not been systematically manipulated inside live T cells . Based on association of cSMAC formation with T cell activation conditions , the cSMAC has been proposed to enhance T cell signaling , terminate it , not be related to signaling or only upon weak stimulation or at late time points ( Čemerski et al . , 2008; Freiberg et al . , 2002; Grakoui et al . , 1999; Lee et al . , 2002; Monks et al . , 1997 ) . cSMAC formation is often associated with efficient T cell activation conditions , fitting with a role in enhancing T cell signaling . Accumulation of signaling intermediates at the T cell:APC interface center is substantially reduced by blockade of the costimulatory receptor CD28 ( Singleton et al . , 2009; Wülfing et al . , 2002 ) , in regulatory T cells ( Zanin-Zhorov et al . , 2010 ) , during thymic selection ( Ebert et al . , 2008 ) , or in the absence of the signal amplifying kinase Itk ( IL-2 inducible T cell kinase ) ( Singleton et al . , 2011 ) . To determine cSMAC properties , we have used stimulated emission depletion ( STED ) super-resolution microscopy and correlative light electron microscopy ( CLEM ) . The cSMAC was composed of multiple complexes of supramolecular dimensions and associated with extensive membrane undulations . To investigate cSMAC function , we have systematically manipulated the localization of three adaptor proteins in live primary T cells: LAT ( Balagopalan et al . , 2015 ) is an integral component of the cSMAC . SLP-76 ( SH2 domain-containing leucocyte protein of 76 kD ) ( Koretzky et al . , 2006 ) is associated with it only during the first minute of T cell activation ( Roybal et al . , 2015 ) . Grb2 ( growth factor receptor-bound 2 ) ( Jang et al . , 2009 ) association with the cSMAC is less prevalent ( Roybal et al . , 2015 ) . Interface recruitment of all three adaptors was diminished upon attenuation of T cell activation by costimulation blockade and Itk-deficiency as was IL-2 secretion , a critical T cell effector function . By fusing these adaptors with various protein domains with a strong interface localization preference we brought them back to the interface under the attenuated T cell activation conditions and restored cSMAC formation . Such restoration enhanced IL-2 secretion but only when executed to the extent and with dynamics seen under full stimulus conditions .
To investigate the function of µm scale protein accumulation at the center of the T cell:APC interface , we first cataloged protein localization events that were consistently associated with efficient T cell activation . We attenuated T cell activation through costimulation blockade ( Singleton et al . , 2009; Wülfing et al . , 2002 ) and Itk deficiency . The 5C . C7 T cell receptor ( TCR ) recognizes the moth cytochrome C ( MCC ) 89–103 peptide presented by I-Ek . In the restimulation of in vitro primed 5C . C7 T cells with CH27 B cell lymphoma APCs and MCC peptide IL-2 amounts in the supernatant were reduced upon blockade of the CD28 ligands CD80 and CD86 ( ‘costimulation blockade’ ) and in T cells from Itk knock out 5C . C7 TCR transgenic mice , in particular at lower peptide concentrations ( Figure 1A ) . As IL-2 amounts in T cell culture supernatants are determined by the difference between IL-2 generation and consumption we also determined IL-2 mRNA levels . Even at an MCC peptide concentration of 10 µM the level of IL-2 mRNA in T cells was significantly ( p<0 . 001 ) reduced to less than 50% upon costimulation blockade and Itk-deficiency ( Figure 1B ) . 10 µM MCC was used for the remainder of the study . To more precisely relate the determination of IL-2 amounts in T cell culture supernatants to IL-2 mRNA generation , we determined the time course of both ( Figure 1—figure supplement 1 ) . IL-2 mRNA generation occurred during the first six hours of T cell activation , consistent with transient nuclear localization of NFkB and previous data establishing that APC contact times of less than one hour are sufficient to commit a primed T cells to proliferation ( Iezzi et al . , 1998 ) . We used IL-2 mRNA generation for the remainder of the study because of its greater sensitivity to stimulus attenuation . We characterized T cell signaling organization as extensively described before ( Ambler et al . , 2017; Roybal et al . , 2015; Singleton et al . , 2009 ) . Briefly , in vitro primed 5C . C7 T cells are retrovirally transduced to express fluorescent signaling intermediates or sensors , FACS sorted to low expression as close as possible to endogenous signaling intermediate concentrations and imaged in three dimensions over time during restimulation with APC and 10 µM MCC peptide ( ‘full stimulus’ ) . In image analysis the frequency of occurrence of geometrically quantified µm scale subcellular distributions that represent underlying cell biological structures is determined ( Figure 1—figure supplement 2A ) ( Roybal et al . , 2013 ) . Of particular interest here are accumulation at the center of the T cell APC interface ( ‘central’ ) , the cSMAC , and accumulation in a µm deep ‘invagination’ at the center of the interface that likely mediates termination of early central signaling ( Singleton et al . , 2006 ) . Upon costimulation blockade most sensors that displayed frequent central accumulation upon full T cell stimulation , in particular during the first two minutes of cell coupling , did less so . In Itk-deficient 5C . C7 T cells sensors with only transient early central accumulation in wild type T cells lost much of that accumulation ( Figure 1C , D; Figure 1—figure supplements 2B–4 , Figure 1—source data 1 ) . Both phenotypes are indicative of reduced cSMAC formation . Efficient IL-2 secretion thus was associated with µm scale central signaling localization . LAT as a key cSMAC component displayed µm scale central localization ( Figure 2A ) in a biphasic pattern . At the time of tight cell coupling under full stimulus conditions 49 ± 6% of cell couples showed central LAT accumulation . After 2 min of cell coupling central LAT accumulation was only found in about 25% of cell couples ( Figure 2B ) and remained stable at that level . In the absence of Itk , the initial peak of central LAT accumulation was significantly ( p=0 . 005 ) diminished to 26 ± 6% of cell couples with central LAT accumulation . Such reduction was more pronounced ( 11 ± 4% , p<0 . 001 ) upon costimulation blockade ( Figure 2B; Figure 2—source data 1 ) . Combining costimulation blockade and Itk deficiency yielded the least interface LAT accumulation ( Figure 2B ) . Impaired activation of cytoskeletal transport processes is a likely contributor to diminished central LAT accumulation upon costimulation blockade and Itk deficiency , as enhancement of actin dynamics with active Rac and Cofilin ( Roybal et al . , 2016 ) significantly ( p<0 . 001 , Figure 2C; Figure 2—source data 1 ) ( Roybal et al . , 2016 ) increased central and overall LAT accumulation . Attenuation of T cell activation was associated with diminished LAT phosphorylation at Y191 ( Figure 2D; Figure 2—source data 1 ) upon costimulation blockade , in particular in combination with Itk deficiency . At 2 , 5 , and 10 min after tight cell coupling LAT phosphorylation was significantly ( p≤0 . 02 ) reduced in Itk-deficient 5C . C7 T cells upon costimulation blockade compared to wild type 5C . C7 T cells under full stimulus conditions by 39% , 61% , and 70% , respectively . To determine cSMAC properties , we stained 5C . C7 T cell:CH27 B cell APC couples for LAT and LAT phosphorylated at Y191 ( ‘pLAT’ ) and imaged them using STED super-resolution microscopy ( Figure 3A ) . Multiple LAT/pLAT complexes formed in the cSMAC and/or beyond , on average four per cell . LAT complexes were significantly larger ( p=0 . 04 ) in the cSMAC region with a supramolecular volume of 0 . 23 ± 0 . 03 µm3 than in T cells without a cSMAC ( 0 . 12 ± 0 . 01 µm3 ) ( Figure 3B ) . pLAT complexes were similarly larger in the cSMAC region with volumes of 0 . 25 ± 0 . 03 µm3 ( cSMAC ) versus 0 . 13 ± 0 . 01 µm3 ( non-cSMAC , p=0 . 007 ) ( Figure 3B ) . To understand how LAT as a transmembrane protein could drive the formation of multiple supramolecular complexes , we related cSMAC formation to plasma membrane topology with CLEM ( Figure 3C; Figure 3—figure supplement 1 ) . During live cell imaging of 5C . C7 T cells activated by CH27 B cell APCs we rapidly fixed samples upon detection of central LAT clustering and processed them for electron microscopy ( EM ) . We used two experimental conditions , a full stimulus and activation of 5C . C7 T cells expressing a fusion protein of LAT with the PKCθ V3 domain upon costimulation blockade as an experimental strategy to achieve enhanced cSMAC formation upon attenuation of T cell activation , as introduced in detail in the next section . In EM sections we measured the extent of membrane undulations in cSMAC and non-cSMAC interface regions as the ratio of the plasma membrane length to the straight-line diameter of the region . This ratio was significantly ( p<0 . 001 ) larger in the cSMAC ( 2 . 2 ± 0 . 3 ) than in the pSMAC of the same T cell ( 1 . 4 ± 0 . 1 ) . In control cell couples without cSMAC formation the ratio as measured across the entire interface was as small ( 1 . 5 ± 0 . 1 ) as that in pSMAC regions of cell couples with central LAT clustering ( Figure 3D ) . The non-cSMAC data are consistent with diminished membrane undulations in 5C . C7 T cells upon costimulation blockade with a ratio of 1 . 5 ± 0 . 1 ( Roybal et al . , 2016 ) . The cSMAC thus is a cellular region where membrane undulations and multiple supramolecular complexes are associated with enhanced proximity of signaling intermediates . To enable the throughput required for functional cSMAC investigation in the remainder of these studies , we used spinning disk confocal microscopy to determine central protein accumulation as a measure of cSMAC formation . To determine cSMAC function , we wanted to restore its formation upon costimulation blockade and in Itk-deficient T cells . To do so , we needed to hypothesize how cSMAC components interact within the complex . Conceptually , a supramolecular complex could function rigidly through formation of stoichiometrically defined protein interactions or flexibly by enhancing signaling proximity through complex formation such that the same proteins can be assembled using varying stoichiometries and protein interaction motifs . Because of the large number of cSMAC components some of which are membrane-bound ( Figure 1 ) we regard the flexible model as most likely . Impaired cSMAC formation upon costimulation blockade and Itk-deficiency should at least in part be driven by a reduction in the number of protein interaction motifs , that is the valence , of key components such as LAT as caused by diminished tyrosine phosphorylation ( Figure 2D ) . We should therefore be able to restore cSMAC formation by enhancing valence through the addition of new protein interaction domains . While this involves slightly divergent stoichiometries and protein interaction motifs the live cell signaling functionality gained through the formation of the central region of increased protein density should be restored . We increased LAT valence by adding three protein domains: PKCθ V3 , Vav1 SH3SH2SH3 , or PLCδ PH . The PKCθ V3 domain is required for central interface accumulation of PKCθ ( Kong et al . , 2011 ) even though it couldn’t drive central localization by itself ( Figure 4—figure supplement 1A ) . The Vav1 SH3SH2SH3 domains drove strong central accumulation only within the first minute of cell coupling ( Figure 4—figure supplement 1A ) , as consistent with the localization of full length Vav1 ( Figure 1—figure supplement 2A ) . The PLCδ PH domain mediated interface accumulation focused on the first two minutes of cell coupling without a central preference ( Figure 4—figure supplement 1A ) . While the addition of protein interaction domains to LAT is predicted to alter their interactome , initial experiments to support this notion remained inconclusive as detailed in the Methods section . Equal expression of LAT and the three LAT fusion proteins was enforced by FACS sorting for the same level of GFP . GFP fusion proteins were expressed at 2 . 1 ± 0 . 7 fold the endogenous level of LAT in non-transduced T cells ( Figure 4—figure supplement 1B , C ) with little change to endogenous LAT levels in the transduced T cells . Fusion of LAT to the PKCθ V3 domain ( LAT V3 ) yielded efficient central accumulation that was well sustained over the entire imaging time frame under all conditions at levels around 50% of cell couples with central accumulation ( Figure 4 ) . Starting 40 s after tight cell coupling such sustained central accumulation was significantly ( p<0 . 05 , Figure 4—source data 1 ) more frequent than central accumulation of non-targeted LAT under full stimulus at almost every time point . Fusion of LAT to the PKCθ V3 domain thus stabilized central LAT accumulation well beyond the levels seen for LAT alone under any physiological condition . Nevertheless , in cSMACs formed upon LAT V3 expression membrane undulations were similarly enhanced as in T cells expressing LAT ( Figure 3D ) supporting comparable cSMAC properties . Fusion of LAT to the Vav1 SH3SH2SH3 domains ( ‘LAT Vav’ ) yielded different effects depending on the T cell activation conditions . Upon a full T cell stimulus LAT Vav resulted in diminished central and overall accumulation compared to LAT alone that was significant ( p<0 . 05 ) across many time points ( Figure 4; Figure 4—source data 1 ) , suggesting that LAT Vav does not enhance properly assembled signaling complexes . Upon the three attenuated stimuli LAT Vav consistently enhanced central accumulation , most dramatically ( p≤0 . 001 , at most time points ) for costimulation blockade in wild type and Itk deficient cells ( Figure 4; Figure 4—source data 1 ) . LAT Vav accumulation upon attenuated T cell stimulation in any pattern was largely indistinguishable from non-targeted LAT accumulation under full stimulus conditions ( p>0 . 05 , Figure 4—source data 1 ) and central accumulation was moderately enhanced only between 1 and 3 min after tight cell coupling . Fusion of LAT to the Vav1 SH3SH2SH3 domain thus allowed for fairly close reconstitution of full stimulus-type LAT localization upon attenuated T cell stimulation . Fusion of LAT to the PLCδ PH domain ( ‘LAT PLCδPH’ ) resembled LAT Vav but was less powerful ( Figure 4; Figure 4—source data 1 ) . Upon the three attenuated T cell stimuli LAT PLCδPH moderately enhanced central and overall LAT accumulation but didn’t consistently reach the same extent as LAT alone under full stimulus conditions . For example , in Itk-deficient 5C . C7 T cells upon costimulation blockade interface accumulation in any pattern was consistently enhanced ( p<0 . 005 for time point 20 and later ) from <32% to>55% upon expression of LAT PLCδPH . However , accumulation at the interface center only moderately increased from a range of 6–17% to 19–35% . To ensure that overall interface accumulation of the targeted LAT constructs was comparable , we measured ( Ambler et al . , 2017; Roybal et al . , 2016 ) their interface recruitment upon costimulation blocked conditions . All constructs showed substantial interface recruitment with moderately less LAT Vav recruitment at the last four time points ( Figure 4—figure supplement 1D , E ) . To ensure functionality of the targeted LAT constructs , we showed that they were tyrosine phosphorylated upon T cell activation ( Figure 4—figure supplement 1F ) . Fusion of LAT with additional protein interaction domains thus allowed us to control central clustering: Fusion with the PKCθ V3 domain yielded consistently enhanced central localization , fusion with the Vav1 SH3SH2SH3 domains largely restored full stimulus-type LAT localization upon costimulation blockade and Itk deficiency and fusion with the PLCδ PH domain resulted in partial restoration . To determine T cell function upon manipulation of LAT valence and localization , we measured IL-2 mRNA induction upon 5C . C7 T cell activation with CH27 APCs and 10 µM MCC peptide , directly mirroring the imaging conditions . Expression of the targeting domains in isolation had only minor effects on IL-2 mRNA amounts ( Figure 5—figure supplement 1 ) . Forcing exaggerated central LAT clustering by fusion with the PKCθ V3 domain did not affect IL-2 mRNA amounts ( Figure 5A , B ) as further discussed below . In contrast , restoring LAT centrality under costimulation blocked and Itk deficient conditions to slightly higher ( LAT Vav ) or slightly lower ( LAT PLCδPH ) levels than seen for non-targeted LAT under full stimulus conditions yielded a consistent and largely significant ( p<0 . 05 ) increase in IL-2 mRNA ( Figure 5A , B ) to levels close to the amounts of IL-2 mRNA in LAT-transduced 5C . C7 T cells under full stimulus conditions . For example , in LAT-expressing 5C . C7 T cells IL-2 mRNA amounts dropped to 18 ± 6% and 18 ± 3% of full stimulus mRNA upon costimulation blockade in wild type and Itk-deficient 5C . C7 T cells , respectively . Expression of LAT Vav restored IL-2 mRNA to 41 ± 9% and 58 ± 10% ( p<0 . 01 ) , respectively , and expression of LAT PLCδPH to 50 ± 9% ( p<0 . 05 ) and 88 ± 23% . µm scale central LAT interface accumulation thus supported efficient IL-2 secretion depending on accumulation extent and dynamics . Next , we investigated with one example to which extent the forced relocalization of one signaling intermediate can drive analogous relocalization of others . We determined the subcellular distributions of Grb2 , Lck and Vav1 in 5C . C7 T cells in the presence of LAT V3 using IRES-containing retroviral vectors for the parallel expression of GFP-tagged versions of the signaling intermediates alongside LAT V3 . Under full stimulus conditions expression of LAT V3 moderately diminished interface recruitment of Grb2 , Lck and Vav1 ( Figure 6; Figure 6—source data 1 ) suggesting that excessive central LAT localization upsets a finely balanced signaling system . Upon costimulation blockade the localization of all three signaling intermediates was largely unaffected by LAT V3 . Grb2 and Lck centrality was moderately enhanced in Itk-deficient 5C . C7 T cells . For example , while the percentage of Itk-deficient 5C . C7 T cells with central Grb2-GFP expression did not exceed 7% at 40 s after cell coupling and thereafter , upon co-expression of LAT V3 this percentage averaged 15% over the same time frame . Such moderate enhancement did not occur at the time of tight cell coupling but was restricted to later time points . In Itk-deficient 5C . C7 T cells upon costimulation blockade , however , Grb2 accumulation was substantially enhanced upon parallel expression of LAT V3 , reaching levels of central Grb2 accumulation not seen under any other experimental condition investigated . The centrality of Vav1 as a signaling intermediate with only minor early central accumulation ( Figure 1—figure supplement 2B ) was not altered under any condition . The forced central localization of LAT thus could only draw in Grb2 and Lck as signaling intermediates with some intrinsic central localization preference to a mostly moderate extent and upon only some attenuated T cell stimuli . We investigated SLP-76 as an adaptor with more transient central accumulation . At the time of tight cell coupling under full stimulus conditions 45 ± 7% of the cell couples displayed SLP-76 accumulation at the interface center , similar to LAT . However , 80 s later this percentage dropped to less than 10% ( Figure 7A , B ) . Also similar to LAT , the peak of central SLP-76 accumulation was significantly ( p≤0 . 01 ) diminished upon costimulation blockade , Itk-deficiency and the combination of both to <27% , <15% and<11% of cell couples with central SLP-76 accumulation , respectively ( Figure 7B; Figure 7—source data 1 ) . To enhance SLP-76 valence and thus control its localization , we fused SLP-76 to the PKCθ V3 ( ‘SLP-76 V3’ ) or the Vav1 SH3SH2SH3 ( ‘SLP-76 Vav’ ) domain . Equal expression of SLP-76 and the two SLP-76 fusion proteins was enforced by FACS sorting for the same level of GFP . Both SLP-76 fusion constructs did not significantly affect SLP-76 centrality under full stimulus conditions , in wild type or Itk-deficient 5C . C7 T cells ( Figure 7B; Figure 7—source data 1 ) . However , accumulation of SLP-76 at the interface center was moderately but significantly ( p<0 . 05 at at least two time points within the first minute of tight cell coupling , the peak of central SLP-76 accumulation ) enhanced upon costimulation blockade in wild type and Itk-deficient 5C . C7 T cells reaching for example 67 ± 7% and 33 ± 7% , respectively , of cell couples with central SLP-76 accumulation upon expression of SLP-76 Vav ( Figure 7B; Figure 7—source data 1 ) . Interestingly , the enhancement of SLP-76 centrality was limited to the first minute of tight cell coupling , the time where non-targeted SLP-76 accumulated at the interface center . Consistent with the enhancement of SLP-76 centrality we observed a modest increase in IL-2 mRNA amounts under costimulation blocked conditions upon expression of SLP-76 V3 and SLP-76 Vav . Upon expression of non-targeted SLP-76 costimulation blockade in wild type and Itk-deficient 5C . C7 T cells reduced IL-2 mRNA amounts to 45 ± 9% and 21 ± 1% , respectively , of full stimulus ( Figure 7C ) . Expression of SLP-76 V3 restored IL-2 mRNA amounts to 87 ± 20% and 74 ± 24% , respectively , expression of SLP-76 Vav to 73 ± 8% and 75 ± 30% without reaching statistical significance in the stringent 2-way ANOVA ( Figure 7C ) . Importantly , enhancement of centrality and IL-2 secretion remained closely linked across multiple T cell activation conditions and spatially targeted SLP-76 constructs ( Figure 5B ) thus corroborating the importance of µm scale central protein accumulation within the first two minutes of cell coupling for IL-2 secretion . As a negative control we enhanced the valence of a signaling intermediate with more tentative central localization preference: We fused Grb2 to PKCθ V3 or Vav1 SH3SH2SH3 . Equal expression of Grb2 and the two Grb2 fusion proteins was enforced by FACS sorting for the same level of GFP . Upon 5C . C7 T cell activation with a full stimulus Grb2 is efficiently recruited to the T cell:APC interface during the first two minutes of tight cell coupling , peaking at 80 ± 5% of cell couples with any interface accumulation . This overall interface accumulation was significantly ( p≤0 . 02 at at least three time points within the first two minutes of tight cell coupling ) diminished upon costimulation blockade , Itk-deficiency and both , remaining below 64% , 35% and 43% , respectively of cell couples with any interface accumulation ( Figure 8A , B; Figure 8—source data 1 ) . Distinguishing Grb2 from LAT and SLP-76 , Grb2 accumulation at the interface center didn’t exceed 20% under any of the T cell activation conditions . Fusion of Grb2 with PKCθ V3 ( ‘Grb2 V3’ ) or Vav1 SH3SH2SH3 ( ‘Grb2 Vav’ ) did not enhance centrality under any of the T cell activation conditions ( Figure 8B , Figure 8—source data 1 ) . Fusion with the PKCθ V3 domain did not substantially alter Grb2 localization at all ( Figure 8B ) . Fusion with the Vav1 SH3SH2SH3 domain enhanced overall Grb2 interface recruitment across many time points under all T cell activation conditions ( Figure 8B ) . However , most of this accumulation was in the peripheral pattern , a pattern common with full length Vav1 ( Figure 1—figure supplement 2B ) . As an important negative control , a protein with a minor intrinsic central preference can thus not be forced to the interface center . Expression of Grb2 V3 or Grb2 Vav did not alter IL-2 mRNA production under any of the T cell activation conditions ( Figure 8C ) despite the substantially enhanced overall interface accumulation upon expression of Grb2 Vav . These data thus provide an important specificity control for the selective functional importance of protein accumulation in the cSMAC .
The cSMAC was characterized by enhanced membrane undulations and the formation of multiple protein complexes of supramolecular dimensions in the 0 . 1–0 . 5 µm3 range . Supramolecular complexes can be driven by the polymerization of a single or few defined proteins ( Cai et al . , 2014; Franklin et al . , 2014; Li et al . , 2012; Marzahn et al . , 2016 ) or , likely more common , they can consists of an agglomeration of large numbers of proteins ( Rodina et al . , 2016; Tarantino et al . , 2014 ) . While supramolecular complexes formed by a small number of components are often characterized by defined structures such as lipid droplets or fibers ( Shin and Brangwynne , 2017 ) , our understanding of supramolecular complexes built from a large number of components is limited . As close to half of the amount of three components of the central signaling complex , LAT , active Rac , and PKCθ , is immobile and the remainder exchanges only slowly with rest of the cell ( Roybal et al . , 2015 ) cSMAC signaling complexes likely have partial solid state properties . T cell signaling has been extensively characterized with super-molecular resolution in T cells activated with planar APC substitutes , supported lipid bilayers and antibody-coated cover slips . Similar to our findings large protein complexes , microclusters , were found . The microclusters are transported to the interface center to form a single µm scale structure ( Varma et al . , 2006; Yokosuka et al . , 2005 ) . However in contrast to T cell activation on planar surfaces , in T cell:APC couples membrane undulations form across the entire interface with F-actin structures perpendicular to the interface plane ( Roybal et al . , 2015 ) and thus impair rather than enhance centripetal transport . A large central invagination may remove proteins from the interface center ( Singleton et al . , 2006 ) and the central membrane undulations characterized here increase the number of transmembrane proteins in contact with a given cytoplasmic volume . Therefore , the cSMAC in T cell:APC couples will likely have similar molecular constituents as the central complex in T cells activated on planar substrates but distinct biophysical properties . It is the signaling functionality arising from these unique properties that we have investigated here . Nevertheless , the more straightforward access to high resolution imaging and synthetic system engineering using planar APC substitutes provides intriguing data that complement our findings . Supramolecular signaling complexes including LAT , SLP-76 and Grb2 have been reconstituted on model membrane and shown to trigger signaling and F-actin assembly ( Su et al . , 2016 ) . Large numbers of exosomes at the center of the interface between T cells and supported lipid bilayers could be related to the cSMAC membrane undulations and contribute to signal attenuation ( Choudhuri et al . , 2014 ) . Highly localized F-actin structures discovered using T cell activation on planar surfaces could mediate transport into supramolecular signaling complexes ( Kumari et al . , 2015 ) with the F-actin uncapping protein RLTPR identified as a regulator of the composition of signaling complexes ( Liang et al . , 2013 ) . Using a large-scale live cell imaging approach , we were able to distinguish between inefficient cSMAC formation and inefficient recruitment of a signaling intermediate to an existing cSMAC: As central interface accumulation of numerous signaling intermediates was diminished upon costimulation blockade and Itk-deficiency ( Figure 1C , D ) cSMAC formation was most likely impaired . Central interface accumulation of spatially targeted LAT and SLP-76 upon the attenuated T cell stimuli thus has to represent cSMAC restoration as confirmed by CLEM ( Figure 3D ) . Complexes of supramolecular dimensions are part of the cSMAC ( Figure 3B ) . In general , supramolecular complex formation becomes more likely with increasing concentrations and valences of the complex components ( Banani et al . , 2016; Li et al . , 2012 ) . Accordingly , replacement of proline-rich regions in Sos that mediate multivalent LAT/Grb2/Sos interactions leads to reduced LAT clustering and phosphorylation ( Kortum et al . , 2013 ) . Similarly , when LAT phosphorylation ( Figure 2C ) , which is required for interactions of LAT with SH2 domain-containing signaling intermediates , was reduced upon costimulation blockade and Itk deficiency LAT clustering at the interface center was also diminished ( Figure 2B ) . Compensation for such diminished LAT valence by fusing LAT with additional protein interaction domains restored µm scale central LAT accumulation . The dependence of LAT clustering on the number of functional protein interaction motifs , that is its valence , supports the supramolecular nature of protein complexes within the cSMAC . The importance of valence is further illustrated in the divergent behavior of fusion proteins versus their components . For example , the PKCθ V3 domain in isolation does not localize to the interface center even under full stimulus conditions ( Figure 4—figure supplement 1A ) nor does LAT upon costimulation blockade ( Figure 2B ) . However , the LAT-V3 fusion protein shows dominant central localization upon costimulation blockade ( Figure 4B ) . How is this possible ? By generating a LAT-V3 fusion proteins we didn’t only add the localization preferences of LAT and the PKCθ V3 domain , we also increased the valence of the fusion protein over its components as a key facilitator of recruitment to supramolecular complexes ( Li et al . , 2012 ) . Interestingly , fusion domains could not drive spatial features of adaptor localization by themselves but could only enhance intrinsic adaptor localization preferences . We could enhance central LAT clustering across all time points , central SLP-76 clustering only during the first minute of cell coupling and central Grb2 clustering not at all . This mirrored the accumulation patterns of the non-targeted adaptors under full stimulus conditions ( Figures 4 , 7 and 8 ) and therefore strongly suggests that localization of the spatially targeted adaptors was driven by a combination of intrinsic localization motifs and the spatial information provided by the fused domains . In support , while both PKCθ V3 and PLCδ PH did not display central localization in isolation , they could enhance central localization of LAT and SLP-76 ( Figures 4 and 7 ) . Similarly , artificially enhanced cSMAC formation through expression of LAT V3 did not lead to the recruitment of signaling intermediates with intrinsically weak cSMAC preference such as Vav1 ( Figure 4C ) . Our inability to force cSMAC localization implies that even upon addition of new protein interaction domains the overall molecular composition of supramolecular complexes in the cSMAC is fairly well conserved: A core of protein-protein interactions may be required for complex formation such that addition of new protein interactions can only enhance complex stability but not generate complexes of fundamentally different composition . Supramolecular assemblies in the cSMAC thus likely exist in a delicate balance between compositional conservation for complex identity and some redundancy in protein interaction motifs and stoichiometry to allow flexible regulation of stability . As a possible exception to this rule it was only upon the strongest attenuation of T cell activation , Itk-deficiency combined with costimulation blockade , that Grb2 as one of three signaling intermediates investigated was recruited to a LAT V3-based cMSAC to an extent not seem for Grb2 interface accumulation under other T cell activation conditions ( Figure 6 ) . We suggest the following scenario . Grb2 doesn’t effectively compete for binding to the cSMAC under strong and even moderate T cell activation conditions . Likely , stronger binding signaling intermediates are activated to a sufficient extent under those conditions to outcompete Grb2 . However , with the strong signaling attenuation provided by the combination of costimulation blockade and Itk-deficiency activation of more competitive signaling intermediates becomes sufficiently inefficient to allow Grb2 recruitment to the LAT V3-based signaling complexes . Reconstitution of central LAT and SLP-76 clustering under attenuated T cell activation conditions could restore IL-2 mRNA generation . The association of lack of Grb2 centrality with lack of an effect on IL-2 mRNA generation provides a specificity control . Protein clustering in the cSMAC thus was a critical component of efficient T cell signaling . For the cSMAC to enhance T cell function experimental reconstitution of cSMAC formation needed to closely mimic cSMAC features observed with non-targeted constructs under full stimulus conditions . Time-dependent roles of the cSMAC have been proposed before ( Freiberg et al . , 2002 ) yet are difficult to compare to the work here as the molecules investigated differed . In our work under full stimulus conditions LAT-GFP and SLP-76-GFP were efficiently recruited to the interface center within the first two minutes of T cell activation with diminished recruitment thereafter . Recruitment of LAT and SLP-76 to the interface center upon attenuation of T cell activation by fusion with the Vav1 SH3SH2SH3 and PLCδ PH domains closely reproduced these dynamics ( Figures 4B and 7 ) and largely restored IL-2 mRNA generation ( Figures 5A and 7 ) . However , recruitment of LAT to the interface center to a greater extent and duration by fusion with PKCθ V3 ( Figure 4B ) could not enhance IL-2 mRNA generation ( Figure 5A ) . We therefore suggest that the cSMAC displays two different time-dependent roles . Within the first one to two minutes of T cell activation it efficiently brings together the large number of proximal T cell signaling intermediates required for efficient T cell activation . Subsequently , a substantial number of key signaling intermediates including SLP-76 , Itk , PLCγ , and Vav1 leave the cSMAC and move to smaller signaling complexes supported by an interface-wide lamellal actin network ( Roybal et al . , 2015 ) . Retention of a more limited subset of signaling intermediates in the cSMAC after this time thus may render them less accessible to their interaction partners and therefore diminish sustained signal transduction . Signal enhancing and attenuating roles of the cSMAC thus may both occur as regulated by the specific time-dependent composition of the complex . Itk and costimulation likely contribute to cSMAC formation by overlapping yet partially distinct means . Recruitment to the interface center peaks within the first two minutes of cell coupling for both ligand-engaged CD28 ( Purtic et al . , 2005; Singleton et al . , 2009 ) and full length Itk ( Roybal et al . , 2015 ) . Both thus can be expected to provide protein-protein interactions during the early signal amplifying stage of the cSMAC . Itk also has enzymatic activity to directly modify cSMAC components . Both costimulation and Itk regulate actin dynamics . However , while costimulation controls core actin turnover through the Arp2/3 complex and Cofilin ( Roybal et al . , 2016 ) , Itk only regulates a SLAT-dependent subset of actin dynamics ( Singleton et al . , 2011 ) . CD28 thus can be expected to contribute more strongly to cSMAC directed transport in complex assembly ( Roybal et al . , 2016 ) ( Figure 2C ) . In summary , our work establishes that the cSMAC formed in the activation of T cells by APCs consists of multiple supramolecular complexes driven by extensive membrane undulations with a dynamically changing composition . Compositionally rich complexes in the first two minutes of cell coupling enhanced T cell activation by facilitating efficient signaling interactions whereas thereafter compositionally poorer complexes may sequester signaling intermediates .
Antibody used for quantitative western blotting were α-LAT pY191 ( Cell Signaling , #3584 , RRID:AB_2157728 ) , α-GAPDH Clone 14C10 ( Cell Signaling , #2118 , RRID:AB_561053 ) , and α-alpha Tubulin Clone DM1A ( ThermoFisher Scientific , #62204 , RRID:AB_1965960 ) . Antibodies used for the blockade of CD80- and CD86-dependent CD28 costimulation were anti-mouse CD80 Clone 16–10-A1 ( BD Pharmingen #553736 ) and anti-CD86 Clone GL1 ( BD Pharmingen #553689 ) . Protein transduction versions of constitutively active cofilin ( S3A ) and Rac1 ( Q61L ) were purified from E . coli and introduced into primary 5C . C7 T cells by 30 min incubation as previously described ( Roybal et al . , 2016 ) . Itk-deficient 5C . C7 mice were generated by crossing B10 . BR 5C . C7 TCR transgenic mice ( Seder et al . , 1992 ) ( RRID:MGI:3799371 ) with Itk-deficient B6 mice ( Schaeffer et al . , 1999 ) ( RRID:MGI:4356470 ) . T cells expanded from the lymph nodes of wild type or Itk-deficient 5C . C7 TCR transgenic mice were used for all experiments . The 5C . C7 TCR recognizes the moth cytochrome c peptide fragment ( amino acid residues 88 to 103 , ANERADLIAYLKQATK ) in the context of I-Ek . Single-cell suspensions were made from the lymph nodes of 6- to 8-week-old mice of either gender . The cells were adjusted to 4 × 106 cells/ml and MCC peptide was added to a final concentration of 3 µM . T cells were transduced with MMLV-derived retroviruses for the expression of signaling sensors , commonly signaling intermediates fused with GFP as described in detail ( Ambler et al . , 2017; Roybal et al . , 2015; Singleton et al . , 2009 ) . All animals were maintained in pathogen-free animal facilities at the University of Bristol under the University mouse breeding Home Office License P10DC2972 . The CH27 B cell lymphoma cell line ( RRID:CVCL_7178 ) was used in all experiments as APCs . It was shown to be mycoplasm-free using the ATCC Universal Mycoplasm Testing Kit and verified by staining for I-Ek , CD80 , CD86 and CD54 ( ICAM-1 ) . To peptide load the APCs , the CH27 cells were incubated in the presence of 10 µM MCC peptide for at least 4 hr . All cells were maintained in medium composed of RPMI with L-glutamine , 10% fetal bovine serum ( FBS , Hyclone ) , penicillin ( 100 IU/mL ) , streptomycin ( 100 µg/ml ) , and 0 . 5 µM β-mercaptoethanol . Interleukin-2 ( IL-2 ) ( TECIN recombinant human IL-2 , NCI Biological Resource Branch ) was added at a final concentration of 0 . 05 U/ml during parts of the retroviral transduction procedure . Our imaging and image analysis protocols have recently been described in great detail in a dedicated publication ( Ambler et al . , 2017 ) . Briefly , time-lapse fluorescence microscopy was performed with retrovirally transduced T cells , FACS-sorted to the lowest detectable sensor expression of 2 µM , and CH27 cells loaded with 10 µM MCC . The T cells and CH27 cells were imaged in imaging buffer ( PBS , 10% FBS , 1 mM CaCl2 , 0 . 5 mM MgCl2 ) on 384-well glass-bottom plates . All imaging was performed on a Perkin Elmer UltraVIEW ERS 6FE spinning disk confocal systems fitted onto a Leica DM I6000 microscope body equipped with full environmental control and a Hamamatsu C9100-50 EMCCD . A Leica 40x HCX PL APO oil objective ( NA = 1 . 25 ) was used for all imaging . Automated control of the microscope was performed with Volocity software ( Perkin Elmer ) . For experiments in which the CD80- and CD86-dependent activation of CD28 was blocked , peptide-loaded CH27 cells were incubated on ice for 30 min in the presence of anti- CD80 Clone 16–10-A1 ( 10 µg/ml ) and anti- CD86 Clone GL1 ( 10 µg/ml ) ( BD Pharmingen ) antibody before the CH27 cells were transferred to the imaging plate with the T cells . For experiments in which cells were reconstituted with protein transduction versions of constitutively active Rac and cofilin , T cells were incubated for 30 min at 37°C with the protein transduction reagents at the indicated concentrations in the imaging plate before the addition of the peptide-loaded CH27 cells . Each time-lapse image sequence was generated by taking a differential interference contrast brightfield image and a 3D image stack of the GFP channel every 20 s for 46 frames at 37°C . Voxels in these 3D images were of size 0 . 34 µm in the horizontal plane and 1 µm along the optical axis . For long-term NFκB-GFP imaging , the glass bottom of the imaging plate was coated with 10 µg/ml anti-CD CD19 antibody ( BD Pharmingen #553784 ) to prevent APC from moving and a Pathological Devices LiveCell chamber was fitted over the imaging plate to prevent evaporation . Nuclear localization of NFκB-GFP was determined as in Roybal et al . ( 2015 ) . The location and frame number of each T cell:APC couple were identified as when either the T cell:APC interface had reached its full width or the cells had been in contact for 40 s , whichever came first . Patterns of signaling sensor enrichment were assessed according to previously established quantitative criteria ( Figure 2 in Singleton et al . , 2009 ) as depicted in the Figure 1—figure supplement 2A . Briefly , the six , mutually exclusive interface patterns were: accumulation at the center of the T cell-APC interface ( central ) , accumulation in a large T cell invagination ( invagination ) , accumulation that covered the cell cortex across central and peripheral regions ( diffuse ) , accumulation in a broad interface lamella ( lamellal ) , accumulation at the periphery of the interface ( peripheral ) or in smaller protrusions ( asymmetric ) . Briefly , for each cell couple at each time point we first determined whether fluorescence intensity in the area of accumulation was >40% above the cellular fluorescence background . If so , the geometrical features of the area of accumulation , fraction of the interface covered , location within the interface , and extension of the area of accumulation away from the interface ( Figure 2 in Singleton et al . , 2009 ) , were used to assign the cell couple to one of the mutually exclusive patterns . Systems-scale cluster analysis was performed with Cluster ( Michael Eisen , UC Berkeley ) as established ( Singleton et al . , 2009 ) . To measure interface enrichment of LAT and spatially targeted versions thereof we used a recently developed computational image analysis routine ( Roybal et al . , 2016 ) . Very briefly , starting with the manual cell couple identification described above T cells were segmented , reoriented with the T cell:APC interface facing up using the ‘two-point synapse annotation’ procedure ( Roybal et al . , 2016 ) , and the cell shape was standardized to a half spheroid to allow voxel-by-voxel comparison across all cell couples analyzed . After transformation to the standard shape , the fluorescence distribution in each cell at a given time point was represented as a standardized vector ( of length 6628 ) formed from the intensity values of each of the voxels within the template shape , where the intensities for each time point were normalized so that the values of the vector were probabilities ( that is , fractions of total intensity ) . To measure interface enrichment , we defined an interface enrichment region as the 10% most fluorescent voxels of the average probability distribution across all cells , for all time points , and for all sensors . We defined enrichment to be the ratio of the mean probability in the distribution of that sensor for that cell at that time point within the interface enrichment region and the mean probability in the entire cell . CH27 APCs were adhered to α-CD19 antibody-coated coverslips . T cells were then allowed to interact with APCs for 4 . 5 min and fixed with 4% PFA for 20 min at 4°C followed by PFA quenching using ammonium chloride for 10 min at 4°C . Cells were permeabilized for 20 min in 0 . 02% Triton X-100 in PBS at 4°C . T cells were blocked in 1% BSA in PBS for 30 min at room temperature and probed with primary antibodies against LAT ( 1:100 , Cell Signalling #9166 , RRID:AB_2283298 ) or phospho-LAT Tyr 191 ( 1:50 , Cell Signalling #3584 , RRID:AB_2157728 ) diluted in 1% BSA with Fc block ( Rat Anti-Mouse CD16/CD32 , #553141 , BD Bioscience , RRID:AB_394656 ) at the same dilution in PBS for overnight at 4°C . Cells were washed with PBS three times and incubated with secondary antibody , Donkey anti-rabbit IgG , Alexa Fluor 488 ( Molecular Probes #R37118 , 1:1000 , RRID:AB_2556546 ) in 1% BSA with Fc block ( 1:500 ) for 1 hr at room temperature . Coverslips were washed with PBS before mounting using ProLong Gold ( Thermo Fisher ) and cured for 24 hr at room temperature . Fixed cells were imaged through a 100x HC PL APO CS2 1 . 4 NA objective on a Leica SP8 AOBS confocal laser scanning microscope . Alexa Fluor 488 was excited using a white light laser with an emission filter between 498–520 nm and STED depletion was achieved using a 592 nm continuous wave fibre laser . Images were first de-convolved using Hyugen Professional followed by automated puncta analysis with ImageJ ( NIH ) . LAT puncta were detected and measured using Wolfson Bioimaging ImageJ plugins ( Modular Image Analysis ) . Individual puncta were identified using the Otsu algorithm ( Otsu , 1979 ) with a threshold multiplier of 3 . 5 A . U . followed by a filtration mode of the Watershed 3D method to identify separate puncta . Small puncta detected in the APCs that don’t express LAT were used to derive a detection size threshold for LAT complexes such that all puncta smaller than the 95 percentile of the APC puncta size distribution were excluded from the analysis . Thus , complexes smaller than 0 . 04/0 . 06 µm3 in the LAT/pLAT data were excluded from the analysis . Repeating the analysis with a 99-percentile cut-off didn’t change the conclusions reached . The closest distance between adjacent puncta detected was 100 nm . The interaction of 5C . C7 T cells with CH27 B cell APCs was imaged live by spinning disk confocal microscopy as described above using a 35 mm glass bottom finder dish ( Mattek ) . Upon formation of a cSMAC , cells were immediately fixed using 2 . 5% Glutaraldehyde ( Agar Scientific ) in 0 . 1M cacodylate buffer , stained in 1% osmium tetroxide ( EMS ) in cacodylate buffer , dehydrated and embedded in Epon812 resin ( TAAB ) for 24 hr at 60°C . Samples were removed from EPON and trimmed to section of interest . Trimmed samples were sectioned at 300 nm using an Ultramicrotome ( Leica , EM UC7 ) with a diamond knife ( DiATOME ) and stained with uranyl acetate and lead citrate ( Agar Scientific ) . Sections were analyzed using a FEI Tecnai 12 120kV BioTwin equipped with a bottom-mount 4*4K Eagle CCD camera . The tomogram data series was acquired using a FRI Tecnai 20 TEM between −50° to +50° with a 2 . 0° increment . The data were reconstructed using IMOD etomo software . Segmentation was made using AMIRA software ( VSG ) , while for visualization , a combination of IMOD , AMIRA and Image J were used . To determine membrane undulations in the cSMAC region , the interface diameter was divided into four equal sections with the central two sections defined as the cSMAC , a conservative assumption as cSMACs commonly were smaller than half of the interface diameter . Live wild type or Itk-deficient 5C . C7 T cells were FACS sorted to generate comparable cell numbers across each assay . CH27 B cells were peptide loaded with 10 μM MCC for four hours or overnight . T and B cells were mixed in round bottom 96-well plates at 1 × 104 T cells to 5 × 104 B cells . For costimulation blockade 10 μg/ml α-CD80 and α-CD86 were added to each well . The cells were then incubated for 18 hr and IL-2 amounts in the supernatant were determined using a mouse IL-2 OptEIA ELISA kit from BD Biosciences as per manufacturer’s instructions . CH27 B cell lymphoma APCs were peptide loaded overnight with 10 μM MCC peptide . Live wild type or Itk-deficient 5C . C7 T cells , non-transduced or expressing adaptor protein-GFP or targeted variants thereof , were FACS sorted to generate comparable cell numbers across each assay . 1 × 104 T cells and 5 × 104 APCs cells were centrifuged for 30 s at 1 , 000 rpm to maximize cell-to-cell contact and incubated at 37°C for 2 hr . For costimulation blockade 10 μg/ml α-CD80 and α-CD86 were added to each well . mRNA was isolated using the Qiagen RNeasy Micro Kit ( Qiagen , UK ) according to manufacturer’s instructions . cDNA was generated using an Invitrogen AMV First-Strand cDNA synthesis kit ( Life Technologies , UK ) according to manufacturer’s instructions . IL-2 mRNA amounts were determined with a SYBR Green PCR master mix from Life Technologies ( 4344463 ) relative to mRNA for β−2 microglobulin on a DNA Engine Opticon II System ( Bio-Rad ) using the following oligonucleotides , IL-2: AGCTGTTGATGGACCTA and CGC AGA GGT CCA AGT TCA T , β−2 microglobulin: GCTATCCAGAAAACCCCTCAA and CGG GTG GAA CTG TGT GTT ACG T . CH27 B cell lymphoma APCs were peptide loaded overnight with 10 μM MCC peptide . Live wild type or Itk-deficient 5C . C7 T cells , non-transduced or expressing LAT-GFP or a targeted variant thereof , were FACS sorted to generate comparable cell numbers across each assay . 1 × 106 T cells and 1 × 106 APCs cells were centrifuged for 30 s at 1 , 000 rpm to maximize cell-to-cell contact and incubated at 37°C for the indicated time . Subsequently , samples were immediately lysed with cold RIPA lysis buffer ( Millipore ) plus protease/phosphatase inhibitor cocktail ( Cell Signaling ) for 30 min on ice . To remove the insoluble fraction , samples were centrifuged at 20 , 000 g for 15 min . Supernatant were run on SDS/PAGE gels , transferred to PDVF membranes and blotted according to standard protocols . Blots were stripped and reprobed with an anti-GAPDH or anti-α tubulin antibody to normalize for sample loading . The addition of protein interaction domains to LAT is predicted to alter their interactome . To identify proteins that selectively interact with the interaction domain-LAT fusion proteins , we used the GFP tag of the fusion proteins in a pull-down experiment . As spatiotemporal organization and expression of key signaling intermediates , prominently phosphatase and tensin homolg ( PTEN ) , commonly differ between primary and immortalized T cells , we performed these experiments using primary murine T cells . LAT-interaction domain-GFP fusion proteins ( Figure 4B ) were expressed in primary T cells by retroviral transduction . As transduction efficiencies of 5C . C7 T cells were not sufficiently high to generate at least 5x106 transduced T cells , we used CD8+ CL4 TCR transgenic T cells ( Jenkinson et al . , 2005 ) . To maximize protein interactions , CL4 T cells were activated with pervanadate ( 0 . 1mM vanadate plus 3mM hydrogen peroxide ) . T cells were lysed in RIPA buffer and GFP fusion proteins were pulled down with GFP-trap beads ( Chromotek , #GTA20 ) according to manufacturer’s instruction . Bead eluates were analyzed on Coomassie-stained gels and specific bands could not be identified on top of a continuous signal . There are possible experimental constraints . T cells are small with 70% of the cell volume comprised by the nucleus ( Roybal et al . , 2015 ) . While we could FACS sort up to 5x106 retrovirally transduced T cells , a strong pull-down signal would have likely required five times as many cells . Protein interactions triggered by pervanadate may less specific than those in the cSMAC . Protein accumulation in the cSMAC enriches less than 10% of the total cellular amount of an accumulated protein ( Singleton et al . , 2009 ) . While such enrichment is readily detectable by imaging , biochemical detection on the background of the larger non-accumulated cellular pool of a protein may be more demanding . The frequency of occurrence of interface accumulation patterns was analyzed pairwise with a proportion’s z-test as reported in figure supplements . Data from independent experiments were pooled after establishing that they were not significantly different . p values were not corrected for multiple comparisons as the corresponding pFDR q-values ( Storey et al . , 2004 ) were similar . Images of 50–100 cell couples are acquired per condition . A difference in pattern occurrence of 30% between two experimental conditions can be detected with a power of 0 . 8 . IL-2 mRNA amounts were first logarithmically transformed to stabilize the variance and approximate to the normal distribution . Outliers were identified using Chauvenet’s criterion . Resulting data were analyzed by 1-way ANOVA with Tukey’s adjustment for multiple comparisons or 2-way ANOVA with the Sidak adjustment for multiple comparisons depending on the number of variables compared . IL-2 ELISA data were first logarithmically transformed and then analyzed by 1-way ANOVA with Tukey’s adjustment for multiple comparisons . LAT phosphorylation data were first logarithmically transformed and then analyzed by 1-way ANOVA with Tukey’s adjustment for multiple comparisons separately for each time point . | Cells receive dozens of signals at different times and in different places . Integrating incoming information and deciding how to respond is no easy task . Signaling molecules on the cell surface pass messages inwards using chemical messengers that interact in complicated networks within the cell . One way to unravel the complexity of these networks is to look at specific groups of signaling molecules in test tubes to see how they interact . But the interior of a living cell is a very different environment . Molecules inside cells are tightly packed and , under certain conditions , they interact with each other by the thousands . They form structures known as ‘supramolecular complexes’ , which changes their behavior . One such supramolecular complex is the ‘central supramolecular activation cluster’ , or cSMAC for short . It forms under the surface of immune cells called T cells when they are getting ready to fight an infection . Under the microscope , the cSMAC looks like the bullseye of a dartboard , forming a crowd of signaling molecules at the center of the interface between the T cell and another cell . Its exact role is not clear , but evidence suggests it helps to start and stop the signals that switch T cells on . The cSMAC contains two key protein adaptors called LAT and SLP-76 that help to hold the structure together . So , to find out what the cSMAC does , Clark et al . genetically modified these adaptors to gain control over when the cSMAC forms . Clark et al . examined mouse T cells using super-resolution microscopy and electron microscopy , watching as other immune cells delivered the signal to switch on . As the T cells started to activate , the composition of the cSMAC changed . In the first two minutes after the cells started activating , the cSMAC included a large number of different components . This made T cell activation more efficient , possibly because the supramolecular complex was helping the network of signals to interact . Later , the cSMAC started to lose many of these components . Separating components may have helped to stop the activation signals . Understanding how T cells activate could lead to the possibility of turning them on or off in immune-related diseases . But these findings are not just relevant to immune cells . Other cells also use supramolecular complexes to control their signaling . Investigating how these complexes change over time could help us to understand how other cell types make decisions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"immunology",
"and",
"inflammation"
] | 2019 | Transient protein accumulation at the center of the T cell antigen-presenting cell interface drives efficient IL-2 secretion |
Opioids are perhaps the most effective analgesics in medicine . However , between 1999 and 2018 , over 400 , 000 people in the United States died from opioid overdose . Excessive opioids make breathing lethally slow and shallow , a side-effect called opioid-induced respiratory depression . This doubled-edged sword has sparked the desire to develop novel therapeutics that provide opioid-like analgesia without depressing breathing . One such approach has been the design of so-called ‘biased agonists’ that signal through some , but not all pathways downstream of the µ-opioid receptor ( MOR ) , the target of morphine and other opioid analgesics . This rationale stems from a study suggesting that MOR-induced ß-arrestin 2 dependent signaling is responsible for opioid respiratory depression , whereas adenylyl cyclase inhibition produces analgesia . To verify this important result that motivated the ‘biased agonist’ approach , we re-examined breathing in ß-arrestin 2-deficient mice and instead find no connection between ß-arrestin 2 and opioid respiratory depression . This result suggests that any attenuated effect of ‘biased agonists’ on breathing is through an as-yet defined mechanism .
More than 180 people died each day in 2018 from depressed breathing following an opioid overdose , a lethal side-effect termed opioid induced respiratory depression ( OIRD ) ( Pattinson , 2008; Scholl et al . , 2018 ) . Nevertheless , opioids remain among the most effective and widely prescribed analgesics , as evidenced by the World Health Organization’s step ladder for pain management . These two contrasting characteristics have created the urgent desire to develop or discover novel µ-opioid receptor ( MOR ) agonists that provide analgesia without depressing breathing . Such a possibility emerged in 2005 when it was reported that mice with germline deletion of ß-arrestin 2 ( Arrb2-/- ) experience enhanced analgesia with attenuated respiratory depression when administered systemic morphine ( Raehal et al . , 2005 ) . This finding inspired a new field of ‘biased agonist’ pharmacology with the goal of engaging MOR-dependent G-protein but not ß-arrestin 2 signaling . This approach remains a central strategy in drug discovery for analgesics ( Turnaturi et al . , 2019 ) . Germline deletion of the µ-opioid receptor gene ( Oprm1 ) completely eliminates OIRD in murine models ( Dahan et al . , 2001 ) and selective deletion of Oprm1 within the brainstem medullary breathing rhythm generator , the preBötzinger Complex ( preBötC ) ( Smith et al . , 1991 ) , largely attenuates OIRD ( Bachmutsky et al . , 2020; Varga et al . , 2019 ) . When engaged , MOR G-protein signaling activates inwardly rectifying potassium channels ( Al-Hasani and Bruchas , 2011 ) and inhibits synaptic vesicle release ( Zurawski et al . , 2019 ) , thereby depressing neural signaling . Additionally , the MOR signals intracellularly via a pathway dependent on MOR internalization by ß-arrestin 2 ( Calebiro et al . , 2010; Luttrell and Lefkowitz , 2002 ) . Indeed , all three pathways have been implicated in OIRD ( Montandon et al . , 2016; Raehal et al . , 2005; Wei and Ramirez , 2019 ) but a role for Arrb2 has been suggested to be important and selective for respiratory depression , relative to analgesia . Since the original OIRD study of Arrb2-/- mice , multiple MOR agonists , dubbed ‘biased agonists’ , have been created that signal through G-protein but not ß-arrestin 2 pathways . These agonists are reported to produce analgesia with reduced respiratory depression ( Manglik et al . , 2016; Schmid et al . , 2017 ) and thereby provide pharmacological support for the proposed role of ß-arrestin 2 in OIRD . However , recently , these studies have been called into question ( Hill et al . , 2018; Kliewer et al . , 2020 ) , prompting us to independently investigate the underlying necessity of ß-arrestin 2 in mediating ORID . Here , in an experiment where we rigorously control for genetic background , we demonstrate that basal breathing and OIRD are similar in Arrb2+/+ , Arrb2+/- , and Arrb2-/- littermates . Furthermore , the in vitro preBötC rhythm is similarly silenced by an MOR agonist in all three genotypes . Our data , together with another recent report , does not show a role of Arrb2 in opioid-induced respiratory depression and suggests that MOR biased agonists attenuate OIRD through a different mechanism .
Breathing behaviors and OIRD severity differ between strains of mice ( Bubier et al . , 2020 ) . Therefore , we sought to compare basal and morphine depressed breathing in mice with the same genetic background . We bred F1 Arrb2+/- mice to generate littermates that were wildtype ( +/+ ) , heterozygous ( +/- ) , and homozygous ( -/- ) for germline deletion of Arrb2 ( Figure 1A ) . Breathing was assessed by whole-body plethysmography following intraperitoneal ( IP ) injection of saline ( for control recordings ) or after IP morphine ( 20 mg/kg ) one day later ( Figure 1B ) . This same breathing protocol was conducted in air with 21% O2 and 0% CO2 ( hereby referred as normoxic ) and air with 21% O2 and 5% CO2 ( hereby referred as hypercapnic ) . The hypercapnic state minimizes potential confounding fluctuations in breathing rate and depth seen in normoxic conditions . We used our previously described analytical pipeline ( Bachmutsky et al . , 2020 ) to assay the two key parameters that define OIRD , slow and shallow breathing . Slow breathing is measured as the instantaneous frequency of each breath and shallow breathing is defined by the peak inspiratory flow since it strongly correlates with the volume of air inspired ( PIF , Figure 1C; Bachmutsky et al . , 2020 ) . In the normoxic condition , the morphology of single breaths , respiratory rate , and peak inspiratory airflow after IP saline appeared similar in Abbr2+/+ and -/- mice ( Figure 2A–B , Table 1 contains mean ± SEM and 95% CI ) . As expected for OIRD , IP morphine decreased the frequency and PIF , but the breathing characteristics remained indistinguishable in Abbr2+/+ versus -/- mice ( Figure 2A–B , Table 1 ) . Consistently , histograms of the instantaneous frequency and PIF for each breath after IP morphine showed overlapping distributions from Arrb2+/+ ( combined from n = 5 mice ) , +/- ( n = 6 ) , and -/- ( n = 7 ) animals ( Figure 2C , E ) . We quantified OIRD as the ratio of the average instantaneous frequency or PIF after IP morphine normalized to IP saline . As expected from the raw data , OIRD was similar among the genotypes ( rate decreased 60% and PIF by 40% , Figure 2D and F , Table 2 contains 95% CI for each mean and the comparisons ) . In fact , there was a small attenuation of respiratory rate depression in Arrb2+/+ compared to Arrb2-/- and +/- mice , inconsistent with the hypothesis that Arrb2 mutation attenuates OIRD . These data lead us to conclude that OIRD in normoxic conditions is not diminished in Arrb2-/- mice . These same breathing assays and analysis were also performed in a hypercapnic state . Hypercapnia eliminates any changes in breathing rate and depth that are simply due to variation in behavioral state , like sniffing versus calm sitting . Thus , although breathing when hypercapnic is faster and deeper ( like in Figure 3A , Table 3 ) , the reduced variability minimizes any chance that the conclusions in the normoxic condition are due to additional effects of opioids on behaviors such as sedation and locomotion , or non-opioid-related differences in arousal . Importantly , OIRD is still robustly observed in the hypercapnic state ( Figure 3A , Table 4 ) . As anticipated , the opioid depression of instantaneous frequency ( by ~30% ) and PIF ( by ~30% ) were the same among all three genotypes ( Figure 3C–F , Table 4 ) . Therefore , OIRD is not diminished in Arrb2-deficient mice in two independent breathing assays . To understand the confidence of these results , we determined the extent that Arrb2+/+ breathing parameters must be depressed ( compared to Arrb2-/- ) in order to produce a significant test statistic more than 80 % of the time , that is , power analysis . Given our cohort sizes and the observed variation in breathing parameters from Arrb2+/+ and -/- littermates , we plotted the relationship between power ( 0–1 ) and percent difference between the Arrb2-/- and hypothetical Arrb2+/+ means ( see methods ) . When comparing the Arrb2-/- measured and Arrb2+/+ hypothetical means , we could confidently distinguish these mean breathing frequencies so long as they differed by at least ~12–20% , and mean PIFs by more than ~10%–22% ( Figure 3—figure supplement 1 ) . Thus , our experimental approach enabled us to only detect mild to large differences in OIRD between Arrb2+/+ and -/- littermates , if they had occurred . It remains possible that a small effect , much smaller than previously reported , was not identified in our study . The most important site for OIRD is the preBötC ( Bachmutsky et al . , 2020 ) . So , we directly measured the effects of the opioid peptide [D-Ala , N-MePhe , Gly-ol]-enkephalin ( DAMGO ) on the preBötC rhythm in all three Arrb2 genotypes . Importantly , DAMGO robustly stimulates both MOR signaling pathways , G-protein and ß-arrestin 2 dependent signaling . The preBötC slice was prepared from Arrb2 littermates and the genotype ( +/+ , +/- , -/- ) was determined post hoc . The rhythm was monitored by measuring electrical activity in the hypoglossal nerve rootlet ( Smith et al . , 1991 ) for 20 minutes at baseline , and then with 20 nM then 50 nM DAMGO ( Figure 4A–B ) . The preBötC rhythms of Arrb2+/+ , +/- , and -/- littermates ( n = 5 , 20 , 6 ) similarly slowed by ~70–80% at 20 nM DAMGO and nearly all were silenced at 50 nM ( Figure 4C–D ) . If anything , it appears the Arrb2-/- slices showed a statistically significant increase in DAMGO sensitivity when compared to Arrb2+/- littermates ( Figure 4C–D ) , although this likely stems from the differences in cohort sizes ( 6 vs 20 ) . In conclusion , a MOR agonist that substantially activates ß-arrestin 2 signaling similarly slows the preBötC rhythm in mice lacking the Arrb2 gene and the littermate controls .
The proposed unique importance of MOR-dependent ß-arrestin 2 signaling in OIRD has motivated the development of biased agonists for analgesia . However , the recent failure to reproduce this result has called model into question ( Kliewer et al . , 2020 ) . Therefore , the goal of our studies was to test the null hypothesis that germline deletion of Arrb2 does not attenuate OIRD . The results from our in vivo studies under normoxic and hypercapnic conditions , as well as our in vitro studies , failed to reject this null hypothesis . In order to directly compare to previous results , we sufficiently powered our cohort size to identify effect sizes reported in Raehal et al . , 2005 ( ~50% less OIRD ) , and the statistical tests were performed conservatively by not correcting for multiple comparisons . Combined , these three independent assays demonstrate that the germline knockout of Arrb2 does not attenuate OIRD . Unlike other studies , we designed ours with five important features to ensure a robust conclusion . First , the OIRD comparisons were made between littermates in an effort to control for any strain specific effects on breathing and opioid sensitivity ( Bubier et al . , 2020 ) . Second , OIRD was defined as the comparison of breathing after IP injection of saline and morphine in the same animal to control for any within-animal specific breathing variation . Third , we analyzed the multiple breathing parameters with and without averaging across long stretches of breathing . Four , OIRD was measured under a hypercapnic state for a more precise quantification . And fifth , we validated our in vivo studies by directly measuring the impact of a MOR ligand on the preBötC rhythm , the key site for OIRD . Future studies should also confirm an unchanged sensitivity to opioids in other brain or peripheral sites that contribute to OIRD . One difference between our study and the original Arrb2 study by Raehal et al . was our method of morphine delivery . Raehal et al . reported that at the maximal concentration of morphine delivered ( 150 mg/kg subcutaneous ) , the respiratory rate in Arrb2 knockouts was depressed to half that of wildtype controls ( ~20% vs 40% depression of breathing rate ) . In our study , although we delivered 20 mg/kg morphine by IP injection , we observed an even larger depression of breathing in all three Arrb2 genotypes ( 60% in normoxic and 30% in hypercapnic conditions ) , indicating our assay was sufficient to observe the reported attenuated OIRD response in Arrb2-/- mice . Beyond this , several important future studies to exhaust other methodological limitations include the measurement of respiratory depression in Arrb2-/- and littermates with other abused opioids like oxycodone and heroin , to temporally or spatially delete Arrb2 using Arrb2flox/flox alleles to overcome possible compensatory mechanisms , and the study of biased agonists in models where potent opioids like fentanyl are lethal . Additional or other mechanisms may underlie the fatal apnea that was not studied here . The core premise for the development of MOR biased agonists is that Arrb2-dependent signaling provides a molecular mechanism to dissociate analgesia from respiratory depression . Our findings do not support this claim . Consistently , a recent study demonstrated that an opioid receptor ligand that does not induce MOR ß-arrestin 2-dependent signaling still temporarily induces OIRD ( Uprety et al . , 2021 ) . Given all of this , how then do some biased agonists show analgesia while minimizing respiratory depression ? Perhaps biased agonists are just partial MOR agonists for activation of G-protein signaling . In this case , we imagine analgesia is more sensitive than respiratory depression , and therefore certain concentrations of MOR ligand enable these two effects to be separated . This would be akin to providing a lower dose of standard opioid-like drugs . Regardless , our results , along with similar data from another recent study ( Kliewer et al . , 2020 ) , refute the foundational model that Arrb2 selectively mediates OIRD and suggest that now we must reconsider and in the future reinvestigate the mechanism of biased agonism in vivo .
Arrb2 -/- mice ( Bohn et al . , 1999 ) were bred to C57BL/6 to generate heterozygous F1 . The F1 littermates were then crossed to make Arrb2-/- , Arrb2-/+ , and Arrb2+/+ ( F2 ) . Mice were housed in a 12 hour light/dark cycle with unrestricted food and water . Mice were given anonymized identities for experimentation and data collection . All animal experiments were performed in accordance with national and institutional guidelines with standard precautions to minimize animal stress and the number of animals used in each experiment . Institutional Animal Care and Use Committee approval number AN181239 . Plethysmography and respiratory analysis were performed as in Bachmutsky et al . , 2020 . Briefly , on the first recording day , adult ( 6–12 weeks ) Arrb2-/- , Arrb2-/+ , and Arrb2+/+ mice were administered IP 100 µL of saline and placed in an isolated recovery cage for 15 min . After , individual mice were then monitored in a 450 mL whole animal plethysmography chamber at room temperature ( 22°C ) in 21% O2 balanced with N2 ( normoxic condition ) or 21% O2 , 5% CO2 balanced with N2 ( hypercapnic condition ) . After 1 day , the same protocol was used to monitor breathing after IP injection of morphine ( 20 mg/kg , Henry Schein 057202 ) . The morphine recordings under normoxic and hypercapnic conditions were separated from saline recordings by at least 3 days . Each breath was automatically segmented based on airflow crossing zero as well as quality control metrics . Respiratory parameters ( e . g . peak inspiratory flow , instantaneous frequency ) for each breath , as well as averages , were then calculated . Reported airflow in mL/sec . is an approximate of true volumes . The analysis was performed with custom Matlab code available on Github with a sample dataset ( https://github . com/YackleLab/Opioids-depress-breathing-through-two-small-brainstem-sites ) . All animals in the study were included in the analysis and cohorts including all these genotypes were run together . A power analysis was performed using the reported effect size from Raehal et al . In this case , 1–4 mice were necessary to observe a statistically significant result . Each cohort ( Arrb2+/+ , +/- , -/- ) exceeded 4 . Statistical tests were performed on the ratio of IP morphine to IP saline for instantaneous respiratory frequency and peak inspiratory flow separately for normoxic and hypercapnic conditions . A Shapiro Wilks test was first done to determine if the data was normally distributed ( Figure 2—source data 1 , Figure 3—source data 1 ) . If normal , a single factor ANOVA was performed to determine any differences among the three genotypes ( alpha <0 . 05 ) . In the instance the p-value was <0 . 05 , the Tukey HSD post-hoc test was done to determine which of the pairwise comparisons were statistically different ( alpha <0 . 05 ) . Additionally , one-way unpaired parametric T-tests were used to compare Arrb2 +/+ and -/- genotypes ( alpha <0 . 05 ) . If the data failed to pass the Shapiro Wilks test , then the non-parametric Kruskal-Wallis test was used to determine if any differences ( alpha <0 . 05 ) . And the Mann-Whitney U test was used to compare Arrb2+/+ and -/- genotypes ( alpha <0 . 05 ) . A two-way ANOVA with regression was used to determine interactions between each of the genotypes IP saline and IP morphine values . To determine the power of our data , we compared hypothetical Arrb2+/+ and measured Arrb2-/- means . The power calculation included our cohort size and the measured standard deviation of the two genotypes . A similar statistical approach was used to analyze the in vitro data ( Figure 4—source data 1 ) . All the above statistics were performed using the publicly available Excel package ‘Real Statistics Functions’ SPSS and Matlab . Rhythmic 550–650 μm-thick transverse medullary slices which contain the preBötC and cranial nerve XII ( XIIn ) from neonatal Arrb2 -/- , +/- , +/+ mice ( P0-5 ) were prepared as described ( Bachmutsky et al . , 2020 ) . Slices were cut in ACSF containing ( in mM ) : 124 NaCl , 3 KCl , 1 . 5 CaCl2 , 1 MgSO4 , 25 NaHCO3 , 0 . 5 NaH2PO4 , and 30 D-glucose , equilibrated with 95% O2 and 5% CO2 ( 4 °C , pH = 7 . 4 ) . Recordings were performed in 9 mM at a temperature of 27°C . Slices equilibrated for 20 min before experiments were started . The preBötC neural activity was recorded from CNXII rootlet . Activity was recorded with a MultiClamp700A or B using pClamp9 at 10 , 000 Hz and low/high pass filtered at 3/400 Hz . After equilibration , baseline activity and then increasing concentrations of DAMGO ( ab120674 ) were bath applied ( 20 nM , 50 nM ) . After the rhythm was eliminated , 100 nM Naloxone ( Sigma Aldrich N7758 ) was bath applied to demonstrate slice viability . The rate was determined from the last 5 min of each 20-min recording and rhythmic activity was normalized to the first control recording for dose response curves . | Opioid drugs are commonly prescribed due to their powerful painkilling properties . However , when misused , these compounds can cause breathing to become dangerously slow and shallow: between 1999 and 2018 , over 400 , 000 people died from opioid drug overdoses in the United States alone . Exactly how the drugs affect breathing remains unclear . What is known is that opioids work by binding to specific receptors at the surface of cells , an event which has a ripple effect on many biochemical pathways . Amongst these , research published in 2005 identified the β-arrestin 2 pathway as being responsible for altering breathing . This spurred efforts to find opioid-like drugs that would not interfere with the pathway , retaining their ability relieve pain but without affecting breathing . However , new evidence is now shedding doubt on the conclusions of this study . In response , Bachmutsky , Wei et al . attempted to replicate the original 2005 findings . Mice with carefully controlled genetic background were used , in which the genes for the β-arrestin 2 pathway were either present or absent . Both groups of animals had similar breathing patterns under normal conditions and after receiving an opioid drug . The results suggest β-arrestin 2 is not involved in opioid-induced breathing suppression . These findings demonstrate that research to develop opioid-like drugs that do not affect the β-arrestin 2 pathway are based on a false premise . Precisely targeting a drug’s molecular mechanisms to avoid suppressing breathing may still be a valid approach , but more research is needed to identify the right pathways . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience"
] | 2021 | ß-arrestin 2 germline knockout does not attenuate opioid respiratory depression |
In order to represent complex stimuli , principle neurons of associative learning regions receive combinatorial sensory inputs . Density of combinatorial innervation is theorized to determine the number of distinct stimuli that can be represented and distinguished from one another , with sparse innervation thought to optimize the complexity of representations in networks of limited size . How the convergence of combinatorial inputs to principle neurons of associative brain regions is established during development is unknown . Here , we explore the developmental patterning of sparse olfactory inputs to Kenyon cells of the Drosophila melanogaster mushroom body . By manipulating the ratio between pre- and post-synaptic cells , we find that postsynaptic Kenyon cells set convergence ratio: Kenyon cells produce fixed distributions of dendritic claws while presynaptic processes are plastic . Moreover , we show that sparse odor responses are preserved in mushroom bodies with reduced cellular repertoires , suggesting that developmental specification of convergence ratio allows functional robustness .
The environmental stimuli animals encounter on a day-to-day basis are extraordinarily numerous . Olfactory systems have evolved to cope with this diversity by maximizing the chemicals that can be detected , through the amplification of chemosensory receptor gene families , and through combinatorial coding , which expands representation capacity from the number of receptors in the genome to the number of combinations among them . The arthropod mushroom body is a cerebellum-like associative learning structure with a well-understood role in representing sensory stimuli and associating sensory and contextual cues ( Farris , 2011; Hige , 2018; Kennedy , 2015 ) . While mushroom bodies of different insect species process information from a variety of sensory modalities , 90% of Kenyon cell inputs in Drosophila melanogaster are olfactory ( Zheng et al . , 2018 ) . The mushroom body of each hemisphere has ~2000 Kenyon cells ( KCs ) , which are two synapses from the sensory periphery . Each olfactory receptor neuron in the antennae of adult flies expresses one or two of 75 olfactory receptor genes encoded in the genome . The axons of neurons expressing the same receptor converge on one of 54 glomeruli in the antennal lobe . ~150 uniglomerular projection neurons ( PNs ) have dendrites in one of the 54 glomeruli and carry signals about distinct receptor channels to two regions of the protocerebrum , the lateral horn and the mushroom body calyx ( Figure 1A ) . PN inputs to the lateral horn are thought to underlie innate behaviors , while inputs to the mushroom body allow flexible learned association of odor stimuli with behavioral context ( Chin et al . , 2018; de Belle and Heisenberg , 1994; Fişek and Wilson , 2014; Jefferis et al . , 2007; Ruta et al . , 2010 ) . In the mushroom body calyx , the presynaptic sites of individual olfactory PNs cluster into multi-synaptic boutons , with PNs of different types ( innervating different glomeruli ) producing consistent , characteristic bouton numbers ( Caron et al . , 2013; Zheng et al . , 2018 ) . Each PN makes 1–20 boutons , and each bouton is wrapped by claws of ~10 KCs , such that each PN sends output to between 10 and 200 of the 2000 KCs ( Leiss et al . , 2009 ) . KCs in turn have 3–10 ( average of five ) claws , which innervate boutons of various PNs ( Caron et al . , 2013; Gruntman and Turner , 2013; Zheng et al . , 2018 ) . Each KC therefore receives innervation from only a minority of the 54 incoming sensory channels , and different individual KCs receive different and relatively unstructured combinations of inputs ( Caron et al . , 2013; Eichler et al . , 2017; Honegger et al . , 2011; Murthy et al . , 2008; Zheng et al . , 2018 ) . The sets of inputs to individual cells vary across hemispheres and likely across individuals ( Caron et al . , 2013; Eichler et al . , 2017; Honegger et al . , 2011; Murthy et al . , 2008 ) . Associative learning mechanisms operate at KC output synapses , in the mushroom body axonal lobes , to re-weight KC outputs depending on experience and shift animal behavior ( Cohn et al . , 2015; Handler et al . , 2019; Hige et al . , 2015; Owald and Waddell , 2015 ) . The mushroom body is a simplified and experimentally tractable example of an expansion layer , in which a set of sensory inputs is mapped combinatorially onto a much larger set of postsynaptic cells , increasing the dimensionality of sensory representations . Like the diversification of antibodies by V ( D ) J recombination , the diversification of sensory input combinations across KCs is thought to allow them to represent arbitrary odors , regardless of evolutionary experience . Neurons of many other expansion layers receive similarly few , or sparse , sensory inputs . These include the cerebellum proper , the electric organ of mormyrid fish , the dorsal cochlear nucleus , and the hippocampus ( Bell et al . , 2008; Keene and Waddell , 2007; Mugnaini et al . , 1980 ) . Cerebellar granule cells have an average of four large , claw-shaped dendrites that are innervated by clustered mossy fiber presynaptic sites in mossy fiber rosettes . The similar convergence ratios in the cerebellum and mushroom body ( 4 or 5 sensory inputs , respectively , per expansion layer cell ) are thought to maximize dimensionality of sensory representations by optimizing the tradeoff between stimulus representation , which is maximized when expansion layer neurons receive large combinations of inputs , and stimulus separation , which is maximized when expansion layer neurons receive few inputs ( Albus , 1971; Cayco-Gajic et al . , 2017; Litwin-Kumar et al . , 2017; Marr , 1969 ) . The number of sensory inputs received by expansion layer neurons is thus a crucial parameter in sensory coding . How the density of inputs to expansion layer neurons is developmentally programmed is not understood in any system . Innervation complexity more generally has been studied in the peripheral nervous system and in the developing mammalian cortex . In peripheral sensory neurons , most prominently those of the Drosophila larval body wall , cell-autonomous mechanisms profoundly influence dendritic complexity ( Corty et al . , 2016; Jan and Jan , 2010; Ziegler et al . , 2017 ) . However , sensory neurons do not need to coordinate their innervation with presynaptic partners . In the vertebrate peripheral nervous system , including the rabbit ciliary ganglion and vertebrate neuromuscular junction , postsynaptic neurons or muscles are thought to dictate wiring complexity ( Gibbins et al . , 2003; Hume and Purves , 1983; Hume and Purves , 1981; Purves and Hume , 1981; Turney and Lichtman , 2012 ) . In contrast , in the developing cortex , extracellular signals including BDNF play a strong role in influencing dendritic complexity , suggesting that presynaptic cells and glia also influence connectivity density ( Kohara et al . , 2003; McAllister et al . , 1997; McAllister et al . , 1995; Purves , 1986 ) . Therefore , while mechanisms acting in both pre- and post-synaptic cells can influence innervation complexity , there is a need to directly compare how pre- and post-synaptic cells influence one another . We sought to ask how convergence ratio is set in the mushroom body calyx . By bidirectionally varying the populations of pre- and post-synaptic cells , we were able to make many different mushroom body aberrations . Across these conditions , we found a consistent pattern of compensations: the number of claws per KC remained largely constant , while the number of presynaptic boutons per olfactory PN varied bidirectionally and in response to changes of both the PN and KC populations . We therefore conclude that in this circuit , connectivity density is set by aspects of KC specification and is accomplished by flexible innervation of the calyx by PNs .
First , we sought to reduce the KC population , to ask whether remaining cells would increase their claw number to fully innervate incoming PN boutons , or whether PNs would scale down their boutons to the KC repertoire . To do this , we took advantage of existing pharmacological techniques for KC neuroblast ablation ( de Belle and Heisenberg , 1994; Sweeney et al . , 2012 ) . In the fly , four KC neuroblasts in each hemisphere produce ~500 KCs each ( Ito et al . , 1997 ) . While most neuroblasts pause their divisions for the first 8 hr after larval hatching ( ALH ) , the KC neuroblasts continue to divide; if larvae are fed the mitotic poison hydroxyurea ( HU ) during this time , KC neuroblasts can be specifically ablated ( de Belle and Heisenberg , 1994 ) . In these animals , lacking all KCs that receive olfactory inputs , olfactory learning is abrogated while innate olfactory behaviors are spared ( de Belle and Heisenberg , 1994 ) . The uniglomerular , excitatory olfactory PNs that innervate the MB calyx are produced by two neuroblasts , called lNB/BAlc and adNB/BAmv3 . Besides the KC neuroblasts , lNB/BAlc is the only other neuroblast in the fly central nervous system dividing from 0 to 8 hr ALH and is thus also susceptible to HU ablation ( Das et al . , 2013; Ito and Hotta , 1992; Lovick et al . , 2016; Lovick and Hartenstein , 2015; Stocker et al . , 1997; Sweeney et al . , 2012 ) . While ablation of all eight KC neuroblasts ( 4/hemisphere ) was used to demonstrate the role of KCs in olfactory learning ( de Belle and Heisenberg , 1994 ) , we sought to perform more subtle manipulations of the mushroom body cell populations . We generated animals in which PNs and KCs were fluorescently labeled to facilitate scoring , and reduced the concentration and duration of HU application , producing mushroom bodies with a variety of Kenyon cell complements , likely due to sporadic loss of different neuroblast complements in each hemisphere ( Figure 1B ) . Together , these neuroblast losses could potentially produce 1–2 PN neuroblasts and mushroom bodies derived from 0 to 4 KC neuroblasts . Using our reporters , we scored the presence or absence of PNs ventrolateral to the antennal lobe , derived from lNB/BAlc , and counted the number of KCs ( Figure 1C ) . In practice , some of the neuroblast states did not occur and some were overrepresented ( Figure 2—figure supplement 1 ) . Hemispheres from the same brain did not generally show the same neuroblast state as one another , but were often similar in severity of neuroblast losses ( Figure 2—figure supplement 1 ) . KCs of different types are produced in a characteristic sequence in development: γ KCs are born in the embryo and early larva , α’β’ KCs are born in late larvae , and αβ KCs are born in pupae . We identified samples in which all 4 KC neuroblasts had been ablated by looking for samples lacking mushroom body vertical lobes , which are composed of axons of α’β’ and αβ KCs and thus do not form if all KC neuroblasts are killed by HU just after hatching . We confirmed previous findings that in animals with all four KC neuroblasts ablated , the PNs fail to target the calyx and project only to the lateral horn ( Stocker et al . , 1997 ) . In these animals , ~100 KCs remained ( Figure 1B ) . These represent the KC population that is born prior to larval hatching , and which is thus unaffected by KC neuroblast ablation after hatching . Recently , these earliest-born cells , called γd KCs , have been shown to receive visual and not olfactory input in adult flies ( Vogt et al . , 2016; Yagi et al . , 2016 ) . These observations suggest that γd KCs are molecularly distinct from later-born cells during pupal calyx wiring , such that even when all the olfactory KCs are absent , olfactory PNs cannot innervate visual KCs . Remarkably , these embryonic-born γd KCs receive olfactory inputs in the early larva and are sufficient for olfactory learning at that time ( Eichler et al . , 2017; Pauls et al . , 2010 ) ; they must therefore switch their partner preferences across developmental stages . We then examined calyx anatomy in animals with intermediate populations of KCs , likely derived from 1 to 3 remaining neuroblasts . We found a progressive decrease in the size of the mushroom body calyx in these animals as measured by the maximum cross-sectional area of the calyx ( Figure 1B , C ) or by calyx volume ( Figure 1—figure supplement 1 , Figure 1—video 1 ) . To ask if this corresponded to a decline in bouton number , we counted the number of PN boutons in each hemisphere ( Figure 1E , Figure 1—figure supplement 2 ) . This revealed a linear relationship between KC number and PN bouton number , suggesting that PNs reduce their population of boutons to match the KC population . The lateral horn appeared normal in these animals and its size did not correlate with KC number , suggesting that PN projections to these two brain areas develop independent of one another ( Figure 1B , D ) . The presence or absence of ventrolateral PNs did not obviously predict bouton number , suggesting individual PNs may be able both to tailor their bouton production to the number of KCs and to adjust to the number of other PNs . However , we were not able to obtain enough samples to draw firm conclusions about whether PN cell number affected PN bouton number in calyces derived from matched KC neuroblast complements . To test this , we developed alternative methods to ablate PNs , described below . To score KC clones directly , we developed a strategy to label only the latest-born , ‘αβ core’ KCs . The somata and principle neurites of these late-born cells remain clustered in the adult , allowing us to count KC clones by counting groups of labeled soma or axon tracts as they leave the calyx ( Figure 2A , B ) . Again , we found that calyx size tracked KC clone number , and that the presence or absence of the PN progeny of lNB/BAlc did not predict calyx size ( Figure 2C ) . KC loss could reduce PN bouton number because PNs die or fail to be born when lacking KC contacts or trophic support . We thus counted the number of anterodorsal PNs ( derived from adNB/BAmv3 ) and ventrolateral PNs ( derived from lNB/BAlc ) in this cohort . The number of anterodorsal PNs was not reduced in animals with some or all KC neuroblasts ablated , and the number of ventrolateral PNs was binary rather than graded , suggesting that vlPN number is determined by whether lNB/BAlc succumbed to HU , rather than by the KC complement ( Figure 2D , E ) . These results suggest that PN neurogenesis and survival do not require KC signals . Like γd KCs , a few PNs derived from lNB/BAlc remained in adults; these likely represent PNs born embryonically from lNB/BAlc . Reduction in the number of PN boutons in the calyx could occur because individual PNs quantitatively reduce their bouton production . To ask whether individual PNs alter their production of presynaptic boutons as the KC repertoire changes , we identified two methods to label single PNs . We focused on anterodorsal PNs , as these are not susceptible to ablation of lNB/BAlc . First , we used 71D09-Gal4 to label the anterodorsal VM6 PN ( Ward et al . , 2015 ) and then subjected labeled animals to HU ablation . Because we found a linear relationship between KC number and calyx area ( Figures 1C , 2C ) , we used calyx area as a proxy for KC number . While VM6 usually produces ~10 boutons , we found only ~5 boutons in animals with reduced calyx size , and 0–1 bouton in animals with severe calyx reductions , likely lacking all olfactory KCs ( Figure 2F–H ) . These results suggest that individual PNs do reduce their bouton production as the KC complement is reduced , and that this reduction is graded rather than binary ( i . e . in animals with some but not all KC neuroblasts ablated , bouton number is in between 0 and the wild type number ) . To characterize additional PNs , we expressed photoactivatable GFP under control of R42D01-Gal4 , which labels ~10 anterodorsal PNs innervating the VM3 and VM4 glomeruli . We photoconverted individual somata using two photon microscopy ( Figure 2—figure supplement 2 ) . Though bouton counts were somewhat more variable than for VM6 as this strategy labels different types of PNs with different characteristic bouton numbers , we again found a correlation between KC loss and bouton reduction ( Figure 2—figure supplement 2 ) . The graded reductions in bouton number we observed for different PN types would preserve the relative representation of these odor channels in the calyx . We also note that the range of bouton numbers we observed are consistent with adjustments to compensate for the presence or absence of ventrolateral PNs . For example , most VM6 neurons innervating a calyx derived from 1 KC neuroblast produced between 3 and 7 boutons , as compared to the median of 10 in unperturbed animals . If KCs are reduced by ¾ without loss of lNB/BAlc , perfect compensation by PNs would reduce boutons from 10 to 2 . 5 . If lNB/BAlc was also lost , perfect compensation would reduce boutons from 10 to 5 . These predicted values are similar to the bouton range we observed experimentally . Finally , we sought to ask if KC claw number changes as KC number changes . We used GFP photoactivation to label individual KCs in animals subjected to HU ablation . The different KC classes have different median claw numbers: γ KCs have ~7 claws , αβ cells have ~5 claws , and α’β’ KCs have ~3 claws ( Caron et al . , 2013 ) ; we therefore used two different strategies to label individual KCs in animals subjected to HU ablation , both of which allow us to assign KC type . First , we expressed PA-GFP broadly in Kenyon cells using MB247-Gal4 , and assigned KC type by imaging the axonal lobe innervation of photoactivated cells ( Figure 2—figure supplement 3 ) . Second , we expressed PA-GFP only in αβ or γ KCs using 58F02-Gal4 ( Figure 2I ) and 89B01-Gal4 , respectively . Combining data from these two methods , we found no change in claw number on individual αβ KCs ( Figure 2J–K ) or γ KCs ( Figure 2—figure supplement 3 ) as the KC population was reduced . From these experiments , we conclude that PNs reduce their bouton production on a cell-by-cell basis as the KC population shrinks , while KC claw number is unaffected by the size of the KC population ( model , Figure 2L ) . In order to assess whether PNs can vary their presynaptic bouton number bidirectionally , we next created a method to specifically amplify the KC population . Recent work has suggested that ectopic or supernumerary populations of neurons can integrate into a variety of fruit fly neural circuits ( Meng et al . , 2019; Pop et al . , 2019; Prieto-Godino et al . , 2019; Seroka and Doe , 2019; Shaw et al . , 2018 ) , and previous studies identified a mutant , called mushroom body defect ( mud ) in which the mushroom body and other brain areas are dramatically expanded ( Guan et al . , 2000; Prokop and Technau , 1994 ) . Mud/NuMA is a spindle orientation protein that ensures neuroblasts divide asymmetrically to produce one neuroblast and one differentiating neuron or neural progenitor ( Siller et al . , 2006 ) . In mud mutants , neuroblasts occasionally divide symmetrically to produce two neuroblasts , amplifying the progeny of that neuroblast . To restrict amplification to KCs , we drove UAS-mud-RNAi ( BL 35044 ) using OK107-Gal4 , which is expressed in KC neuroblasts as well as in KCs ( Liu et al . , 2015 ) . Importantly , in our RNAseq data from 45 hr APF and adult animals , we observe no expression of mud in PNs , KCs , or other brain cells , suggesting mud is not expressed in differentiating neurons ( EJ Clowney , unpublished ) . As OK107 labels mature KCs as well as KC neuroblasts , we included UAS-CD8GFP in these animals to count KCs . We observed potent amplification of KCs that was variable across animals , sometimes more than doubling KC number ( Figure 3A , B ) . We measured calyx size , which expanded dramatically , and used GH146-driven fluorescence to count PN bouton number , which doubled in mushroom bodies with the largest KC expansions ( Figure 3C , Figure 3—video 1 ) . These results suggest that just as PNs scale down their boutons if the KC population is reduced , PNs scale up their boutons when the KC population is expanded . To test whether PNs increased boutons on a cell-by-cell basis in these animals , we labeled VM6 using 71D09-LexA , and PNs innervating DA1 , VA1d , and DC3 using MZ19-QF . First , we observed no change in VM6 or MZ19+ cell numbers in brains with KC expansions ( Figure 3E , H ) , suggesting that PN numbers do not increase as KC number increases . We observed no change in VM6 bouton numbers , perhaps because the VM6 PN already has a relatively large number of boutons ( Figure 3H ) . In contrast , MZ19 cells increased their number of presynaptic boutons as the KC population expanded , from an average of 3 . 5 boutons per cells in the control genotype to an average of 4 . 75 boutons per cell in the KC-expansion genotype . In brains with measurably enlarged calyces , indicative of KC population expansion , MZ19+ boutons doubled ( Figure 3D , F , G ) . To rule out reductions in the number of claws per KC as the KC population expands , we expressed photoactivatable GFP under the control of OK107 in animals with UAS-mud-RNAi or control and subjected individual or small groups of KCs to photoactivation . We scored the number of claws per labeled KC in each calyx . As above , calyx size predicted KC number in these animals , so we used calyx size as a proxy for KC number . Rather than a decrease in claw number per KC in animals with expanded KC complements , we found a slight increase in claw number , from a median of 5 to a median of 6 claws . As different types of KCs produce different characteristic numbers of claws , this could represent a change in claw production , a change in the distribution of KC types in these animals , or bias in which KC types we labeled in expanded versus control brains . While we were unable to identify LexA lines to specifically label αβ or γ KCs in OK107>mudRNAi brains , we found a LexA line , R41C07 , that labeled α’β’ KCs , and targeted these specifically for photoactivation ( Figure 3—figure supplement 1 ) . In these KCs of matched type , we found no change in KC claw number per cell , though we cannot rule out subtle differences . Together , these experiments suggest that in animals with an amplified KC complement , PNs dramatically increase their bouton production to match the KC population , while KCs may moderately increase , and do not reduce , their individual claw production . We have shown that olfactory PNs bidirectionally vary their presynaptic boutons to suit populations of KCs derived from 0 to 8 KC neuroblasts . To ask whether PNs can also vary their bouton production in response to changes in their own numbers , or in response to changes in the ratio between PNs and KCs that do not affect KC clone number , we identified methods to vary the PN population . We found that just as we could expand the KC population by knocking down mud in KC neuroblasts , we could expand the PN population by knocking down mud in PN neuroblasts . This strategy produced GH146-labeled PN complements of up to 600 cells , as compared to ~120 GH146-labeled PNs we counted in controls . While the PN neuroblast Gal4 we used , 44F03 , can label both lNB/BAlc and adNB/BAmv3 progeny ( Awasaki et al . , 2014 ) , we found that ventrolateral PNs were much more susceptible to expansion than anterodorsal PNs . Supernumerary PNs innervated the antennal lobe , causing a massive expansion of antennal lobe area ( Figure 4A , C ) . The normal glomerular map in the AL appeared distorted , but glomerular divisions were still present . These PNs also projected to the calyx , observed by thickening of the PN axon tract crossing the calyx ( red , Figure 4D ) , and increased the density of PN innervation of the lateral horn ( Figure 4D ) , though not lateral horn area ( Figure 4—figure supplement 1 ) . However , there was no change in calyx area or total bouton number in these animals compared to wild type ( Figure 4E–H ) . At a population level , the average number of boutons per PN was half that in controls ( Figure 4I ) . To ask how small groups of PNs adjusted to PN expansion , we again used MZ19-QF , which labels 15–20 PNs in controls ( Figure 4J , K ) . We counted MZ19+ cells and boutons in control versus animals of the expansion genotype ( Figure 4K ) . As the expansion was highly variable and we rarely observed more than 25 MZ19+ cells in controls , we divided the expanded genotype into hemispheres with <25 PNs or ≥25 PNs . There was a modest but insignificant increase in MZ19+ boutons in the ≥25 PNs groups ( Figure 4L ) , attributable to a strong and significant reduction of average number of boutons per PN ( Figure 4M ) . Thus PNs scale down their individual bouton repertoire to compensate for increases in the size of the PN population . Like MZ19 PNs , the VM6 PN also maintained a stable number of total boutons as the PN population expanded ( Figure 4—figure supplement 1 ) . This was possible because the VM6 cell ( s ) had similar numbers of total boutons as the VM6 population expanded . In order to reduce the PN repertoire independent of KC number , we identified two Gal4 lines , VT033006 and VT033008 , that label large complements of uniglomerular PNs in the adult ( Figure 5A ) . We drove diphtheria toxin A ( Berdnik et al . , 2006 ) and fluorescent reporters under control of these lines in animals where additional PNs were labeled by GH146-LexA ( Figure 5B , G ) . In each case , DTA eliminated all Gal4+ cells as well as additional GH146+ PNs , suggesting that these lines drive DTA expression in broader PN populations at earlier stages of development . Driving DTA under control of VT033008 reduced the total PNs labeled by VT033008 and/or GH146 by nearly half ( Figure 5C ) , while VT033006>DTA reduced PNs labeled by VT03006 and/or GH146 by >90% ( Figure 5H ) . PN labeling by VT033006 and VT033008 , and cell loss when used to drive DTA , were already evident in third instar larvae , suggesting PNs are lost well before pupal calyx wiring ( Figure 5—figure supplement 1 ) . While PN dendrites have been shown to occupy glomerular territories in pupal stages before OSNs innervate the antennal lobe ( Jefferis et al . , 2004 ) , we found that antennal lobes lacking the vast majority of uniglomerular PNs retained glomerular divisions . This suggests that these PNs may not be required for the formation of glomerular divisions in the adult antennal lobe . We measured maximum calyx cross-sectional area and bouton number across these conditions and found that in each case , the effect on the calyx was measurable but much less severe than the reduction in PN number . For VT033008>DTA , loss of nearly half the PNs yielded no change in calyx area and a 20% reduction in overall bouton number ( Figure 5D–F ) . Remarkably , while we estimate that only ~10 PNs remain in VT033006>DTA animals ( ~7% the normal complement , Figure 5H ) , the number of boutons in the calyx was reduced by only half ( Figure 5H , L ) . This suggests that individual PNs could have expanded their bouton production by as much as 6-fold . While remaining PNs sometimes appeared to make exuberant collections of boutons ( e . g . Figure 6D ) , distortion of the antennal lobes in these animals prevented us from assigning identities to these spared PNs and comparing bouton numbers to matched individual PNs in controls . This severe loss of inputs could lead to KC death . We therefore labeled and counted KCs in VT033006>DTA animals with severe PN loss . While we cannot rule out a small reduction in KC complement , at least 80% of KCs remained in these brains . We also subjected control and VT033006>DTA brains to flow cytometry and found that PN ablation brains had at least 85% as many KCs as did control brains ( Figure 5—figure supplement 2 ) . We expected that while KC claw number was largely invariant in the face of increases and decreases of the KC population , KCs would be forced to reduce their claw number when PN bouton production was reduced by 50% , as in VT033006>DTA animals . We therefore compared individual KCs in these animals versus controls . As different KC types produce different numbers of claws in wild type , we sought to simplify the analysis by labeling only a single KC type . To do this , we targeted Texas red dextran dye electroporation to the tips of the mushroom body α lobes , which are innervated by αβ KCs ( Figure 6A ) . This allowed us to label 1–12 KCs per calyx . We could discern discrete dye-labeled claws and somata in animals with eight or fewer KCs labeled , and restricted our analyses to these samples ( Figure 6B–D ) . For each hemisphere we counted labeled somata and claws , generating one average value , claws per labeled KC , per separable group of cells ( Figure 6F ) . Surprisingly , despite 90% reduction in the number of PNs and 50% reduction in their boutons , we observed no reduction in claw number per KC . Do these enduring claws receive inputs ? While labeled KCs in VT033006>DTA brains sometimes made clear contacts with spared PNs , as in Figure 6D , E , in other samples we also observed KCs that avoided the one or few GH146-labeled , spared PNs ( data not shown ) . Besides PNs , GH146 labels an inhibitory neuron , APL , that innervates the mushroom body lobes and calyx . APL is presynaptic to KCs in the calyx , but innervates portions of KCs that are not claws ( Lin et al . , 2014; Zheng et al . , 2018 ) . Labeled KCs in VT033006>DTA brains still formed contacts with APL ( Figure 6C ) . A recent study in the Drosophila larval body wall found that neurons manipulated to produce excess dendrites can obtain input from atypical presynaptic partners ( Tenedini et al . , 2019 ) . In wild type animals , 10% of bouton inputs to Kenyon cells in the main ( olfactory ) calyx are from gustatory or thermal modalities ( Caron et al . , 2013; Eichler et al . , 2017; Frank et al . , 2015; Kirkhart and Scott , 2015; Zheng et al . , 2018 ) . It is possible that thermal or gustatory neurons projecting to the calyx could have also expanded their production of boutons to make up for the losses of olfactory PNs . We cannot yet determine whether the excess KC claws observed in this condition get inputs from PN boutons , from APL , or from some other source . Nevertheless , these data suggest that KCs are inflexible in their production of claws , even in the face of severe reduction of their typical , favored presynaptic partners . The disparate and numerous strategies we used above to vary the ratio between PNs and KCs all produced the same result , that PNs adjusted to changes in PN::KC ratio by altering their presynaptic boutons , while the distribution of KC claw numbers changed little even when perturbations to PNs were so severe as to prevent production of the normal bouton complement . This finding suggests that developmental mechanisms prioritize the sparseness of olfactory inputs to KCs , and might therefore preserve sparse odor coding and high-dimensional odor representations despite stark changes to the underlying circuit constituents . We therefore sought to ask how KC odor responses were affected by perturbations to the PN::KC ratio . We used OK107 to drive GCaMP6s expression in all KCs , and subjected animals to HU just after larval hatching . We then imaged KC somatic odor responses in vivo in 24 adult HU-treated and 14 adult sham-treated animals . We observed robust odor responses in both conditions , with 8/28 sham-treated and 4/46 HU-treated calyces completely unresponsive . To identify HU-treated hemispheres with reduced KC complement , following functional imaging we imaged the anatomy of the calyx by two-photon or by immunostaining and confocal . Among ablated animals , we identified 14 of 46 imaged hemispheres in which calyces were clearly still present but of reduced size , which we designate as ‘reduced-KC’ calyces . We subjected these datasets and 17 responsive controls to motion correction using Suite2p ( Pachitariu et al . , 2017 ) . After motion correction , seven sham and seven reduced KC calyces were sufficiently still to allow us to define ROIs for individual somata . For each cell , we calculated peak change in fluorescence following odor stimulus ( Figure 7—source data 1 ) . Example responses for sham-treated and reduced-KC calyces are shown in Figure 7A , C , Figure 7—figure supplement 1 , and in Figure 7—video 1 , 2 . In order to compare response magnitudes across conditions , we first pooled all cells from each condition and compared overall Δf/f in response to each odor ( Figure 7D ) . The distributions were statistically indistinguishable between conditions for mineral oil ( mechanosensory control ) , ethyl acetate , and isobutyl acetate , and significantly different for benzaldehyde and methylcyclohexanol . Using these aggregate responses , we set a threshold for ‘responsive’ cells of 20% increase in fluorescence over baseline , as we observed bifurcation of the cellwise response distribution at this cutoff . Using these criteria , a median of ~50% of cells from sham calyces responded to ethyl acetate and isobutyl acetate , and ~20–30% of cells to benzaldehyde and methylcyclohexanol ( Figure 7E ) . These response rates are higher than has been reported with GCaMP3 and GCaMP1 ( Honegger et al . , 2011; Wang et al . , 2004 ) , but correspond with spiking response rates observed electrophysiologically ( Murthy et al . , 2008 ) . We observed variation in odor responses in both control and reduced-KC calyces; interindividual variation is expected given the stochastic innervation of KCs by PNs . We next compared the proportions of KCs per calyx responding to each odor between sham and reduced-KC calyces . These distributions were statistically indistinguishable ( Figure 7E ) . Finally , we asked how many odors each cell responded to ( Figure 7F ) . In each condition , ~60% of cells responded to 0 or one odors . Together , these findings suggest that the developmental plasticity mechanisms operating to set the density of olfactory inputs to KCs preserve qualitatively sparse KC odor responses when the KC population is experimentally reduced . Further experiments will be required to assess quantitative effects of KC reduction on responses to odors that typically stimulate fewer cells , like benzaldehyde and methylcyclohexanol .
For many animals , brain development continues while juveniles are already living freely--searching for food , navigating , and learning about the environment . Developmental transitions in brain structure are particularly stark in holometabolous insects , who build the brain twice . In D . mel , neurons generated during embryonic and early larval stages wire the larval nervous system . These circuits support larval behaviors , while neural stem cells continue to generate additional neurons that will be used to build the adult circuit . In keeping with this , the ratio between PNs and KCs in the larval olfactory circuit is starkly different from the adult: 21 embryonically-born PNs wire to early-born KCs to construct the larval mushroom body ( Eichler et al . , 2017; Masuda-Nakagawa et al . , 2005; Ramaekers et al . , 2005 ) . Connections among these populations dissolve in early pupae , and are then re-wired during pupal development ( Lee et al . , 1999; Marin et al . , 2005 ) , joined by ~100 more larvally-born PNs , and >1000 more KCs per hemisphere that continue to be born until immediately before adult eclosion . The 21 PNs in the early larva connect to ~75 KCs , a 1:3 ratio , while in the adult , ~150 PNs connect to ~2000 KCs , a 1:10 ratio . ( Eichler et al . , 2017; Zheng et al . , 2018 ) . We found that unlike cells in many other systems , including the vertebrate cerebellum , PNs and KCs did not rely on each other for survival signals ( Fan et al . , 2001 ) . This may be due to the constantly changing ratio between these cell types across developmental time . Instead , setting connectivity density cell-autonomously in KCs could allow KCs to obtain the appropriate number of inputs at the different life stages of the animal , when cellular constituents are very different from one another . Similarly , while PN neurogenesis ceases well before PNs and KCs begin to contact one another in the pupa , we estimate that ~10% of KCs are born after PN:KC synapsis has already initiated ( Muthukumar et al . , 2014 ) . Strict , cell-autonomous dendrite structuring and flexible PN bouton production could together ensure that late-born KCs obtain the inputs appropriate to support coding . Olfactory PNs of different types are specified in a predictable temporal order , have characteristic numbers of boutons , and overlap in their innervation of the calyx ( Lai et al . , 2008; Yu et al . , 2010; Zheng et al . , 2018 ) . Differences in bouton number across different PNs allow different odor channels to be differentially represented in the calyx and in KC odor responses ( Caron et al . , 2013; Honegger et al . , 2011; Murthy et al . , 2008 ) . Several classes of PNs also differ in number between the sexes ( Grabe et al . , 2016 ) . While PNs changed their individual bouton repertoires in response to changes in cell repertoires , we found that to some extent , the representation level of different PNs in the calyx was preserved . For example , in Figure 2F and Figure 2—figure supplement 2 , we show the effect of reducing KC number on bouton production by the VM6 PN and 42D01 PNs . While each population of PNs reduced individual bouton number in this condition , they retained their typical relative representation . The VM6 PN reduced its boutons from 10 to 5 , while the 42D01 PNs decreased their boutons from 4 to 2 . Similarly , in Figure 4 , we expanded the PN population by inducing ectopic PN neuroblast duplication . In these experiments , we mainly observed amplification of the ventrolateral clone . We found that individual anterodorsal VM6 cells did not scale down their boutons when the ventrolateral PN clone expanded , but only when VM6 itself was duplicated . Again , this could maintain the relative wild type representations of different odor channels in the calyx . A recent analysis suggests that the spontaneous activity of different ORs correlates with number of boutons representing that odor channel in the calyx ( Kennedy , 2019 ) . One possible model for how PNs scale to KC numbers while maintaining their relative representations in the calyx is thus that KC number limits total bouton number across all PNs , while allocation of these boutons to individual PNs is determined by activity-based competition among PN types . Qualitative aspects of sparse coding in the mushroom body appear robust to severe perturbations to the circuit . Alternative developmental compensatory mechanisms would be much less likely to preserve sparse coding . For example , we increased the ratio of PNs to KCs in two ways , by increasing the number of PNs and by decreasing the number of KCs . In both cases , PNs dialed down their bouton number , making 25–50% of the boutons they make in wild type . This allowed the KCs to receive their typical number of inputs . If in contrast bouton number was rigid and claw number flexible , in these cases KCs would have expanded their claw production 2–4 fold to innervate all the incoming PN boutons . Individual KCs with for example 20 instead of 5 claws would receive input from ~40% of glomeruli , increasing the overlap in tuning across different KCs and degrading the ability of the network to distinguish different stimuli from one another ( Litwin-Kumar et al . , 2017 ) . In two other cases , we increased the ratio of KCs to PNs , by increasing the number of KCs and by decreasing the number of PNs . Again , KCs retained their typical claw number . If instead PNs had maintained a static production of boutons while KCs had adjusted their claw production , KCs would receive very few inputs . While increasing the number of inputs per KC is theorized to reduce dimensionality of odor responses by making different KCs more similar to one another , decreasing the number of inputs per KC is theorized to reduce dimensionality by reducing the number of different possible KC input combinations ( Litwin-Kumar et al . , 2017 ) . That this sweet spot maximizing dimensionality , ~5 inputs per cell , is programmed into KC identity testifies to the evolutionary importance of maintaining connectivity density in associative brain centers that rely on combinatorial coding . The olfactory receptors are the largest insect gene family and have been subject to frequent and extreme gains and losses in many clades . Similarly , brain centers devoted to learning are radically different across species , as exemplified by the diversity in KC repertoire across arthropods ( Strausfeld et al . , 2009 ) . In order to acquire a novel olfactory circuit , many different evolutionary changes are required: A new receptor must evolve , an OSN type that uniquely expresses the receptor needs to arise , that OSN needs to synapse onto PNs , and a new PN type and new glomerulus must arise . For these events to accrue over time , each individual change must be compatible with continued circuit function and olfactory behavior . While development of a dedicated circuit that assigns an innate meaning to a newly-detectable odor would require many further changes , the signal could add to fitness immediately through representation in the mushroom body . We have described two mechanisms of developmental robustness that maintain coherent mushroom body wiring in the face of a broad range of phenotypic alterations . First , we observe that olfactory PNs can adjust to gain and loss of PNs while maintaining the balance of odor channel representations in the calyx . Plastic development of PN presynaptic sites that makes room for additional players in the repertoire would allow immediate access of evolutionarily expanded PNs to the calyx and the use of their signals for olfactory learning , thus making time for the evolution of hardwired circuits for innate interpretations . Second , we show here that developmental programs wiring the calyx can accommodate variation in KC number from at least ¼ to 2-fold the endogenous complement . Again , this flexibility could support continued MB function on the path to the evolution of mushroom body innovations . Future experiments will ask how KC claw number is developmentally programmed , and what mechanisms operate in olfactory PNs to allow them to tailor bouton production to the KC repertoire .
Flies were maintained on cornmeal-molasses ( ‘R’ ) food ( Lab Express , Ann Arbor , MI ) in a humidified incubator at 25C on a 12:12 light:dark cycle . Genotypes were as follows: Figure 1 , and Figure 1—video 1Figure 1—figure supplement 1GH146Gal4 , 10XUAS-IVS-myr::tdTomato/CyO; MB247-lexA , lexAop-CD2-GFP/TM2Figure 1—figure supplement 2GH146LexA , lexAop-mRFP-nls/P ( caryP ) attp40; VT033006-Gal4 , UAS-CD8GFP/+Figure 2A–EGH146LexA , lexAop-mRFP-nls/CyO; 58F02-Gal4 , UAS-GCaMP6s/TM2Figure 2F–HMB247-DsRed/CyO; 71D09-Gal4 , UAS-CD8GFP/"Figure 2I60A10-LexA , LexAop-tdTomato . Myr/CyO; 58F02-Gal4 , UAS-C3PA/"Figure 2J , KGH146Gal4 , 10XUAS-IVS-myr::tdTomato/CyO; MB247-lexA , 13XLexAop2-IVS-Syn21-mC3PA-GFP-p10/"60A10-LexA , LexAop-tdTomato . Myr/CyO; 58F02-Gal4 , UAS-C3PA/"Figure 2—figure supplement 1GH146LexA , lexAop-mRFP-nls/CyO; 58F02-Gal4 , UAS-GCaMP6s/TM2Figure 2—figure supplement 2MB247-DsRed/CyO; 42D01-Gal4 , UAS-C3PA/"Figure 2—figure supplement 3A , BGH146Gal4 , 10XUAS-IVS-myr::tdTomato/CyO; MB247-lexA , 13XLexAop2-IVS-Syn21-mC3PA-GFP-p10/"Figure 2—figure supplement 3C , DGH146Gal4 , 10XUAS-IVS-myr::tdTomato/CyO; MB247-lexA , 13XLexAop2-IVS-Syn21-mC3PA-GFP-p10/"60A10-LexA , LexAop-tdTomato . Myr/CyO; 89B01-Gal4 , UAS-C3PA/"Figure 3A–CGH146QF , QUAS-mtdTomato-3xHA /+; UAS-CD8GFP/UAS-mud-RNAi; OK107Gal4/+GH146QF , QUAS-mtdTomato-3xHA /+; UAS-CD8GFP/P ( caryP ) attp2; OK107Gal4/+Figure 3D–GMZ19QF , QUAS-mtdTomato-3xHA/Bl; UAS-mud-RNAi/TM2 or TM6B; OK107Gal4/+MZ19QF , QUAS-mtdTomato-3xHA/Bl; P ( caryP ) attp2/TM2 or TM6B; OK107/+Figure 3H71D09-LexA , LexAop-tdTomato . Myr/Bl; UAS-mud-RNAi/TM2; OK107 Gal4/+71D09-LexA , LexAop-tdTomato . Myr/Bl; P ( caryP ) attp2/TM2; OK107 Gal4/+Figure 3IUAS-C3PA/+; UAS-C3PA/UAS-mud-RNAi; OK107Gal4/+UAS-C3PA/+; UAS-C3PA/P ( caryP ) attp2; OK107Gal4/+Figure 3J , KUAS-C3PA/+; UAS-C3PA/UAS-mud-RNAi; OK107Gal4/+UAS-C3PA/+; UAS-C3PA/P ( caryP ) attp2; OK107Gal4/+UAS-C3PA/41C07-LexA , LexAop-tdTomato . Myr; UAS-C3PA/UAS-mud-RNAi; OK107Gal4/+UAS-C3PA/41C07-LexA , LexAop-tdTomato . Myr; UAS-C3PA/P ( caryP ) attp2; OK107 Gal4/+UAS-C3PA/89B01-LexA , LexAop-tdTomato . Myr; UAS-C3PA/UAS-mud-RNAi; OK107 Gal4/+UAS-C3PA/89B01-LexA , LexAop-tdTomato . Myr; UAS-C3PA/P ( caryP ) attp2; OK107 Gal4/+Figure 3—figure supplement 141C07-LexA , LexAop-tdTomato . Myr/UAS-C3PA; UAS-mud-RNAi 35044/UAS-C3PA; OK107 Gal4/+41C07-LexA , LexAop-tdTomato . Myr/UAS-C3PA; P ( caryP ) attp2/UAS-C3PA; OK107 Gal4/+Figure 4A–IGH146LexA , LexAop-SPA-T2A-SPA/+; 44F03-Gal4/UAS-mud-RNAiGH146LexA , LexAop-SPA-T2A-SPA/+; 44F03-Gal4/P ( caryP ) attp2Figure 4J–MGH146LexA , LexAop-SPA-T2A-SPA/MZ19QF , QUAS-mtdTomato-3xHA; 44F03-Gal4/UAS-mud-RNAiGH146LexA , LexAop-SPA-T2A-SPA/MZ19QF , QUAS-mtdTomato-3xHA; 44F03-Gal4/P ( caryP ) attp2MZ19QF , QUAS-mtdTomato-3xHA/+; 44F03-Gal4/UAS-mud-RNAiMZ19QF , QUAS-mtdTomato-3xHA/+; 44F03-Gal4/P ( caryP ) attp2Figure 4—figure supplement 1A , BGH146LexA , LexAop-SPA-T2A-SPA/+; 44F03-Gal4/UAS mud-RNAiGH146LexA , LexAop-SPA-T2A-SPA/+; 44F03-Gal4/P ( caryP ) attp2Figure 4—figure supplement 1C–F71D09-LexA , LexAop-tdTomato . Myr/+; 44F03-Gal4/UAS-mud-RNAi71D09-LexA , LexAop-tdTomato . Myr/+; 44F03-Gal4/P ( caryP ) attp2Figure 5ACyO/+; VT033006-Gal4/UAS-CD8GFPCyO/+; VT033008-Gal4/UAS-CD8GFPFigure 5B , CGH146LexA , lexAop-mRFP-nls/UAS DTA; VT033008-Gal4 , UAS-CD8GFP/+GH146LexA , lexAop-mRFP-nls/P ( caryP ) attp40; VT033008-Gal4 , UAS-CD8GFP/+Figure 5D–FUAS-DTA/+; VT033008-Gal4/TM6BUAS-DTA/+; P ( caryP ) attp2/TM6BFigure 5G–HGH146LexA , lexAop-mRFP-nls/UAS DTA; VT033006-Gal4 , UAS-CD8GFP/+GH146LexA , lexAop-mRFP-nls/P ( caryP ) attp40; VT033006-Gal4 , UAS-CD8GFP/+Figure 5I–LMB247-DsRed/UAS DTA; VT033006-Gal4 , UAS-CD8GFP/+MB247-DsRed/+; VT033006-Gal4 , UAS-CD8GFP/+Figure 5—figure supplement 1AGH146LexA , lexAop-mRFP-nls/UAS DTA; VT033008-Gal4 , UAS-CD8GFP/+GH146LexA , lexAop-mRFP-nls/CyO; VT033008-Gal4 , UAS-CD8GFP/+Figure 5—figure supplement 1BGH146LexA , lexAop-mRFP-nls/UAS DTA; VT033006-Gal4 , UAS-CD8GFP/+GH146LexA , lexAop-mRFP-nls/CyO; VT033006-Gal4 , UAS-CD8GFP/+Figure 5—figure supplement 2MB247-DsRed/UAS DTA; VT033006-Gal4 , UAS-CD8GFP/+MB247-DsRed/+; VT033006-Gal4 , UAS-CD8GFP/+Figure 6GH146LexA , LexAop-SPA-T2A-SPA/UAS DTA; VT033006-Gal4/+GH146LexA , LexAop-SPA-T2A-SPA/P ( caryP ) attp40; VT033006-Gal4/+Figure 7 , Figure 7—figure supplement 1 , Figure 7-videos 1 , 2UAS-GCaMP6s/UAS-GCaMP6s; UAS-GCaMP6s/UAS-GCaMP6s; OK107Gal4/ OK107Gal4 or + Protocol was adapted from Sweeney et al . ( 2012 ) . Briefly , we set up large populations of flies in cages two days prior to ablation and placed a 35 or 60 mm grape juice agar plate ( Lab Express , Ann Arbor , MI ) in the cage with a dollop of goopy yeast . One day prior to the ablation , we replaced the grape juice/yeast plate with a new grape juice/yeast plate . On the morning of the ablation , we removed the plate from the cage and discarded the yeast puck and any hatched larvae on the agar . We then monitored the plate for up to four hours , until many larvae had hatched . Larvae were washed off the plate using a sucrose solution , and eggs were discarded . Larvae were then strained in coffee filters , and submerged in hydroxyurea ( Sigma , H8627 ) in a yeast:AHL mixture , or sham mixture without HU . Ablation conditions were as follows: Larvae were then strained through coffee filters again , rinsed , and placed in a vial or bottle of R food until eclosion . We opened a new container of hydroxyurea each month as it degrades in contact with moisture and we found its potency gradually declined . We typically analyzed 5–10 animals ( 10–20 hemispheres ) of each condition per batch , except in in vivo imaging experiments , where we analyzed 1–5 animals per condition per batch . Applying HU for four hours produced a somewhat U-shaped distribution , with many samples unaffected and many with all four KC neuroblasts lost . Shifting from 3 , to 5 , to 10 mg over this long time scale shifted how many brains were completely ablated versus unaffected , but did little to increase the proportion of brains with intermediate phenotypes . Applying the stronger concentration , 10 mg , for a shorter time produced more sporadic effects , with more brains showing 1 , 2 , or 3 KC neuroblasts remaining . lNB/BAlc was more affected by HU concentration than time of application , with stronger effects seen applying 10 mg/mL for one hour than applying 3 or 5 mg/mL for four hours . Overall , combining these different protocols allowed us to observe a broad range of mushroom body states . Brains were dissected for up to 20 min in external saline ( 108 mM NaCl , 5 mM KCl , 2 mM CaCl2 , 8 . 2 mM MgCl2 , 4 mM NaHCO3 , 1 mM NaH2PO4 , 5 mM trehalose , 10 mM sucrose , 5 mM HEPES pH7 . 5 , osmolarity adjusted to 265 mOsm ) , before being transferred to 1% paraformaldehyde in PBS , on ice . All steps were performed in cell strainer baskets ( caps of FACS tubes ) in 24 well plates , with the brains in the baskets lifted from well to well to change solutions . Brains were fixed overnight at 4C in 1% PFA in PBS . On day 2 , brains were washed 3 × 10’ in PBS supplemented with 0 . 1% triton-x-100 on a shaker at room temperature , blocked 1 hr in PBS , 0 . 1% triton , 4% Normal Goat Serum , and then incubated for at least two overnights in primary antibody solution , diluted in PBS , 0 . 1% triton , 4% Normal Goat Serum . Primary antibody was washed 3 × 10’ in PBS supplemented with 0 . 1% triton-x-100 on a shaker at room temperature , then brains were incubated in secondary antibodies for at least two overnights , diluted in PBS , 0 . 1% triton , 4% Normal Goat Serum . When used , Alexa 568-conjugated phalloidin ( 1:80 ) and/or DAPI ( one microgram/mL ) were included in secondary antibody mixes . Primary antibodies used were mouse anti-ChAT 9E10 ( DSHB , 1:200 ) , Rabbit anti-dsRed ( Clontech , 1:500-1:1000 ) , Sheep anti-GFP ( Bio-Rad , 4745–1051 , 1:250-1:1000 ) , Chicken anti-GFP ( 1:5000 , gift from Dawen Cai ) . Secondary antibodies were Alexa 488 , 568 , and 647 conjugates ( 1:500 , Invitrogen ) . Brains were mounted in 1x PBS , 90% glycerol supplemented with propyl gallate in binder reinforcement stickers sandwiched between two coverslips . Samples were stored at 4C in the dark prior to imaging . The coverslip sandwiches were taped to slides , allowing us to perform confocal imaging on one side of the brain and then flip over the sandwich to allow a clear view of the other side of the brain . This allowed us to score features on the anterior and posterior sides of each sample . Scanning confocal stacks were collected along the anterior-posterior axis on a Zeiss 880 or Leica SP8 with one micrometer spacing in Z and ~200 nm axial resolution . Determination of sample size: Brains were prepared for imaging in batches of 5–10 . In initial batches , we assessed the variability of the manipulation , for example if we were trying to change Kenyon cell number , we looked at how variable the size of the Kenyon cell population was following the manipulation . We used this variability to determine how many batches to analyze so as to obtain enough informative samples . To avoid introducing statistical bias , we did not analyze the effect of the manipulation on the other cell type until after completing all batches; for example if the manipulation was intended to alter Kenyon cell numbers , we did not assess projection neuron bouton phenotypes until completing all samples . Genotypes or conditions being compared with one another were always prepared for staining together and imaged interspersed with one another to equalize batch effects , and we used at least two batches for each type of experiment . Criteria for exclusion , treatment of outliers: In Figures 1–6 , we only excluded from analysis samples with overt physical damage to the cells or structures being measured . In figures and analyses we treated outliers the same way as other data points . For Figure 7 , full criteria for inclusion and exclusion of each sample are shown under Analysis of KC somatic odor responses , below . Recent discussions among experts have suggested that biological studies over-rely on statistical tests , report p values when statistical differences are self-evident , and emphasize statistical significance rather than effect size ( Goedhart , 2018a; Goedhart , 2018b; Ho et al . , 2019 ) . These choices can cloud findings instead of clarifying them . In order to communicate our findings in the simplest and most complete way , we have displayed each data point for each sample to allow readers to assess effect size and significance directly . In general , we analyzed mixed-sex populations , where sex ratios were carefully balanced across experimental conditions . Because MZ19 labels the sexually dimorphic DA1 PNs , in MZ19 experiments we analyzed the two sexes independently in pilot experiments . As we observed no correlation with sex ( not shown ) , in subsequent MZ19 experiments we used mixed sexes . Researchers performing quantification could not generally be blinded to experimental condition due to the overt changes in neuron numbers and brain structures induced by our manipulations . However , analysis was performed blind to the goals of the experiment when possible , and quantitation of features on the anterior and posterior sides of the brain were recorded independent of one another and merged after all quantifications were completed . Moreover , many of our analyses make use of variation within an experimental condition or genotype , providing an additional bulwark against observational bias . To measure neuropil structures such as the mushroom body calyx , lateral horn , or antennal lobe , we used markers such as ChAT and phalloidin to visualize the structure , identified its largest extent in Z ( i . e . along the A-P axis ) , outlined it in FIJI ( as in white outlines in Figure 1B ) and then calculated the cross-sectional area using the ‘Measure’ command . To analyze calyx volumes we used ImageJ . Briefly , we drew Regions of Interest ( ROIs ) around the MB247 ( Figure 1 ) or OK107 ( Figure 3 ) calyx signal for each slice . Other imaging channels displaying projection neurons and nc82 staining were used as a reference . The drawn ROIs for each slice were saved using the ROI manager plugin and the area of each ROI in a stack was measured using the Measure tool . The measured areas were added together and then multiplied by the Z-spacing between slices ( one micron ) to get the volume of the combined ROI through the stack . The 3D viewer plugin was used to make movies of the kenyon cell and calyx volumes . To produce 3D Movies of the calyx volumes without somata , areas outside the selected ROIs were cleared and then the 3D viewer plugin was used to display only the 3D ROI volume that was measured for the volume analysis . To facilitate comparison in 3D movies , stacks in a set were all given the same depth in Z by adding blank slices at the end of the stack when necessary . To count aggregate boutons , we used ChAT signal with reference to phalloidin ( which stains actin-rich structures , including KC dendrites ) , except in Figure 1 and Figure 3A , where we counted GH146-positive boutons . We counted as separate structures ChAT signals that were compact and appeared discontinuous with one another and that were 2+ micrometers in diameter ( Figure 1—figure supplement 2 ) . When phalloidin signal or KC fluorescence was available , two boutons would be counted separately if ChAT signals were separated by phalloidin signal or KC signal . In initial analyses , we found that boutons in slice 0 often appeared in slices −1 and +1 as well , but never in slices −2 or +2 . In order to avoid counting the same boutons more than once , we therefore began counting at the most superficial slice in the stack where boutons were visible , and counted every other slice , i . e . every second micron . To count boutons of small groups of PNs , we drove fluorescent reporters under control of 42D01 , 71D09 , or MZ19 and counted coherent and compact fluorescent signals , with reference to ChAT signals and to phalloidin , when available . As GH146 labels ~2/3 of PNs , our counts of 450–650 total GH146+ boutons in Figures 1 , 3 , which would correspond to 675–975 total boutons per calyx , correspond with estimates established by EM ( 578 boutons ) and light microscopy ( 768 total boutons , or five boutons per PN ) ( Leiss et al . , 2009; Turner et al . , 2008; Zheng et al . , 2018 ) . To count cell populations , we used genetically-encoded fluorescence as indicated . We counted labeled somata in every third slice in the stack ( every third micron along the A-P axis ) , with reference to DAPI to distinguish individual cells from one another . As we did for boutons , in analyzing somata we initially determined that somata in slice 0 could also be seen in slices −2 , –1 , +1 , and +2 but not in slice −3 or +3 . To avoid double-counting , we therefore counted every third micron . To count KC claws , we scanned image stacks for cup-shaped terminal structures of the appropriate size ( 1–2 μm in diameter ) . We sometimes labeled small groups of KCs . We scored independently any single KC or small group of KCs that could be visually separated from one another ( like the 3 cells shown in Figure 6C ) . For Figure 2 and Figure 2—figure supplement 3 , we did not score any cells which we could not separate from one another because we did not want to mix counts for KCs of different types . For Figure 3I , where we did not attempt to separate KCs of different types , and Figure 6F , where our dye labeling strategy was specific to αβ KCs , we present an average claw/KC data point for small groups of 2–4 cells whose dendritic structures were interwoven . Finally , in Figure 3I , we found that in calyces with expanded KC populations , KC axons sometimes could not enter the pedunculus and instead wandered around at the base of the calyx . As CNS neurons in D . mel are unipolar , these wandering axons could be difficult to distinguish from dendritic claws . To determine whether these wandering axons should be counted as claws , we stained a subset of photoactivated brains with anti-ChAT ( not shown ) , to determine which labeled KC neurites were in the bouton region of the calyx , and should thus be scored for claws , and which were ‘lost’ axons . Though photoactivated GFP signals were only poorly preserved following staining , relating ChAT-stained images to images of the same brain by two photon allowed us to learn to identify the bouton region of the calyx in subsequent experiments due to its ‘swiss cheese’ appearance , where the presence of PN boutons produces holes in KC fluorescence . To determine KC and PN neuroblast state--PN lNB/BAlc: Whenever possible , we included GH146 or 60A10 driving a fluorescent report in our animals and scored the presence or absence of a group of somata lateral to the antennal lobe on the anterior side of the brain for each analyzed hemisphere . In Figure 2F–H , we examined ChAT+ somata ventrolateral of the antennal lobe to determine the presence or absence of lNB/BAlc progeny . KC neuroblasts: In Figure 2A–C , we used 58F02 to fluorescently label late-born KCs and counted clumps of labeled somata surrounding the calyx as well as groups of labeled neurites leaving the calyx and entering the pedunculus . These estimates usually matched; in the few cases where they did not , we used the number of axon clumps , as somata are closer to the surface of the brain and more susceptible to mechanical disruption during dissection . As seen in Figure 2B , 58F02 labels additional neurons; these were easy to discriminate from KCs as they did not enter the calyx or pedunculus . For Figure 2J , K and Figure 2—figure supplement 3C , D , we combined data from two different types of experiments . In some experiments , we expressed PA-GFP in all KCs using MB247 . We traced the axons of labeled KCs into the lobes to assign KC type ( Figure 2—figure supplement 3 ) . In a second group of experiments , we used 58F02-Gal4 to express PA-GFP exclusively in αβ core KCs ( Figure 2I ) , or 89B01 Gal4 to express PA-GFP only in γ KCs and thus knew KC type a priori . 89B01 also labels γd KCs , which can be distinguished from olfactory γ KCs as they do not innervate the main calyx . We do not include γd KCs in our analysis . GFP photoactivation and Texas red dextran dye labeling were performed as previously described ( Clowney et al . , 2015; Ruta et al . , 2010 ) , using a Bruker Investigator microscope and Spectra-Physics MaiTai laser with DeepSee module . Dye filling electrodes were pulled using a Brown/Flaming puller ( Sutter Instruments ) and were guided to the mushroom body α lobes using visible light illumination and/or two photon autofluorescence . To identify Gal4 and/or LexA lines labeling PN populations , we screened the Janelia FlyLight collection online database and then crossed lines with potential to fluorescent reporters . While 42D01 ( ~10 PNs ) and 60A10 ( 60 PNs ) were the most useful PN lines for us here , we also identified other lines , 33C10 and 37H08 , that each label small groups of PNs and may be of use to others . To identify Gal4 lines labeling individual KC types , we screened expression of ‘regular’ FlyLight Gal4 constructs made using the same regulatory fragments incorporated in KC-type-specific split Gal4s ( Aso et al . , 2014 ) . We found 58F02 useful for labeling α/β core KCs , 89B01 for γ KCs ( labels both γmain and γd ) , and 41C07 for α’/β’ KCs . 89B01 and 58F02 LexA did not recapitulate Gal4 expression patterns driven by the same fragments . Brains were dissected in Schneider’s medium ( Sigma S0146 ) supplemented with 1% BSA and placed on ice . After all dissections were completed , collagenase ( Sigma C0130 ) was added to a final concentration of 2 mg/mL and samples were incubated at 37C for 20 min . Samples were dissociated by trituration and spun down at 300g , 4C , for 5 min . Collagenase solution was removed and replaced with PBS+0 . 1% BSA supplemented with 1:1000 DyeCycle Violet . Cells were incubated with dye for 30 min on ice and passed through a cell strainer cap , before being subjected to flow cytometry on an Attune Flow Cytometer . Forward scatter and side scatter measurements were used to gate single cells , and DyeCycle Violet signal was used to gate cell bodies from membrane debris . We prepared 2–7 day old adult flies subjected to HU or sham treatment just after larval hatching for in vivo two photon calcium imaging on a Bruker Investigator essentially as described previously ( Ruta et al . , 2010 ) , affixing the fly to packaging tape ( Duck EZ Start ) with human hair and UV glue ( Loctite 3106 ) . The fly was tilted to allow optical access to KC somata , at the dorsal posterior surface of the brain . In some experiments , we waxed the proboscis in an extended position to reduce motion . Imaging was performed in external saline . For odor delivery , half of a 1200 mL/min airstream , or 600 mL/min , was directed toward the antennae through a Pasteur pipette mounted on a micromanipulator . At a trigger , 25% of the total air stream was re-directed from a 10 mL glass vial containing mineral oil to a vial containing odorants diluted 1:10 in mineral oil , or a second vial of mineral oil ( mechanosensory control ) . Final odor dilution was therefore 1:40 . Filter paper ‘wicks’ were inserted into each vial to allow odor to saturate the vial headspace . Odors were delivered for two seconds , with 30–60 s in between stimulations . We used a simplified olfactometer capable of delivering five different odorants in which overall airflow was metered by analogue flowmeters ( Brooks Instruments ) and valve switching controlled by an Arduino . Odor delivery was initially optimized using a mini-PID placed in the half of the air stream not directed at the fly ( Aurora Biosciences ) . Images were collected at 1 . 5–5 Hertz , and we imaged a single plane for each sample . We collected data from a total of 14 sham-treated animals ( 28 hemispheres imaged ) and 24 HU-treated animals ( 46 hemispheres imaged ) . The two hemispheres were imaged separately . Occasionally , only one hemisphere was imaged due to the preparation and a few hemispheres were excluded from analysis because of poor image quality . Mineral oil was delivered first , and then each of four odors was delivered twice , in sequence . Following mineral oil , odor order varied . Because experiments in locust found that the first presentation of certain odors can cause distinct KC responses to all subsequent presentations of that same odor , we analyzed the second presentation of each odor , according to convention ( Murthy et al . , 2008; Stopfer and Laurent , 1999 ) . We also observed that in some animals , enduring increases in fluorescence followed the first presentation of isobutyl acetate . Following functional imaging , we either collected a Z-stack of the mushroom body on the two-photon , or dissected out the brain and subjected it to immunostatining and confocal imaging . We used these images to categorize the extent of KC reduction . All samples are accounted for here: Total hemPoor imageDamaged/not responsiveUnknownFull ablationCan’t tell if ablatedDoesn’t look ablatedToo much motionAnalyzedSham28281N/AN/AN/A107: four fem , three maleHU46401351077: four male , three fem We excluded 11 hemispheres from the control dataset due to damage or poor image quality . We also excluded 34 hemispheres from the ablation dataset because of any of the following: the samples were fully ablated , unaffected by ablation , ablation status was ambiguous , or the mushroom body calyx appeared damaged . The remaining samples were motion-corrected using Suite2p ( Pachitariu et al . , 2017 ) . Using FIJI , we blinded ourselves to the remaining datasets and excluded datasets if there was too much motion to be able to follow the same cell over time . After this gate , there were seven sham and seven reduced-KC hemispheres for the analysis . ROIs were chosen in each of the fourteen samples and we measured Δf/f ( i . e . ( f-f0 ) /f0 ) , comparing stimulus frames with pre-stimulus frames . ROIs were manually chosen and we were careful about ensuring that the cell remained within its ROI for all frames . For two of the ablated samples , we had to choose different ROIs for two of the odor deliveries due to motion . These samples were only used to calculate the proportion of cells that were responsive to each odor , and not in calculating the number of odors each cell responded to . The baseline was determined for each odor delivery by taking the average of the frames within the three seconds before the trigger . The stimulus frame was determined by taking the peak response between 0–4 . 5 s after stimulus onset , to account for the valve opening and the two second odor delivery . To define ‘responsive’ cells , we chose to use the same cutoff , Δf/f > 0 . 2 ( 20% increase in fluorescence over baseline ) across samples . This was in order to avoid obscuring overall differences in responsiveness across ablated and sham-treated calyces . For visualization , we used the Pretty Heatmap package in R to cluster cells with similar odor responses and display them . | Despite having a limited number of senses , animals can perceive a huge range of sensations . One possible explanation is that the brain combines several stimuli to make each specific sensation . The olfactory learning system in the fruit fly Drosophila melanogaster is in a part of the brain called the mushroom body . It allows fruit flies to associate a specific smell with a reward ( e . g . food ) or a punishment ( e . g . poison ) and behave accordingly . Two groups of neurons process stimuli from sensory receptors in the mushroom body: olfactory projection neurons carry information from the receptors and pass it on to neurons called Kenyon cells . The system relies on Kenyon cells receiving the combined input of multiple olfactory projection neurons , and therefore information from multiple receptors . The number of inputs each Kenyon cell receives is thought to determine the number of sensations that can be told apart , and thus , the number of signals that can be used for learning . While many mechanisms dictating the complexity of a neuron’s shape have been described , the logic behind how two populations of neurons become connected to combine several inputs into a single sensation has not been addressed . A better understanding of how these connections are established during development can help explain how the brain processes information , and the D . melanogaster mushroom body is a good system to address these questions . Elkahlah , Rogow et al . manipulated the number of olfactory projection neurons and Kenyon cells in the mushroom body of fruit flies during development . They found that despite there being a varying number of cells , the number of connections into a post-synaptic cell remained the same . This indicates that the logic behind the combinations of inputs required for a sensation depends on the Kenyon cell , while olfactory projection neurons can adapt during their development to suit these input demands . Thus , if there are fewer Kenyon cells , the olfactory projection neurons will each provide connections to fewer cells to compensate , and if there are fewer olfactory projection neurons , each of them will input into more Kenyon cells . To show that the developing mushroom body could indeed adapt to different numbers of olfactory projection neurons and Kenyon cells , the modified flies were tested for olfactory perception: their responses to odor were largely normal . These results underline the robustness of neuronal circuits . During development , the mushroom body can compensate for missing or extra neurons by modifying the numbers of connections between two groups of neurons , thus allowing the olfactory system to work normally . This robustness may also predispose the system to evolutionary change , since it allows the system to continue working as it changes . These findings are relevant to any area of the brain where neurons rely on combined input from many sources . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2020 | Presynaptic developmental plasticity allows robust sparse wiring of the Drosophila mushroom body |
The intergenic IRES of Cricket Paralysis Virus ( CrPV-IRES ) forms a tight complex with 80S ribosomes capable of initiating the cell-free synthesis of complete proteins in the absence of initiation factors . Such synthesis raises the question of what effect the necessary IRES dissociation from the tRNA binding sites , and ultimately from all of the ribosome , has on the rates of initial peptide elongation steps as nascent peptide is formed . Here we report the first results measuring rates of reaction for the initial cycles of IRES-dependent elongation . Our results demonstrate that 1 ) the first two cycles of elongation proceed much more slowly than subsequent cycles , 2 ) these reduced rates arise from slow pseudo-translocation and translocation steps , and 3 ) the retarding effect of ribosome-bound IRES on protein synthesis is largely overcome following translocation of tripeptidyl-tRNA . Our results also provide a straightforward approach to detailed mechanistic characterization of many aspects of eukaryotic polypeptide elongation .
Initiation of protein synthesis in eukaryotic cells proceeds via two well-established pathways . The cap-dependent pathway involves recognition of 7-methyl-guanosine at the 5’-terminus of mRNA by a preinitiation complex of 40S ribosomal subunit and a host of initiation factors prior to a scanning step that results in initiator aminoacyl-tRNA ( aa-tRNA ) pairing with a cognate start codon , followed by 60S binding to form the 80S initiator complex ( Jackson et al . , 2010; Aitken and Lorsch , 2012 ) . The second pathway involves binding of the ribosome to an internal ribosome entry site ( IRES ) , a structure that is present in many virus-encoded mRNAs , as well as in some cellular mRNAs ( Fitzgerald and Semmler , 2009 ) . Initiation of protein synthesis from an 80S·IRES complex can take place in the absence of some or even all of the initiation factors required in the cap-dependent pathway ( Filbin and Kieft , 2009 ) , depending on the IRES source . The intergenic IRES of Cricket Paralysis Virus ( CrPV-IRES ) forms a complex with 80S ribosomes that is capable of initiating the synthesis of complete proteins in cell-free assays completely lacking initiation factors ( Jan et al . 2003; Pestova and Hillen , 2003 ) . More recently , high resolution structural studieshave shown that , prior to polypeptide chain initiation , the closely related Dicistroviridae IRES structures from CrPV ( Fernandez et al . , 2014; Muhs et al . , 2015 ) and Taura syndrome virus ( Koh et al . , 2014 ) occupy all three tRNA binding sites ( E , P , and A ) on the ribosome , with the protein coding region beginning immediately downstream from IRES segment occupying the A-site ( Figure 1 ) . 10 . 7554/eLife . 13429 . 003Figure 1 . Structure of CrPV-IRES bound to the 80S ribosome superposed on A , P , and E tRNA binding sites . The position of the first codon is indicated . Adapted from Fernandez et al . ( 2014 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 00310 . 7554/eLife . 13429 . 004Figure 1—figure supplement 1 . In vitro translation of firefly luciferase with WT and mutated F-IRES mRNA . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 004 CrPV-IRES binds with high affinity ( Kd ~ 10 nM ) to the 80S ribosome ( Jang and Jan , 2010 ) , raising the question of what effect the necessary IRES dissociation from the tRNA binding sites , and ultimately from all of the ribosome as well , has on the rates of initial peptide elongation steps as nascent peptide is formed ( Muhs et al . , 2015 ) . Since prior to the work reported in this paper nothing had been published concerning the rate of initial oligopeptide synthesis by an 80S·CrPV-IRES complex , it has been unclear whether there is a retarding effect due to the presence of IRES on the ribosome , and , if so , how many cycles of peptide elongation are required before the ribosome begins to form peptide bonds at a higher rate . In considering this question , we make use of the simplified 12-step scheme of initial tetrapeptide synthesis shown in Figure 2 , which provides a useful framework for presenting the results described in this paper . In this scheme Steps 1–3 show the reactions required for initial binding of the first tRNA to the A site followed by translocation to the P-site , and reactions 4–6 , 7–9 , and 10–12 represent three elongation cycles , ending with P-site bound tetrapeptidyl-tRNA , completing the third cycle of polypeptide synthesis . This model makes the reasonable assumption that binding of successive aminoacyl-tRNAs ( aa-tRNAs ) cognate to the mRNA requires the progressive removal of IRES structures from each of the tRNA binding sites , such that translocation of dipeptidyl tRNA to the P-site ( structure 7 ) requires removal of the IRES from the last of the three tRNA binding sites . In the work reported below , we demonstrate first , that the initial elongation steps are indeed quite slow and are limited by the translocation step of the elongation cycle , and second , that the rate of elongation accelerates following translocation of tripeptidyl-tRNA to the P-site . 10 . 7554/eLife . 13429 . 005Figure 2 . Proposed scheme for initial tetrapeptide synthesis on CrPV IRES-programmed ribosomes . This simplified scheme neglects the several substeps , including GTP hydrolysis , Pi release , and elongation factor release , that accompany both productive binding of ternary complex to the ribosome ( Steps 2 , 4 , 7 , 10 ) and tRNA translocation ( Steps 3 , 6 , 9 , 12 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 005
We previously have utilized two assays to measure binding of the ternary complex Phe-tRNAPhe·eEF1A·GTP ( Phe-TC ) to the 80S·CrPV-IRES ( 80S·IRES ) complex ( Ruehle et al . , 2015 ) . The increase in proflavin-labeled Phe-tRNAPhe fluorescence anisotropy measures binding to either the A- or P-site ( structures 3 and 4 , respectively , Figure 2 ) . [3H]-Phe-tRNAPhe cosedimentation with the 80S·IRES complex measures accumulation of 4 only , since A-site binding is too labile to survive the ultracentrifugation step ( Yamamoto et al . , 2007 ) . In Figure 3 we present time-resolved application of the anisotropy assay that allows us to measure the rates of Phe-TC binding to form Structure 3 from 1 . These resultswere fit to the scheme shown in Figure 2 , giving values for k1 , k-1 , and k2 in both the presence and absence of eEF2·GTP that are summarized in Table 1 . In the absence of eEF2 ( blue trace ) , the equilibrium position of Step 1 , a so-called pseudo-translocation step ( Muhs et al . , 2015 ) in which the IRES vacates the A-site , favors Structure 1 over Structure 2 by approximately 20-fold , consistent with recent structural studies ( Fernandez et al . , 2014; Koh et al . , 2014; Muhs et al . , 2015 ) . Phe-TC binds to Structure 2 yielding Structure 3 , in a process where the rate-limiting step is the conversion of Structure 1 to Structure 2 . Preincubation of 80S·IRES complex with 1 µM or 3 µM eEF2·GTP leads to clear biphasic binding of Phe-TC , with the more rapid and slower phases each accounting for ~50% of binding , respectively ( red and black traces ) . These results indicate that , consistent with recent results of Petrov et al . ( 2016 ) , the equilibrium between Structures 1 and 2 is shifted in the presence of eEF2·GTP , such that approximately half of 80S·IRES is present as 2 . Phe-TC binding to 2 , resulting in the formation of Structure 3 , accounts for the rapid phase in the red and black traces . Further formation of 3 is limited by the slower rate of 1 to 2 conversion . Although added eEF2·GTP decreases all three apparent rate constants , the effect is much greater on k-1 ( ~50-fold reduction ) than on either k1 ( ~twofold reduction ) or k2 ( ~fourfold reduction ) . The near identity of the red and black traces , performed at different eEF2·GTP concentrations , suggests that this factor interacts with both 1 and 2 , with a dissociation constant significantly less than 1 µM . The large inhibitory effect of eEF2·GTP on k-1 is consistent with its role as a translocase , and with recent results demonstrating that a principal role of EF-G , the prokaryotic equivalent of eEF2 , is to inhibit back-translocation ( Adio et al . , 2015 ) . eEF2·GTP inhibition of k2 may be due , at least in part , to a requirement for eEF2·GDP dissociation prior to Phe-TC binding . 10 . 7554/eLife . 13429 . 006Figure 3 . Rates of initial Phe-tRNAPhe binding measured by fluorescence anisotropy or Phe-tRNAPhe cosedimentation . Fluorescence anisotropy changes were monitored after rapid mixing of Phe-tRNAPhe ( Prf ) ternary complex ( 0 . 1 µM final concentration , containing 1 mM GTP ) with 80S·FVKM-IRES complex ( 0 . 1 µM final concentration ) either in the absence of eEF2 ( blue line ) or with 80S·FVKM-IRES complex that was pre-incubated with either 3 µM ( black line ) or 1 µM eEF2·GTP ( red line ) for 1–2 hr . These long times ensured full equilibration prior to TC addition . In the latter cases , eEF2 concentration was kept constant by including 3 µM or 1 µM eEF2 , respectively , in the TC solution . eEF2 displays virtually no GTPase activity when it is not bound to the ribosome ( Nygård and Nilsson , 1989 ) . Rates of Phe-tRNAPhe binding to the P site , as determined by cosedimentation , were measured by rapidly mixing Phe-TC ( 1 . 6 µM final concentration ) with 80S·FVKM-IRES complex ( 0 . 8 µM final concentration ) pre-incubated for 5’ – 60’ in the presence ( 1 µM ) ( □ ) or absence of eEF2·GTP ( ○ ) . In both cases , eEF2 final concentration after mixing was adjusted to 1 µM , by including 1 µM or 2 µM eEF2·GTP , respectively , in the TC solution . After quenching with 0 . 5 M MES buffer ( pH 6 . 0 ) , ribosome bound Phe-tRNAPhe was measured by cosedimentation . In the preincubation experiment , three-fold increases of both eEF2·GTP and Phe-TC concentrations , or of just eEF2·GTP concentration , had little effect on the cosedimentation results . Results in this Figure are corrected for IRES-independent changes in fluorescence anisotropy or Phe-tRNAPhe cosedimentation ( Figure 3—figure supplements 1 , 2 ) . All three solid green lines are best fits of the results obtained to the scheme in Figure 2 , using the numerical integration program Scientist . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 00610 . 7554/eLife . 13429 . 007Figure 3—figure supplement 1 . Corrected IRES-dependent time courses for initial Phe-tRNAPhe binding as measured by fluorescence anisotropy . Experiments were carried out as described in Figure 3 , but in the presence or absence of added IRES . Fluorescence anisotropy changes were monitored after rapid mixing of Phe-tRNAPhe ( Prf ) ternary complex ( 0 . 1 µM final concentration ) with either 80S·FVKM-IRES complex ( 0 . 1 µM final concentration ) or just 80S ( 0 . 1 µM final concentration ) . These experiments were carried out either in the absence of eEF2 ( 80S·FVKM-IRES complex , dotted blue line; 80S , dotted purple line ) or with 0 . 5–2 . 5 hr preincubation with eEF2 [80S·FVKM-IRES complex , dotted red line , 1 μM eEF2; dotted black line , 3 μM eEF2; 80S , 1 μM eEF2 , dotted green line ) . In the latter cases , eEF2 concentration was kept constant by including 1 or 3 µM eEF2 in the TC solution . Subtraction of the results obtained with 80S alone from the results obtained with 80S·FVKM-IRES complex yields the corrected time courses for IRES-dependent fluorescence anisotropy change with eEF2 preincubation ( solid red ( 1 μM ) and black ( 3 μM ) lines ) or in the absence of eEF2 ( solid blue line ) . These solid lines are presented in Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 00710 . 7554/eLife . 13429 . 008Figure 3—figure supplement 2 . Corrected IRES-dependent time courses for initial Phe-tRNAPhe binding as measured by Phe-tRNAPhe cosedimentation . Phe-TC ( 1 . 6 µM final concentration ) was rapidly mixed with either 80S·FVKM-IRES complex ( 0 . 8 µM final concentration ) or just 80S ( 0 . 8 µM final concentration ) . These experiments were carried out either in the absence of preincubation with eEF2 or with 5–60 min preincubation with 1 µM eEF2 . In both cases , eEF2 final concentration after mixing was adjusted to 1 µM by adding the appropriate amounts to the TC solution . After quenching with 0 . 5 M MES buffer ( pH 6 . 0 ) , ribosome bound Phe-tRNAPhe was measured by cosedimentation . As preincubation with eEF2 gave no significant difference on either IRES-dependent or IRES-independent Phe-tRNAPhe cosedimentation , the results with and without eEF2 preincubation were averaged . Subtraction of the averaged results obtained with 80S alone ( l ) from the averaged results obtained with 80S·FVKM-IRES complex ( O ) yields the corrected time course for IRES-dependent Phe-tRNAPhe cosedimentation ( Δ ) . The corrected results and solid line which is a best fit of the results obtained to the scheme in Figure 2 , using the numerical integration program Scientist , are presented in Figure 3 . The final corrected stoichiometry was 0 . 29 Phe/40S . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 00810 . 7554/eLife . 13429 . 009Table 1 . Apparent rate constants for Steps 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 009Apparent rate constants ( s-1 ) -eEF2+eEF2k10 . 0071 ± 0 . 00330 . 0033 ± 0 . 0001k-10 . 15 ± 0 . 040 . 0034 ± 0 . 0001k2 ( [Phe-TC] = 0 . 1 µM ) 0 . 11 ± 0 . 040 . 0256 ± 0 . 0002 Formation of Structure 4 from Structure 1 , as measured by the co-sedimentation assay , requires the presence of eEF2·GTP and proceeds at a considerably slower rate than formation of Structure 3 from Structure 1 ( Figure 3 ) , allowing estimation of a t1/2 for Step 3 , a second pseudo-translocation step involving conversion of 3 to 4 , of 210 ± 10 s . It is this further slow step that accounts for the lack of significant effect of preincubation with eEF2·GTP ( 5’ or 60’ ) on the rate of formation of 4 from 1 ( Figure 3 ) . Using ribosomes programmed with the appropriate coding sequence mutants ( Supplementary file 1 ) and [35S]-Met-TC , we employ a rapid mixing and quench assay to measure rates of PheMet , PheLysMet , and PheLysValMet synthesis , with detection and quantification of product by thin layer electrophoresis ( TLE ) ( Figure 4A and Figure 4—figure supplement 1 ) . For PheMet synthesis ( Figure 4B ) we preform Structure 4 and measure its conversion to Structure 6 . We measurePheLysMet synthesis , Structure 9 , starting from either Structure 4 or Structure 7 ( Figure 4C ) and PheValLysMet synthesis , Structure 12 , starting from either Structure 7 or Structure 10 ( Figure 4D ) . In all three cases , reactions involving only TC binding and a single peptide bond formation ( 4 to 6; 7 to 9; 10 to 12 ) proceed in remarkably similar fashion , each showing biphasic behavior with a rapid phase accounting for 65 ± 10% of reaction proceeding with a t1/2 of ~6–9 s and a slower , minor phase proceeding much more slowly ( t1/2 ~220–240 s ) , possibly corresponding to defective ribosomes . Reactions involving formation of two peptide bonds , as in the conversion of 4 to 9 or 7 to 12 are well approximated as single phase reactions with t1/2 values of 90–110 s . Conversion of 4 to 9 proceeds via Steps 4 – 8 , allowing the t1/2 value for the translocation Step 6 to be estimated as 84 s , from the difference between the t1/2 value for the 4 to 9 reaction and the sum of the t1/2 values for the 4 to 6 and 7 to 9 reactions ( major phases ) . Similarly , the t1/2 value for the translocation Step 9 can be estimated as 110 s from the difference between the t1/2 value for the 7 to 12 reaction and the sum of the t1/2 values for the 7 to 9 and 10 to 12 reactions . Since the di- , tri- and tetrapeptides synthesized in the results reported in Figure 4 use different coding sequence mutants , these estimates of translocation t1/2 values depend on the not unreasonable assumption that the identities of the tRNAs undergoing translocation do not have a major influence on the translocation rate . With this caveat , the results presented in Figure 4 lead to the clear conclusion that translocation is the rate limiting step in each of the first two cycles of polypeptide elongation , proceeding from 4 to 10 . 10 . 7554/eLife . 13429 . 010Figure 4 . Kinetics of peptide synthesis and Met-tRNAMet cosedimenting with ribosomes . Reaction mixtures were quenched at various times after mixing . Peptide synthesis aliquots were quenched with 0 . 8 M KOH , and the released [35S]-containing peptide was resolved and quantified by TLE and autoradiography ( Materials and methods ) . Cosedimentation assay aliquots were quenched with with 0 . 5 M MES buffer ( pH 6 . 0 ) and [35S] cosedimenting with ribosomes was determined . For all the reactions shown , final concentrations of reactants after mixing were: 80S·IRES complexes ( 0 . 8 μM ) ; all added TCs ( 1 . 6 µM ) ; eEF2·GTP ( 1 µM ) . The numbers in blue in parts ( B–D ) refer to the Structures in Figure 2 whose rates of conversion are measured . For example , the peptide synthesis result in part ( B ) labeled 4 – 6 measures conversion of Structure 4 to Structure 6 . ( A ) Time course for formation of PheValLysMet tetrapeptide as determined by TLE . 80S·FVKM-IRES complex was mixed with Phe-TC , Val-TC , Lys-TC and [35S]-Met-TC . The migration positions of [35S]-Met and [35S]-PheValLysMet ( * ) are indicated . ( B ) 80S·FM-IRES complexes with Phe-tRNAPhe at the P site were mixed with [35S]-Met-TC . Dipeptide synthesis ( □ ) ; cosedimentation assay ( ■ ) . ( C ) Tripeptide synthesis: 80S·FKM-IRES complexes with either Phe-tRNAPhe ( O ) in the P site ( Structure 4 ) or PheLys-tRNALys ( Δ ) in the P site ( Structure 7 ) were mixed with either Lys-TC and [35S]-Met-TC or with just [35S]-Met-TC , respectively . Cosedimentation assay: 80S·FKM-IRES complex with PheLys-tRNALys in the P site was mixed with [35S]-Met-TC ( ■ ) . ( D ) Tetrapeptide synthesis: 80S·FVKM-IRES complexes with either PheVal-tRNAVal ( O ) in the P site ( Structure 7 ) or PheValLys-tRNALys ( Δ ) in the P site ( Structure 10 ) were mixed with either Lys-TC and [35S]-Met-TC or with just [35S]-Met-TC , respectively . Cosedimentation assay: 80S·FKM-IRES complex with PheValLys-tRNALys in the P site was mixed with [35S]-Met-TC ( ■ ) . Solid lines are best fits using single ( B , 4–7; C , 4–9; D , 7–12 ) or double ( B , 4–6; C , 7–8 , 7–9; D , 10–11 , 10–12 ) exponentials . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 01010 . 7554/eLife . 13429 . 011Figure 4—figure supplement 1 . Time courses for formation of PheMet dipeptide and PheLysMet tripeptide as determined by TLE . ( A ) Dipeptide synthesis: 80S·FM-IRES complex with Phe-tRNAPhe in the P site was mixed with [35S]-Met-TC . ( B ) Tripeptide synthesis: 80S·FKM-IRES complex with Phe-Lys-tRNALys ( Δ ) in the P site was mixed with [35S]-Met-TC . The migration positions of [35S]-Met and [35S]-labeled peptides ( * ) are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 01110 . 7554/eLife . 13429 . 012Figure 4—figure supplement 2 . Added 30S carrier does not significantly change the amount of FVKM-tRNAMet co-sedimenting with 80S ribosomes in the presence and absence of FVKM-IRES . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 01210 . 7554/eLife . 13429 . 013Table 2 . t1/2 values* . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 013Step ( s ) t1/2 ( s ) 1† 1 ( +eEF2 ) †230 ± 5 237 ± 52‡ 2 ( +eEF2 ) ‡15 ± 9 30 ± 53210 ± 104 + 58 ± 24-898 ± 156 = ( 4-8 ) – ( 4+5 ) – ( 7+8 ) §84 ± 1673 ± 18 = ( 7+8 ) – 7§4 ± 27 + 86 ± 27-11128 ± 269 = ( 7-11 ) – ( 7+8 ) – ( 10+11 ) §110 ± 30102 ± 111 = ( 10 + 11 ) – 10§7 ± 310 + 119 ± 212<10* Error ranges shown are based on the variances of fits to single or double exponentials of the results presented in Figure 4 , unless otherwise noted . † Calculated as 0 . 69 ( k-1 + k2 ) /k1k2 ( see Table 1 ) . ‡ Calculated as 0 . 69 ( k-1 + k2 ) /k22 ( see Table 1 ) . § Error ranges for these steps , which are not observed directly , are based on the error ranges of the directly observed steps . In an attempt to resolve the TC binding step ( reactions 4 , 7 , and 10 ) from the peptide formation step ( reactions 5 , 8 , and 11 ) we also employed a rapid mixing and quench assay to determine the rates with which [35S]-Met-tRNAMet is able to cosediment with the ribosome following mixing of [35S]-Met-TC with structures 4 , 7 , or 10 . This strategy was successful for [35S]-Met–TC reaction with structure 7 ( containing P-site bound PheLys-tRNALys , Figure 4C ) or structure 10 ( containing P-site bound PheValLys-tRNALys Figure 4D ) , in which the [35S]-Met-TC cosedimentation rates outpace the rates of peptide bond formation with Met-TC . These rate differentials permit estimates to be made for the t1/2 values of TC binding ( Step 7 , 3 s; Step 10 , 2 s ) and peptide bond formation ( Step 8 , 4 s; Step 11 , 7 s ) . They also provide a clear indication that , within Structures 8 , 9 , 11 and 12 , Met-tRNAMet , PheLysMet-tRNAMet , and PheValLysMet- tRNAMet , whenbound to the A-site , efficiently cosediment with ribosomes , which is typical for A-site bound tRNAs in conventional ( non-IRES ) elongation complexes ( Warner and Rich , 1964; Nwagwu , 1975 ) . However , for [35S]-Met–TC reaction with structure 4 ( containing P-site bound Phe-tRNAPhe ) , the [35S]-Met-TC cosedimentation rate is much slower than the dipeptide formation rate ( Figure 4B ) . This indicates that PheMet-tRNAMet , and possibly Met-tRNAMet as well , are not bound stably to the ribosome in Structures 5 and 6 , and that only PheMet-tRNAMet bound to the P-site ( Structure 7 ) is fully recovered by cosedimentation . As a result , the cosedimentation assay does not permit estimation of the t1/2 values for Steps 4 and 5 . It is possible that the lability of the A-site tRNAs in structures 5 and 6 is due to IRES binding to the E-site , which is absent in structures 8 , 9 and 11 , 12 , and may reflect an allosteric A-site: E-site interaction . Evidence for allosteric A-site/E-site interactions has been presented for both bacterial and eukaryotic ribosomes ( Nierhaus 1990; Chen et al . , 2011; Ferguson et al . , 2015 ) , although the general validity of this interaction has been questioned ( Semenkov et al . , 1996; Petropoulos and Green , 2012 ) . The results presented in Figure 4 show that translocation proceeds slowly through the first two elongation cycles of nascent protein synthesis , raising the question of how far nascent protein synthesis has to proceed to overcome the retarding effect of ribosome-bound IRES . In Figure 5 we present the results of two experimental approaches demonstrating that translocation of tetrapeptidyl-tRNA proceeds much more rapidly than translocation of tripeptidyl-tRNA . 10 . 7554/eLife . 13429 . 014Figure 5 . Tetrapeptide translocation ( Step 12 ) is faster than tripeptide translocation ( Step 9 ) . ( A ) Puromycin reaction with PheValLys-tRNALys bound either at the A site ( D ) or at the P-site ( O ) of the 80S·FVKM-IRES complex or being translocated from the A site to the P site ( □ ) . ( B ) Puromycin reaction with PheValLysMet-tRNAMet either bound at the P-site ( O ) of the 80S·FVKM-IRES complex or being translocated from the A site to the P site ( □ ) . Lines in A . and B . Are fits to single exponentials . ( C ) Time dependence of PheLysValArgGlnTrpLeuMet octapeptide synthesis from the 80S·FKVRQWLM-IRES complex containing various peptidyl-tRNAs pre-bound at the P site , as indicated . The pre-bound peptidyl tRNAs were prepared using the standard procedure ( see Complex Preparations in Materials and methods ) by incubating the 80S-IRES complex with the relevant TCs for 15 min . The remaining TCs needed for octapeptide synthesis , including [35S]-Met-TC , were then added , each at a concentration of 1 . 6 µM , for the indicated times prior to quenching . PheLysValArgGlnTrpLeuMet octapeptide synthesis was measured by [35S]-Met cosedimenting with 80S ribosome . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 01410 . 7554/eLife . 13429 . 015Figure 5—figure supplement 1 . Octapeptide synthesis: 80S·FKVRQWLM-IRES complex with FKVRQWLM-tRNAMet in the P-site was prepared using the standard procedure ( see Complex Preparations in Materials and methods ) and incubating the 80S-IRES complex with the eight relevant TCs ( including [35S]-Met-TC ) for 40 min . The resulting labeled octapeptide , released by base hydrolysis , was analyzed by TLE . Migration positions of [35S]-Met and [35S]-labeled FKVRQWLM ( * ) are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 13429 . 015 The first approach makes use of the fact that formation of peptidyl-puromycin proceeds more rapidly with peptidyl-tRNA bound to the P-site than to the A-site , permitting puromycin reactivity to distinguish A-site from P-site peptidyl-tRNA . As shown in Figure 5A , puromycin ( 1 mM ) reacts with A-site bound PheValLys-tRNALys , Structure 9 , about 20times more slowly ( t1/2 1400 ± 300 s ) than it reacts with P-site bound PheValLys-tRNALys ( t1/2 76 ± 16 s ) . The corresponding t1/2 value for puromycin reaction with PheValLys-tRNALys undergoing translocation from the A- to P-site is 170 ± 30 s . This increase of approximately 100 s for translocating PheValLys-tRNALysvs . translocated PheValLys-tRNALys closely matches the t1/2 value of 110 ± 30 s estimated above for the translocation of tripeptidyl-tRNA ( Table 2 ) and can be assigned to the translocating step . In contrast , the rates of puromycin reaction with translocating and translocated PheValLysMet-tRNAMet ( Structure 13 ) are indistinguishable from one another ( t1/2 values of 37 ± 4 s and 46 ± 7 s , respectively , Figure 5B ) , a clear demonstration that translocation of PheValLysMet-tRNAMet proceeds rapidly with respect to puromycin reaction . Our results allow us to estimate an upper limit value of t1/2 for the translocation Step 12 of ≤10 s . Puromycin reacts at similar rates with translocated PheValLys-tRNALys ( Structure 10 , t1/2 76 ± 16 s ) and PheValLysMet-tRNAMet ( Structure 13 , t1/2 46 ± 7 s ) . These rates , while consistent with those reported by others for puromycin reaction with eukaryotic P-site bound Met-tRNAMet ( Lorsch and Herschlag , 1999 ) , N-AcPhe-tRNAPhe ( Ioannou et al . , 1997 ) , and Cy3-Met-tRNAMet ( Ferguson et al . , 2015 ) , are several hundred-fold slower than those measured for puromycin reaction with prokaryotic P-site bound peptidyl- or fMet-tRNA . This largely explains why the rate reduction for puromycin reaction with A-site vs . P-site bound peptidyl-tRNA is so much more modest for eukaryotic ribosomes ( ~20-fold , Figure 5A ) than for prokaryotic ribosomes ( 103–104-fold , Pan et al . , 2007; Semenkov et al . , 1992 ; Sharma et al . , 2004; Peske et al . , 2004 ) . Above we have demonstrated that , under our conditions , aa-tRNA binding and peptide bond formation proceed with an overall t1/2of 6 – 9 s for each of the three elongation steps we have studied . This relative constancy , coupled with the much slower translocation of tripeptidyl-tRNA ( Step 9 ) vs . tetrapeptidyl-tRNA ( Step 12 ) , leads to the prediction that synthesis of a longer peptide that required the tripeptidyl-tRNA translocation step ( Step 9 ) would proceed significantly more slowly than synthesis not requiring this step . In the second approach we verified this prediction by demonstrating that octapeptide FKVRQWLM formation , as measured by the cosedimentation assay , is much slower when synthesis is initiated with P-site bound PheLys-tRNALys ( Structure 7 ) vs . P-site bound PheLysVal-tRNAVal ( Structure 10 ) ( Figure 5C ) . Indeed , the rates of FKVRQWLM synthesis are only marginally increased when reaction is initiated with P-site bound tetrapeptidyl-tRNA or pentapeptidyl-tRNA as compared with tripeptidyl-tRNA , reinforcing the notion that the retarding effect of ribosome-bound IRES on protein synthesis is largely overcome following translocation of tripeptidyl-tRNA .
The results presented in this paper constitute the first time that rates of reaction have been determined for the initial cycles of IRES-dependent elongation . They demonstrate quite clearly that the first two cycles of elongation proceed much more slowly than subsequent steps , and that these reduced rates arise from slow , rate-determining , pseudo-translocation and translocation steps . Translocation during the first elongation cycle ( Step 6 ) clearly requires displacement of the IRES from the E-site , so it is not unexpected that it would be slow . Less predictable is the slow translocation in the second elongation cycle , ( Step 9 ) after the IRES structure has , presumably , already left the E-site ( Figure 1 ) . The slow rate of Step 9 might be due to a full dissociation of IRES from the ribosome during this step , a suggestion that could be tested by appropriately designed structural studies . In any case , our results do clearly demonstrate that , following translocation of tripeptidyl-tRNA from the A- to P-site , the pace of nascent peptide chain elongation picks up dramatically . Further work , comparing quantitatively the rates of successive cycles of nascent peptide elongation following tetrapeptide formation ( i . e , cycles 4 , 5 , 6 , 7 , etc . ) will be required to determine how many cycles are required before any retarding influence of bound CrPV-IRES is completely eliminated . Our results also clarify an aspect of the initial binding of the first aa-tRNA to the 80S·CrPV-IRES complex . Prior results have shown that initial aa-tRNA binding , in the form of a ternary complex , to an 80S·IRES complex , as measured either by cosedimentation ( Fernandez et al . , 2014 ) , or by filter binding and toeprinting ( Yamamoto et al . , 2007 ) , requires eEF2·GTP , leading to the conclusion that initial aa-tRNA binding can only bind to the 80S·IRES complex after an eEF2-dependent translocation event ( Fernandez et al . , 2014 ) . While we agree with the experimental results , and have in fact reproduced the cosedimentation result in our own work , we disagree with the conclusion . This is because these earlier experiments only measured stable aa-tRNA binding , corresponding to formation of Structure 4 in which aa-tRNA binds to the P-site . However , it is clear from the anisotropy experiment conducted in the absence of added eEF2·GTP ( Figure 3 , blue trace ) that ternary complex binding measured in situ , which can monitor labile binding to the A-site ( Structure 3 ) does not require eEF2·GTP . This is easily understood as an example of Le Chaltelier’s principle , in which the equilibrium between Structure 1 ( closed A- site ) and Structure 2 ( open A- site ) , which strongly favors Structure 1 , is pulled to the right by aa-tRNA binding . Preincubation with eEF2 also shifts the equilibrium to the right , leading to an initial rapid phase of reaction with Phe-TC ( Figure 3 ) . This latter shift , for which the results presented in Figure 3 provide strong inferential evidence , appears to be at odds with earlier toeprinting results showing no shift in IRES position within the 80S·CrPV-IRES complex on addition of eEF2 alone ( Pestova et al . , 2003; Jan et al . , 2003 ) . In agreement with the suggestion of Muhs et al . ( 2015 ) , we believe it likely that this apparent inconsistency arises from eEF2 dissociation from the ribosome during the toeprinting assay ( Pestova et al . , 1996 ) , with the consequent favoring of Structure 1 . This is because GTP is required for tight binding of eEF2 to the ribosome ( Nygård and Nilsson , 1984 ) , but the toeprinting assay is carried out for an extended period of time ( 45 min ) under non-denaturing conditions in the absence of added GTP , conditions that would eventually deplete GTP due to ribosome-dependent eEF2·GTP hydrolysis ( Nygård and Nilsson , 1989 ) . In addition , the toeprinting assay is performed at a Mg2+ concentration of 10 . 5 mM , considerably higher than the 5 mM used in our kinetic studies , which could also affect the 1 to 2 equilibrium position . How relevant are the present results for in vivo initiation of IRES-dependent protein synthesis ? We note three potential concerns . First , our in vitro systemis quite heterogeneous , with ribosomes derived from shrimp cysts , yeast elongation factors , and yeast and E . coli charged tRNAs . However , as reviewed in Koh et al . ( 2014 ) , IRESs can initiate translation on ribosomes from many eukaryotic organisms , including shrimp ( Cevallos and Sarnow , 2005 ) , indicating that the molecular mechanism is not species-specific . CrPV IRESs in particular can initiate translation on ribosomes from yeast ( Thompson et al . , 2001 ) to human ( Spahn et al . , 2004 ) . Furthermore , eukaryotic elongation factors have structures that are very strongly conserved ( Soares et al . , 2009; Jørgensen et al . , 2006 ) , and there is strong evidence that charged tRNAs from one species form functional complexes with both eEF1A and ribosomes from a different species ( Jackson et al . , 2001; Ferguson et al . , 2015 ) . Second , the coding sequences employed in this work are different from that immediately downstream of wt-CrPV-IRES ( Supplementary file 1 ) . This is also unlikely to pose a major difficulty , given the strong evidence that mutations in the downstream sequence are , in general , tolerated without substantial effect on initiation of translation ( Tsukiyama-Kohara et al . , 1992; Wang et al . , 1993; Hellen and Sarnow , 2001; Rijnbrand et al . , 2001 ) , although mutations of some downstream sequences do give rise to relatively minor changes in IRES activity ( Kim et al . , 2003; Shibuya et al . , 2003; Wang et al . , 2013 ) . Third , the elongation rate of even the later cycles of IRES-dependent elongation ( Figure 5C ) is quite slow ( ~0 . 1 s-1 ) . Although this rate is essentially identical to that reported for tripeptide synthesis in a cap-dependent yeast-based in vitro translation system which requires five initiation factors and eEF3 in addition to eEF1A and eEF2 ( Acker et al . , 2007; Eyler and Green , 2011; Gutierrez et al . , 2013 ) , it is 1 . 5–2 orders of magnitude slowerthan rates of peptide elongation that have been estimated for intact eukaryotic cells at 37°C ( 3–10 s-1 ) ( Boehlke and Friesen , 1975; Hershey , 1991 ) . There is evidence that , in many eukaryotic cells , the protein synthesis machinery is highly organized , containing several components , including ribosomes , a multi-aminoacyl-tRNA synthetase complex , eEF-1A , and several auxiliary proteins ( Negrutskii et al . , 1994; Negrutskii and El’skaya , 1998; David et al . , 2011 ) . It has been suggested that this organized structure optimizes translation rate by coordinating synthetase activities to facilitate channeling of aa-tRNAs to the elongating ribosomes . Thus , protein synthesis in a permeabilized mammalian cell , in which this structure is likely to be preserved , proceeds 40-fold faster than what is obtained in a cell-free system prepared from the same cells which presumably lacks this structure ( Negrutskii et al . , 1994 ) . The slow rates measured for both the IRES-dependent and cap-dependent in vitro systems could be due , at least in part , to their lack of aa-tRNA channeling . Such channeling would be unlikely to accelerate the very slow translocation rates in the initial peptide elongation cycles reported in this work , although we cannot exclude the possibility that other proteins present in vivo might have such effects . Future efforts will address this issue . Here , incorporation of some of the features of a recently introduced in vitro protein synthesissystem in which initiation is carried out using the IRES from hepatitis C virus could be useful ( Machida et al . , 2014 ) . Detailed mechanistic characterization of many aspects of eukaryotic polypeptide elongation has been held back by the lack of a convenient system for its study . The very simple in vitro IRES-dependent elongation system described here should be useful in overcoming this limitation . As one example , it is generally assumed , based on extensive structural similarities ( Jørgensen et al . , 2006 ) , that eEF2 functions in catalyzing eukaryotic elongation in much the same way that EF-G catalyzes prokaryotic elongation , but this assumption does not take into account some important structural differences , including the fact that eEF2 is subject to post-translational modifications not found in EF-G , with clear consequences for activity but , as yet , little understanding of mechanism ( Dever and Green , 2012; Mittal et al . , 2013 ; Greganova et al . , 2011; Liu et al . , 2012 ) . The CrPV-IRES based system should permit detailed rate and structural dynamic studies of eEF2 catalytic function , of the kind that have proved so useful in elucidating EF-G function in bacterial protein synthesis ( Pan et al . , 2007; Chen et al . , 2011; Chen et al . , 2013; Holtkamp et al . , 2014; Salsi et al . , 2015 ) .
The wt CrPV Phe-IRES vector , as well as several variants in which the first Ala codon is replaced by a Phe codon , were the kind gifts of Dr . Eric Jan . This replacement , which has little effect on the initiation of translation ( see Discussion and Figure 1—figure supplement 1 ) , was made as a matter of convenience , since tRNAPhe was available to us and the appropriate tRNAAla acceptor was not . The vectors encoding the PheMet , PheValMet , PheValLysMet , and PheLysValArgGlnTrpLeuMet were generated by PCR insertion of corresponding sequences ( Supplementary file 1 ) into the CrPV Phe-IRES vector . All cloned sequences were verified by standard sequencing methods using appropriate primers . For in vitro transcription of full-length mRNA for the Luciferase assay , the WT and mutated Phe-IRES plasmids were linearized with XbaI , which cleaves the plasmids after the firefly luciferase coding region . mRNA was transcribed in vitro using the AmpliScribe T7 transcription kit ( EPICENTRE ) according to the manufacturer . For in vitro transcription of short-length mRNAs , the mutated IRES plasmids were linearized with NarI , which cleaves 33 nt downstream of the ATG start codon of the luciferase coding region . In vitro translation of firefly luciferase with WT and mutated F-IRES mRNA ( 1 μg in 50 μL of reaction mixture ) was performed using the Flexi Rabbit Reticulocyte Lysate System ( Promega ) according to the manufacturer . IRES mRNA was omitted in the control reaction . Fluc activities ( Figure 1—figure supplement 1 ) were determined using a plate reader ( Envision 2103 , Perkin-Elmer ) to detect the luminescence signal . Shrimp ( A . salina ) 80S ribosomes were prepared from dried , frozen cysts as previously described ( Iwasaki and Kaziro , 1979 ) with some modifications . After the shrimp cysts were ground open , debris was removed by centrifugation at 30 , 000xg for 15 min and crude 80S ribosomes were precipitated from the supernatant by addition of 4 . 5% ( w/v ) PEG 20K ( Ben-Shem et al . , 2011 ) . 40S and 60S subunits were resolved on 10–30% sucrose gradients after puromycin treatment . E . coli 30S subunits were prepared as described ( Grigoriadou et al . , 2007 ) . eEF1A was purified from yeast according to published methods ( Thiele et al . , 1985 ) . His6-eEF2 was isolated from an overexpressing yeast strain ( TKY675 ) generously provided by Dr . Terri Kinzy , and purified as described ( Jørgensen et al . , 2002 ) . Proflavin-labeled Phe-tRNAPhe , denoted Phe-tRNAPhe ( prf ) , was prepared as previously described ( Wintermeyer and Zachau 1974 , Betteridge et al . , 2007 ) . Yeast tRNAPhe was purchased from Sigma . Other isoacceptor tRNAs were prepared from bulk tRNA ( Roche ) from either E . coli ( tRNAGln , tRNALys , tRNAMet ) or yeast ( tRNAArg , tRNALeu , tRNATrp , tRNAVal ) by hybridization to immobilized complementary oligoDNAs , as described ( Barhoom et al . , 2013 ; Liu et al . , 2014 ) . E . coli and yeast tRNAs were charged with their cognate amino acids as described ( Pan et al . , 2006 , 2009 ) . All complexes were prepared in buffer 4 ( 40 mM Tris-HCl pH 7 . 5 , 80 mM NH4Cl , 5 mM MgOAc2 , 100 mM KOAc , 3 mM 2-mercaptoethanol ) at 37°C . For the preparation of ternary complexes ( TC , aa-tRNA·eEF1A·GTP ) and 80S·IRES complexes containing either Phe-tRNAPhe or peptidyl-tRNA bound in the P-site , buffer 4 was supplemented with 1 mM GTP and 1 mM ATP . All TC complexes were prepared by incubating the relevant charged tRNA ( 1 . 6 μM , based on amino acid stoichiometry ) with eEF1A ( 8 μM ) for 5 min . 80S·IRES complexes were formed by incubation of shrimp 40S ( 0 . 8 µM ) and 60S ( 1 . 6 µM ) subunits with the appropriate IRES ( 2 . 4 µM ) for 5 min . 80S·IRES complexes containing Phe-tRNAPhe or peptidyl-tRNA bound in the P-site were formed by mixing 80S·IRES complexes ( 0 . 8 µM ) with 1 μM eEF2 and the appropriate TCs ( 1 . 6 µM for each ) for 15–40 min . To determine radioactively labeled aa-tRNA binding stoichiometries , 40 µL samples were subjected to ultracentrifugation at 4°C ( 540 , 000xg ) for 40 min through a 1 . 1 M sucrose cushion . Excess bacterial 30S bacterial ribosome subunits ( 600 pmol/15 ± 5 µL ) were added as carrier to enhance pelleting and allow facile calculation of complex recovery . Control experiments carried out in the absence of IRES or of both IRES and 80S ribosomes demonstrated that only negligible amounts of labeled peptidyl-tRNA cosedimented due to binding to 30S subunits ( Figure 4—figure supplement 2 ) . The pellets were gently washed twice with buffer 4 and dissolved in 100 μL of buffer 4 for A260 determination . Ribosome recoveries typically varied between 60 and 80% . Unless otherwise noted , all reactions were performed at 37°C in buffer 4 supplemented with 1 mM GTP . All kinetic results reported are the averages of 2–4 independent determinations , performed on different days . No systematic effort was made to carry out duplicate experiments using independently made stock reagent solutions , although this was sometimes done . Error bars in figures are shown as average deviations . | Inside cells , machines called ribosomes make proteins using instructions carried by molecules of messenger RNA ( or mRNA ) . The ribosomes bind to the mRNA and then move along it to assemble the proteins in a process called translation . The first step of translation – when the ribosome binds to the mRNA – is known as initiation . In human and other eukaryotic cells , initiation mainly occurs through a mechanism that requires many proteins called initiation factors to recruit the ribosome to a cap structure formed at one end of the mRNA . When viruses infect cells , they hijack the ribosomes of the host cell to produce large quantities of viral proteins . However , unlike their host cells , many viruses use a different pathway to initiate translation of their mRNAs . The mRNAs of these viruses have regions known an internal ribosome entry sites ( IRESs ) that host cell ribosomes can bind to instead . After initiation , the ribosome progressively assembles the building blocks of proteins ( amino acids ) into a chain until the new protein is complete . Molecules called transfer RNAs bind to individual amino acids and bring them to the ribosome . Previous research has shown that , prior to initiation , IRESs on Cricket Paralysis Virus mRNAs bind to the ribosome and occupy sites where transfer RNAs would normally bind . However , it was not clear how this affects the elongation process . Zhang et al . now address this question using a cell-free system that allowed them to recreate and observe translation outside of the normal cell environment . Zhang et al . found that the binding of an IRES to a ribosome slows down the early steps of elongation . A likely explanation for this is that the IRES elements have to be displaced from the ribosome before the incoming transfer RNAs can occupy the three tRNA sites . However , as elongation progresses , the effects of the IRES elements are overcome and the pace of elongation increases significantly . Zhang et al . ’s findings provide a convenient approach that could be used for future studies of elongation . This approach could also help researchers find out how abnormalities in translation contribute to human diseases , including muscle-wasting disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology"
] | 2016 | Kinetics of initiating polypeptide elongation in an IRES-dependent system |
Constrained , cyclic peptides encoded by plant genes represent a new generation of drug leads . Evolution has repeatedly recruited the Cys-protease asparaginyl endopeptidase ( AEP ) to perform their head-to-tail ligation . These macrocyclization reactions use the substrates amino terminus instead of water to deacylate , so a peptide bond is formed . How solvent-exposed plant AEPs macrocyclize is poorly understood . Here we present the crystal structure of an active plant AEP from the common sunflower , Helianthus annuus . The active site contained electron density for a tetrahedral intermediate with partial occupancy that predicted a binding mode for peptide macrocyclization . By substituting catalytic residues we could alter the ratio of cyclic to acyclic products . Moreover , we showed AEPs from other species lacking cyclic peptides can perform macrocyclization under favorable pH conditions . This structural characterization of AEP presents a logical framework for engineering superior enzymes that generate macrocyclic peptide drug leads .
Asparaginyl endopeptidases ( AEPs ) are a group of asparagine/aspartic acid ( Asx ) specific proteases that have been classified as belonging to the C13 family of cysteine proteases based on the presence of a His-Gly-spacer-Ala-Cys motif ( Hara-Nishimura et al . , 1993; Chen et al . , 1997; Mathieu et al . , 2002; Shafee et al . , 2015 ) . First described in plants as vacuolar processing enzymes based on their propensity for processing seed proteins stored in vacuoles , AEPs have since been described in a variety of organisms and shown to be involved in a wide range of processes including , cell death antigen processing and hemoglobin degradation ( Hara-Nishimura et al . , 1993; Manoury et al . , 1998; Hatsugai et al . , 2004; Kuroyanagi et al . , 2005; Yamada et al . , 2005; Sojka et al . , 2007 ) . In addition to the proteolytic function observed in these processes , AEP has become well known for its curious ligation reactions ( Min and Jones , 1994; Sheldon et al . , 1996; Mylne et al . , 2012; Nguyen et al . , 2014; Zhao et al . , 2014; Dall et al . , 2015 ) . The ability of endoproteases to perform ligation reactions was first observed by Bergmann and Fruton in 1938 with chymotrypsin ( Bergmann and Fruton , 1938 ) . Later , in vitro ligation reactions were performed with AEP from jack bean ( Canavalia ensiformis ) seeds ( Bowles et al . , 1986; Min and Jones , 1994 ) . The recent discovery that evolutionarily distinct plant families have repeatedly recruited AEPs to catalyze the formation of ribosomally synthesized and post-translationally modified peptides ( RiPPs ) , through the macrocyclization of linear precursor sequences , has caught the attention of drug designers keen to overcome the current inefficiencies in native chemical ligation that limit the therapeutic use of cyclic peptides ( Pattabiraman and Bode , 2011; Mylne et al . , 2012; Arnison et al . , 2013 ) . Such therapeutic cyclic peptides are viewed by many to have the potential to capitalize on a niche in the current pharmaceutical market by virtue of their intermediate size between small molecule drugs and large protein structures , and their unique capacity to combine favorable bioavailability and stability characteristics with high target specificity facilitated by tolerance to site-directed mutagenesis ( Clark et al . , 2005 , 2010; Gould et al . , 2011; Ji et al . , 2013; Poth et al . , 2013; Truman , 2016 ) . Moreover , as computational techniques for the discovery of RiPPs improve and the number of cyclic peptides described continues to expand , an ever-wider array of scaffolds might be exploited to tailor molecules to specific drug targets ( Bhardwaj et al . , 2016; Truman , 2016; Hetrick and van der Donk , 2017 ) ( Figure 1 ) . Sunflower trypsin inhibitor-1 ( SFTI-1 ) is a 14-residue , bicyclic peptide with a cyclic backbone and an internal disulfide bond ( Luckett et al . , 1999 ) . Its biosynthesis is rather unusual as its sequence is buried within a precursor that also encodes seed storage albumin . Seed storage albumins are a major class of seed storage protein that constitute over 50% of total seed protein and become a source of nitrogen and sulfur during seed germination ( Youle and Huang , 1978; Shewry and Halford , 2002 ) . The common sunflower ( Helianthus annuus ) has many genes encoding precursors for these napin-type or 2S seed storage albumins that are synthesized in the rough endoplasmic reticulum before undergoing cleavage maturation by AEP and localizing to storage vacuoles ( Bollini and Chrispeels , 1979; Franke et al . , 2016; Jayasena et al . , 2016 ) . Along with an adjacent albumin , SFTI-1 is post-translationally processed by AEP from within a unique seed storage albumin precursor called Preproalbumin with SFTI-1 ( PawS1 ) ( Mylne et al . , 2011 ) . SFTI-1 is a potent inhibitor of serine proteases , and its intrinsic stability and cellular penetration capabilities have led to its application as a bioactive scaffold ( White and Craik , 2016; Swedberg et al . , 2017 ) . AEP-catalyzed macrocyclization of SFTI-1 is hypothesized to proceed via a cleavage-coupled intramolecular transpeptidation reaction whereby catalysis begins with the deprotonation of the active site Cys by a localized His of the catalytic center , facilitating the nucleophilic attack on the carbonyl carbon of the Asp by the activated Cys thiol of AEP ( Bernath-Levin et al . , 2015 ) . This attack culminates in the formation of a thioacyl intermediate between the substrate and AEP , and the removal of the C-terminus . The reaction is proposed to be subsequently concluded by the nucleophilic attack of this intermediate by a Gly at the N-terminus of the substrate , resulting in the macrocyclization of SFTI-1 in a head-to-tail manner ( Mylne et al . , 2011; Bernath-Levin et al . , 2015 ) . Notably , this reaction proceeds in competition with nucleophilic attack upon the thioacyl intermediate by any nearby water molecules which would produce hydrolyzed , acyclic-SFTI . In vitro studies revealed the ratio of acyclic to cyclic SFTI-1 to be in the order of 5 . 8:1 , with the less stable acyclic products hypothesized to be quickly degraded in vivo ( Bernath-Levin et al . , 2015 ) . Evidence for AEP-mediated and hydrolysis-independent transpeptidation was demonstrated through the exclusion of a heavy atom O18 in the cyclic SFTI-1 product from an in vitro jack bean AEP ( CeAEP1 ) catalyzed reaction ( Bernath-Levin et al . , 2015 ) . The formation of a much larger cyclic peptide by an AEP from Oldenlandia affinis ( OaAEP1 ) was also shown to lack O18 incorporation , suggesting a conserved mechanism of macrocyclization despite differences in substrate sequences ( Harris et al . , 2015 ) . However , the suggestion that an AEP from butterfly pea ( Clitoria ternatea ) termed butelase one functions only as a ligase ( Nguyen et al . , 2014 ) , combined with the proposal of a succinimide-driven , cleavage-independent ligation event based on the crystal structure of human AEP ( hAEP ) and its ability to ligate substrate in the absence of the catalytic Cys , has cast uncertainty on the mechanism of AEP-catalyzed macrocyclization ( Dall et al . , 2015 ) . In order to elucidate a structural explanation why plant AEPs have been recruited by distinct plant lineages to perform macrocyclization and to understand the catalytic and structural nuances that might allow preferences towards cleavage or ligation reactions , we sought the crystal structure of a sunflower AEP . Herein , we describe the first structure of an active plant AEP; one capable of performing peptide macrocyclization . This AEP , the most abundant AEP of five AEPs in the common sunflower ( HaAEP1 ) , displays structural similarity to previously published active AEPs from mammals , with subtle differences at residues involved in substrate recognition . Our characterization by site-directed mutagenesis of HaAEP1 residues integral to macrocyclization will facilitate the bioengineering of plant AEPs for improved macrocyclization efficiency , diversifying the scaffolds usable as cyclic therapeutic leads .
AEPs are synthesized as inactive precursors that have been shown to undergo irreversible auto-activation into their mature form on exposure to a low pH environment that resembles the acidic pH in the vesicles/vacuole where these proteins are active in vivo ( Dall and Brandstetter , 2013; Shafee et al . , 2015 ) . In order to obtain an active form of a plant AEP , a ~51 kDa pro-HaAEP1 ( residues 28–491 ) lacking an endoplasmic reticulum signal sequence and including a N-terminal His-tag was expressed in Escherichia coli and purified by nickel affinity chromatography before being activated at pH 4 . 0 overnight . The activated form of HaAEP1 was then further purified by size exclusion chromatography , enabling separation of the core domain from the ‘cap’ domain ( Zhao et al . , 2014 ) , and crystal trials were undertaken ( Figure 2—figure supplement 1 ) . SDS-PAGE analysis of the ~38 kDa core domain peak revealed the disassociation of the core domain from the cap domain but also showed several bands of HaAEP1 , suggesting the presence of several cleavage sites at the termini of the core domain , as seen previously for several AEPs ( Hara-Nishimura et al . , 1998; Nguyen et al . , 2014; Zhao et al . , 2014; Harris et al . , 2015 ) . Indeed sequence comparison reveals the conservation of several of these predicted cleavage sites but notably lacks a previously described C-terminal di-Asp motif ( Hiraiwa et al . , 1999 ) ( Figure 2—figure supplement 2 ) . This led us to hypothesize that Asp52 , Asn57 , Asn338 , Asp356 , and Asp358 might represent the dominant autocatalytic cleavage sites in HaAEP1 . Crystallization trials of the ~38 kDa activated HaAEP1 yielded diffraction quality crystals that diffracted to a resolution of 1 . 8 Å ( Table 1 ) . The crystal structure was solved by molecular replacement yielding a single monomer in the asymmetric unit , and revealed an active monomeric HaAEP1 ( residues 58–338 with weak electron density for Asn338 ) that forms a canonical C13 caspase structure , with a central six-stranded β-sheet region confined by five α-helices ( Hara-Nishimura et al . , 1993; Yamada et al . , 2005 ) ( Figure 2A , Figure 2—figure supplement 3 ) . The structure of HaAEP1 lacks the C-terminal cap domain and N-terminal His-Tag , displaying dimensions of approximately 44 Å x 42 Å x 39 Å . Sequence analysis of this core domain suggests that the aforementioned pro-domains are likely to have been auto-catalytically processed during maturation as the previously predicted Asn cleavage sites precede and follow the defined active structure . HaAEP1 displays structural similarity to hAEP and a recently published structure of inactive OaAEP1 ( PDB ID: 4N6O and 5H0I ) with an r . m . s . d . value of 1 . 0 and 0 . 7 Å over 262 and 267 carbon alpha residues , respectively ( Holm and Rosenström , 2010; Yang et al . , 2017 ) ( Figure 2B ) . Due to such topological conservation it is expected that subtle differences around the substrate binding pocket will define substrate specificity and catalytic efficiency . Indeed , comparison of these three structures reveals HaAEP1 exhibits a unique flexible extension , reflected by weak electron density , in the α5-β6 loop and differences between the residues that are local to the catalytic His and Cys and those that delineate the S3-S5 pockets ( following the protease nomenclature defined by Schechter and Berger where the cleavage site residue is termed P1 and residues prior to and following the cleavage site are labeled P5-P2 and P1′-P2′ , respectively , and where the corresponding binding sites on the protease are described as S5-S2′ ) ( Schechter and Berger , 1967 ) . Specifically , differences in the substrate pocket include residues YGT 249–251 in HaAEP1 ( hAEP: YAC 217–219 , OaAEP1: WCY 246–248 ) , a bulky Trp232 in hAEP versus Leu271 in HaAEP1 and Leu268 in OaAEP1 , and the presence of an additional proline prior to the βIV-βV polyproline loop which orients E257 away from the S4 region in HaAEP1 ( Figure 2B ) . In OaAEP1 , the C247 residue at the entrance to the S4 pocket was recently proposed to function as a ‘gate keeper for ligation’ with large bulky side chains inhibiting ligation ( Yang et al . , 2017 ) . Differences local to the active site include residues P181 , Q245 , N247 in the βI sheet and β5-βIV loop ( OaAEP1: A178 , T242 , S244 , hAEP: T151 , R213 , S215 ) and G185 , E189 , H191 in the βII-βIII region ( OaAEP1: G182 , K186 , Y188 , hAEP: V155 , N158 , D160 ) ( Figure 2B ) . A 6-residue insertion in the α5-β6 loop results in the disruption of a potential N-linked glycosylation site that was hypothesized to affect substrate binding upstream to P5 in hAEP ( Dall and Brandstetter , 2016 ) . Interestingly , none of the four conserved potential N-linked glycosylation sites in hAEP and mouse AEP ( mAEP ) are found in HaAEP1 ( Figure 2—figure supplement 2 ) . Given that these glycosylation sites have been predicted to protect AEP from non-specific protease activation it is intriguing to find that the only two potential N-linked glycosylation sites in HaAEP1 ( N138 and N143 ) are located on the opposite side of the protein to the activation peptide in mammalian AEPs , at the beginning and center of the α2 helix , respectively ( Dall and Brandstetter , 2016 ) . Moreover , only N138 is prominently surface exposed , suggesting that N-linked glycosylation in HaAEP1 is not utilized for mitigation against non-specific premature activation . The 1 . 8 Å resolution of our HaAEP1 structure allowed us to observe a succinimide moiety ( Snn ) below the catalytic His in a location identical to human AEP ( Figure 3A ) that was hypothesized to play a role in peptide ligation ( Dall et al . , 2015 ) . We distinguished dual conformations of the catalytic His178 and Cys220 residues which we hypothesize represent conformational changes that occur during catalysis and correspond to substrate free and reaction intermediate states ( Figure 3A ) . In the intermediate state the catalytic Cys Sδ is oriented ~95° towards the Nδ of the catalytic His imidazole ring which is orientated ~3 . 8 Å closer to this residue in the corresponding intermediate state . Moreover , flexibility of His in a relatively open pocket free of steric hindrance ( Figure 3—figure supplement 1 ) suggests an additional role for conformational shifts of the His in catalysis as seen with a range of proteases ( McLuskey et al . , 2012; Clark , 2016; Chekan et al . , 2017 ) . In the proposed alternate resting state the catalytic Cys Sδ is oriented towards the backbone amine of the highly conserved Gly179 , reducing the distance between them from 5 . 4 Å to 3 . 6 Å . In this orientation the backbone amine might function in stabilizing proton abstraction from the Cys thiol . Close examination of the electron density in the active site of HaAEP1 suggested a small peptide chain is intermittently bound to the intermediate conformation of Cys220 ( Figure 3B ) . Given AEPs specificity for Asx and the nature of the electron density , we built and refined a tetrahedral complex of a three-residue peptide ligand ( AAN ) bound to the active site Cys of HaAEP1 with partial occupancy ( Figure 3B ) . Residues Ala-Ala were modeled upstream of the P1 Asn due to the weak electron density away from the peptide backbone at these residues . The AAN peptide ligand allowed characterization of interactions likely to exist between HaAEP1 and P1 Asn and main chain of residues P2-P3 during substrate recognition . Due to the high sequence conservation of the HaAEP1 active site with mammalian AEPs and conserved substrate orientation , many of the interactions with the substrate match those observed for inhibitors of hAEPs ( Dall et al . , 2015 ) ( Figure 3C ) . The presence of this unexpected substrate in the HaAEP1 active site is likely to be a product of auto-activation . The continuous electron density between Cys220 and the P1 Asn supports the interpretation of the formation of a tetrahedral intermediate that is stabilized by the presence of an oxyanion hole formed by His178 , Gly179 and the backbone amine of Cys220 that is more congruous to the electron density than an acyl intermediate or free peptide ( Figure 3B , Figure 3—figure supplement 2 ) . Moreover , the observed short distance between the tetrahedral intermediate carbon atom and catalytic cysteine sulfur atom is incompatible with the absence of a covalent interaction . Main chain amino groups of Gly residues have previously been proposed to function in the creation of an oxyanion hole stabilizing the formation of a tetrahedral intermediate with the substrate in a wide range of cysteine proteases ( Dall and Brandstetter , 2016 ) . Furthermore , previous observations of tetrahedral intermediates and enzyme-product complexes with serine proteases have been shown to exhibit a pH-dependent equilibrium ( Wilmouth et al . , 2001; Radisky et al . , 2006; Lee and James , 2008 ) . Similarly , the trapping of this intermediate state may have been fortuitously facilitated by the activation of HaAEP1 at pH 4 , followed by crystallization at pH 7 . Although alternative conformations of the active site Cys in hAEP have been described previously , this tetrahedral intermediate state has not been described before ( Dall et al . , 2015 ) . The similarity of the HaAEP1 active site to that of OaAEP1 and the revelation that it contained a reactive succinimide prompted us to test an enzyme preparation similarly taken to pH 4 . 0 against the modified SFTI-1 precursor substrate SFTI ( D14N ) -GLDN substrate . HaAEP1 had previously been unable to create a macrocyclic product from SFTI ( D14N ) -GLDN , but had been shown to efficiently cleave it at a rate , kcat/Km value of 610 M−1 S−1 , similar to rates published for other AEPs ( Bernath-Levin et al . , 2015 ) . To our surprise , our new preparations of HaAEP1 taken to pH 4 . 0 produced cyclic SFTI ( D14N ) when the reactions were conducted at pH 6 . 5 ( Figure 4B WT , Figure 4—figure supplement 1A WT ) . Previously , HaAEP1 had been purified at pH 8 , activated at pH five and then used in reactions at pH 5 . The HaAEP1 preparations that were able to macrocyclize were similarly purified at pH 8 , but activated at pH four then returned to pH 6 . 5 . Activation at lower pH has been shown to be more effective at removing the cap domain , which in mammalian AEPs had a propensity to re-ligate ( Zhao et al . , 2014 ) . Higher pH could also favor macrocyclization by facilitating the deprotonation of the Gly N-terminus at the active site , priming it to attack its C-terminus , which is held in the thioacyl intermediate at the active site . To investigate the mechanism of macrocyclization by HaAEP1 we identified several residues for site directed mutagenesis . Firstly , we hoped to clarify the roles of Cys220 and Snn177 in macrocyclization through Ser and Ala mutations , respectively . Furthermore , we also mutated Snn177 to Gly , as the C13 protease GPI8 has also been proposed to carry out an intramolecular transpeptidation reaction yet displays a Gly residue at a location equivalent to Snn177 ( Zacks and Garg , 2006 ) ( Figure 2—figure supplement 2 ) . Secondly , we hoped to alter the ability of HaAEP1 to macrocyclize its native substrate SFTI-1 by altering the residues that fine tune this catalysis . By modeling the binding of an NMR structure of PawS1 ( Franke et al . , 2017 ) and the N-terminally cleaved SFTI-1 precursor ( SFTI-GLDN ) to HaAEP1 we were able to hypothesize the location of a hydrophobic S2ʹ binding region encompassing strand βII with G185 located at the center of this region ( Figure 4—figure supplement 2 ) . In addition , previous studies have also implicated that Asn , Glu and Asp residues proximal to the catalytic Cys and His function in catalysis ( Dall and Brandstetter , 2013; Zhao et al . , 2014 ) . Inspection of the HaAEP1 structure revealed that several of these residues ( including Asn73 and Glu221 ) are conserved in HaAEP1 . Glu221 is oriented away from Cys220 , in a direction similar to that seen in hAEP bound to human cystatin E , it might assume alternate conformations due to its solvent exposure and high B-factor . The mutant HaAEP1 proteins were expressed in E . coli and analyzed by circular dichroism ( Figure 4—figure supplement 3 ) , with wild-type ( WT ) HaAEP1 displaying a melting temperature of ~52°C similar to previous reports for hAEP ( Dall and Brandstetter , 2013 ) , and mutants displaying similar spectra to WT . Incubation of WT HaAEP1 with a fluorophore labeled ( BODIPY ) activity-based probe ( Lu et al . , 2015 ) illustrated its heterogeneity in size following activation at pH 4 and incubation with substrate at pH 6 . 5 , which has previously been observed for AEP proteins both in vivo and in vitro and speculated to be the result of processing of non-glycosylated forms ( Zhao et al . , 2014 ) ( Figure 4a ) . This probe also revealed a C220S mutation in HaAEP1 to result in the enzyme becoming incapable of pH-dependent activation ( Figure 4a ) . Further analysis of AEP mutants activity using the BODIPY probe illustrates the substantial heterogeneity in size between active WT and N73A , N73D , D177G , D177A or E221K mutant AEPs , likely due to autocatalytic processing , yet indicates that they remain active ( Figure 4—figure supplement 1D ) . The use of a seleno-modified synthetic SFTI ( D14N ) -GLDN substrate , which we have previously shown to be processed by HaAEP1 and CeAEP1 ( Bernath-Levin et al . , 2015 ) , allowed for a comparison of activity profiles of HaAEP1 mutants through the quantification of distinctive isotopic cyclic , acyclic and unprocessed peak areas by MALDI-MS ( Figure 4B–C , Figure 4—figure supplement 1A ) . This comparison reveals subtle changes between mutants in the ratio of cyclic , acyclic and unprocessed peptides and confirmed that the HaAEP1-C220S mutant is unable to cleave as evidenced by the lack of a peak at mass 1608 or 1626 ( Figure 4 , Figure 4—figure supplement 1 ) . As expected , mutation of the second residue of the catalytic dyad ( H178A ) also results in a drastic reduction in activity , based on SFTI ( D14N ) -GLDN processing and activity based BODIPY probe results , confirming the significance of C220 and H178 in AEP activity ( Figure 4 , Figure 4—figure supplement 1 ) . Interestingly , the H178A mutation does not abolish HaAEP1 activity , as previously seen with mAEP , and suggests a third residue could facilitate proton transfer at the active site ( Zhao et al . , 2014 ) . Similar results to the H178A mutation were also observed for G185S that was directed at altering the hydrophobic S2ʹ binding region . The effect of the G185S mutation suggests G185 has a role in substrate recognition and could sterically alter the conformation of H178 and Snn177 ( Figure 4 , Figure 3—figure supplement 1 ) . The mutation E221K , which has previously been shown to increase endopeptidase activity in hAEP , resulted in a loss of cyclic product as shown by an absence of a peak of mass 1608 ( Figure 4B , Figure 4—figure supplement 1A ) . Mutation of N73A leads to a higher ratio of cyclic to acyclic product as shown by an increased peak area relative to WT of mass 1608 and a reduced acyclic product peak of mass 1626 . N73D and D177G mutants appear to process SFTI ( D14N ) -GLDN in a manner similar to WT . Whereas the large fraction of unprocessed SFTI ( D14N ) -GLDN , mass 2025 , after incubation with D177A illustrates a reduction in processing efficiency , as previously noted for the mAEP D149A mutant ( Zhao et al . , 2014 ) . The effect of the N73A , E221K and D177G mutations on AEP catalyzed ligation was probed further by investigating each mutant’s ability to revert to its inactive form via re-ligation of its cap domain upon shift to neutral pH . As described for mAEP and equivalent mutants ( Zhao et al . , 2014 ) , incubation of HaAEP1 at pH 4 led to an irreversible dissociation of its cap domain ( Figure 4—figure supplement 4 ) . However , following activation at pH 5 . 5 and 6 . 5 , upon shifting to pH 8 the WT , N73A , E221K and D177G mutants were able to re-ligate the cap onto the core domain resulting in the formation of the inactive pro-enzyme with a Mw ~52 kDa as evidenced by SDS-PAGE and activity based probes ( Figure 4—figure supplement 4 ) . These results confirm the importance of the catalytic dyad in AEP function and show that HaAEP1 , like mAEP and the closely related C13 family member GPI8 , is able to perform its ligation reaction in the absence of a Snn residue . Moreover , subtle mutations affecting the stability of the catalytic dyad might favor either hydrolysis or macrocyclization . To discern the structural determinants that favored AEPs to be recruited independently by evolution multiple times for macrocyclization , we compared the structure of HaAEP1 and its predicted binding mode with the crystal structures of closely related cysteine proteases: sortase A , papain and metacaspase MCA2 ( Suree et al . , 2009; Chu et al . , 2011; McLuskey et al . , 2012 ) ( Figure 5 ) . Sortase A is a Staphylococcus aureus cysteine protease which catalyzes a similar transpeptidation reaction to AEP , ligating proteins bearing a LPXTG motif to peptidoglycan precursors in the bacterial cell wall ( Mazmanian et al . , 1999 ) . NMR studies have shown that the resolution of the transpeptidation thioacyl intermediate reaction occurs through nucleophilic attack of a lipid terminal amine in a steep valley between the β7-β8 and β4-β5 loops ( Suree et al . , 2009 ) . In comparison to HaAEP1 , the sortase A β7-β8 loop protrudes much further from the active site and the aromatic residues F122 , Y128 , W194 in these loops orient over the catalytic Cys ( Figure 5B ) . Together these loops , despite reported flexibility , would likely impart considerable steric hindrance for macrocyclization by inhibiting both the resolution of the intermediate by the N-terminal amine group and binding at the S2′ substrate ‘tail’ . Papain from the melon tree Carica papaya is the archetypal plant C1 family cysteine protease and has been found to bind to a cystatin homolog , tarocystatin , in a manner analogous to that of hAEP binding cystatin ( Otto and Schirmeister , 1997; Chu et al . , 2011; Dall et al . , 2015 ) . However , inspection of the papain active site reveals a topology that is not conducive for macrocyclization with the catalytic triad buried deep within the protein and steric hindrance for peptide N-terminal attack likely to be imparted by W177 , D158 and the extended alpha helical loops of the α3 and α7 regions ( Figure 5C ) . Metacaspases are expressed in plant , fungi and protozoa and display a C14 caspase domain that is structurally homologous to human caspases ( Tsiatsiani et al . , 2011 ) . Metacaspases reside within the same CD clan as AEP , but exhibit a strict specificity for a cleavage following Arg or Lys ( Vercammen et al . , 2004 ) . Currently no crystal structure is available for a plant metacaspase , however the crystal structure of the protozoan metacaspase MCA2 reveals an architecture that like papain would likely be unfavorable for macrocyclization due to steric hindrance around the active site from several prominent loops; including the β1-α1 , βA-βB and 280-loop ( McLuskey et al . , 2012 ) ( Figure 5D ) . The crystal structure of active HaAEP1 suggests that the combination of a relatively open reaction interface with space around the active site allowing for catalytic residue flexibility has resulted in the convergence upon AEPs for macrocyclization ahead of the other 30 families of cysteine proteases in plants ( Rawlings et al . , 2016 ) . Given the sequence similarity between AEPs and the conservation of residues involved in catalysis we hypothesized that the ability to macrocyclize peptides might be inherent to AEPs . To test this hypothesis we recombinantly expressed two AEPs from species which are currently not thought to contain cyclic peptides of any kind; Arabidopsis thaliana ( AtAEP2 ) and Ricinus communis ( RcAEP1 ) , respectively . These AEPs were purified and activated as described for HaAEP1 ( pH 4 ) and then incubated with non-native substrates; SFTI-GLDN and SFTI ( D14N ) -GLDN , at a pH that favors ligation ( pH 6 . 5 ) . Under these conditions RcAEP1 was able to macrocyclize both SFTI-GLDN and SFTI ( D14N ) -GLDN substrates whereas AtAEP2 was able to macrocyclize only SFTI ( D14N ) -GLDN ( Figure 6 ) . These findings further support our hypothesis that the structural features of AEPs described above have allowed for the convergence upon AEP for peptide macrocyclization reactions .
Herein , we have described the structure of an active plant AEP containing a peptide ligand with partial occupancy bound to the active site catalytic Cys as a tetrahedral intermediate , illustrating conformational flexibility in the AEP catalytic dyad . This structure has enabled us to predict a model for SFTI-1 macrocyclization by HaAEP1 where the GLDN tail of the SFTI-1 precursor orients in a manner analogous to cystatin binding to hAEP , over the catalytic His to a hydrophobic patch on the βII region , and where the N-terminus attack occurs between the diminutive βI sheet and β5-βIV loop in a manner analogous to the attack of thioacyl intermediates by peptidoglycan precursors in sortase-catalyzed transpeptidation reactions ( Suree et al . , 2009; Clancy et al . , 2010 ) . This model complements the requirement for a small amino acid followed by a hydrophobic residue at P1′ and P2′ due to their orientation over the catalytic His and towards the hydrophobic βII-βIII region , respectively . For several Cys proteases , including AEPs , a minor role for a third Asx/Arg residue in catalysis has previously described ( Vernet et al . , 1995; Dodson and Wlodawer , 1998; Dall and Brandstetter , 2016; Clancy et al . , 2010; Buller and Townsend , 2013; Dall and Brandstetter , 2013 ) . This third residue has been suggested to function in stabilization and orientation of the His imidazole ring . Interestingly , functional analysis of HaAEP1 catalytic triad residue N73 revealed that an N73A mutation resulted in the production of a higher ratio of cyclic SFTI-1 . The observation of potential conformational shifts in the crystal structure of HaAEP1 during catalysis suggests that the loss of an orientating N73 side chain allows for further conformational flexibility of H178 . This increased flexibility of His in a relatively open pocket that might accommodate a range of rotamers could reduce steric hindrance of an N-terminus entry towards the acyl intermediate and also facilitate deprotonation of the substrate N-terminus . The requirement for space and flexibility between the catalytic dyad favors the convergence for peptide macrocyclization upon Cys proteases over Ser proteases due to the close proximity between Ser-His residues in these proteases , an inherent requirement based upon the reduced nucleophilic properties of Ser ( Buller and Townsend , 2013 ) . Indeed , a recent structure of a macrocyclizing Ser protease , PCY1 , suggested that a shift in the catalytic His away from Ser is required for macrocyclization , based on a comparison to their hydrolytic relatives , and suggested a role for the catalytic His in deprotonation of the attacking peptide N-terminus ( Chekan et al . , 2017 ) . Furthermore , comparison with other Cys proteases suggests that AEPs have been converged upon for macrocyclization based on their relatively flat and open catalytic site . In contrast to a recent hypothesis for efficient macrocyclization based on a comparison of AEP structures focused on a region closer to the S4 pocket , here we suggest that differences in AEP catalytic efficiency and substrate specificity will be defined by subtle amino acid differences in the βII-βIII region , βI sheet and β5-βIV loop and the S1-S5 pocket ( Yang et al . , 2017 ) . Moreover , this orientation of the tail away from the βI sheet and β5-βIV loop is hindered in other caspases which exhibit a relatively straight substrate channel ( Figure 5 ) and thus prevent simultaneous attack of the thioacyl intermediate by the N-terminus loop upon scissile bond cleavage . Following activation , AEP functionality has been illustrated to be dependent on a delicate balance between pH and stability , where endopeptidase activity is favored at a low pH ( ~pH 4 ) and ligase activity is favored at a higher pH ( ~pH 6 ) ( Dall et al . , 2015 ) . A pH-dependent activity has been well documented for cysteine proteases and is a function of the formation of a thiolate and imidazolium ions on the catalytic dyad of Cys and His residues ( Dall and Brandstetter , 2016; Frankel et al . , 2005 ) . Given the hypothesized role of His in deprotonating the attacking N-terminal peptide chain in peptide macrocyclization it would therefore be expected that at a low pH the catalytic His would be more readily protonated and hence this reaction might be less frequent resulting in more hydrolysis at low pH . This hypothesis is supported by the low level of endopeptidase activity observed with Ala mutations of the catalytic His in HaAEP1 and other proteases , illustrating that thiolates might form and that hydrolytic resolution of the thioacyl intermediate might occur in the absence of His ( Frankel et al . , 2007; Zhao et al . , 2014 ) . Previous reports have also suggested that a local Gly amide backbone might facilitate catalysis via a transfer of a proton to the leaving group of the tetrahedral intermediate from a hydrogen bonded water molecule; analysis of the HaAEP1 structure reveals Gly179 could potentially perform this role in the absence of His ( Brady et al . , 1999 ) . Moreover , mutation of the acidic residue following the catalytic Cys to a basic residue in hAEP has previously been shown to enhance catalytic efficiency by decreasing the local pKa of the Cys residue ( Dall and Brandstetter , 2013 ) . Interestingly , we found an equivalent mutation in HaAEP1 to result in the loss of peptide macrocyclization indicating that E221 could also aid in deprotonation of the incoming N-terminus or that the residue’s larger side chain and expected orientation in activating the catalytic Cys could impart steric hindrance upon the N-terminus attack of the thioacyl intermediate ( Figure 4 , Figure 4—figure supplement 1 ) . However , given the finding that E221K mutation is unable to prevent re-ligation of the cap domain to the active core domain upon pH shift in HaAEP1 , there is likely to be redundancy between the local residues in creating a nucleophilic amine group to complete a transpeptidation reaction . The finding that AEPs from species that lack cyclic peptides may be coaxed into performing peptide macrocyclization of a linear peptide under favorable conditions significantly expands the potential use of AEPs for the production of cyclic peptides ( Figure 6 ) . Moreover , further investigation into the subtle nuances that define substrate specificity and catalytic activity between AEPs is warranted , with differences in the βII-βIII region , βI sheet , β5-βIV loop and around the S1-5 pocket including the variable α5-β6 and βIV-βV loops likely to be key ( James et al . , 2017 ) . Indeed , in comparison to HaAEP1 and OaAEP1 , the efficient peptide macrocyclizing AEP butelase one displays several different amino acids that could be responsible for this AEPs efficacy . These differences include shorter sidechains in the βI sheet and β5-βIV loops that may reduce steric hindrance on a peptides N-terminus during attack on a thioacyl intermediate ( HaAEP1: P181 , Q245 , N247 OaAEP1: A178 , T242 , S244 Butelase-1: A168 , G232 , S234 ) and differences in the βIV-βV region that could generate substrate specificity ( Figure 5—figure supplement 1 ) . Active HaAEP1 exhibits a succinimide at the same position as hAEP which has previously been postulated to perform a Cys-independent ligation reaction through cyclic rearrangement with the P1 Asx side chain ( Dall et al . , 2015 ) . Functional analysis reveals that HaAEP1 is able to perform peptide macrocyclization despite D177G/A mutations that cannot form a succinimide group ( Figure 4B–C ) . Moreover , these mutants continued to perform re-ligation of the cap domain when activated at pH >5 . 5 ( Figure 4—figure supplement 4 ) , a result that has previously been shown with mAEP ( Zhao et al . , 2014 ) . In the absence of a critical requirement for succinimide formation in macrocyclization the question remains as to why this relatively unstable aspartimide appears stable in AEP crystal structures and is largely conserved in the C13 proteases . Succinimides have been shown to form more readily when adjacent to His residues as the His Nδ abstracts a proton from the His backbone NH allowing the deprotonated main chain amine to attack the Asx side chain ( Brennan and Clarke , 1995; Takahashi et al . , 2016 ) . Once formed this succinimide could enhance the activity of the catalytic His by virtue of reducing stabilizing hydrogen bonding interactions with the carboxyl terminus and be involved in the proper positioning of the catalytic His . A further potential role for succinimides could present upon their hydrolysis through racemization and favored formation of iso-Asp ( Geiger and Clarke , 1987; Reissner and Aswad , 2003 ) . This iso-Asp could potentially represent a subtle mode of AEP regulation as the orientation of this side chain towards the S1 pocket is likely to disrupt substrate binding . Such iso-Asp residues have previously been proposed to regulate protein activity by a time-dependent molecular switch ( Geiger and Clarke , 1987; Ritz-Timme and Collins , 2002 ) . In the absence of caspases , plants have evolved a wide range of cysteine proteases to ensure functional redundancy in a myriad of highly regulated programmed cell death pathways in response to environmental and developmental cues ( Fagundes et al . , 2015 ) . Furthermore , plants have developed a variety of strategies to control the destructive prowess of these proteases including the expression of proteases as inactive zymogens with cofactor dependency , compartmentalization and the production of protease inhibitors ( Martínez et al . , 2012 ) . Of these cysteine proteases the AEPs have recently attracted considerable interest due to their ability to carry out peptide macrocyclization and their potential application in the synthesis of pharmacoactive cyclic peptides . Herein , the structural and functional analysis of HaAEP1 has revealed residues that are able to favor the production of cyclic or acyclic products from SFTI ( D14N ) -GLDN . Moreover , we have modeled a binding mode for productive macrocyclization of the HaAEP1 natural ligand SFTI-1 , based on a comparison with related cysteine proteases that is likely to be conserved between AEPs , where substrate specificity is defined by the amino acids around the binding site for respective AEPs . Furthermore , we have shown that AEPs from diverse species lacking cyclic peptides are able to perform macrocyclization under favorable pH conditions . These findings provide the foundation for further optimization of AEPs , potentially widening the array of peptide substrates that could be cyclized by AEPs .
DNA sequence encoding residues 28–491 of HaAEP1 ( accession code: KJ147147 ) , Ricinus communis AEP ( RcAEP1 ) residues 58–497 ( accession code: D17401 ) and Arabidopsis thaliana AEP ( AtAEP2 ) residues 47–486 ( accession code: Q39044 ) were cloned into a pQE30 ( QIAGEN , Hilden , Germany ) expression vector with an N-terminal six-histidine tag and expressed in the SHuffle strain E . coli ( New England Biolabs ) transformed with pREP4 ( QIAGEN ) . Briefly , cultures were grown at 30°C to an OD600 of 0 . 8–1 . 0 in Luria Broth medium containing 100 μg/mL ampicillin and 35 μg/mL kanamycin with expression induced at 16°C with 0 . 1 mM isopropyl β-D-1-thiogalactopyranoside then cells cultured overnight . Cell pellets were collected by centrifugation and lysed by ultrasonication in 50 mM Tris ( pH 8 . 0 ) , 100 mM sodium chloride , 0 . 1% Triton X-100 . The soluble fraction was then harvested by centrifugation and the supernatant was incubated ( batch wise ) with Ni-NTA resin overnight at 4°C . The resin was then washed with 20 mL of 50 mM Tris ( pH 8 . 0 ) and 20 mL of 50 mM Tris ( pH 8 . 0 ) 20 mM imidazole and protein was eluted stepwise with 20 mL of 50 mM Tris ( pH 8 . 0 ) 300 mM imidazole . For purification for crystal screens , six-histidine tagged inactive protein was purified , using an ÄKTA FPLC platform , by anion-exchange chromatography ( HiTrap Q HP 5 mL ) with a gradient of 0 to 500 mM sodium chloride over 60 min and then either activated by dialysis in 100 mM citric acid - sodium citrate buffer ( pH 4 . 0 ) containing 50 mM sodium chloride and 5 mM DTT , overnight at 16°C , or directly concentrated and further purified by size-exclusion chromatography ( HiLoad 16/600 Superdex 200 ) in 50 mM Tris ( pH 8 . 0 ) , 50 mM sodium chloride . Following dialysis active AEP was separated from insoluble material by centrifugation and purified by size-exclusion chromatography ( HiLoad 16/600 Superdex 200 ) in 100 mM citric acid – sodium citrate buffer ( pH 4 . 0 ) containing 50 mM sodium chloride . Protein was assessed for purity by SDS-PAGE . HaAEP1 site-directed mutations were made following the Stratagene QuickChange protocol with the following primers: The N73A mutation was made with forward primer 5′-GTA GCA AAG GTT ATG GTG CTT ATC GTC ATC AGG CC-3′ and reverse primer 5′-GGC CTG ATG ACG ATA AGC ACC ATA ACC TTT GCT AC-3′; the N73D mutation with 5′-GTA GCA AAG GTT ATG GTG CTT ATC GTC ATC AGG CC-3′ and 5′-GCC TGA TGA CGA TAA TCA CCA TAA CCT TTG CTA C-3′; The D177A mutation with 5′-CTG TTT TAT AGC GCT CAT GGT GGT CCG G-3′ and 5′-CCG GAC CAC CAT GAG CGC TAT AAA ACA G-3′; the D177G mutation with 5′-CTG TTT TAT AGC GGC CAT GGT GGT CCG GG-3′ and 5′-CCC GGA CCA CCA TGG CCG CTA TAA AAC AG-3′; the H178A mutation with 5′-CTG TTT TAT AGC GAT GCT GGT GGT CCG GGT G-3′ and 5′-CAC CCG GAC CAC CAG CAT CGC TAT AAA ACA G-3′; the G185S mutation with 5′-GTC CGG GTG TTC TGA GTA TGC CGA ATG AAC-3′ and 5′-GTT CAT TCG GCA TAC TCA GAA CAC CCG GAC-3′; the C220S mutation with 5′-TGA TTT ATC TGG AAG CAT CTG AAA GCG GCA GCA T-3′ and 5′-ATG CTG CCG CTT TCA GAT GCT TCC AGA TAA ATC A-3′; and the E221K mutation with 5′-GAT TTA TCT GGA AGC ATG TAA GAG CGG CAG CAT TTT TGA AGG-3′ and 5′-CCT TCA AAA ATG CTG CCG CTC TTA CAT GCT TCC AGA TAA ATC-3′ . Mutations were verified by sequencing and expressed as described for WT . Protein was concentrated to 10–15 mg/mL and crystal screening performed using the sitting-drop-vapor diffusion method with 80 μL of reservoir solution in 96-well Intelli-Plates at 16°C . Protein to mother-liquor ratios for the sitting drops were varied in each condition: 0 . 1:0 . 1 , 0 . 1:0 . 2 , and 0 . 2:0 . 1 μL . Crystals of active HaAEP1 were obtained in 0 . 1 M HEPES ( pH 7 . 5 ) , 1 . 4 M sodium citrate tribasic dihydrate . Single crystals were soaked in mother-liquor supplemented with 20% glycerol as a cryoprotectant prior to being flash-frozen and stored in liquid nitrogen . Data collection was performed at 100 K on the Australian MX2 ( micro-focus ) beamline using a wavelength of 0 . 9537 Å and diffraction data for crystals were collected to a resolution of 1 . 8 Å . Diffraction data were processed using iMOSFLM and scaled with AIMLESS from the CCP4 program suite ( Battye et al . , 2011; Winn et al . , 2011 ) in space group P3121 with unit cell dimensions a = b = 77 . 03 , c = 108 . 17 . A sequence alignment of HaAEP1 and human AEP1 was generated using ClustalO and used to create a search model of HaAEP1 based on the last common atom of human AEP ( 4FGU ) using CHAINSAW . The structure of HaAEP1 was solved by molecular replacement with PHASER using this search model , followed by automatic building with ARP/WARP . Manual building and refinement was performed in iterative cycles with COOT and REFMAC5 using the CCP4 program suite . Structural analysis and validation were carried out with COOT and MolProbity ( Emsley and Cowtan , 2004; Chen et al . , 2010 ) . Crystallographic data and refinement statistics are summarized in Table 1 with Ramachandran plot values calculated from COOT . The peptide AAN modeled into the HaAEP1 active site was oriented into the active site based on the similar mode of cystatin binding to human AEP ( 4N6O ) ( Dall et al . , 2015 ) . The tetrahedral intermediate was evidenced from initial visualization in Fo-Fc electron density difference maps using a polder OMIT map as implemented in phenix . polder ( Liebschner et al . , 2017 ) . Coordinates and structure factors were deposited into the Protein Data Bank ( PDB ) under accession code 6AZT . Tetrahedral complex of a three residue peptide ligand ( AAN ) bound to the active site Cys220 defined as CX9 in PDB file . Figures illustrating structures were generated using PyMol , electrostatic surface potentials were contoured at ±10 kT/e using an APBS PyMol plugin ( Dolinsky et al . , 2007; Schrodinger , 2010 ) . Models of C . ensiformis and C . ternatea AEPs were generated using the I-TASSER server ( Roy et al . , 2010 ) . Proteins purified in 50 mM Tris ( pH 8 . 0 ) 300 mM imidazole were concentrated in an 30 kDa Amicon centrifugal filter and buffer exchanged with an excess of 10 mM sodium phosphate buffer ( pH 8 ) . Concentrations were checked by absorbance at 280 nm with a NanoDrop using the extinction coefficient of Pro-AEP and samples diluted to 0 . 1 mg/ml or 0 . 2 mg/ml . CD measurements were made in triplicate using a Jasco J-810 CD spectrometer with a 0 . 1 cm quartz cuvette using a 1 nm bandwidth on a 0 . 1 mg/ml sample . CD wavelength spectra were collected from between 200–260 nm a rate of 1 nm/sec at 20°C . Melt curves were collected using the same bandwidth at 222 nm with temperature increasing at a rate of 1 °C/min from 20–95°C on a 0 . 2 mg/ml sample . WT HaAEP1 melting temperature was interpolated from melt curve using a sigmoidal four parameter logistic regression fit ( GraphPad Prism , La Jolla , USA ) . AEPs were activated by dialysis for four hours at room temperature in activation buffer ( 20 mM sodium acetate pH 4 . 0 , 5 mM DTT , 100 mM sodium chloride , 1 mM EDTA ) followed by a second dialysis into ligation buffer ( 20 mM MES pH 6 . 5 , 0 . 5 mM DTT , 100 mM sodium chloride , 1 mM EDTA acid ) . Protein concentrations were determined by measuring absorbance at A280 using a NanoDrop ( 1 Abs = 1 . 0 mg/mL ) . For mass spectrometry analysis of the processing of SFTI ( D14N ) -GLDN or native SFTI-GLDN by WT HaAEP1 , RcAEP , AtAEP2 and mutant HaAEPs , AEPs at a concentration of 40 µg/mL were incubated with 0 . 25 mM peptide with a diselenide bond and 25 mM native ( i . e . disulfide ) SFTI-1 as an internal standard in activity buffer ( 20 mM MES pH 6 . 5 , 5 mM DTT , 1 mM EDTA ) . Reactions were carried out at 37°C for 16 hr . Three independent reactions were performed for each protein tested . The reactions were stopped by dilution 100-fold in 50% acetonitrile , 0 . 1% formic acid and spotted with an α-cyano-4-hydroxycinnamic acid matrix onto a plate for analysis by MALDI-MS . Quantification of peak area by MALDI-MS was calculated using the internal standard to normalize for ionization efficiency as described previously ( Bernath-Levin et al . , 2015 ) . Briefly , for activity probe analysis 50 µL of AEP at 10 µg/mL was incubated with 1 µM of the BODIPY probe JOPD1 ( Lu et al . , 2015 ) at room temperature overnight and protected from light . The labeling reaction was stopped by the addition of 10 µL of 6x SDS-PAGE loading buffer containing β-mercaptoethanol and proteins separated using 4–12% Bis-Tris SDS-PAGE gels as previously described ( Lu et al . , 2015 ) . Labeled proteins were visualized in gel with excitation and emission wavelengths of 532 and 580 nm using a Typhoon 9500 ( GE Healthcare , Paramatta , Australia ) . WT , D177G , N73A and E221K proteins were purified by affinity chromatography as described above and activated by overnight ( 4°C ) dialysis at pH 4 . 0/5 . 5/6 . 5 in 100 mM citric acid - sodium citrate buffer containing 50 mM sodium chloride and 5 mM DTT , with pH adjusted by addition of 1 . 0 M Tris-HCl pH 8 . Prior to pH-shift , sample aliquots were flash frozen and stored at −80°C for subsequent activity probe analysis and SDS-PAGE analysis . Remaining activated protein was then returned to the previous buffer but with pH adjusted to pH 8 . 0 by adding 1 . 0 M Tris-HCl pH 8 . 0 , as previously described , and incubated overnight at 4°C ( Hara-Nishimura et al . , 1998 ) . Following dialysis at pH 8 . 0 samples were flash frozen as described above . Proteins were then analyzed for activity , as described above , using the BODIPY activity based probe . Following in gel visualization of active protein gels were immediately fixed in Coomassie Blue stain for comparison of active and inactive protein . | Most proteins are long , chain-like molecules that have two ends respectively called the N-terminus and C-terminus . However , certain proteins can close on themselves to become circular . This requires a chemical reaction between the N- and C-termini , which creates a strong bond between the two extremities . To go through this ‘cyclization’ process , a straight protein attaches to a certain type of protease , a class of enzyme that usually cuts proteins into smaller pieces . In plants that are distantly related , the same group of enzymes – called AEPs – has been selected to perform cyclization . Here , Haywood et al . study an AEP enzyme from sunflowers: they identify what about this enzyme’s structure is important to drive the complex chemical reaction that results in the protein being cyclized rather than simply cut . Using a technique called X-ray crystallography to see the positions of individual atoms in the enzyme , Haywood et al . caught a snapshot of the enzyme . Its structure explained how the enzyme’s shape can guide cyclization . In particular , the part of the enzyme that binds to the proteins , the active site , was relatively flat and open , but also flexible: this helped the N and C-termini react with each other and close the protein . Further experiments artificially mutated specific areas of the enzyme , which helped determine exactly which elements guide this succession of chemical reactions . The activity of AEPs is influenced by their local environment , such as acidity . In fact , Haywood et al . showed that certain AEPs , which do not normally carry out cyclization , can start performing this role when exposed to a different level of acidity . The pharmaceutical industry is increasingly interested in circular proteins , as these are stable , easily used by the body , and can be genetically customized to act only on specific targets . If the cyclization process is better understood , and then harnessed , new drug compounds could be produced . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2018 | Structural basis of ribosomal peptide macrocyclization in plants |
Mollicutes , a widespread class of bacteria associated with animals and plants , were recently identified as abundant abdominal endosymbionts in healthy workers of attine fungus-farming leaf-cutting ants . We obtained draft genomes of the two most common strains harbored by Panamanian fungus-growing ants . Reconstructions of their functional significance showed that they are independently acquired symbionts , most likely to decompose excess arginine consistent with the farmed fungal cultivars providing this nitrogen-rich amino-acid in variable quantities . Across the attine lineages , the relative abundances of the two Mollicutes strains are associated with the substrate types that foraging workers offer to fungus gardens . One of the symbionts is specific to the leaf-cutting ants and has special genomic machinery to catabolize citrate/glucose into acetate , which appears to deliver direct metabolic energy to the ant workers . Unlike other Mollicutes associated with insect hosts , both attine ant strains have complete phage-defense systems , underlining that they are actively maintained as mutualistic symbionts .
Bacterial endosymbionts , defined here as comprising both intra- and extra-cellular symbionts ( Bourtzis and Miller , 2006 ) , occur in all eukaryotic lineages and range from parasites to mutualists ( Bourtzis and Miller , 2006; Martin et al . , 2017 ) . Their genomes tend to evolve faster than those of free-living bacteria ( Delaney et al . , 2012; Moran et al . , 1995 ) and they often rely on recombination and horizontal gene transfer when their tissue localizations allow frequent DNA exchange with other bacteria , which tends to purge deleterious mutations when effective population sizes are small ( Naito and Pawlowska , 2016; Takeuchi et al . , 2014 ) . Host-level selection can also induce radical changes in the gene content of endosymbionts ( Wernegreen , 2002 ) . When they are pathogens such changes can be adaptations to prevail against host defenses or competing bacteria ( Didelot et al . , 2016 ) , as expected from arms races with Red-Queen dynamics ( Mallo et al . , 2002; Paterson et al . , 2010 ) . However , when symbionts are mutualists and provide nutritional services , they may become so tightly co-adapted to their hosts that they resemble organelles ( Douglas , 1996; Brinza et al . , 2009 ) . In such cases natural selection is expected to have purged any genes that mediated functions that could be provided more productively by the hosts , a process that has been referred to as Black-Queen dynamics ( Morris et al . , 2012 ) . The increasing availability of sequenced genomes and accurate molecular phylogenies ( e . g . Leclercq et al . , 2014; Gerth et al . , 2014 ) has allowed a number of intricate endosymbioses between bacteria and arthropods to be understood at functional metabolic levels well beyond qualitative assessments based on 16S ribosomal sequencing . Comparative genomics studies have detected gains and losses of genes or pathways when specialized endosymbionts co-evolve with arthropod hosts ( Wernegreen , 2002; Didelot et al . , 2016; Moran et al . , 2008 ) , and have shown that bacterial endosymbionts are particularly useful when their metabolites complement nutrient-poor diets of hosts . Examples are Buchnera ( γ-Proteobacteria ) providing aphids with essential amino acids ( Baumann et al . , 1995; Shigenobu et al . , 2000 ) , Wolbachia ( α-Proteobacteria ) producing vitamin B for Cimex lectularius bedbugs ( Hosokawa et al . , 2010; Nikoh et al . , 2014 ) , Baumannia and Sulcia ( γ-Proteobacteria ) providing sharpshooters ( Homalodisca coagulata ) with vitamins and amino acids ( Wu et al . , 2006 ) , Nardonella ( γ-Proteobacteria ) providing beetles with tyrosine required for cuticle formation ( Anbutsu et al . , 2017 ) , and Stammera bacteria allowing leaf beetles to decompose pectin ( Salem et al . , 2017 ) . A recent comparative analysis confirmed that nutrient supplementation often drives evolution towards host dependence especially when symbionts are vertically transmitted ( Fisher et al . , 2017 ) . The social insects with superorganismal colonies , characterized by permanent physiologically and morphologically differentiated castes , appear particularly amenable for hosting specialized bacterial symbionts . Previous studies have documented amino acid provisioning by Blochmannia ( γ-Proteobacteria ) hosted by Camponotus carpenter ants ( Feldhaar et al . , 2007 ) , and suggested that bacterial symbionts provision Cephalotes turtle ants with essential amino acids ( Hu et al . , 2018 ) and Cardiocondyla ants with useful intermediate metabolites ( Klein et al . , 2016 ) . Honeybees were further suggested to rely on several specialized gut bacteria for carbohydrate breakdown of ingested pollen and nectar ( Engel et al . , 2012 ) and fungus-growing termites were discovered to have caste-specific microbiomes depending on whether individuals ingest plant material ( mainly decaying wood ) or only farmed fungus ( Poulsen et al . , 2014 ) . Finally , both bees and termites rely on gut microbes to provide them with acetate that can cover up to 100% of their metabolic needs ( Odelson and Breznak , 1983; Zheng et al . , 2017 ) . The leaf-cutting ants are the crown group of the attine fungus-growing ants , a monophyletic tribe that evolved 55–60 MYA when their ancestor switched from a hunter-gatherer lifestyle to an exclusive fungal diet ( Nygaard et al . , 2016; Branstetter et al . , 2017 ) . The evolutionarily derived attine lineages rear fully domesticated and co-adapted fungal cultivars that provide the ant farmers with specialized hyphal tips ( gongylidia ) containing mostly carbohydrates and lipids that the workers harvest and digest ( De Fine Licht et al . , 2014; Quinlan and Cherrett , 1979 ) . The ant brood is completely dependent on the ingestion of fungal biomass ( Hölldobler and Wilson , 1990 ) , but workers may ingest and assimilate liquids as well ( Littledyke and Cherrett , 1976; Shik et al . , 2018 ) . However , similar to other ants , they cannot ingest solid plant or animal fragments that they collect to provision their fungus gardens because a sieve in the infrabuccal cavity filters out any particles in excess of ca . 100 µm ( Mueller et al . , 2001 ) . This obligate reciprocity between cultivation and nutrition facilitated further innovations in the terminal clade of Acromyrmex and Atta leaf-cutting ants , which evolved 15-20 MYA ( Nygaard et al . , 2016; Branstetter et al . , 2017 ) . These two genera obtained functionally polyploid cultivars ( Kooij et al . , 2015 ) , adopted multiple queen-mating so their colonies became genetic chimeras ( Villesen et al . , 2002 ) , and became herbivores with massive ecological footprints in Latin America ( Schultz and Brady , 2008; Mehdiabadi and Schultz , 2010; Schiøtt et al . , 2010; Leal et al . , 2014; Shik et al . , 2014 ) . Previous studies have shown that Acromyrmex and Atta leaf-cutting ants harbor low-diversity microbiomes , which include Wolbachia ( only in Acromyrmex ) , Mollicutes and hindgut Rhizobiales ( Van Borm et al . , 2002; Andersen et al . , 2012; Sapountzis et al . , 2015; Meirelles et al . , 2016 ) , symbionts that were inferred to possibly complement the nitrogen-poor diets of Acromyrmex leaf-cutting ants ( Sapountzis et al . , 2015 ) . Depending on the actual species studied , Mollicutes – tiny bacteria that lack a cell-wall – can often be found as abundant endosymbionts in up to 100% of leaf-cutting ant colonies ( Sapountzis et al . , 2015; Meirelles et al . , 2016; Zhukova et al . , 2017 ) , but the absence of in-depth genomic studies has precluded more than speculation about their putative roles as either parasites ( Meirelles et al . , 2016 ) or mutualists ( Sapountzis et al . , 2015 ) . To clarify the functional metabolic properties of attine-associated Mollicutes , we mapped the abundances of the two most common strains , EntAcro1 and EntAcro10 ( cf . Sapountzis et al . , 2015 ) , in thirteen Panamanian fungus-growing ant species and compared these abundances with the typical spectrum of forage-material that different fungus-farming ants collect and use as compost to manure their fungus-gardens ( Kooij et al . , 2014a; Leal and Oliveira , 2000; Shik et al . , 2016 ) . The Panamanian fauna of attine ants encompasses nine of the 17 known genera , including the three most basal genera ( Apterostigma , Mycocepurus and Myrmicocrypta ) , two other basal genera ( Cyphomyrmex and Mycetophylax ) being more closely related to the Trachymyrmex and Sericomyrmex lineages that arose and diversified while rearing gongylidia-bearing cultivars , and finally the Atta and Acromyrmex leaf-cutting ants who came to practice fungus-farming at an ‘industrial’ scale ( Branstetter et al . , 2017; Schultz and Brady , 2008; Mueller et al . , 1998 ) . To explore nutritional mechanisms underlying putative mutualistic functions of these bacteria , and their association with changes in scale of farming over evolutionary time , we isolated and sequenced the EntAcro1 and EntAcro10 symbionts ( Sapountzis et al . , 2015; Zhukova et al . , 2017 ) . We subsequently compared their draft genomes with ten published genomes of Mollicutes associated with insect hosts having specialized diets , as well as with several other Mollicutes genome-sequences to assess enrichments and losses of gene categories and metabolic pathways . For EntAcro1 , where the genomic data suggested the most advanced mutualistic functions , we measured expression levels of bacterial transporter genes related to the decomposition of plant-derived compounds and ant genes related to the uptake of exogenous acetate , an end-product of Mollicutes’ anaerobic metabolism . Our genome comparisons also allowed us to evaluate arginine decomposition functions and defense mechanisms against bacteriophage attack , assuming that: i ) variable food-borne arginine supplementation by the fungal cultivar ( Nygaard et al . , 2016; Nygaard et al . , 2011 ) may have offered a niche to both Mollicutes symbionts to convergently evolve similar mutualistic interactions with attine ants , and ii ) the abundant and specific bacteriophage sequences that we obtained in the libraries of the EntAcro1 symbiont indicate that these bacteria have been under selection to maintain costly defenses because extracellular life in the gut lumen likely exposes them to frequent phage encounters .
De-novo assembly , annotation and phylogenetic binning of contigs generated from Ac . echinatior fecal fluid and Ap . dentigerum fat body produced a single bin of contigs for each symbiont , confirming them to be valid bacterial species that we will henceforth refer to as EntAcro1A and EntAcro10A; Supplementary file 1A ) . The predicted coding sequences gave top matches with previously sequenced Mollicutes strains ( Supplementary file 1B ) and each of the bins had a single rRNA operon organized as 16S , 5S , 23S ( Figure 1—figure supplement 1 ) , similar to closely related Spiroplasma ( Ku et al . , 2013; Lo et al . , 2013; Chang et al . , 2014 ) and identical to OTUs in a previous 16S phylogeny ( Sapountzis et al . , 2015 ) . Additional bins ( B; see Supplementary file 1B and 1C ) did not contain relevant bacterial sequences and were not considered further , similar to the EntAcro1C bin that contained bacteriophage sequences with similarity to members of the Gokushovirinae subfamily ( Microviridae family ) , a phage lineage known to infect Spiroplasma ( Chipman et al . , 1998; Supplementary file 1 ) . Further analyses of the annotated coding sequences confirmed that EntAcro1A and EntAcro10A represented discrete draft genomes of EntAcro1 and EntAcro10 with no or very few missing genes . These genomes had 758 and 776 coding sequences , respectively , and genome sizes of less than 0 . 9 Kb based on the annotation features ( Supplementary file 1A; Figure 1—figure supplement 1 ) . A total of 59 published Mollicutes genomes were used for phylogenomic reconstructions after their predicted proteins gave clear matches to the EntAcro1 and EntAcro10 amino acid sequences ( Supplementary file 1B ) . This produced nearly identical trees after maximum likelihood ( Figure 1 ) and Bayesian analysis of nucleotide and amino acid sequences ( Supplementary file 2; Figure 1—figure supplement 2 ) and revealed that EntAcro1 and EntAcro10 belong to the Entomoplasmatales group , confirming earlier 16S assignments ( Sapountzis et al . , 2015 ) . This clade contains Spiroplasma and Mesoplasma bacteria associated with insects and plants and Mycoplasma bacteria known to be mammalian pathogens . Sister-group relationships showed that EntAcro1 and EntAcro10 are not closely related and thus likely to have been independently acquired as attine ant symbionts . EntAcro10 is a relatively basal Spiroplasma-like species , a genus with pathogenic , mutualistic and yet unknown interactions with mostly arthropod hosts . However , EntAcro1 is sister to the Mesoplasma/Mycoplasma clade and relatively similar to one of the very few Entomoplasmatales known to be associated with plant hosts in a clade that otherwise consists of vertebrate pathogens . The more distant sister clades are also predominantly pathogenic . We restricted our comparative evaluations and hypotheses testing to the Spiroplasma and Mesoplasma symbionts associated with insect hosts ( the attine symbionts highlighted in dark yellow in Figure 1 and the ones in between ) and plants ( M . florum ) . Functional annotations ( eggNOG database ) showed that strains clustered primarily based on metabolic genes and secondarily according to shared identity for informational genes ( transcription , translation and recombination/repair processes ) , which both correlated with host associations ( Figure 1—figure supplement 3; Figure 1—figure supplement 3—source data 1 ) . Functional gene-similarities were confirmed by Mantel tests showing that the Euclidean dissimilarity matrix of orthogroups was more strongly associated with phylogenomic distances between insect hosts ( r2 = 0 . 298 , p=0 . 036 ) than with phylogenomic distances between bacterial species ( Figure 1; r2 = 0 . 201 , p=0 . 133 ) , suggesting that many genes that adapt Entomoplasmatales symbionts to their hosts have been horizontally acquired . Metabolic reconstructions ( KEGG ) further suggested that all Entomoplasmatales are facultative anaerobes , because they lack the genes encoding TCA cycle enzymes and are thus universally incapable of oxidative phosphorylation . However , EntAcro1 likely lost its aerobic abilities completely because pyruvate dehydrogenase genes are also missing . Differences in metabolism between the two attine symbionts and other insect-associated Spiroplasma/Mesoplasma strains were primarily found in pathways mediating catabolism of glycerol , dihydroxyacetone ( DHA ) , citrate , arginine , and N-Acetylglucosamine ( GlcNAc ) ( Figure 2 and Supplementary file 3 ) . Metabolic pathway reconstructions were consistent with EntAcro10 being a less specifically adapted symbiont than EntAcro1 ( Figure 2 ) . Inferences of this kind , based on direct similarity between bacterial genomic databases , may not be fully accurate because gene-families encoding metabolic transporters evolve rather rapidly so the actual transported substrates may no longer be identical . However , the draft genomes that we obtained were sufficiently complete to provide reasonable confidence for reconstruction of operational metabolic transporters through the plasma membrane , the associated metabolic pathways inside the bacterial cells , and the metabolic end-/by-products involved ( Figure 2 ) . We found that EntAcro10 can utilize glycerol from ant host cells and monosaccharides which are likely derived from fungus-garden metabolites or juices ( glucose/fructose ) ingested during foraging ( Figure 2 ) . We also found an arginine transporter and metabolic genes indicating that EntAcro10 can decompose arginine . This finding is of interest because the attine ants lost the ability to synthesize this nitrogen-rich amino acid ( highest nitrogen to carbon ratio of all amino acids ) when fungus-farming evolved , so they obtain arginine from the fungal cultivar and potentially also from ingested fruit juice and plant sap in the herbivorous crown-group leaf-cutting ants ( Kooij et al . , 2014a; Winter et al . , 2015 ) . The evolutionarily derived EntAcro1 symbiont had all the transporters and pathways identified for EntAcro10 but also novel ones that would appear to be particularly adaptive for an abundant extracellular endosymbiont in the gut lumen ( Sapountzis et al . , 2015 ) of leaf-cutting ants . First , it has a GlcNAc transporter likely importing chitin monomers , a relatively common metabolic pathway in Mollicutes found in 85 of the 177 genomes that we considered . The possession of a chitin importer is relevant because chitin is one of the most abundant compounds in the ants’ fungal diet ( Figure 2 ) , particularly in the leaf-cutting ants whose cultivars have modified and likely thicker cell walls than the cultivars of phylogenetically more basal attine ants ( Nygaard et al . , 2016 ) . Second , EntAcro1 has a rather unique citrate transporter indicative of a rare catabolic pathway observed in only four of the 177 available Mollicutes genomes; Figure 2—figure supplement 1 ) . Further examination of the five or six citrate utilization genes involved in this pathway ( citS , citC , citD , citE , citF and potentially citG; see Figure 2—figure supplement 1 ) showed that all these genes are extremely rare across the Mollicutes genomes and present similarities to genes from bacterial classes outside the Mollicutes ( e . g . Firmicutes and Clostridia ) suggesting they were horizontally obtained ( Figure 2—figure supplement 1 ) . Anaerobic citrate fermentation , or co-fermentation of glucose/citrate , using the citS-citF operon will produce acetate ( Pudlik and Lolkema , 2011; Starrenburg and Hugenholtz , 1991 ) , which is likely imported by eukaryote cells to fuel metabolism ( Figure 2—figure supplement 2 ) or stored in the fat body cells ( Figure 2—figure supplement 3 ) . The citrate pathway thus appears to reflect that leaf-cutting ants can utilize the citrate metabolite that they are known to ingest in substantial quantities as plant sap when cutting fresh leaves ( Littledyke and Cherrett , 1976 ) and in the form of other juices when drinking from freshly fallen fruit ( De Fine Licht and Boomsma , 2010; Evison and Ratnieks , 2007 ) . Screening EntAcro abundances in worker bodies across the Panamanian attine ants showed distinct patterns of prevalence . EntAcro10 was present in most attine species investigated such that there were no significant differences across the entire set of 13 species ( planned contrasts , z = −0 . 62 , p=1 . 00 ) . However , EntAcro1 was almost exclusively found in the leaf-cutting ants ( planned contrasts , z = 2 . 88 , p=0 . 016 ) with their closest Panamanian relative T . cornetzi and yeast-growing C . rimosus as the only ( partial ) exceptions . At the same time EntAcro10 abundances in leaf-cutting ants were lower than EntAcro1 abundances albeit only marginally so ( planned contrasts , z = −2 . 56 , p=0 . 045 ) ( Figure 3A; Figure 3—source data 1 ) . This pattern implies that the appearance of EntAcro1 is correlated with changes in the spectrum of substrates that the farming ants provide to their fungus gardens ( Figure 3B;Figure 3—source data 2 ) , with fresh leaf , fruit and flower provisioning dominating in the leaf-cutting ants and the phylogenetically more basal attines collecting primarily detritus-based substrates such as insect frass and wood chips ( Figure 3—figure supplement 1 ) . These differences were variably ( non ) significant per forage category ( Figure 3—figure supplement 1 ) , but bacterial abundances and foraging preferences generally covaried for EntAcro1 ( Mantel test; r2 = 0 . 447 , p=0 . 012 ) but not for EntAcro10 ( Mantel test; r2 = −0 . 127 , p=0 . 882 ) . These results suggest that the derived metabolic pathways of EntAcro1 ( Figure 2 ) may have been associated with the expansion of the scale of fungus farming and the adoption of functional herbivory in the leaf-cutting ants . The abundances of EntAcro1 bacteria are known to be highly variable , between lab and field colonies , between replicate colonies that seem otherwise fully comparable , and between nestmate workers of both Atta and Acromyrmex leaf-cutting ants ( Sapountzis et al . , 2015; Zhukova et al . , 2017 ) . To evaluate the potential mutualistic benefit of acetate production by EntAcro1 ( Figure 2—figure supplement 2 ) , we experimentally manipulated EntAcro1 abundances in lab colonies of Ac . echinatior and measured rates of acetate uptake in the same ants . We compared ants maintained on their normal fungus garden diet with ants on a sugar/citrate replacement diet with or without antibiotics , known from pilot experiments to remove most Mollicutes from the guts and associated organ systems . We quantified EntAcro1 abundances by measuring the number of transcripts of the bacterial housekeeping gene ftsZ and the rates of acetate uptake by ant host cells by measuring the expression of MCT1 ( Figure 4; Figure 4—source data 1 ) , a gene encoding a plasma membrane protein that imports acetate in eukaryotic cells; Figure 2—figure supplement 2 ) . We found that these variables were positively correlated ( ρ = 0 . 57 , p<0 . 001 ) , suggesting that acetate production by EntAcro1 boosts acetate uptake by the ants who likely convert acetate directly into ATP ( Figure 2—figure supplement 2 ) . This inference is somewhat tentative because tetracycline can impair mitochondrial function ( Moullan et al . , 2015 ) and thus overall metabolic functionality , a confounder that could have been measured by tracking a specific mitochondrial protein . The fact that the antibiotics data point in Figure 4 directly extended the trend obtained from the treatments without antibiotics ( sugar/citrate diet ) and the control ( fungal diet ) suggests that this confounding effect has been minor but further work will be needed to validate this result . Conversion of citrate to acetate by the EntAcro1 symbiont would be consistent with the general observation that leaf-cutting ants sustain much higher levels of worker activity , both inside nests and while foraging , than phylogenetically more basal attine ants ( Kooij et al . , 2014a ) . To better understand the function of the arginine processing genes ( Figure 2 ) , we returned to our comparative genomic data for the attine symbionts EntAcro1 and EntAcro10 and the ten Spiroplasma species associated with other insects ( Figure 5 ) . The observation ( Figure 4—figure supplement 1; Figure 4—source data 1 ) that the arginine transporter of EntAcro1 was most highly expressed in the hindgut lumen suggests that EntAcro1 cells may need to decompose arginine in exceedingly low pH conditions ( ≤4 ) . This physiological tolerance may represent a fine-tuned mutualistic service , as became clear when we evaluated patterns of gene expression of all four transporters mediating catabolism of key resources in the gut system and associated ant tissues . The predicted transporters of arginine , citrate , GlcNAc and glycerol or DHA ( Figure 4—figure supplement 1; Figure 4—source data 1 ) were expressed throughout the guts and associated organs of Ac . echinatior workers , but their expression levels differed across abdominal tissues possibly in response to a steep gradient from pH seven in the midgut and fat body cells where the glycerol transporter is highly expressed , via pH five in the ileum and pH four in the rectum ( Erthal et al . , 2004 ) where the arginine transporter is highly expressed ( Figure 4—figure supplement 1; Figure 4—source data 1 ) . Earlier work has indicated that utilization of the citS-citF operon is most efficient at or just above pH 5 . 5 ( Magni et al . , 1999; Sánchez et al . , 2008 ) , suggesting that citrate catabolism by EntAcro1 cells happens primarily in the midgut and in ileum where the pH is optimal for that function ( Figure 4—figure supplement 1 ) , which leaves arginine decomposition as the main terminal digestion process in the hindgut where pH is low . Decomposition of all available hindgut arginine into NH3 just before the anaerobic EntAcro1 symbionts would die from exposure to aerobic conditions via the ants’ fecal fluid would ensure that manure of the fungus-garden provides nitrogen in its most readily available form for fungal protein synthesis . The conversion of excess arginine to ammonia in the ant hindgut may thus resolve a potential mutualistic mismatch because the attine fungal cultivars have generic amino acid transporters , but they lack specialized arginine transporters to process environmental arginine , similar to other basidiomycete fungi ( Figure 5—figure supplement 1 ) . In general , ammonia is the preferred nitrogen source for fungal growth ( Ahmad et al . , 1990; Abril and Bucher , 2004 ) , so any increase in the ammonia to arginine ratio of fecal fluid manure would benefit the farming symbiosis as a whole . At the same time this conversion prevents nitrogen waste , as would happen when excess arginine were to be deposited on a fungal cultivar primarily adapted to using simpler nitrogen sources . We found clear evidence for both EntAcro1 and EntAcro10 having two intact bacterial defense systems to ward off phage attack , a Type-1 R-M system and a CRISPR pathway ( Figure 5 ) . Genes belonging to both defense systems are often horizontally transmitted among bacteria ( Labrie et al . , 2010 ) and maintaining them is costly ( Stern et al . , 2010; Vasu and Nagaraja , 2013; Vale et al . , 2015; Burstein et al . , 2016 ) , so these defenses are primarily expected in bacterial species that face consistent threats of phage attack without being severely resource-constrained . Gene-level comparisons with the other Spiroplasma symbionts showed that none of them had the same dual defense system against phage attack . Type I R-M system genes were present only in non-pathogenic Mollicutes strains , similar to arginine catabolism pathways being restricted to Spiroplasma strains with non-pathogenic host associations , except for the reputedly pathogenic S . apis ( Figure 5; Supplementary file 4 ) . The generally rarer CRISPR system was complete ( i . e . both CRISPR repeat/space regions and cas genes being present ) only in S . chrysopicola , S . apis , and the two ant associated EntAcro strains . Our finding that the two attine ant symbionts are unusually well protected is consistent with them being vulnerable to phage attack when they reach high densities in the gut lumen . We did indeed find Spiroplasma-specific phages of the Gokushovirinae in the contig bin ‘C’ of the EntAcro1 symbiont ( Supplementary file 1C ) isolated from fecal fluid . Not finding these phage sequences in EntAcro10 might reflect that these bacteria were isolated from the fat body of A . dentigerum where they are intracellular symbionts and that titers of EntAcro10 were very low ( Figure 3 ) . These functional inferences are tentative , but potentially of significant interest , so we will return to them below .
The loss of the arginine synthesis pathway in the basal attine ants ( Nygaard et al . , 2011; Suen et al . , 2011 ) has been instrumental in making their fungus-farming symbiosis obligate ( Nygaard et al . , 2016 ) . The selection regime that caused this loss remains unknown ( Nygaard et al . , 2016; Ješovnik et al . , 2016 ) , but it is reasonable to assume that outsourcing the production of this most nitrogen-rich amino acid to fungal cultivars gave complementary efficiency benefits even though it also generated symbiotic dependency . For symbiotic division of labor to be sustainable under variable environmental conditions , average levels of fungal arginine production would have to be higher than the minimally sufficient level to avoid occasional windows of fatal shortage in the symbiosis as a whole . Symbiotic dependency may thus have created a niche for Mollicutes symbionts to ensure that surplus arginine is recycled as NH3 to provide the most efficient manure for new garden growth . Tissues of adult insects no longer grow and may thus only need small amounts of nitrogen for maintenance , so passing on excess NH3 to fungus gardens via fecal fluid would have unambiguous benefits for the complex mutualism as a whole ( Schiøtt et al . , 2010 ) . This conjecture was recently confirmed for Atta workers in an independent study showing that workers fed with ammonium nitrate ( the protonated form of ammonia ) transfer nitrogen via their fecal fluid to the fungus garden ( Shik et al . , 2018 ) . Mollicutes-assisted garden manuring would thus imply that any surplus nitrogen remains a stable resource for new fungal protein synthesis and thus growth of the ant brood that only ingests fungal food . This underlines that driving-agency in obligate farming mutualisms is ambiguous . A shorter explanation of the prudency of this co-adaptation is that garden fungi domesticated ants to maintain and disperse them , and that they benefitted from the ants domesticating Mollicutes to ensure not a single nitrogen atom is wasted and their keepers could ( later ) utilize external resources such as citrate to work harder to enhance fungal growth . The only other ant lineage in which Entomoplasmatales ( Mollicutes ) endosymbionts have so far been abundantly found are the army ants ( Funaro et al . , 2011 ) . These ants are exclusive predators of mostly invertebrate prey ( Kronauer , 2009 ) and 16S rDNA sequences of their Mollicutes symbionts suggested they are closely related to EntAcro1 but rather distantly to EntAcro10 ( Funaro et al . , 2011 ) . It is intriguing that the Dacetine sister lineage of the fungus-farming ants are also specialized predators ( Branstetter et al . , 2017; Ward et al . , 2015 ) . It would thus be interesting to clarify whether also Dacetine ants have Entomoplasmatales symbionts , how ( un ) related they would be to the EntAcro symbionts of the fungus-growing ants , and whether army ants acquired their Mollicutes horizontally from preying upon on attine ants ( Powell and Clark , 2004 ) . Broader consideration of the general feeding ecology of all insects hosting Mollicutes symbionts revealed that they are mostly specialized on nutritionally deficient but protein-rich diets , such as vertebrate blood ( flies and mosquitoes ) and pollen/nectar ( bees and flies ) . The pollen/nectar feeders all had Spiroplasma symbionts with complete arginine catabolism pathways , suggesting they might similarly convert excess dietary arginine into NH3 , although the absence of functional studies precludes speculation about the type of mutualistic advantage yielded by this conversion ( de Groot , 1952; Vrzal et al . , 2010; Uchida , 1993; Honeybee Genome Sequencing Consortium , 2006; Nene et al . , 2007; Figure 5—figure supplement 2 ) . Thus , while excreted NH3 would appear to have a clear mutualistic function for manuring attine fungus-gardens ( Figure 5—figure supplement 3 ) , the benefits of NH3 production in the gut system of bees and some mosquitoes is less clear . Overall , it seems that primarily non-pathogenic Spiroplasma strains may have been selected to catabolize host-food-associated arginine ( Figure 5; Supplementary file 3; Supplementary file 4 ) , but this provisional inference needs explicit functional verification by artificial diet experiments and selective removal of symbionts to quantify putative changes in arginine and ammonia titers . Our results indicate that EntAcro1 was acquired as additional symbiont to EntAcro10 relatively shortly before the leaf-cutting ants evolved and that EntAcro1 supplements already available arginine recycling with novel pathways allowing ant workers to process non-fungal metabolites ( Figure 2; Figure 4; Figure 5—figure supplement 3 ) . Because citrate catabolism genes do not exist in EntAcro10 the lower attine ants may be generally unable to convert plant-derived citrate or glucose/citrate into acetate . The acquisition of EntAcro1 thus likely allowed the ant farmers to tap into additional non-fungal resources to maintain higher metabolic rates . These differences match what is generally known about the increases in farming scale , foraging activity , and garden growth-rates when moving from the basal attine ants to the derived branches of the attine phylogenetic tree ( Kooij et al . , 2015; Shik et al . , 2014; Shik et al . , 2016 ) . It is interesting that one of the closest relatives of EntAcro1 ( M . florum ) is associated with plants ( Figure 1 ) , which could suggest that this symbiont was acquired when the ancestors of the leaf-cutting ants started to forage on live plant material . However , confirmation of this hypothesis would require the closest relatives of EntAcro1 to be associated with American Angiosperms . The few current records are from citrus trees of Asian origin ( Liu et al . , 2012 ) , so substantial sampling effort will be needed to investigate this possible plant association . The timing of the acquisition of EntAcro1 is intriguing . It was recently shown ( Branstetter et al . , 2017 ) that a monophyletic crown group of the attine ants evolved in Central/North America following colonization of this subcontinent by a single South-American attine ancestor 22–27 MYA , well before the isthmus of Panama closed . The timing of isthmus closure is controversial with some maintaining that it happened as recently as ca . 3 MYA ( O'Dea et al . , 2016 ) , while other studies indicate it may have been as early as the mid-Miocene ca . 13–15 MYA ( Bacon et al . , 2015; Montes et al . , 2015 ) . A recent study on army ants , whose queens are wingless and thus dependent on solid land-bridges for dispersal , indicated colonization of Central-North America 4–7 MYA ( Winston et al . , 2017 ) . Also this dating is much later than the inferred timing of the first attine ant arrival in what was then the Central-North American subcontinent ( Branstetter et al . , 2017 ) . The implication is that most of the higher attine ant radiation happened on a novel subcontinent that was devoid of attine ants . In this context it is interesting that we also found EntAcro1 in some field colonies of T . cornetzi ( Figure 3 ) , a higher non-leaf-cutting ant representing the most basal attine branch that colonized Central-North America ( Branstetter et al . , 2017 ) . This suggests that EntAcro1 may have been domesticated in Central-North America in response to the founding lineage encountering new ecological opportunities perhaps including plants with EntAcro1-like symbionts . Also here , larger sampling efforts will be needed to verify whether endosymbiotic microbiome signatures of this major biogeographic vicariance event continue to be found across the extant attine ants of Central/North and South America . If EntAcro1 was domesticated just before or during the transition to active herbivory it may have somehow facilitated the transition to industrial-scale farming as presently found in the Atta and Acromyrmex leaf-cutting ants throughout the Americas . We found that both EntAcro1 and EntAcro10 have dual , fully intact cellular defenses against phage attack , consisting of a R-M ( Restriction Modification ) Type one system and a CRISPR pathway ( Figure 5 ) . We obtained a substantial number of phage sequences specific for Spiroplasma-like bacteria ( the EntAcro1C bin; Supplementary file 1C ) and our data show that the abundances of particularly EntAcro1 inside ant bodies can be very high ( Zhukova et al . , 2017 ) ( Figure 3 ) . The mere presence of phage defensive mechanisms is not surprising because high clonal bacterial densities make bacteriophage attack efficient and rewarding . However , what makes the intactness of two complete pathways of the two EntAcro symbionts interesting is that none of the related Spiroplasmas associated with other insects has two such intact pathways . Both types of phage defenses are last resort systems , operating only when a phage has already broken through the bacterial cell membrane and has released its DNA into the bacterial cell . Recent work has clarified that these bacterial defenses are analogous to a non-specific innate immune system ( R-M ) and an adaptive ( trainable ) immune system for recognizing specific viral DNA ( CRISPR ) ( Vasu and Nagaraja , 2013; Seed , 2015 ) , and that the two systems can operate synergistically ( Dupuis et al . , 2013 ) . While ca . 90% of all bacterial genomes have at least one Restriction Modification system ( Vasu and Nagaraja , 2013 ) , less than 44% of bacterial genomes appear to have a CRISPR system ( Burstein et al . , 2016; Makarova et al . , 2011 ) and these specific defenses are typically absent in obligate symbionts ( Burstein et al . , 2016 ) . It is increasingly documented that both types of phage-defense systems are likely to have fitness costs ( Stern et al . , 2010; Vasu and Nagaraja , 2013; Vale et al . , 2015; Burstein et al . , 2016 ) . These costs may be expressed as slower growth in the absence of phages , somewhat analogous to the costs of autoimmune errors ( Stern et al . , 2010 ) and would be consistent with many bacterial lineages losing CRISPR genes relatively easily ( Burstein et al . , 2016 ) . Lack of exposure explains why intracellular symbionts rarely have phage defenses compared to gut-lumen symbionts with much higher exposure to phages , which would also explain the presence of these systems in EntAcro1 , an abundant gut lumen symbiont ( Sapountzis et al . , 2015 ) . However the presence of these systems may also depend on a general trade-off between maintenance and growth . Preservation by active defense is much more likely to be a naturally selected priority for a vertically transmitted mutualist than for a pathogen selected to infect other colonies at the highest possible rate . Although tissue localization and proximate mechanisms such as the need to maintain chromosomal stability and recombination are important in determining the likelihood of acquisition and loss of phage-defense genes ( Vasu and Nagaraja , 2013 ) , the ultimate evolutionary cost-benefit argument is compelling enough to be spelled out for explicit testing in the future . The endosymbiont-host interactions that we document include several feedback loops that should allow the ants to regulate EntAcro symbiont densities upwards or downwards depending on the overall costs and benefits of their services , similar to other hosts such as aphids which are able to control their intracellular Buchnera symbionts ( Wilkinson et al . , 2007; Russell et al . , 2014 ) . This underlines that phenotypic mechanisms for using symbiont services based on immediate cost-benefit ratios apply for both intra- and extra-cellular symbionts . Providing EntAcro1 and EntAcro10 symbionts with sufficient resources to maintain a full complement of phage-defense systems when they reach high densities would then appear to be a cost-efficient strategy to secure mutualistic services . This is because the only available route for propagation to future generations of a Mollicutes strain is to help maximize the colony’s production of dispersing virgin queens ( Meirelles et al . , 2016 ) . This would be achieved by host-induced optimization of bacterial titers rather than by maximal rates of bacterial cell division , in contrast to commensal or pathogenic bacteria that remain under selection to primarily maximize their rates of horizontal transmission ( Frank , 1996 ) . Comparative experimental tests measuring the phage-attack-sensitivity of intracellular and extracellular symbionts with and without phage defenses could be a way to verify these expectations that are consistent with the results presented here and with general evolutionary theory on levels-of-selection , efficiency of transmission , and the expression of competitive symbiont traits ( e . g . Frank , 2012 ) . A study by Gupta et al . ( 2018 ) that came online while our article was in the final stage of proof-checking performed a phylogenomic analysis of 121 conserved protein sequences in 32 Mollicutes genomes , of which some overlapped with the Mollicutes compared in our analyses . The Gupta et al . study complements the results presented in our paper by: 1 . Confirming that EntAcro1 and EntAcro10 have very different origins , 2 . Confirming that EntAcro10 is the most basal branch of the Entomoplasmacaeae and Spiroplasmatacaeae consistent with our Figure 1 , 3 . Confirming that the closest relative of EntAcro1 is associated with plants – in their study a Mesoplasma lactucae isolated from lettuce corroborating the suggestion that the ancestors of the leafcutter ants may have acquired EntAcro1 from plants on which they foraged , 4 . Showing that citrus-associated Mesoplasma florum belongs to a more derived lineage closer to Mycoplasma strains consistent with our Figure 1 , and 5 . Placing the Mesoplasma lactucae strain in a new genus Edwardiiplasma together with EntAcro1 .
Apterostigma dentigerum and Acromyrmex echinatior colonies were collected in Gamboa , Panama and maintained in rearing rooms at 25 ˚C and 70% relative humidity during a 12:12 hr photoperiod at the Centre for Social Evolution , University of Copenhagen , Denmark . The Acromyrmex colony ( Ae331 ) used in this study had been kept under laboratory conditions for 8 years ( collected in 2007 ) , while the Apterostigma field colonies were all collected in May 2015 , brought to the lab and sampled within the first two weeks from their field collection date . Some aspects of the composition of attine ant microbiota associated with the intestinal system may change after they are reared in the lab for a number of years , but the EntAcro1 and EntAcro10 symbionts are little affected and remain the two dominant Mollicutes symbionts across the attine ants at our field site in Gamboa , Panama also after colonies are transferred to the lab ( Sapountzis et al . , 2015; Zhukova et al . , 2017 ) . Colony Ae331 from which we isolated the EntAcro1 symbiont had been screened previously by 16S-Miseq sequencing and targeted 16S-PCR reactions , which showed its workers had high titers of EntAcro1 and no detectable traces of the possible alternative strains EntAcro2 or EntAcro10 ( Sapountzis et al . , 2015 ) . To obtain a pure sample of EntAcro10 bacteria we used the lower attine ant Ap . dentigerum and performed an initial survey on 20 freshly collected colonies by extracting DNA from whole workers and performing PCR with EntAcro10 specific primers ( Sapountzis et al . , 2015 ) . This showed that workers from one colony ( RMMA150520-03 ) were carrying the EntAcro10 strain without any other Mollicutes being detectable . Prior to the further isolation of the two symbiont strains , a series of Acromyrmex and Apterostigma workers from these two colonies were anesthetized and surface sterilized by submergence in 70% ethanol for 1 min , after which they were rinsed twice in autoclaved MilliQ water , submerged in 50% bleach for 2 min , and rinsed again twice in autoclaved MilliQ water . For the bacterial isolations we used a previously described protocol with some modifications ( Ellegaard et al . , 2013; Iturbe-Ormaetxe et al . , 2011 ) . For EntAcro1 we obtained ca . 50 fecal droplets from Ac . echinatior workers under a laminar flow hood ( Kooij et al . , 2014b ) and deposited them in sterile petri dishes using sterile forceps , after which they were jointly suspended in 1000 µL cold SPG Buffer ( 218 mM sucrose , 3 . 8 mM KH2PO4 , 7 . 2 mM K2HPO4 , 4 . 9 mM l-glutamate , pH 7 . 2 ) and transferred to 1 . 5 ml Eppendorf tubes . To isolate EntAcro10 , we dissected fat body cells from ca . 25 surface sterilized Ap . dentigerum workers under a stereomicroscope and immediately transferred them to a sterile 15 mL glass homogenizer ( Wheaton ) on ice , along with 1000 µL of cold SPG buffer . Using a glass pestle , we disrupted the tissues on ice and immediately transferred them into a new 1 . 5 ml Eppendorf tube . The samples from both ant species were centrifuged at 4 ˚C for 15 min at 3 , 200 g after which the supernatant was transferred to new 1 . 5 ml microcentrifuge tubes and centrifuged again with the same settings . The supernatant was subsequently purified through a 5 µm ( Acrodisc ) and a 2 . 7 µm ( Whatman ) syringe filter , and finally through a 1 . 3 µm ( Acrodisc ) filter before transfer to a new 1 . 5 ml tube followed by centrifugation for 20 min at 18 , 000 g at 4 ˚C . The supernatant was discarded and the pellets ( bacterial cells ) were re-eluted in 5 µl SPG buffer . Approximately , 1 µl was then used for Multiple Displacement Amplification ( MDA ) to obtain whole genomic DNA using the Qiagen REPLI-g Midi Kit following the manufacturer’s instructions . A blank reaction using sterile water as template instead of 1 µl of bacterial cell suspension was included in the same protocol to check for bacterial contaminations with eubacterial 515F/806R primers after the entire procedure was completed , which showed no detectable 16S amplicons . The purified bacterial pellets used for MDA and subsequent dilution of the amplified DNA were subjected to PCR using the 16S generic primers 515F and 806R ( Caporaso et al . , 2012 ) as previously described ( Bourtzis and Miller , 2006 ) , purified using the Invitek kit ( Westberg , Germany ) , and sent to MWG ( Germany ) for Sanger sequencing . After we had confirmed that the 16S amplicons were of Mollicutes’ origin and that the chromatographs showed no signs of other bacterial 16S rDNA sequences , DNA was further purified using the Qiagen mini spin kit following the manufacturer’s instructions . The extracted DNA was then quantified for both the Ac . echinatior and Ap . dentigerum sample using a Nanodrop spectrophotometer and sent to seqIT ( Germany ) where libraries were generated from 100 to 200 ng of DNA using the Nextera XT kit ( Illumina , USA ) . Finally , MiSeq sequencing was performed at 2 times 250 bp read length , which generated approximately 3 , 000 , 000 reads per sample . The Nextera adaptors used for the library construction were removed from the fastq files using Trim Galore ( Babraham Institute ) and the filtered reads were checked with FastQC ( Andrews , 2016 ) . We then used the SPAdes Genome Assembler ( version 3 . 5 . 0 ) to generate a de novo assembly using the ‘--careful’ option which reduced the number of mismatches and short indels before running MismatchCorrector with kmer sizes of 21 , 33 , 55 and 77 to obtain a consensus assembly based on four individual assemblies ( Bankevich et al . , 2012 ) . We then used the Burrows-Wheeler Aligner ( BWA ) to map reads to the assembled contigs ( Li and Durbin , 2009 ) , which produced a SAM file that was further analyzed using SAMTOOLS and converted to a BAM file that could be analyzed with Bamviewer v1 . 2 . 11 ( Carver et al . , 2010 ) . The assembled contigs were further checked for errors using the Reapr v1 . 0 . 18 software ( Hunt et al . , 2013 ) and contigs that had less than 9x coverage or were smaller than 250 bp were removed . The final set of assembled contigs Supplementary file 1 ) was deposited in the NCBI Genome submission portal under accession numbers SAMN06251630 and SAMN06251631 . Genes for each contig were predicted using the RAST annotation server ( Aziz et al . , 2008 ) after which predicted amino acid sequences were compared to a local database Uniref100 using BLASTP v2 . 2 . 28+ ( evalue <1e-15 , percentage identities > 30% ) , and top matches with the assembled contigs were phylogenetically binned ( Supplementary file 1 ) . This grouped the contigs belonging to Mollicutes together in what we refer to as ‘A’ bins allowing further evaluation of the strain-specific RAST annotations ( Simão et al . , 2015 ) . We functionally annotated the protein sequences using a standalone version of InterproScan v-5 . 17–56 . 0 with SUPERFAMILY , Pfam , ProSiteProfiles , Coils , ProSitePatterns , TIGRFAM , Hamap and ProDom ( Jones et al . , 2014 ) and Phobius ( Käll et al . , 2007 ) ( http://www . cbs . dtu . dk/services/SignalP/ and http://phobius . sbc . su . se/ ) to predict signal peptide and transmembrane domains . To identify and compare metabolic pathways we used the KAAS tool ( Moriya et al . , 2007 ) provided by the KEGG database ( Kanehisa and Goto , 2000; Kanehisa et al . , 2010 ) together with the BLAST algorithm and the single best hit ( SBH ) procedure with default settings . For the phylogenomic reconstructions , we first downloaded all 176 available Mollicutes genomes that were present in the Ensembl database ( page accessed in January 2016; ( Yates , 2016 ) ) and created a merged customized BLAST database , which we used to compare the predicted amino acid proteins of EntAcro1A and EntAcro10A . We included only complete genomes and thus excluded the partially sequenced Entomoplasma melaleucae genome . The BLAST comparisons ( Altschul et al . , 1990 ) using an e-value of 1e-15 as cutoff and a percentage identity of 30% , revealed clear similarities between our two EntAcro symbionts and 59 previously genome-sequenced Mollicutes strains . We therefore used their genome sequences to define the orthologous single-copy protein-coding genes using the Orthofinder software ( Emms and Kelly , 2015 ) which resulted in 73 genes being available for phylogenomic analyses . Both the nucleotide and amino acid sequences of these genes were extracted for each of the two EntAcro symbionts , which allowed the construction of gene-specific alignments using MUSCLE v3 . 8 . 31 ( Edgar , 2004 ) . These alignments were subsequently refined using the trimAl software , which removed all positions with gaps in 10% or more of the sequences unless this left fewer than 50% of the original sequences ( Capella-Gutiérrez et al . , 2009 ) . The filtered alignments were further tested for recombination using the Phipack software ( Bruen et al . , 2006 ) and for nucleotide saturation using the Xie test implemented in DAMBE5 ( Xia and Xie , 2001 ) . For the 65 genes that remained ( Supplementary file 2 ) individual alignments were concatenated using Amas 0 . 98 ( Borowiec , 2016 ) and the appropriate substitution models for the nucleotide and protein alignments were selected after testing them with jmodeltest v2 . 1 . 7 and prottest v3 . 4 ( Darriba et al . , 2011 ) . We used the nucleotide and amino acid alignments for the two EntAcro symbionts and the other 59 Mollicutes genomes to reconstruct phylogenomic trees with maximum likelihood ( ML ) and Bayesian methods . For the ML analyses we used the RaxML software v . 8 . 2 . 10 ( Stamatakis et al . , 2008 ) with specified partitions of the concatenated alignments , which produced 65 partitions ( one for each gene ) using the GTR model with invariable sites for the nucleotide alignment and the LG + IG model for the amino acid alignments ( Le and Gascuel , 2008 ) after 1000 bootstrap sampling replications . For the Bayesian inferences , we used the MrBayes v3 . 2 . 6 software ( Ronquist et al . , 2012 ) applying the same nucleotide and amino acid models as above . The concatenated alignments were specified for each gene-specific partition and a variable rate of sequence evolution ( ratepr ) was allowed for each of them . Initially , five chains were run for one million generations and statistical samples were taken every 1000 generations , after which the analyses were repeated to cover a total of 10 million generations , because analysis of the effective sample sizes showed under-sampling in the initial trials . This produced an appropriate deviation of split frequencies ( a standard measure in mrbayes which allows examination of how similar the calculated trees of two independent runs were ) for the nucleotide ( 0 . 0014 ) and protein alignments ( 0 . 00001 ) , with values well below the 0 . 01 threshold recommended as evidence for sufficient convergence and effective sample sizes exceeding 100 in all cases . All trees were further processed using FigTree v1 . 4 . 2 implemented in Geneious R7 . 1 . 1 ( Kearse et al . , 2012 ) . The clipart images used in Figure 1 were either obtained from wikimedia commons ( https://commons . wikimedia . org/ ) , or phylopic ( http://phylopic . org/ ) , available under a Public Domain License or drawn in Adobe Photoshop CS6 . We used the bactNOG database v4 . 5 ( page accessed April 2016 ) to find clusters of orthologous genes and to compare the predicted proteins with HMMER v3 . 1 . b1 ( Eddy , 2011; Huerta-Cepas et al . , 2016 ) . To obtain specific comparisons among Mollicutes genomes based on different numbers of genes assigned to distinct functional categories via bactNOG , we obtained ratio estimates for the number of genes in each functional category by counting the number of genes assigned to specific orthogroups and dividing by the total number of annotated genes ( proportional abundances ) , and used these data as input for Principal Component Analysis ( PCA ) using the ‘stats’ package in RStudio ( v . 1 . 0 . 136 ) . We used Mantel tests to compare phylogenomic distances based on orthologous genes for insect hosts or bacterial symbionts with the overall genome dissimilarities . We focused on the eleven Mollicutes genomes ( and their insect hosts ) that we used for most of our genomic comparisons , because they were closely related . These were EntAcro1 , EntAcro10 , S . sabaudiense , S . diminutum , S . culicicola , S . taiwanense , S . apis , S . melliferum , S . atrichopogonis , S . chrysopicola and S . syrphidicola . We originally also included the S . poulsonii genome but the extremely high number of transposases ( identified as ENOG410907Q , ENOG4105YCW , ENOG4105Y09 , ENOG4105DQ6; Paredes et al . , 2015 ) made this genome a clear outlier ( also in the PCA ordinations presented in Figure 1—figure supplement 3 ) so we excluded it from further comparative analyses . We thus set out to compare 11 Spiroplasma and EntAcro strains with respect to: 1 ) their gene functional categories , 2 ) their bacterial phylogeny , and 3 ) their hosts’ phylogeny . To compare gene functional annotations of the Mollicutes genomes , we used their orthogroup proportional abundances ( see above ) and created a dissimilarity matrix using Euclidean distance in R . For the bacterial phylogeny matrix , we used the phylogenomic distances constructed for this study based on the amino acid sequences of 65 orthologs ( see above and Figure 1 ) . For the genome-based host phylogeny matrix , we used a recent publication that constructed a global phylogeny for the holometabolous insects ( Song et al . , 2016 ) . Whenever possible we used the same host species that the Mollicutes bacteria are associated with ( Apis mellifera , Aedes albopictus , Culex pipiens ) , but when that was not possible we used the closest respresentatives within the same taxonomic clade for which a sequenced genome was available: Simosyrphus grandicornis instead of Eristalis arbustorum , Cydistomyia duplonata instead of Chrysops . sp . , Culicoides arakawae instead of Atrichopogon , and Solenopisis geminata instead of Ac . echinatior and Ap . dentigerum ( Ward et al . , 2015 ) . A host phylogeny for our matrix comparisons was then reconstructed using the selected mitochondrial genome sequences retrieved from NCBI ( Accession numbers available in Supplementary Material of ( Song et al . , 2016 ) using MUSCLE v3 . 8 . 31 ( Edgar , 2004 ) and FastTree v1 . 0 ( Price et al . , 2009 ) . The final Mantel tests were performed in R using 10 , 000 permutations , which produced correlations between: 1 . the functional annotation matrix and the bacterial phylogeny matrix , 2 . the functional annotation matrix and the host phylogeny matrix , and 3 . the host and the bacterial phylogeny matrices . To verify the putative mutualistic functions of the EntAcro1 symbiont , we pursued two experimental approaches . First , we experimentally reduced the abundance of the symbionts in the bodies of ant workers and measured whether this had a negative effect on the acetate uptake activity of ant cells . We achieved this reduction by keeping ants on artificial diets that we knew from pilot experiments were either marginal or outright discouraging for the maintenance of Mollicutes endosymbionts . For these experiments we used four different colonies of A . echinatior: Ae150 , Ae331 , Ae360 and Ae507 , which had been kept in the lab for 15 , 9 , 8 and 5 years respectively , before the experiment . Ant workers were removed from their fungus gardens and placed in sterile petri dishes with an inverted screw-lid in the middle filled with the nutrition-medium of choice . Using the known composition of mango fruit juice ( Medlicott and Thompson , 1985 ) , a resource that leaf-cutting ants at our field site regularly utilize , we created an artificial diet consisting of 140 mM sucrose , 130 . 6 mM fructose , 70 mM glucose and 4 mM sodium citrate ( Sigma aldrich , Denmark ) . This diet was offered to the ants for seven days either in pure form or with 1 mg/ml tetracycline and 1 mg/ml rifampicin added . As controls we used workers picked directly from their fungus-garden . We examined the expression of relevant genes in three type of tissues of ant workers: ( 1 ) approximately 50 fecal droplets from the rectum; ( 2 ) fat body , midgut and part of the ileum tissues of 20 ants; and ( 3 ) heads and thoraxes of 5 ants to represent the remaining body parts as controls . We excluded all samples originating from the heads and thoraxes from further analyses , because we never detected any bacterial gene expression in them except for a single sample ( ftsZ gene expression in colony Ae360 ) . We validated the effect of our diet manipulation by measuring the expression of ftsZ , a single copy bacterial gene that we amplified with qRT-PCR using EntAcro1-specific primers ( Supplementary file 5 ) . To evaluate acetate import activity by ant cells , we measured the expression of an MCT-1 ortholog that has been demonstrated to mediate acetate uptake in animals ( Kirat and Kato , 2006; Moschen et al . , 2012 ) as well as nine other MCT-like genes in the Ac . echinatior genome that are predicted to import multiple short-chain fatty acids ( SCFAs ) potentially including acetate ( www . uniprot . org; Supplementary file 6 ) . To measure the expression of MCT-like genes and the number of EntAcro1 cells ( by measuring expression of the bacterial housekeeping gene ftsZ ) , we only used fat body and midgut tissue samples ( which also included part of the ileum ) , since we could not detect any expression of MCT-like genes in the rectum lumen ( not surprising because the rectum lumen has no ant cells ) . To evaluate the statistical significance of the correlation between expression of each MCT or MCT-like gene and ftsZ , we used non-parametric Spearman correlation tests on the deltadelta CT values . In a second set of experiments using the same type of ant tissues as above , we compared the expression of four predicted EntAcro1 transporter genes ( arginine , GlcNAc , citrate , glycerol or DHA; see Figure 4—figure supplement 1 ) in the midgut and fat body tissues relative to the ( hindgut ) rectum lumen to obtain insight in the metabolic activity of EntAcro1 symbionts throughout the alimentary tract where pH is known to change from ca . neutral to acid ( ca . pH 4 ) conditions just before defecation ( Erthal et al . , 2004 ) . For this experiment , we also sampled 50 ant workers from the same four Ac . echinatior colonies ( Ae150 , Ae331 , Ae360 and Ae507 ) . To test the effect of the gene identity and tissue compartment on the expression data , we used an ANOVA linear model with the deltadeltaCt values ( ratio of target gene expression/ftsZ gene expression ) from the qPCR as response variable and the predicted transporter gene identity ( arginine , citrate , GlcNAc and glycerol/DHA ) , the tissue identity ( midgut/fat body or hindgut ) and their interaction as fixed factors using the function ‘aov’ in R ( Chambers and Hastie , 1992 ) . We evaluated significant differences across groups ( transporter gene and gut compartment ) using post-hoc comparisons as planned contrasts and Bonferroni corrections based on the ‘glht’ function in the ‘multcomp’ package ( Hothorn et al . , 2008 ) . Before performing these tests , we visually examined normality of the data distribution ( 'hist' command ) and confirmed impressions of no significant deviations with Shapiro–Wilk tests , which gave no indications of heteroskedasticity or deviations from normality that would compromise the validity of parametric statistics . Technical procedures in both experiments were as follows: All tissues were dissected and collected in ice-cold RNAlater and stored at −80°C until extraction after which total RNA was extracted using an RNeasy minikit ( Qiagen , Germany ) . Fifty-five ng of extracted RNA was treated with RQ1 RNase-free DNase I ( Promega Corporation , Madison , WI ) , and 50 ng of the resulting product was reverse transcribed with an iScript RT kit ( Bio-Rad Hercules , CA , USA ) to obtain first-strand cDNA . As a negative control , the remainder of the DNase-treated RNA was examined by PCR under the same conditions . All gene-specific PCRs were performed on cDNA , DNase-treated RNA , ant DNA , and water as in previous procedures ( Andersen et al . , 2012; Sapountzis et al . , 2015 ) . For the qPCR reactions we used the SYBR Premix Ex Taq PCR mix ( TaKaRa Bio Inc . , St . Germain en Laye , France ) on the Mx3000P system ( Stratagene , Santa Clara , CA , USA ) . Reactions took place in a final volume of 20 µl containing 10 µl buffer , 8 . 3 µl sterile double-distilled water ( ddH2O ) , 0 . 4 µl of each primer ( 10 mM ) , 0 . 4 µl ROX standard , and 0 . 5 µl template cDNA . PCR conditions were as follows: denaturation for 2 min at 94°C , followed by 40 cycles of 30 s at 94°C , 30 s at the annealing temperature ( see Supplementary file 5 for primers used ) , and 30 s at 72°C , followed by dissociation curve analysis . All quantitative PCRs ( qPCRs ) were replicated and each run included two negative controls with no added template . To generate the delta values for the qPCR analyses , we used the fold change method , based on a standard curve with PCR products in a tenfold dilution series of known concentrations to calculate the PCR efficiency of each primer pair using the REST software ( Pfaffl , 2002 ) . Data were then imported into R and expressed as deltadelta CT values , that is , as fold changes relative to the Ac . echinatior specific rpl7 housekeeping gene for the ant host and the ftsZ gene specific for the EntAcro1 symbiont ( Pfaffl , 2001 ) . Pooled abdominal tissues from five ant workers were collected from five lab colonies of At . colombica , four colonies of At . cephalotes , four colonies of At . sexdens , five colonies of Ac . echinatior , four colonies of Ac . octospinosus , five colonies of T . cornetzi , five colonies of T . zeteki , five colonies of S . amabilis and four colonies of Ap . dentigerum . These samples were supplemented with ca 5–15 whole worker gaster samples ( abdomens minus the first segments that are integrated in the thorax or form the well-known hymenopteran constriction that has no organs ) of attine ant species that were too small to dissect: five lab colonies of C . costatus , three colonies of C . rimosus , three colonies of Myr . ednaella and five colonies of Myc . smithii . All samples were stored in −20°C and DNA was extracted based on previously described methods ( Sapountzis et al . , 2015 ) . We estimated the abundances of Mollicutes symbiont DNA using qPCR with primers targeting the 16S gene of EntAcro1 ( Sapountzis et al . , 2015 ) and EntAcro10 ( Supplementary file 5 ) . For each 16S gene that we analyzed , the initial template concentration was calculated from a standard curve with PCR product in tenfold dilution series of known concentration , as quantified by nanodrop . Since our genomic data showed that one 16S copy corresponded to one EntAcro1 or one EntAcro10 cell , we first calculated the results as numbers of bacterial symbiont cells per ant . We chose not to normalize the data using a specific single copy ant gene ( such as EF-1a used in other studies; Sapountzis et al . , 2015 ) , because samples were from different tissues , that is , either dissected fat bodies and midguts for the ant species with larger body size or whole gasters for the small-bodied species . However , we did obtain data on species-specific fresh-weight body mass of workers and used them to approximately scale the bacterial cell counts . To estimate mean worker body mass per attine ant species , we weighed five random workers from three colonies from each species to obtain the following average values: A . colombica 4 mg , A . cephalotes 4 . 1 mg , A . sexdens 4 . 7 mg , A . echinatior 9 . 7 mg , A . octospinosus 12 . 5 mg , T . cornetzi 1 mg , T . zeteki 1 . 9 mg , S . amabilis 1 . 2 mg , C . costatus 0 . 1 mg , C . rimosus 0 . 4 mg , M . ednaella 0 . 3 mg , M . smithii 0 . 2 mg and A . dentigerum 2 . 2 mg . To compare the abundance levels of EntAcro1 and EntAcro10 per unit of worker body mass , we used a generalized ( negative binomial ) linear model ( GLM ) with the function ‘glm . nb’ in the package ‘MASS’ ( Venables and Ripley , 2002 ) . This model was a better fit than a GLM model with gamma or Poisson distribution when we compared models according to the Akaike Information Criterion ( AIC ) . We used the absolute abundance values ( bacterial counts normalized per unit of ant biomass ) as response variable , and bacterial strain ( i . e . EntAcro1 or EntAcro10 ) , phylogenetic host group ( i . e . leaf-cutting or non-leaf-cutting ants ) , and the statistical interaction between these predictors as fixed categorical variables . We evaluated significant differences across groups using post-hoc comparisons as planned contrasts and Bonferroni corrections based on the ‘glht’ function in the ‘multcomp’ package ( Hothorn et al . , 2008 ) . Foraging substrate preference data were collected in the field from nine attine ant species ( T . cornetzi , T . zeteki , S . amabilis , C . costatus , C . longiscapus , C . rimosus , Myc . smithii , Myr . ednaella and Ap . dentigerum ) in Gamboa , Panama and represent 101 hr of observation time on 103 colonies ( T . cornetzi ( n = 48 ) , T . zeteki ( n = 12 ) , S . amabilis ( n = 6 ) , C . costatus ( n = 8 ) , C . longiscapus ( n = 5 ) , C . rimosus ( n = 8 ) , Myc . Smithii ( n = 14 ) and Ap . dentigerum ( n = 2 ) . Colonies were located and marked in several field sites within lowland Panamanian rainforest near Gamboa after placing polenta baits in the leaf litter and then tracking workers back to their nests when vouchers of workers were collected in EtOH to allow identification . After at least a week , laden returning foragers were observed with a headlamp during set observation periods . Colonies were typically observed during 60 min intervals ( 59 ± 14 min ) , although observations were cut short in the case of rain , or extended in the case of very slow foraging ( e . g . 120 min intervals for Ap . dentigerum ) . Harvested substrates were carefully removed from the mandibles of the workers , collected in eppendorf tubes , and returned to the lab where they were dried at 60°C for 24 h and then sorted under a dissecting microscope and catalogued . Substrates collected by the ants were split in seven categories ( leaves , fruits , flowers , seeds , wood fragments , insect frass and ‘other’ , that is , small dead insect fragments , pieces of unidentifiable detritus , and putative bird feces ) for which counts per observed colony were generated . Similar data were extracted from a recently published study conducted at the same field site at the same time of year ( May ) but focusing on six leaf-cutting ant species ( At . colombica , At . cephalotes , At . sexdens , Ac . echinatior , Ac . octospinosus and A . volcanus ) in Gamboa , Panama ( Kooij et al . , 2014a ) . Merged datasets were normalized by converting all observations to total observed counts in one hour . To visualize foraging substrate preferences across the attine ant species , we converted the count data to proportions and performed an unscaled Principal-Component Analysis ( PCA ) in R using the ‘ade4’ package . To independently verify the statistical significances obtained , we used the mean of the replicated count values per colony . We then fitted the normalized hourly count data in a zero-inflated negative binomial ( ZINB ) regression model with the function ‘zeroinfl’ in the package ‘pscl’ ( Kleiber et al . , 2008 ) . ZINB regression is typically used for true count variables to model positively-skewed data with an abundance of zeros and it fitted our data better than a zero-inflated Poisson or a negative binomial generalized linear model ( GLM ) without zero-inflation when we compared them using the Akaike Information Criterion ( AIC ) and the Vuong's non-nested test ( ‘vuong’ function in ‘pscl’ package ) . We used the absolute itemized foraging substrate preference values as response variable and the interaction of the foraging substrate types ( i . e . leaves , fruits , flowers , seeds , insect frass and other ) and phylogenetic group ( i . e . leaf-cutting versus non-leaf-cutting ants ) as fixed categorical variables . We conducted Tukey’s HSD post hoc tests for each substrate type between leaf-cutting and non-leaf-cutting using the ‘lsmeans’ package ( Lenth , 2016 ) . We used Mantel tests to compare differences in absolute EntAcro1 or EntAcro10 abundances ( calculated for each species using the qPCR data ) with the overall dissimilarities in their foraging substrate preferences . We used only data from the 12 attine ant species which were common in both datasets ( At . colombica , At . cephalotes , At . sexdens , Ac . echinatior , Ac . octospinosus , T . cornetzi , T . zeteki , S . amabilis , C . costatus , C . rimosus , Myc . smithii and Ap . dentigerum ) . We created a bray-curtis distance matrix for each of the EntAcro strains and a similar dissimilarity matrix based on the the seven foraging substrate categories using bray-curtis distance in R . The final Mantel tests were performed in R using 10 , 000 permutations . | Bacteria live inside the gut of most creatures . Some are harmful , some beneficial , and some have no clear effects . Studying the genetic material ( the genome ) of gut bacteria has revealed how they can improve the health , efficiency , and reproductive success of their hosts . For example , studies show that insects with low nutrient diets often benefit from gut bacteria that make vitamins or help them convert food into energy . Panamanian leafcutter ants live in large colonies and farm fungus for food . They harvest leaves to feed their fungus farms and many are major crop pests in Latin America . How they evolved to be so successful is unclear . Recent studies have shown that huge numbers of bacteria called Mollicutes live in the leafcutter ants’ guts . These bacteria do not make the ants sick , so they were thought to be somehow beneficial . Now , Sapountzis et al . show that the two most common types of Mollicutes found in leafcutter ants evolved to make fungus farming more efficient . The complete genomes of two Mollicutes strains were analyzed and compared to the ones found in other insects . The results showed that both types of Mollicutes can turn excess quantities of the amino acid arginine into a nitrogen-rich fertilizer the ants deposit on their fungal gardens as feces . This helps the ants produce more food . One of the two types can also decompose citrate from plant sap and fruit juice consumed by the ants . This produces acetate , which supplements the ants’ fungal diets and provides extra energy . The unique energy-producing Mollicutes may explain why leafcutter ants evolved larger colonies and sustain higher levels of worker activity than other species of fungus-growing ants . The genome data also showed that both types of Mollicutes have costly defense systems to protect themselves against bacteria-destroying viruses . Many bacteria do not invest is such systems , but the Mollicutes may be able to afford them because their ant hosts provide them with plenty of food . This suggests that both the ants and the Mollicutes benefit from their symbiotic relationship . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"evolutionary",
"biology"
] | 2018 | Reconstructing the functions of endosymbiotic Mollicutes in fungus-growing ants |
Cell surface Fc receptors activate inflammation and are tightly controlled to prevent autoimmunity . Antibodies also simulate potent immune signalling from inside the cell via the cytosolic antibody receptor TRIM21 , but how this is regulated is unknown . Here we show that TRIM21 signalling is constitutively repressed by its B-Box domain and activated by phosphorylation . The B-Box occupies an E2 binding site on the catalytic RING domain by mimicking E2-E3 interactions , inhibiting TRIM21 ubiquitination and preventing immune activation . TRIM21 is derepressed by IKKβ and TBK1 phosphorylation of an LxxIS motif in the RING domain , at the interface with the B-Box . Incorporation of phosphoserine or a phosphomimetic within this motif relieves B-Box inhibition , promoting E2 binding , RING catalysis , NF-κB activation and cytokine transcription upon infection with DNA or RNA viruses . These data explain how intracellular antibody signalling is regulated and reveal that the B-Box is a critical regulator of RING E3 ligase activity .
Antibodies are an essential component of protective immunity and their induction is a major aim of vaccination . They mediate a sufficiently rapid inflammatory response upon pathogen re-exposure that infection is halted before host colonization can occur . However , the consequences of inappropriate antibody-induced inflammation are severe , including lethal cytokine storms , autoimmunity and chronic inflammatory disease ( Suntharalingam et al . , 2006 ) . It is therefore essential that antibody immunity is precisely regulated and tightly controlled . Antibody Fc receptors ( FcRs ) are essential for the diverse effector responses mediated by antibodies ( Woof and Burton , 2004 ) . Classical FcRs are expressed on the surface of immune cells . Their signalling induces phagocytosis , antigen presentation , antibody dependent cellular cytotoxicity ( ADCC ) , degranulation and pro-inflammatory cytokine production ( Guilliams et al . , 2014 ) . Regulating FcR signalling controls the antibody response . FcR immune signalling is influenced at multiple levels by; varying cell type expression and abundance , balancing the activity of inhibitory vs activatory receptors , requiring receptor crosslinking , setting minimal activation thresholds and by synergising with other pattern recognition receptors ( PRRs ) ( Vogelpoel et al . , 2015 ) . These diverse mechanisms , deciphered over many decades , operate individually and synergistically to precisely regulate extracellular antibody immunity ( Bruhns and Jönsson , 2015 ) . Recent work from our lab has shown that , in addition to classical FcR immunity mediated at the cell-surface , antibodies exert potent immune function from inside cells via a unique cytosolic IgG receptor called TRIM21 ( James , 2014; Mallery et al . , 2010; McEwan et al . , 2013 ) . TRIM21 is structurally , functionally and evolutionarily distinct from surface FcRs and binds antibodies with a significantly higher affinity ( James et al . , 2007 ) . TRIM21 is also distinct in being constitutively and ubiquitously expressed in most cell types and tissues . TRIM21 protects cells from pathogen infection by stimulating a dual effector and sensor response . Upon infection , antibody-coated pathogens are detected in the cytosol by TRIM21 . TRIM21 recruits cellular degradation machinery , including the AAA ATPase VCP and the proteasome , resulting in the destruction of viral particles and neutralization of infection . TRIM21 simultaneously activates immune transcription pathways , including NF-κB , leading to potent upregulation of pro-inflammatory cytokines including TNF , CXCL10 , IL-6 and IFN ( McEwan et al . , 2013 ) . TRIM21 activation results in cytokine upregulation in mice within hours of infection and is a significant component of the inflammatory response induced by protective antibodies ( Watkinson et al . , 2015 ) . Deletion of TRIM21 compromises the efficiency of protective antibody immunity and causes virus-induced mortality ( Vaysburd et al . , 2013 ) . In contrast to cell-surface FcRs , and despite its capacity for such rapid and potent inflammatory signalling , it is unclear exactly how TRIM21 is regulated . The regulatory mechanisms that exist for classical FcRs are not applicable to TRIM21 . TRIM21 is not expressed at the cell surface and therefore cannot be activated by clustering in lipid rafts or employ this as a mechanism to discriminate monomeric IgG from immune complex . Furthermore , as TRIM21 is dimeric , it is capable of binding both heavy chains of a single IgG simultaneously and forming a 1:1 complex . This is in contrast to other FcRs , which bind asymmetrically to a single heavy chain . There is also no known complementary inhibitory receptor to TRIM21 as exist for surface FcRs . Finally , TRIM21 lacks the ITAM and ITIM motifs that drives the function of other FcRs , and whose signalling can be precisely regulated by recruitment of Src family kinases and inositol-lipid phosphatases ( Vogelpoel et al . , 2015 ) . TRIM21 is a member of a large family of ~100 proteins with a common ‘tripartite motif’ architecture , comprising N-terminal RING , B-Box and coiled-coil domains . Instead of recruiting Src and Syk kinases , TRIM21 stimulates immune signaling via the activity of its RING domain , which catalyses the synthesis of polyubiquitin chains . TRIM21 first recruits the E2 Ube2W to modify itself with an N-terminal monoubiquitin ( Fletcher and James , 2016 ) . This acts as a primer for the synthesis of an anchored K63-chain catalysed by the heterodimer Ube2N/2V2 . TRIM21 is further modified with K48-ubiquitin chains , culminating in proteasomal recruitment . The 19S-resident proteasomal deubiquitinase Poh1 subsequently liberates K63 chains from TRIM21 ( Fletcher and James , 2016 ) , and these then activate signal transduction pathways via TBK1 , TAB/TAK and NEMO . We therefore hypothesized that TRIM21 immune signaling must be regulated at the fundamental level of its ubiquitination activity . Here we show that initiating ubiquitination is a key step in regulating TRIM21 activity and that a novel mechanism of autoinhibition and kinase-induced E2 enzyme recruitment controls its potent pro-inflammatory activity .
To investigate the regulation of TRIM21 we tested whether over-expression in the absence of antibody stimuli triggers NFκB . As a positive control , we compared its activity to that of TRIM5α , a related antiviral TRIM protein whose over-expression is sufficient to drive spontaneous NFκB signalling ( Pertel et al . , 2011 ) . While transfection of full-length TRIM5α caused dose-dependent activation of NFκB , full-length TRIM21 failed to activate NFκB at any dose ( Figure 1A ) . When activated , both TRIM21 and TRIM5α autoubiquitinate , thereby making themselves a degradative substrate for the proteasome . Consistent with the NFκB result , while endogenous TRIM5α underwent rapid recycling we found little evidence of TRIM21 turnover even after several hours ( Figure 1B and Figure 1—figure supplement 1 ) . Together these observations suggest either that TRIM21 is constitutively inactive and its ubiquitination activity tightly regulated or it is a less active E3 ligase than TRIM5α . However , comparing the ubiquitination activity of TRIM5α and TRIM21 isolated RING domains ( T5-R or T21-R ) in vitro revealed that TRIM21 was the more active E3 , capable of forming ubiquitin chains within minutes ( Figure 1C ) . RING dimerization via higher order assembly is the proposed mechanism by which TRIMs autoregulate ( Wagner et al . , 2016 ) . TRIM5α self-assembles and mutations that disrupt RING dimerization inhibit catalytic ability ( Yudina et al . , 2015 ) . We therefore introduced structurally corresponding mutations into TRIM21 ( Figure 1D and Figure 1—figure supplement 2 ) to test whether RING dimerisation is required for activity . SEC MALS and AUC demonstrated that while the wild-type RING is in monomer-dimer equilibrium , mutants M10E and M10E/M72E do not undergo detectable dimerization ( Figure 1E and F ) . However , both mutants remained active and could catalyse unanchored K63 chain synthesis , Ube2W-primed anchored K63 chain extension and ubiquitin discharge ( Figure 1G–I ) , albeit less efficiently than wild-type . These results demonstrate that a pre-formed TRIM21 RING dimer is not a pre-requisite for ubiquitination activity . RING dimerization is not required for E2 enzyme binding , as all interactions occur within a single monomer ( Yudina et al . , 2015 ) . Instead , dimerization of RING E3s is thought to be needed for additional contacts with the E2-charged ubiquitin . Mutations in the partner RING at this interface reduce TRIM25 catalytic activity in vitro ( Figure 2A ) ( Sanchez et al . , 2016 ) . Equivalent mutations in TRIM21 had no impact on ubiquitination , confirming that RING dimerization is not an intrinsic requirement for TRIM21 activity ( Figure 2B–D ) . The mutation E10R reduces ubiquitination by TRIM25 , possibly by disrupting a hydrogen bond with ubiquitin ( Koliopoulos et al . , 2016 ) . We hypothesised that the corresponding glutamate ( E13 ) in TRIM21 may allow a monomeric RING to contact both E2 and bound ubiquitin simultaneously , obviating the requirement for dimerization . Consistent with this hypothesis , E13R reduced TRIM21 catalysis ( Figure 2D ) . Taken together these data show that the monomeric TRIM21 RING is a highly active E3 . Thus , the lack of constitutive TRIM21 NFκB activity must be the result of tight regulation , independent of higher order oligomerisation . The next domain in the tripartite motif of TRIM21 after the RING is the B-Box2 ( hereafter referred to as ‘B-Box’ ) , a domain largely unique to TRIM proteins but whose function ( besides an ability to oligomerise in TRIM5α ) is unknown . Comparing the activities of RING , RING-Box and MiniTRIM21 proteins ( based on a monovalent MiniTRIM5 that preserves the Box-coiled coil interface ( Wagner et al . , 2016 ) ) , revealed that the presence of the B-Box significantly reduces ubiquitination ( Figure 3A ) . This is due to direct inhibition of catalytic activity , as the B-Box prevented ubiquitin discharge from either Ube2D1 or Ube2N ( Figure 3B ) . To understand how the B-Box modulates TRIM21 catalysis we solved a 2 Å crystal structure of the RING-Box protein ( Supplementary file 1 ) . Surprisingly , this revealed that the B-Box occupies the E2 binding site on the RING domain ( Figure 3C ) . Comparison to the TRIM5α RING:E2 complex shows that the B-Box is located where helix 1 of the E2 would normally be positioned during RING binding ( Figure 3D ) . Crucially , the B-Box is able to occupy the E2 binding site on the RING by acting as an E2 mimic , forming similar electrostatic interactions . The RING surface contains a negatively charged patch , which is complementary to a strongly positively charged patch on both the E2 and the B-Box ( Figure 3E ) . In particular , Box-RING interaction is stabilized through a salt bridge between residues R118 and E12 , which is analogous to the interaction between E2 residue R14 and RING residue E11 in the TRIM5:Ube2N structure ( Figure 3D ) . To test directly whether the B-Box prevents E2 binding we assigned 15N HSQC spectra of monomeric ( M10E ) RING and RING-Box constructs and measured chemical shift perturbations ( CSPs ) upon titration with Ube2N ( Figure 4 and Figure 4—figure supplement 1 ) . Titration of Ube2N into the RING resulted in significant CSPs in E2 interface residues ( denoted by the green shaded regions in Figure 4A ) . In contrast , no evidence of Ube2N binding to the RING-Box was observed . Taken together , these data suggest that the B-Box may be an autoinhibitory domain whose function is to regulate RING activity by preventing E2 recruitment . Next , we investigated what relieves B-Box inhibition and promotes TRIM21 ubiquitination activity . Recently it was shown that unrelated innate immune adaptors including MAVS , STING and TRIF all contain a short pLxIS motif ( where p is hydrophilic , x is non-aromatic and S is the target serine ) , which recruits kinases IKKβ or TBK1 resulting in serine phosphorylation and signal potentiation ( Liu et al . , 2015 ) . Remarkably , TRIM21 contains a very similar motif at the end of its RING domain . Moreover , the target serine in this motif is located at the centre of the B-Box:RING interface ( Figure 5A ) and undergoes phosphorylation in cells ( Figure 5B ) . We hypothesized that this could provide a mechanism to activate TRIM21 during an immune response . To investigate this further , we raised specific antisera against the phosphoserine peptide 67RQLANMVNNLKEISQ81 ( Figure 5C ) . Using this anti-pS80 serum , we detected cellular phosphorylation of TRIM21 but not an S80A mutant , upon IKKβ overexpression ( Figure 5D ) . To ensure that IKKβ was acting directly rather than inducing the expression of a second kinase via NFκB , we repeated the experiment in the presence of proteasome inhibitor Bortezomib , which inhibits the degradation of IκBα and blocks NFκB activation . Efficient IKKβ phosphorylation of TRIM21 was observed even under conditions where NFκB signaling was abolished ( Figure 5E ) . To confirm that IKKβ directly phosphorylates TRIM21 , we incubated recombinant RING-Box and IKKβ proteins in vitro , observing ATP-dependent IKKβ phosphorylation of TRIM21 that was lost upon mutation of motif residues or addition of an IKKβ inhibitor ( Figure 5F–H ) . Robust IKKβ-mediated phosphorylation was also observed using full-length TRIM21 protein ( Figure 5I ) . Finally , knock-out of IKKβ by CRISPR/Cas9 abolished modification of overexpressed TRIM21-His ( Figure 5J ) , suggesting that endogenous kinase phosphorylates TRIM21 . Previously , it has been shown that while MAVS is phosphorylated by IKKβ and TBK1 , STING and TRIF are only phosphorylated by TBK1 suggesting that there is differential kinase dependence even though signaling adapters share the consensus pLxIS motif ( Liu et al . , 2015 ) . To test whether TRIM21 is a substrate for TBK1 phosphorylation , we incubated the recombinant full-length protein with the kinase in the presence of ATP . TBK1 was capable of phosphorylating TRIM21 in vitro comparably with IKKβ ( Figure 6A ) . Overexpressed TBK1 was also capable of strongly phosphorylating cellular TRIM21 and causing significant stabilization of the protein ( Figure 6B ) . To overcome limitations of kinase overexpression we sought to stimulate TRIM21 phosphorylation by activing the upstream pathway . First , we stimulated the TBK1 pathway by expressing one of its upstream adaptors , MAVS . MAVS expression significantly increased phosphorylation of TRIM21 ( Figure 6C ) . Second , we stimulated cells with the MDA5/RIG-I ligand poly ( I:C ) ( Kato et al . , 2008 ) . We observed a rapid increase in TRIM21 phosphorylation upon poly ( I:C ) stimulation , consistent with TBK1 activation ( Figure 6D ) . Finally , we monitored changes in TRIM21 phosphorylation following infection with antibody-coated adenovirus , under conditions where we have previously demonstrated TRIM21-dependent antiviral activity ( McEwan et al . , 2013 ) . The data show that in equivalent levels of immunoprecipitated TRIM21 there is a significant increase in the proportion of phosphorylated TRIM21 ( Figure 6E ) . We next investigated whether serine phosphorylation can act as a switch to relieve B-Box inhibition and activate TRIM21 . To test this , we introduced the phosphomimetic mutation S80E into the RING-Box and measured its ability to catalyse ubiquitin discharge . Remarkably , introduction of S80E into the RING-Box was sufficient to completely restore catalytic activity to the same level as the RING alone ( Figure 7A–B ) . While glutamate is a routinely used phosphomimetic it does not fully recapitulate a phosphoserine . We therefore utilized an evolved orthogonal aminoacyl-tRNA synthetase/tRNACUA pair ( Rogerson et al . , 2015 ) to incorporate phosphoserine co-translationally at position 80 in MiniTRIM21 protein . Comparison with wild-type protein revealed that S80 phosphorylation potentiates TRIM21 ubiquitination activity ( Figure 7C ) . Phosphorylation-dependence was confirmed by treating the phosphoprotein with phosphatase , which dephosphorylated S80 and restored autoinhibition ( Figure 7C ) . Taken together , this show that modification of S80 is sufficient to control ubiquitination activity of TRIM21 in the presence of the B-Box . Moreover , mutant S80E can be used as a functional mimic of serine phosphorylation at position 80 . The above data suggest that TRIM21 is kept in a constitutively inactive state by the B-Box domain , which prevents E2 binding and whose autoinhibition is released upon phosphorylation . To test this , we compared NFκB activation by TRIM21 mutants S80E or S80A with TRIM5α . Neither S80A nor wild-type TRIM21 triggered NFκB , but S80E conferred robust activation ( Figure 8A ) . In contrast , mutation of residue S80 did not alter virus neutralization ( Figure 8B ) . This is consistent with previous data that TRIM21 signalling has an activation threshold whereas neutralization does not ( Foss et al . , 2016 ) . To demonstrate the importance of TRIM21 phosphorylation in immune sensing during infection , we challenged cells expressing different TRIM21 mutants with human adenovirus ( Adv ) ± human serum IgG and measured TNFA transcription after 4 hr . Significant TNFA induction was only observed during infection in the presence of antibody , and not in TRIM21 knockout cells ( K21 , Figure 8C and Figure 8—figure supplement 1 ) . Normal TNFA induction was restored in K21 cells by TRIM21 overexpression , confirming TRIM21-dependence . Importantly , expression of S80A failed to rescue TNFA induction , while S80E enabled TNFA induction beyond wild-type levels . A change in TRIM21 phosphorylation upon infection could not be detected but this may be because only a fraction of cellular TRIM21 is recruited and modified during the response . This would be consistent with previous data showing that TRIM21 ubiquitination and degradation is also undetectable , despite being required for activity ( Mallery et al . , 2010 ) . To confirm the importance of S80 and B-Box inhibition in regulating TRIM21 immune signaling , we repeated our infection experiments using an unrelated RNA virus , human rhinovirus 14 ( HRV ) . As with AdV , expression of wild-type TRIM21 in knockout cells increased TNFA induction upon infection with HRV + antibody ( Figure 8D ) . Mutant S80E greatly potentiated TNFA transcription while S80A failed to reconstitute TRIM21 activity . Similar results were obtained for cytokines CXCL10 and IL6 and the interferon-stimulated gene IFIT1 . To obtain direct evidence that S80 phosphorylation activates TRIM21 by displacing the B-Box domain , we compared the E2 binding of wild-type and S80E RING-Box proteins by NMR . In contrast to wild-type , we detected significant CSPs in canonical E2 binding site residues upon titration of Ube2N into S80E ( Figure 9A ) . There were also changes in residues at the base of helix 2 , within the linker and on B-Box residues that contact the RING . This is consistent with restoration of E2 binding and the stabilization of an active conformation in which the B-Box domain is displaced . To further demonstrate B-Box displacement , we compared the dynamics of wild-type and S80E proteins . The {1H}15N NOE ratios reveal that the most dynamic region is the linker ( residues 83–102 ) between the RING and B-Box ( Figure 9B ) . This is in agreement with the higher temperature factors observed for these residues in the crystal structure ( Figure 9—figure supplement 1 ) . Importantly , mutant S80E has markedly lower enhancement factors both in the linker region and in helix 1 ( residues 1–17 ) , against which the B-Box packs ( Figure 9B ) . This increase in local motion is consistent with reduced interaction between RING and B-Box in the S80E mutant . To demonstrate that B-Box displacement promotes access to the E2 binding site on the RING , we carried out paramagnetic relaxation enhancement experiments ( solvent PREs ) . In these experiments , addition of the paramagnetic metal complex Gd ( DTPA–BMA ) selectively attenuates signals from solvent-exposed surfaces ( Pintacuda and Otting , 2002 ) . The S80E mutation caused a dramatic change in surface accessibility precisely localized to the RING-Box interface ( Figure 9C and Supplementary file 2 ) . These data confirm that phosphomimetic mutation at serine 80 disrupts the RING:Box interaction and uncovers the E2 binding site .
Our results suggest intracellular antibody immune signalling is kept in check via tight regulation of the cytosolic antibody receptor TRIM21 . TRIM21 activates immune transduction pathways upon sensing an antibody-coated pathogen by synthesizing K63-ubiquitin chains . Controlling synthesis of these chains allows TRIM21 signalling to be regulated at the fundamental level . Here we have shown that this is accomplished by keeping TRIM21 in a constitutively silent state through an autoinhibitory mechanism that blocks E2 enzyme recruitment . Autoinhibition is mediated by the B-Box , providing a function for this ubiquitous but hitherto enigmatic TRIM protein domain . The B-Box inhibits E2 recruitment by acting as an E2 mimic and competing for RING binding . We also show that TRIM21 contains a variant of an IKKβ and TBK1 phosphorylation motif and is phosphorylated at residue S80 . Phosphorylation of S80 or its replacement by a phosphomimetic displaces the B-Box , allowing E2 recruitment and potentiating TRIM21 ubiquitination activity , NF-κB signalling and cytokine transcription upon infection with DNA or RNA viruses . Although TRIM21 regulation is seemingly distinct from that of cell-surface FcRs , parallels exist . For instance , signalling by both types of antibody receptor involves regulation by kinases . In the case of TRIM21 , this involves a pLxxIS motif ( where p is hydrophilic ) and the kinases IKKβ and TBK1 while in other FcRs this involves an YxxL/I motif and recruitment of Src kinases . Importantly , it is not clear for either receptor type exactly how kinase activation occurs . For surface FcRs , this is thought to involve receptor clustering via cross-linking but why this causes Src recruitment and activation is unclear ( Huang et al . , 1992 ) . TRIM21 may also undergo clustering , via cross-linking or higher order assembly , when it is recruited to opsonized pathogens . This may provide a basal activity that is amplified by kinase phosphorylation . The co-option of IKKβ and TBK1 as amplifying kinases suggests that TRIM21 could participate in its own positive-feedback loop . Again there are parallels to surface FcRs , where tyrosine phosphorylation has been shown to promote receptor clustering , which promotes phosphorylation and so on ( Sobota et al . , 2005 ) . Both types of Fc receptor also synergise with other PRRs and immune signalling pathways to amplify and diversify the resulting inflammatory response . Surface FcRs are known to exhibit cross-talk with Toll-like receptors ( TLRs ) , greatly amplifying the production of cytokines such as TNFα during phagocytosis of opsonized bacteria ( den Dunnen et al . , 2012; Vogelpoel et al . , 2014 ) . Similarly , TRIM21 has been shown to synergise with cytosolic nucleic acid receptors such as RIG-I and cGAS , resulting in pathogen-specific responses and multiple signalling waves ( Watkinson et al . , 2015 ) . Whether there is direct collaboration between TRIM21 and surface FcRs is currently unknown and will be important to determine in future work . Both immune signalling and antigen processing functions of TRIM21 are well placed to synergise with FcR immune complex capture and sorting activities , for instance to promote class I presentation . Precedence exists in the form of collaboration between FcRn and surface FcRs to drive antigen processing by dendritic cells ( DCs ) ( Baker et al . , 2011; Qiao et al . , 2008 ) . TRIM21 belongs to a large family of structurally related proteins , many of which have known roles in cell signalling . They include other proteins with roles in immune signalling , such as the antiretroviral TRIM5 ( 14 ) and RIG-I activator TRIM25 ( Gack et al . , 2007 ) . While progress has been made in understanding the structure and function of TRIMs , it is still unclear how TRIM proteins are activated and regulated . Existing models of activation were solely based on substrate-induced aggregation or higher order assembly . This was thought to be necessary for dimerization of the RING domains located at either end of the molecule , which was assumed to be a pre-requisite for ubiquitination activity ( Koliopoulos et al . , 2016 ) . However , our data show that this is not the case for TRIM21 , and so cannot be assumed for other members of the TRIM family . Furthermore , higher order assembly alone does not explain the discrepancy between in vitro and cellular TRIM ubiquitination activity . TRIM RINGs , including those of TRIM5 , 25 and 32 , are active in vitro at concentrations where they are monomeric ( Yudina et al . , 2015; Sanchez et al . , 2016; Koliopoulos et al . , 2016 ) . Given that TRIMs are not constitutively active in cells , there must be another level of regulation . It has been proposed that binding of a ubiquitin-conjugated E2 may drive RING equilibrium towards the dimeric form , explaining the in vitro activity . However , this would not explain the absence of activity in cells where the same phenomenon would occur . Moreover , if E2-Ub can drive TRIM RING dimerization then a pre-existing dimer is by definition no longer a pre-requisite for function and higher order assembly cannot be the sole regulator . It seems likely that the activation and regulation of TRIMs is a complex process that is regulated at multiple levels . This is exemplified by TRIM21 whose neutralization and signalling functions are both mediated by ubiquitination , but only signalling requires release of B-Box inhibition . Previously we have observed that TRIM21 signalling has a distinct threshold whereas neutralization takes place at very low antibody occupancy ( <2 antibodies ) ( McEwan et al . , 2012 ) and under conditions of sub-optimal TRIM21:antibody and antibody:antigen kinetics ( Foss et al . , 2016; Bottermann et al . , 2016 ) . Therefore , basal activity via higher order assembly may be sufficient for neutralization but not signalling . Strict regulation of signalling is intuitive , as it ensures a proportional response and avoids inappropriate triggering . Setting activation thresholds and regulating signalling may mitigate against any endosomal leakage of antibodies into the cytosol . The activation of TRIM21 by leaked immune complexes in lysosome-maturation-defective macrophages of lupus-prone mice illustrates the consequences of such an event ( Monteith et al . , 2016 ) . However , while the requirement for tight regulation of TRIM21 signalling is intuitive the mechanism behind this has remained elusive . Our data suggest that it is achieved by a novel autoinhibitory mechanism in which the B-Box domain prevents ubiquitination by acting as an E2 mimic and the kinases IKKβ and TBK1 promote ubiquitination by phosphorylating the RING to relieve B-Box inhibition . Use of innate immune kinases intimately associates TRIM21 with the antiviral state not only at the level of gene expression as an ISG but also post translationally and in a similar manner to key immune adaptors MAVS , STING and TRIF ( Liu et al . , 2015 ) .
Ube1 , Ube2N , Ube2V2 , Ube2D1 , TRIM21RING-Box ( residues 1–129 ) were produced as previously described ( Fletcher et al . , 2015 ) . TRIM21RING ( residues 1–85 ) , TRIM5αRING ( residues 1–88 ) and Ube2W were sub-cloned into pOP-TG for expression with a TEV cleavable GST tag . Point mutations were introduced using the site-directed mutagenesis protocol of Liu and Naismith ( Liu and Naismith , 2008 ) . Gibson cloning was used to remove the N-terminal GST tag and introduce a non-cleavable C-terminal STREP tag ( Gibson et al . , 2009 ) . The miniTRIM21 construct was designed using the strategy of ( Wagner et al . , 2016 ) . Residues 225–265 of a bacterial seryl-RNA synthetase , encoding an anti-parallel coiled-coil hairpin was inserted between two regions of the TRIM21 N-terminus , including residues 1–154 ( RING-B-Box-coil ) and residues 221–260 ( coil ) . The resulting gene was cloned into pOP-TG for expression with a TEV cleavable GST tag at the N-terminus . The Escherichia coli strain C41 , was used for recombinant protein expression . Cells were grown in 2TY media supplemented with 100 µg/mL ampicillin at 37°C until an OD600 of 0 . 7 . The cells were induced with 1 mM IPTG and incubated over-night at 18°C . For expression of TRIM proteins , 1 mM IPTG was included upon induction . After harvesting , cells were resuspended in 50 mM Tris pH 8 , 150 mM NaCl , 10 μM ZnCl2 , 1 mM DTT , 20% ( vol/vol ) Bugbuster ( Novagen ) and cOmplete protease inhibitors ( Roche , Switzerland ) . The cell suspension was lysed by sonication , clarified by centrifugation at 18000 rpm for 40 min and passed over glutathione-sepharose resin ( GE Healthcare ) . The fusion proteins were cleaved with TEV protease overnight at 4°C and were left with an N-terminal scar consisting of the tripeptide , Gly-Ser-His . Size exclusion chromatography was carried out on a High load 26/60 Superdex 75 prep grade column ( GE Healthcare ) as final purification step . All proteins were stored in 50 mM Tris pH 8 , 150 mM NaCl , 1 mM DTT . Isotopically labeled TRIM21RING-M10E , TRIM21RING-Box-M10E , TRIM21RING-Box-M10E/M72E proteins were produced in the Escherichia coli strain Rosetta 2 ( DE3 ) . Cells were grown in K-MOPS minimal medium supplemented with 20 mM 15NH4Cl and/or 0 . 4% [13C]-glucose , 4 mM K3PO4 pH 8 . 0 , vitamin mix ( 0 . 1% Thiamine , 0 . 02% each d-Biotin , Choline chloride , Folic acid , Niacinamide , D-pantothenic acid , Pyridoxal , 0 . 002% Riboflavin ) and 100 µg/ml pantothillin . Expression and purification conditions were as for unlabelled proteins , described above . Polyacrylamide gel slices ( 1–2 mm ) containing the proteins were prepared for mass spectrometric analysis using the Janus liquid handling system ( PerkinElmer , UK ) . Briefly , the excised protein gel pieces were placed in a well of a 96-well microtitre plate and destained with 50% v/v acetonitrile and 50 mM ammonium bicarbonate , reduced with 10 mM DTT , and alkylated with 55 mM iodoacetamide . After alkylation , proteins were digested with 6 ng/μL trypsin ( Promega , UK ) overnight at 37°C . The resulting peptides were extracted in 2% v/v formic acid , 2% v/v acetonitrile . Digests were analysed by nano-scale capillary LC-MS/MS using an Ultimate U3000 HPLC ( ThermoScientific Dionex , San Jose , USA ) to deliver a flow of approximately 300 nL/min . A C18 Acclaim PepMap100 5 μm , 100 μm x 20 mm nanoViper ( ThermoScientific Dionex , San Jose , USA ) , trapped the peptides prior to separation on a C18 Acclaim PepMap100 3 μm , 75 μm x 250 mm nanoViper ( ThermoScientific Dionex , San Jose , USA ) . Peptides were eluted with a 60 min gradient of acetonitrile ( 2% to 80% ) . The analytical column outlet was directly interfaced via a nano-flow electrospray ionisation source , with a hybrid quadrupole orbitrap mass spectrometer ( Q-Exactive Plus Orbitrap , ThermoScientific , San Jose , USA ) . Data dependent analysis was carried out , using a resolution of 30 , 000 for the full MS spectrum , followed by ten MS/MS spectra . MS spectra were collected over a m/z range of 300–2000 . MS/MS scans were collected using a threshold energy of 27 for higher energy collisional dissociation ( HCD ) . LC-MS/MS data were then searched against a protein database ( UniProt KB ) using the Mascot search engine programme ( Matrix Science , UK ) ( Perkins et al . , 1999 ) . Database search parameters were set with a precursor tolerance of 10 ppm and a fragment ion mass tolerance of 0 . 8 Da . One missed enzyme cleavage was allowed and variable modifications for oxidized methionine , carbamidomethyl cysteine , pyroglutamic acid , phosphorylated serine , threonine and tyrosine , ubiquitylation at lysine and the N-terminus were included . MS/MS data were validated using the Scaffold programme ( Proteome Software Inc . , USA ) . All data were additionally interrogated manually . Samples of TRIM21RING , and mutants TRIM21RING-M10E and TRIM21RING-M10E/M72E at concentrations of 7 . 5 mg/ml were subjected to velocity sedimentation at 50 , 000 rpm at 20 ˚C in 50 mM Tris-HCl , pH 8 . 0 , 150 mM NaCl , 1 mM DTT using 12 mm double sector cells in an An50Ti rotor using an Optima XL-I analytical ultracentrifuge ( Beckmann ) . The sedimentation coefficient distribution function , c ( s ) , was analyzed using the SEDFIT program , version 14 . 0 ( Schuck , 2003 ) . The partial-specific volumes ( v-bar ) , solvent density and viscosity were calculated using Sednterp ( Dr . Thomas Laue , University of New Hampshire ) . Size-exclusion chromatography ( SEC ) was performed with inline multi-angle laser light scattering ( MALLS ) using a Wyatt HELEOS-II 18-angle photometer coupled to a Wyatt Optilab rEX differential refractometer ( Wyatt Technology Corp ) . Samples of 100 µL were injected at 19 . 5 mg/mL and separated over a Superdex 75 10/300 GL ( GE Healthcare ) equilibrated in 50 mM Tris , pH 8 , 150 mM NaCl , 1 mM DTT . In vitro ubiquitination reactions were carried out in 50 mM Tris at pH 8 , 2 . 5 mM MgCl2 , 0 . 5 mM DTT with 0 . 2 mM Ub , 2 mM ATP , 1 µM Ube1 , 0 . 5 µM Ube2N , 0 . 5 µM Ube2V2 1 . 5 or 1 µM TRIM21 RING-Box or RING protein respectively . The reaction was started upon incubation at 37°C for the time points indicated in the text . The reaction was stopped via the addition of LDS sample buffer at 4°C , followed by boiling at 90°C for 2 min . To observe the synthesis of TRIM21-anchored ubiquitin chains , 1 . 5 µM Ube2w was included in the reaction . The reactions were resolved by LDS-PAGE and TRIM21 or ubiquitin were detected by immunoblot using anti-TRIM21 [raised against human TRIM21RING-Box or TRIM21RING] , or anti-Ub-HRP ( Santa Cruz , sc8017-HRP P4D1 , 1:1 , 000 ) . To prevent auto-ubiquitination , lysine 92 of Ube2N was replaced with arginine ( McKenna et al . , 2001 ) . Ube2NK92R or Ube2D1 were loaded with ubiquitin by mixing 40 µM of the E2 with 1 µM Ube1 , 0 . 37 mM Ub and 3 mM ATP in 50 mM HEPES pH 7 . 5 , 150 mM NaCl , 20 mM MgCl2 , and incubating the reaction at 37°C for 30 min . The reaction was transferred to ice and used immediately . To observe E3 mediated discharge of ubiquitin , 2 µM ubiquitin loaded E2 was mixed with 1 . 5 µM TRIM21RING-Box or TRIM21RING in 50 mM HEPES pH 7 . 5 , 150 mM NaCl , 20 mM MgCl2 , 50 mM L-lysine . For discharge of ubiquitin from ubiquitin loaded Ube2N , 2 . 5 µM Ube2V2 was added to the reaction . Samples were taken at the time points indicated in the text and mixed immediately with LDS sample buffer at 4°C . The samples were boiled for exactly 20 s , resolved by LDS-PAGE and observed by immunoblot using anti-Ube2D ( Boston Biochem , A-615 , 1:1 , 000 ) or anti-Ube2N ( Bio-Rad , AHP974 , 1:1 , 000 ) . For Ube2N immunoreactivity was observed by near-infrared detection ( Odyssey , LI-COR ) . Gibson cloning was used to sub-clone the TRIM21RING-Box and MiniTRIM21 sequences into the pNHD1 . 3 plasmid ( Rogerson et al . , 2015 ) with the addition of a non-cleavable C-terminal Strep tag ( GSWSHPQFEK ) . An S80TAG mutation introduced using the site-directed mutagenesis protocol of Liu and Naismith to create pNHD-T21RB ( S80TAG ) and pNHD-T21MT ( S80TAG ) ( Liu and Naismith , 2008 ) . Phosphorylated proteins were expressed in BL21 ΔserB ( DE3 ) transformed with pKW2 EF-Sep and pNHD-T21RB ( S80TAG ) or pNHD-T21MT ( S80TAG ) . Cells were grown at 37°C in TB media supplemented with 25 µg/mL chloramphenicol and 25 μg/ml tetracycline . At OD600 = 0 . 6 , the cells were induced with 1 mM IPTG and 2 mM phosphoserine and incubated overnight at 18°C . Cells were harvested by centrifugation and resuspended in 100 mM Tris pH 8 , 150 mM NaCl , 2 mM DTT , 10 μM ZnCl2 , 20% ( vol/vol ) Bugbuster ( Novagen ) and Complete protease inhibitors ( Roche ) . The cells were lysed by sonication and centrifuged ( 18 000 rpm , 40 min , 4° C ) to remove insoluble material . The soluble lysates were applied to 5 mL of StrepTactin Sepharose High Performance resin ( GE Healthcare ) that had been pre-equilibrated in wash buffer ( 100 mM Tris , pH 8 , 150 mM NaCl , 1 mM DTT ) . The resin was washed in 10 column volumes of wash buffer and the protein was eluted in wash buffer with the addition of 2 . 5 mM D-desthiobiotin . Pooled protein fractions were subjected to size exclusion chromatography over a High load 26/60 Superdex 75 prep grade column ( GE Healthcare ) equilibrated in 50 mM Tris pH 8 , 150 mM NaCl , 1 mM DTT . Incorporation of phosphoserine into the recombinant protein was confirmed by ESI-MS or LC-MS/MS and western blot . Phosphorylated miniTRIM21 or the unmodified control were mixed with 400 U of Lambda protein phosphatase and buffered with 1 X NEBuffer ( 50 mM HEPES , 100 mM NaCl , 2 mM DTT , 0 . 01% Birj 35 , pH 7 . 5 @ 25°C ) , supplemented with 1 mM MnCl2 . The reactions were incubated at 30°C for up to 30 min with samples taken at the times indicated in Figure 4 . Dephosphorylation was confirmed by immunoblot using anti-TRIM21 and pS80 sera In vitro phosphorylation reactions were carried out in 10 μL reactions with 50 mM Tris pH 7 . 4 , 10 mM MgCl2 , 0 . 5 mM DTT , 1 mM ATP , 430 ng IKKβ ( Life Technologies ) , 100 or 400 ng TBK1 ( Promega ) , 1 μM TRIM21RING-Box or LipoylTRIM21 , incubated at 37°C for 2 hr , then quenched by addition of LDS sample buffer and boiling at 95°C for 5 min . Samples were resolved by LDS-PAGE and TRIM21 detected by immunoblot using anti-TRIM21 ( clone D12 , Santa Cruz ) or pS80 sera at 1:10000 and 1:1000 dilutions , respectively . 106293 T cells were transfected with 300 ng TRIM21-His ( WT or S80A ) and 3 μg empty vector or IKKβ ( S177E/S181E ) . 48 hr later cells were washed , pelleted , denatured in 500 μL 6 M GuHCl , 0 . 1 M Na2HPO4/NaH2PO4 ( pH 8 ) , 10 mM Imidazole ( pH 8 ) and samples rotated for 3 hr at room temperature with 30 μL equilibrated NiNTA agarose ( Qiagen ) . The agarose matrix was washed twice with 500 μL lysis buffer , twice with 500 μL 3:1 wash buffer:lysis buffer , once with 500 μL wash buffer ( 25 mM Tris , 20 mM imidazole pH 6 . 8 ) , resuspended in 2 × LDS sample buffer supplemented with 300 mM Imidazole to elute bound His-tagged proteins and heated for 5 min at 95°C before LDS-PAGE . For assaying IKKβ activity in the absence of NF-κB signaling , 106293T were transfected with 500 ng pGL4 . 32[luc2P/NF-κB-RE/Hygro] , 200 ng TRIM21-His and 2 μg empty vector or IKKβ . 6 hr later , cells were treated with DMSO or 30 nM bortezomib for 16 hr . Cells were washed , resuspended in 500 μL cold PBS . 20 μL cells were mixed with 100 μL SteadyLite ( Perkin Elmer ) and luminescence measured ( Pherastar ) . The remaining cells were pelleted , denatured and mixed with NiNTA agarose as described above . For TBK1 and MAVS transfection experiments , 200 ng TRIM21-His was transfected with 2 μg empty vector or TBK1 or MAVS . Cells were harvested at 24 hr post transfection . 293 T cells were seeded in 24 well plates 24 hr before transfection . Each well was transfected with , typically , 5 ng TRIM cDNA , with 10 ng pGL4 . 32[luc2P/NF-κB-RE/Hygro] ( NF-κB response element-dependent firefly luciferase ) and 5 ng pRL-TK ( thymidine kinase promoter-dependent Renilla luciferase ) . 24 hr post transfection , cells were lysed in Passive Lysis Buffer and sequential firefly and renilla luminescence measured ( BMG Pherastar plate reader ) , according to manufacturers instructions ( Promega ) . Firefly luciferase luminescence was normalised to Renillia luciferase luminescence , and these values normalised to those of empty vector transfected cells . Single-guide RNA ( sgRNA ) against IKKβ ( GCTGACCCACCCCAATGTGG ) was incorporated into the Lenti CRISPR v2 plasmid ( Addgene ) and VSV-G pseudotyped lentiviral particles were generated by three-plasmid transfection of 293T with Fugene-6 ( Promega ) , using 1 μg HIV-1 Gag-Pol expression plasmid , 1 μg VSV-G expression plasmid pMD2 . G ( GenScript ) , and 1 . 5 μg Lenti CRISPR v2 ( IKKβ ) , or empty Lenti CRISPR v2 as a Cas9-encoding but sgRNA-negative control . 3 × 105 293T were transduced with 50 μL 293T viral supernatant in the presence of 8 μg/mL polybrene ( Santa Cruz ) , and selected with 2 . 5 μg/mL puromycin . Single cell clones were derived by limiting dilution . Loss of protein expression was confirmed by immunoblot . Generation of TRIM21 knockout 293Ts ( T21KO ) was achieved by electroporation ( Neon ) of Cas9/gRNA ribonucleoprotein complex ( Cas9 RNP ) . The recombinant Cas9-2NLS-GFP protein was produced as described ( Jinek et al . , 2012 ) . Synthetic tracrRNA and crRNA against TRIM21 ( ATGCTCACAGGCTCCACGAA ) were obtained from Sigma-Aldrich . tracrRNA-crRNA complex was assembled by incubating at 20°C for 10 min . The RNA complex was combined with recombinant Cas9 protein at a molar ratio of 1:1 . 2 to form the Cas9 RNP complex following an incubation at 37°C for 10 min . Cas9 RNP against TRIM21 was introduced into 8 × 105 293 T cells using the Neon Transfection System ( Invitrogen ) with 2 pulses of 1400 V for 20 ms . 48 hr post electroporation the cells were cloned by fluorescence-activated cell sorting into 96-well plates ( 1 cell/well ) . Loss of protein expression was confirmed by immunoblot . 293T/T21KO/T21His were made by transducing T21KOs with T21His in pHR’ . C57BL/6 wild-type mice were obtained from Jackson Laboratories . Thirteen week old-mice were used for immunisation experiment , which was conducted in accordance with the 19 . b . 6 moderate severity limit protocol and Home Office Animals ( Scientific Procedures ) Act ( 1986 ) . All animal work was licensed under the UK Animals ( Scientific Procedures ) Act , 1986 and approved by the Medical Research Council Animal Welfare and Ethical Review Body . Mice were immunised subcutaneously ( s . c ) with 100 ug TRIM21 phophoS80-peptide ( 67RQLANMVNNLKEISQ81 ) conjugated to KLH in PBS mixed with complete Freund’s adjuvant followed by 2 rounds of s . c boosting with 50 ug peptide mixed with incomplete Freund’s adjuvant . Serum prepared from tail bleeds and cardiac blood was analysed by ELISA for binding to pS80-TRIM21 or TRIM21 . WT and TRIM21-/- ( K21 ) MEF cell lines were maintained in Dulbecco’s Modified Eagle Medium ( DMEM ) supplemented with 10% fetal calf serum , penicillin at 100 U/ml and streptomycin at 100 μg/ml . MEF cells were obtained from WT or TRIM21 knockout C57/B6 mice and have previously been described and authenticated ( Guilliams et al . , 2014 ) . Human adenovirus type five vector ( ΔE1 , ΔE3 ) expressing GFP ( AdV ) was purchased from ViraQuest . Human rhinovirus type 14 ( HRV ) was produced by infection of HeLa cells , and virus was purified by 2 rounds of CsCl centrifugation . 293T CRL-3216 cells were purchased from ATCC and authenticated by the supplier . All cells used are regularly tested and are mycoplasma free . 107293T/T21KO/T21His were transfected with 25 ug/mL poly ( I:C ) using Lipofectamine LTX ( Life Technologies ) and incubated for 5 or 10 min , or left untransfected . Cells were then washed , pelleted and denatured in 500 μL 6 M GuHCl , 0 . 1 M Na2HPO4/NaH2PO4 ( pH 8 ) , 10 mM Imidazole ( pH 8 ) and pull-downs performed as in the ‘cellular kinase assay’ described above . 2 . 5 × 1011 pts/mL AdV vector was mixed with 10 μg/mL monoclonal anti-hexon 9C12 and incubated for 1 hr at room temperature . 107293T/T21KO/T21His were transduced with virus/antibody mixture , or treated with equal volume of PBS , and incubated at 37C for 7 hr . Cells were then washed , pelleted and denatured in 500 μL 6 M GuHCl , 0 . 1 M Na2HPO4/NaH2PO4 ( pH 8 ) , 10 mM Imidazole ( pH 8 ) and pull-downs performed as in the ‘cellular kinase assay’ described above . Human TRIM21 WT and mutants S80E , S80A were PCR amplified , cloned into lentiviral vector pHR’ ( kind gift of Dr . Adrian Thrasher ) using NotI and SalI restriction sites and sequence verified . Lentiviral particles were generated by co-transfection of 10 cm2 dishes of 293Ts with 1 ug pCRV GagPol , 1 ug pMD2G VSVg and 1 . 5 ug TRIM21 pHR’ using Fugene-6 ( Promega ) . K21 MEF cells were infected with lentivirus in the presence of 8 μg/mL polybrene and stably transduced cells were selected using puromycin at 2 μg/ml . For the generation of stably expressed TRIM21-His , we cloned TRIM21-His into pHR’ , prepared lentivectors as above , transduced 293T TRIM21KO cells in the presence of 8 μg/mL polybrene , and selected cells with 2 . 5 μg/ml puromycin . Cells were plated at 1 × 104 per well in 96 well plates . After 24 hr , viruses were mixed 1:1 with antibody and incubated for 30 mins at room temperature prior to infection . AdV was used at 1 . 25 × 109 pts per well mixed with 10 mg/ml human serum IgG ( Sanquin ) . HRV was used at 2 . 5 × 102 TCID50 units per well mixed with 5 mg/ml human serum IgG ( Sanquin ) . Cells were incubated at 37° C for 3 . 5 hr , before washing with PBS and preparation of cDNA using Cells to CT Kit ( Ambion ) . Gene expression was monitored by TaqMan Gene Expression Assays ( Applied Biosystems ) on a StepOnePlus Real Time PCR System ( Life Technologies ) . Gene expression assays were: mouse β-actin ( 4352933E ) , CxCL10 ( Mm00445235_m1 ) , TNF ( Mm00443260_g1 ) , IL-6 ( Mm00446190_m1 ) , IFIT1 ( Mm00515153_m1 ) . Relative expression was quantified using the 2−ΔΔCt method . Cells were seeded into 24-well plates at 5 × 104 cells/well and allowed to adhere overnight . Virus was diluted to 1 . 5 × 108 pts in 6 μl in phosphate-buffered saline ( PBS ) , mixed with an equal volume of 9C12 antibody variant at the stated concentration , and incubated for 30 min to allow binding to reach equilibrium . NAb-virus mixes were diluted 100-fold into cell culture medium and added to cells . GFP-positive cells were analyzed by flow cytometry after 24 hr of incubation at 37°C . Levels of infection in the absence of NAb were in the range of 50% to 70% . Relative infection was calculated by dividing the percent GFP-positive cells in the presence of antibody to that of a PBS-treated control infection . TRIM21RING-Box protein at 10 mg/mL was buffer exchanged into 50 mM Tris pH 8 , 150 mM NaCl , 1 mM DTT and subjected to sparse matrix screening in sitting drops at 17°C . Crystals grew reproducibly in 10 % w/v PEG 4000 , 20% v/v glycerol , 0 . 03 M diethylene glycol , 0 . 03 M trithylene glycol , 0 . 03 M tetraethylene glycol , 0 . 03 M pentaethylene glycol , 0 . 055 M MES , 0 . 045 M Imidazole , pH 6 . 5 . The crystals were cryoprotected in 20% Ethylene Glycol and flash frozen in liquid Nitrogen . Diffraction experiments were performed at the Diamond light source on the I04-1 beamline , which was equipped with a PILATUS 2M detector . A 1 . 95 Å resolution data set was collected at a wavelength of 0 . 92 Å . TRIM21RING-Box protein contained four zinc atoms per monomer , allowing for the collection of SAD data . Data indexing and scaling were performed with MOSFLM and AIMLESS respectively ( Winn et al . , 2011 ) . Experimental phasing , density modification and initial model building was performed using AutoSol from the PHENIX suite ( Adams et al . , 2010 ) . Further model building was performed in Coot ( Emsley and Cowtan , 2004 ) . REFMAC 5 . 7 was used for model refinement , with initial rounds of restrained refinement including phase information from Hendrickson-Lattman coefficients ( Murshudov et al . , 1997 ) . In later rounds , Translation/liberation/screw ( TLS ) refinement was introduced with individual polypeptide chains defined as TLS groups . MolProbity , which was used for model validation , gave ramachandran statistics of: 95 . 2% of residues were found in favoured regions and 100% in allowed regions ( Chen et al . , 2010 ) . The structure factors and coordinates can be found in the Protein Data Bank , with the accession number 5OLM . 13C/15N or 15N isotopically labelled TRIM21 proteins were buffer exchanged into 50 mM [2H11]-tris , pH 7 . 0 , 150 mM NaCl , 1 mM [2H10]-DTT , H2O/2H2O 95:5 . Spectra were recorded at 25°C on Bruker DRX 500 and Avance I 600 spectrometers ( Bruker BioSpin GmbH ) , each equipped with a ( 1H/15N/13C ) 5 mm cryoprobe , and were processed using the program Topspin ( Bruker BioSpin GmbH ) and analysed using the program Sparky ( Goddard ) . Assignments were made using 13C/15N labelled protein and a standard suite of NMR experiments ( [15N , 1H] HSQC , [13C , 1H] HSQC , HNCA , HNCOCA , CBCANH , CBCA ( CO ) NH , HBHANH and HBHA ( CO ) NH ) , allowing 93% of TRIM21RING-M10E , 97% of TRIM21RING-Box-M10E and 95% of TRIM21RING-Box-M10E/S80E amide resonances of non-proline residues to be assigned . Chemical shift perturbations ( CSPs ) were measured using samples of 15N-labelled 90 µM TRIM21RING-M10E , 200 µM TRIM21RING-Box-M10E or 200 µM TRIM21RING-Box-M10E/S80E , to which were added either 0 . 5 or 1 molar equivalents of unlabeled Ube2N; the reported CSP values are all for 1:1 complexes and were calculated according to the formula ( ( Δδ ( 1H ) ) 2 + ( Δδ ( 15N ) /5 ) 2 ) 1/2 . 15N{1H} heteronuclear NOE ratios for the amide signals of TRIM21RING-Box-M10E and TRIM21RING-Box-M10E/S80E were measured using 200 µM 15N labelled samples and a pre-saturation time of 7 s , essentially as described by ( Skelton et al . , 1992 ) . Solvent paramagnetic relaxation enhancement ( PRE ) data were measured by the addition of the soluble contrast agent , Gd-DTPA . A 100 mM stock solution of Gd-DTPA was made in 50 mM [2H11]-tris and the pH adjusted to 7 . 0; this solution was added to a final concentration of 4 mM to samples of 200 µM TRIM21RING-Box-M10E or 200 µM TRIM21RING-Box-M10E/S80E . Peak intensities for each amide signal in the [15N , 1H] HSQC spectra were recorded before ( IREF ) and after ( IPRE ) addition of the contrast agent; for Figures 4H and 6 , the reported PRE ratio is calculated as ( [IPRE/IREF]WT ) / ( [IPRE/IREF]S80E ) . | Antibodies are molecules made by the immune system that protect us from infections . They were discovered over 100 years ago , and for most of that time scientists thought they only worked in the bloodstream . Yet recent research showed that when a virus infects our cells it also carries antibodies in with it . Once inside the cell , a protein called TRIM21 recognises the antibody-covered virus and – together with other proteins called ubiquitin enzymes – targets it for destruction via the cell’s waste disposal system . At the same time TRIM21 sends a signal to the cell’s nucleus to activate certain genes that protect cells across the body from subsequent infection . The genes activated by TRIM21 have potent antiviral activity . Yet they can also damage the body’s own tissues if they are switched on at the wrong time , which may lead to autoimmune diseases like rheumatoid arthritis and multiple sclerosis . It is thus critical that TRIM21 is carefully regulated and only activated during an infection , but it was not clear how this is achieved . Dickson , Fletcher et al . now show that although TRIM21 is made all the time and is always ready to detect an incoming virus it is made in an inactive state . This is because part of TRIM21 , called a B-Box , inhibits the protein’s own activity . This was an unexpected discovery because , although the B-Box domain is found in around 100 other human proteins , it was unclear what it did . Dickson , Fletcher et al . then combined structural biology and biochemical approaches to show that the B-Box mimics specific enzymes that work with TRIM21 , and blocks them from binding to it . This keeps TRIM21 in an inactive state . Next , Dickson , Fletcher et al . discovered that TRIM21 becomes active when a chemical tag , specifically a phosphate group , is added to the protein . This modification displaces the B-Box , allowing ubiquitin enzymes to bind to TRIM21 and switch on its activity . Further experiments then showed that this process helps regulate TRIM21 during infections with different viruses , including rhinovirus – the virus behind the common cold – and adenovirus – a common cause of respiratory infection . Antibodies are now used to treat many medical conditions , but present technologies are based on our understanding of how antibodies work outside cells . By revealing the basis of antibody immunity inside cells , these new findings may lead to new treatments for a range of conditions . Future studies could also explore how failures in the TRIM21 system contribute to autoimmune diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2018 | Intracellular antibody signalling is regulated by phosphorylation of the Fc receptor TRIM21 |
Robust sleep/wake rhythms are important for health and cognitive function . Unfortunately , many people are living in an environment where their circadian system is challenged by inappropriate meal- or work-times . Here we scheduled food access to the sleep time and examined the impact on learning and memory in mice . Under these conditions , we demonstrate that the molecular clock in the master pacemaker , the suprachiasmatic nucleus ( SCN ) , is unaltered while the molecular clock in the hippocampus is synchronized by the timing of food availability . This chronic circadian misalignment causes reduced hippocampal long term potentiation and total CREB expression . Importantly this mis-timed feeding resulted in dramatic deficits in hippocampal-dependent learning and memory . Our findings suggest that the timing of meals have far-reaching effects on hippocampal physiology and learned behaviour .
The circadian system is a finely tuned network of central and peripheral oscillators headed by a master pacemaker , the suprachiasmatic nucleus ( SCN ) , which governs daily rhythms in physiology and behaviour , including cognition . This network regulates cognitive processes ( Holloway and Wansley , 1973; Chaudhury and Colwell , 2002 ) , and the neural circuits involved in learning and memory also exhibit circadian rhythms in gene expression and synaptic plasticity ( Eckel-Mahan et al . , 2008; Fropf et al . , 2014; Lamont et al . , 2005; Lyons , 2006 ) . Genetic disruption of these molecular oscillations has severe consequences on cognition ( Van der Zee et al . , 2008; Wang et al . , 2009; Wardlaw et al . , 2014 ) . Environmental perturbations also have the capacity to disrupt synchrony and misalign this clock network ( Fekete et al . , 1985; Cho et al . , 2000; Devan et al . , 2001; Ruby et al . , 2008; Loh et al . , 2010; Gibson et al . , 2010; Karatsoreos et al . , 2011; Fonken et al . , 2012; LeGates et al . , 2012; Fernandez et al . , 2014 ) and are problematic as many people in our modern society extend their work and recreation into the night hours . There has been mounting evidence that the timing of when we eat is critical for our metabolic health ( Bass , 2012; Mattson et al . , 2014 ) . At this point , timing of food intake is well-established to have a major impact on the phase of the molecular oscillations in peripheral organs such as the liver and pancreas ( Damiola , 2000; Stokkan , 2001 ) . Mis-timed meals during the sleep phase accelerates weight gain compared with animals fed during their wake phase ( Arble et al . , 2009; Bray et al . , 2013 ) , whereas wake-phase feeding has a protective effect against the cardiac and metabolic dysfunction caused by high fat diets ( Hatori et al . , 2012; Gill et al . , 2015 ) . Similar disruptive effects are seen in humans , where misaligned mealtimes produce cardiac and metabolic deficits , leading to a pre-diabetic state ( Scheer et al . , 2009 ) . We thus became interested in the possibility that these ill consequences of eating at inappropriate phases of the daily cycle may also be maladaptive for cognitive function . In this study , we sought to determine the effects of chronic but stable misalignment of the circadian network by scheduling access to food at an inappropriate phase of the daily cycle . We demonstrate that this simple manipulation has far-reaching consequences for learning and memory .
Mice were allotted a 6 hr window in which food was made available either during the middle of their active ( aligned ) or inactive ( misaligned ) phase ( Figure 1 ) . Mice adapted to the feeding protocol within 6 days ( p = 0 . 9 , Figure 2A ) and there were no significant differences in body weight between the two groups at the time of testing ( p = 0 . 5 , Figure 2B ) . Daytime activity was increased in mice subjected to misaligned feeding ( p < 0 . 001; Figure 3A , B ) , and the strength of daily rhythms of activity was reduced by misaligned feeding ( p = 0 . 003; Figure 3C ) . Similarly , the temporal pattern of sleep was altered by misaligned feeding ( Figure 3D ) . Immobility-defined video monitoring of sleep behaviour of misaligned mice showed decreased time spent asleep during the day ( p < 0 . 001 ) and a corresponding increase in sleep during the night ( p < 0 . 001; Figure 3E ) . Misaligned mice no longer exhibit a day vs night difference in sleep , sleeping equally during the day and night ( p = 0 . 5 ) . Crucially , the total time spent asleep over a 24 hr day was not reduced by misaligned feeding ( p = 0 . 2 ) . The change in temporal pattern of sleep quantity is also reflected in sleep fragmentation , where misaligned mice exhibit a greater number of sleep bouts ( p < 0 . 05 ) with a corresponding decrease in average sleep bout duration ( p < 0 . 001 ) during the day compared to aligned mice ( Figure 3—figure supplement 1 ) . This increased day-time sleep fragmentation is compensated by fewer ( p < 0 . 05 ) and longer ( p < 0 . 001 ) night sleep bouts in the misaligned animals , resulting in no significant change in the total number ( p = 0 . 9 ) and average duration ( p = 0 . 6 ) of sleep bouts over a 24 hr period . 10 . 7554/eLife . 09460 . 003Figure 1 . Cartoon schematic of experimental design . Standard mouse cages were modified to restrict access to the food chamber . Access was controlled by a motorized gate controlled by timed relay switches . Mice could only access food pellets when the gate was lifted ( 6 hr ) , and positive drive of the motor kept the gate closed for the remaining 18 hr . The scheduled feeding protocol was maintained for a minimum of 2 weeks prior to and during sample collections and behavioural tests , indicated by * . PER2::LUC , PER2-driven bioluminescence . LTP , long term potentiation . FC , fear conditioning . NOR , novel object recognition . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 00310 . 7554/eLife . 09460 . 004Figure 2 . Food pellets in the automated feeding chambers were weighed daily to determine the amount of food consumed . ( a ) Misaligned mice caught up with aligned mice in daily food consumption by day 6 of scheduled feeding ( p = 0 . 9 ) and did not differ in food consumption subsequently ( day 6–14 post hoc p > 0 . 2 ) . ( b ) Mice were weighed daily prior to food access . Body weights between treatment groups did not differ significantly through the duration of scheduled feeding ( two way ANOVA p = 0 . 3 ) . Line graphs represent the mean ± SEM ( n = 16 per treatment ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 00410 . 7554/eLife . 09460 . 005Figure 2—source data 1 . Food consumption and body weights of mice subjected to aligned and misaligned feeding . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 00510 . 7554/eLife . 09460 . 006Figure 3 . Altered temporal patterns of activity and sleep in mice subjected to misaligned feeding . ( a ) Mice ( n = 8 per treatment ) were given 6 hr windows of food access during the middle of the night ( aligned , grey ) or day ( misaligned , red ) . Representative double-plotted actograms show the increased daytime activity of the misaligned mice throughout the treatment in a 12 hr:12 hr light:dark ( LD ) cycle . Grey shading in the actograms indicates lights off . ( b ) Nocturnality ( % activity in the night ) is reduced in misaligned mice ( ***p < 0 . 001 ) . ( c ) Rhythm strength measured by the amplitude of a chi-square periodogram ( %V ) is reduced in misaligned mice ( **p < 0 . 01 ) . Box and whisker plots display the 25th to 75th percentiles , and the 10th to 90th percentiles respectively , with the median indicated by a line . ( d ) Sleep was measured by video monitoring after 2 weeks of scheduled food access ( n = 10 per treatment ) . Running averages of immobility-defined sleep are shown for mice given aligned ( black ) and misaligned ( red ) access to food . The grey shading indicates lights off in a 12:12 LD cycle . ( e ) Total time spent asleep during the day ( 12 hr ) , night ( 12 hr ) , or over 24 hr are shown . ***p < 0 . 001 as determined by t-tests of aligned vs misaligned groups; # p < 0 . 001 day vs night within treatment . Bar graphs represent the mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 00610 . 7554/eLife . 09460 . 007Figure 3—source data 1 . Activity and sleep rhythm parameters of mice subjected to aligned and misaligned feeding . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 00710 . 7554/eLife . 09460 . 008Figure 3—figure supplement 1 . The temporal pattern of sleep fragmentation is altered by misaligned feeding without affecting overall sleep fragmentation over the 24 hr period . ( a ) The number of sleep bouts during the day is increased in misaligned mice , decreased at night , and unchanged between groups when considered over 24 hr ( total ) . ( b ) The average duration of a sleep bout is shorter in misaligned mice during the day , longer at night , and unchanged between groups over 24 hr ( total ) . ( c , d ) The distribution of sleep bout lengths is similar between groups ( p = 0 . 3 ) . *p < 0 . 05 , ***p < 0 . 001 as determined by t-tests of aligned vs misaligned groups . Bar graphs represent the mean ± SEM . Circles represent individual sleep bouts rounded to an integer minute . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 008 Using ex vivo organotypic cultures of explants from PER2::LUC reporter mice subjected to aligned or misaligned feeding , we confirmed that daytime feeding shifts the phase of the liver oscillator ( p < 0 . 001 ) without altering the phase of the SCN oscillator ( p = 0 . 07; Figure 4A , C , D ) . Importantly , we demonstrate that daytime feeding significantly misaligns the hippocampal oscillator by 14 . 1 hr ( p < 0 . 001; Figure 4B , D ) , with small but significant effects on the intrinsic properties of the oscillator , including period ( p = 0 . 05; Figure 4E ) and damping rate ( p = 0 . 02; Figure 4F ) . 10 . 7554/eLife . 09460 . 009Figure 4 . Differential impact of misaligned feeing on PER2-driven rhythms in bioluminescence of the SCN , hippocampus and liver ( n = 8 per treatment ) . Representative examples of baseline-subtracted traces of PER2-driven bioluminescence in the SCN ( a ) , hippocampus ( HP; b ) , and liver ( c ) explants from aligned ( black ) and misaligned ( red ) mice . ( d ) Phase relationship between the first calculated peaks of ex vivo bioluminescence plotted against time of the prior lighting cycle ( ZT ) shows a significant phase change in the HP and liver . ( e ) Period of bioluminescence rhythms were determined by sine wave fitting . ( f ) Damping rates were determined from 6 days in culture . *denotes significant differences ( p < 0 . 05 ) between aligned and misaligned samples . Box and whisker plots display the median as a line , the 25th to 75th percentiles , and the 10th to 90th percentiles respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 00910 . 7554/eLife . 09460 . 010Figure 4—source data 1 . PER2-driven bioluminescence rhythms of mice subjected to aligned and misaligned feeding . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 010 We measured expression of phosphorylated CREB ( pCREB ) and total CREB ( tCREB ) in the hippocampus of aligned and misaligned mice at 6 hr intervals through a 24 hr day . pCREB protein levels were reduced in the misaligned animals at each time point , although significant differences were not detected ( Figure 5A , B ) . Strikingly , expression of tCREB was significantly reduced by misaligned feeding ( p < 0 . 01; Figure 5A , C , D ) , with the strongest effects in the day ( post hoc p < 0 . 05 for ZT 2 , 8 , and 20 ) . This decrease in tCREB levels was uniformly observed throughout the hippocampus ( Figure 5D ) . 10 . 7554/eLife . 09460 . 011Figure 5 . Reduced magnitude of hippocampal tCREB expression with a corresponding reduction in long term potentiation ( LTP ) in misaligned mice . ( a ) Representative immunoblots show the decrease in tCREB expression levels in misaligned vs . aligned mice , sampled at 6 hr intervals through the daily cycle ( ZT ) . ( b ) pCREB levels are not significantly altered . ( c ) Misaligned feeding led to significant decreased levels of tCREB ( p < 0 . 01 ) . * indicates significant differences between groups at each time point ( post hoc p < 0 . 05 ) . Protein levels are expressed as arbitrary units ( a . u . ) . Bar graphs represent the mean ± SEM of aligned ( n = 4 ) and misaligned ( n = 5 ) animals per time point . ( d ) tCREB immunoreactivity is decreased throughout the hippocampus in misaligned mice . Scale bar = 500 µm . ( e ) LTP was induced by high-frequency stimulation of the Schaffer collateral fibres , 2 x 100 Hz , 1 sec duration , 10 sec inter-train interval , delivered at time = 0 . Daytime LTP responses recorded from the CA1 region were significantly decreased in misaligned mice ( n = 6; p < 0 . 05 ) compared to aligned mice ( n = 6 ) . The inset shows fEPSPs recorded during baseline and 55–60 min post-HFS in aligned ( left ) and misaligned slice ( right ) . ( f ) Paired-pulse facilitation ratios changed with intervals ( p < 0 . 001 ) , but were not significantly different between aligned and misaligned mice ( p = 0 . 5 ) and no interactions between both factors were detected ( p = 0 . 8 ) . Line graphs represent the mean ± SEM ( n = 6 per treatment ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 01110 . 7554/eLife . 09460 . 012Figure 5—source data 1 . Hippoca mpal CREB levels and LTP in mice subjected to aligned and misaligned feeding . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 01210 . 7554/eLife . 09460 . 013Figure 5—figure supplement 1 . Expression levels of β-actin did not vary with time of day in both groups ( p = 0 . 9 ) , nor between aligned and misaligned mice ( p = 0 . 9 ) . Bar graphs represent the mean ± SEM of aligned ( n = 4 ) and misaligned ( n = 5 ) animals per time point . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 01310 . 7554/eLife . 09460 . 014Figure 5—figure supplement 2 . Aligned feeding does not significantly alter LTP magnitude compared to mice under ad libitum feeding ( p = 0 . 4 ) . LTP was induced in both groups using the same protocol; data from aligned mice is the same as reported in Figure 5e . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 014 Long term potentiation ( LTP ) responses of the dorsal hippocampus were measured during the day from mice subjected to aligned or misaligned feeding . LTP was significantly impaired in the misaligned group ( p< 0 . 05; Figure 5E ) , indicating significant deficits in synaptic plasticity in the misaligned hippocampus . This decrease in LTP is specific to the misaligned feeding treatment and not due to an LTP-boosting effect by aligned feeding ( Figure 5—figure supplement 2 ) . To determine the impact of misaligned feeding on presynaptic neurotransmitter release probability , paired pulse facilitation ( PPF ) experiments were performed at 25 , 50 , 100 and 250 msec intervals . There were no significant differences between the aligned and misaligned PPF ratios ( p = 0 . 5; Figure 5F ) , and EPSP profiles were similar . To test our hypothesis that misaligning the hippocampal oscillator from the SCN oscillator is detrimental to learning and memory , we subjected aligned and misaligned mice to hippocampal-dependent contextual fear conditioning . Mice were trained to associate a specific novel context to a fearful stimulus in the form of a mild shock . Both aligned and misaligned mice acquired freezing behaviour during the training trial ( Figure 6—figure supplement 1 ) . When tested 24 hr later by replacing the mice in the same context , the misaligned mice exhibited a significant reduction in fear-conditioned behaviour ( ZT 2 , p < 0 . 0001; Figure 6A , left ) , indicating that circadian misalignment affects long term memory . Performance on the fear conditioning test is dependent on time of day and performance peaks in the early day ( Chaudhury and Colwell , 2002; Eckel-Mahan et al . , 2008; Loh et al . , 2010 ) . To test for the possibility that misaligned mice have an inverted peak performance time , we trained and tested a separate cohort of mice at the opposite time of day ( night , ZT 14 ) . Both aligned and misaligned groups acquired freezing behaviour ( Figure 6—figure supplement 1 ) . Two way comparisons revealed a time of day effect on recall ( p < 0 . 001 ) and an interaction between time of day and feeding condition ( p < 0 . 001 ) . Recall was significantly reduced in the aligned mice compared to the daytime tests ( post hoc p < 0 . 001 ) , but did not significantly change with time of day in misaligned mice ( post hoc p = 0 . 3 ) . Night-tested aligned and misaligned mice exhibited equally poor recall ( post hoc p = 0 . 4; Figure 6A , right ) , ruling out the possibility that the misaligned mice have an altered peak phase of cognition . 10 . 7554/eLife . 09460 . 015Figure 6 . Memory deficits arise from misaligned feeding in mice . ( a ) Recall of the fear conditioned ( FC ) context is measured by the percentage of freezing when re-exposed to the fearful context . Misaligned mice ( n = 8 ) show significant deficits in recall of contextual ( FC ) compared to aligned mice ( n = 8;***p < 0 . 001 ) when trained and tested during the day ( ZT 2 ) . Circadian regulation of learning and memory is demonstrated by the decreased recall in aligned mice trained and tested at night ( ZT 14; n = 8; #p < 0 . 001 ) . This time of day effect is lost in misaligned mice ( n = 8 ) , which perform equally poorly at both times . ( b ) Novel object recognition ( NOR ) is reported using a discrimination index of Tnovel/ ( Tnovel+Tfamiliar ) , and mice are considered to exhibit NOR at values of 0 . 5 and above ( dotted red line ) . NOR is impaired in misaligned mice trained and tested at night ( ZT 21; n = 8 ) compared to aligned mice ( n = 8; ***p < 0 . 001 ) . Time of day effects were also found for NOR in aligned mice ( n = 8; #p < 0 . 01 ) , which perform better during the night than day ( ZT 9 ) . Misaligned mice fail to show a time of day effect ( n = 7 ) , again showing equally poor performance at both times . Box and whisker plots display the median as a line , the 25th to 75th percentiles , and the 10th to 90th percentiles respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 01510 . 7554/eLife . 09460 . 016Figure 6—source data 1 . Performance on long term memory tests in mice subjected to aligned and misaligned feeding . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 01610 . 7554/eLife . 09460 . 017Figure 6—figure supplement 1 . Acquisition of FC freezing in the day ( ZT 2 , n = 8/group ) and night ( ZT 14; n = 8/group ) are unaltered by misaligned feeding . ( a ) Misaligned mice did not show significant changes in baseline freezing ( p = 0 . 3 ) , freezing in response to the first foot shock ( US-1; p = 0 . 9 ) , or freezing in response to the second foot shock ( US-2; p = 0 . 1 ) . Recall was significantly reduced in misaligned mice as reported in the main text and Figure 6 . ( b ) Misaligned feeding did not cause differences in baseline freezing ( p = 1 . 0 ) , nor in acquisition of freezing in response to US-1 ( p = 0 . 9 ) and US-2 ( p = 0 . 5 ) . There was a time of day effect on freezing in response to US-2 ( #p < 0 . 05 ) , as previously reported ( 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09460 . 017 We applied the novel object recognition test to examine cognition that peaks during the night in nocturnal rodents ( Ruby et al . , 2008 ) . The misaligned mice exhibited significantly reduced novel object recognition ( NOR ) during night-time tests ( ZT 21 , p < 0 . 001; Figure 6B , right ) , indicating decreased cognitive performance . Day-time tests of a separate cohort of aligned and misaligned found no differences between both groups ( Figure 6B , left ) , ruling out the possibility that peak performance of misaligned mice had shifted to the opposite phase . Two factor analysis revealed a significant effect of feeding condition ( p = 0 . 003 ) and an interaction between time of day and feeding condition ( p = 0 . 02 ) . Specifically , aligned mice showed reduced NOR during the day ( post hoc p = 0 . 04 ) , and day-tested misaligned mice were not significantly different from aligned ( post hoc p = 0 . 6 ) .
Many people in our society find themselves working or playing during their normal sleep times . Due to these schedules , we eat around the clock with well-established literature of metabolic consequences ( Mattson et al . , 2014 ) . In this study , we sought to determine if temporally restricted feeding schedules in mice could impact cognition . We found that time-restricted feeding led to dramatic impairments in learned behaviours such as hippocampal-dependent contextual fear conditioning and novel object recognition . Not all behavioural tests were similarly affected , as amygdala-dependent cued-fear conditioning was not altered by the misalignment . This implies that some learned behaviours are more vulnerable to the impact of misaligned feeding . Crucially , the total amount of sleep was not altered by the scheduled feeding and the mice did not lose weight as might be expected if their caloric intake was restricted . Hence , by simply adjusting the time of food access , we demonstrated that we can alter the cognitive performance of mice . We do not know yet if this equally applies to humans but shift work has been associated with decreased performance on cognitive tests ( Wright et al . , 2006; Marquie et al . , 2015; Zhou et al . , 2011 ) . The importance of sleep for memory is well-documented , where both the amount of sleep and the quality of sleep have been found to be critical for memory consolidation ( reviewed in Diekelmann and Born , 2010 ) . The misaligned feeding treatment did not result in an overall decrease in amount of sleep , but instead had a severe impact on the temporal pattern , suggesting that this treatment acts via disruption of the circadian timing of sleep . While sleep quality as assessed by polysomnography was not measured in this study , we were able to examine the degree of sleep fragmentation as determined by the number and duration of individual sleep bouts . Misaligned feeding again has a greater impact on the temporal pattern of sleep fragmentation , where the misaligned mice appear to “catch up” on consolidated sleep during what should be their active phase . Mechanistically , we used the clock-driven rhythms in bioluminescence to demonstrate that the altered feeding time not only shifted the peripheral clocks in the liver , but also altered the circadian clock in the hippocampus . This finding is consistent with earlier literature suggesting the phasing of the molecular clock in the corticolimbic system and hippocampus are strongly influenced by the time of food availability ( Amir , 2004; Ángeles-Castellanos et al . , 2007; Wakamatsu et al . , 2001 ) . We have thus created a situation in which the molecular clocks within the nervous system are running at different phases , i . e . internal desynchronisation of the circadian system . A consequence of this misalignment is altered synaptic plasticity within the hippocampal circuit ( Schaffer Collaterals/CA1 ) as revealed by the reduction in LTP . This is the first demonstration that the time of eating can impact the physiological underpinning of learned behaviour . Importantly , our manipulation of feeding caused a significant decline in CREB in the hippocampus , which has been demonstrated to be critical for memory allocation in mice ( e . g . Sano et al . , 2014; Zhou et al . , 2009 ) . The levels of tCREB were reduced at all phases that we sampled ( ZT 2 , 8 , 14 and 20 ) , and although we did not carry out cognitive tests at each of these phases , this finding indicates that memory would be impaired throughout the daily cycle . The reduction in tCREB provides a biochemical explanation for our finding , as well as providing a guidepost for future analysis of the effects of the misaligned feeding in the hippocampus . Therefore , this work raises the possibility that the timing of when we eat alters the physiological and biochemical events underlying learning and memory .
All experimental protocols used in this study were approved by the UCLA Animal Research Committee ( Protocol 1998–183 ) . UCLA Division of Laboratory animal recommendations for animal use and welfare , as well as National Institutes of Health guidelines were followed . Adult ( 2–4 month old ) male C57BL/6N wild-type mice ( UCLA ) were housed in a 12:12-hr lighting ( LD ) cycle with ad libitum access to food and water . For the bioluminescence experiments , PER2::LUC knock-in homozygotes on the C57Bl/6J background were used ( Yoo et al . , 2004 ) . Mice were first entrained to a 12:12 LD cycle for a minimum of 2 weeks prior to further manipulations . For measurements of activity , sleep , and food consumption , mice were individually housed in cages . For bioluminescence measurements , learning and memory tests , hippocampal physiology , biochemistry and anatomy , mice were housed in groups of 3–5 mice per cage . Group-housed animals showed similar activity and food consumption patterns as singly housed animals . Both strains of mice exhibited similar activity and food consumption patterns . Mice entrained to a 12:12 LD cycle were randomly sorted into two groups per experiment . Cages were topped with wire grids adapted to restrict access to the food chamber , which enabled us to provide access to food during specific times of day ( Figure 1 ) . We visually verified that food pellets or powder were not being hoarded in nesting or bedding materials , but did not empty the cage bottoms to avoid excessive handling stress to the animals . Both groups were given a 6 hr window per 24 hr day to access the food chamber . Mice in one group were given access to the food chamber during the middle of the dark phase from Zeitgeber Time ( ZT; ZT 0 equates to lights on ) 15 to 21 , aligned to their active phase . Mice in the second group were given automated access to the food chamber during the middle of the light phase from ZT 3 to 9 , misaligned from what should have been their active phase . Scheduled food access was applied for 2 weeks prior to and for the entire duration of all experiments . The exceptions were for the data described in Figure 2 and Figure 3A: food consumption and activity , for which we monitored the mice from day 1 of scheduled food access . Mice were individually housed in automated feeding cages with a top-mounted passive infrared motion sensor ( Honeywell IS-215T ) to detect cage activity ( aligned n = 8 , misaligned n = 8 ) . Data was collected using the Vital View system ( Mini Mitter , Bend , OR ) and analysed as previously described for wheel running activity using the El Temps software ( Loh et al . , 2013; A . Diez-Noguera , Barcelona , Spain ) . Specifically , nocturnality was calculated by determining the percentage of activity conducted during the dark phase ( ZT 12–24 ) from 7 days of activity monitoring ( days 8 to 14 of scheduled feeding ) . The strength of the daily rhythms was determined from the same 7 days , and is reported as the power of the circadian harmonic of a Fourier analysis ( %V ) . Following activity monitoring , sleep-wake behaviour from the same cohort of aligned ( n = 10 ) and misaligned ( n = 10 ) mice was measured using continuous video recording and automated mouse tracking as previously described ( Loh et al . , 2013 ) . Mice were maintained in the same automated feeding cages under the same lighting cycle and feeding schedule , and continuous video recording was performed from days 15 to 17 of scheduled feeding . Side-on views of the cage were acquired using CCTV cameras ( Gadspot , GS-335C , City of Industry , CA ) , and the ANY-maze software ( Stoelting Co . , Wood Dale , IL ) was used to track the animals . Prior work by Fisher and colleagues established the parameters for immobility detection ( 95% immobility of the area of the animal , for a minimum of 40 sec ) by correlation with EEG/EMG recordings ( Fisher et al . , 2012 ) . Immobility-defined sleep in 1 min bins from days 16 to 17 of scheduled feeding were averaged , and day ( ZT 0–12 ) and night ( ZT 12–24 ) sleep were compared . Average waveforms for display purposes were generated by smoothing the data using 1 hr running averages . PER2::LUC male mice ( 2–3 mo ) were subjected to either aligned ( n = 8 ) or misaligned ( n = 8 ) feeding for 2 weeks prior to sampling . Mice were sacrificed after anaesthesia ( isoflurane ) between ZT 10 and 11 , and 1–2 mm3 liver explants were immediately dissected in ice-cold Hanks’ balanced salt solution ( HBSS; Sigma , St Louis , MO ) supplemented with 4 . 5 mM NaHC03 , 10 mM HEPES and 100 U/ml penicillin-streptomycin as previously described ( Loh et al . , 2011 ) . Brains were incubated in ice-cold slice solution ( in mM: 26 NaHCO3 , 1 . 25 NaH2PO4 , 10 glucose , 125 NaCl , 3 KCl , 5 MgCl2 , 1 CaCl2 ) aerated with 95% O2/5% CO2 for 5 min , and 300 µm coronal sections were collected using a vibratome and further microdissected in HBSS under a 10X dissecting microscope . The SCN was cut away from the rest of the section using two cuts with a surgical scalpel ( No . 21 blade , Fisher Sci . , Waltham , MA ) . To acquire the hippocampal explant , the dorsal-most section ( Bregma -1 . 2 to -1 . 6 mm ) was microdissected to liberate the hippocampus and dentate gyrus by gently teasing away the cortex with scalpels ( No . 11 blade , Fisher Sci . ) . All explants were individually transferred to Millicell membranes ( 0 . 4 μm , PICMORG50 , Millipore , Bedford , MA ) resting on 1 . 2 ml of recording media: ( 1X DMEM ( Sigma ) , 1X B27 supplement ( Gibco , Carlsbad , CA ) , 4 . 5 mM NaHCO3 , 10 mM HEPES , 40 mM Glutamax ( Gibco ) , 4 . 5 mg/ml D-glucose , 25 U/ml penicillin , 25 U/ml streptomycin , 0 . 1 mM sodium salt monohydrate luciferin ( Biosynth , Staad , Switzerland ) ) in a 35 mm dish sealed with autoclaved vacuum grease ( Dow Corning , Midland , MI ) . SCN , hippocampus and liver explants were inserted into the Lumicycle photometer ( Actimetrics , Wilmette , IL ) , incubated at 37°C , and bioluminescence was continuously monitored for 7 consecutive days . Raw bioluminescence values were normalized by baseline subtraction ( 24 hr running average ) and smoothed with 2 hr windows to prepare the representative bioluminescence traces . The phase and damping rate of each explant were determined as previously described ( Loh et al . , 2011 ) . Period was determined using the sine-wave fit function in Lumicycle Analysis ( Actimetrics ) . Hippocampi were rapidly dissected and homogenized in lysis buffer ( 50 mM Tris-HCl , 0 . 25% ( w/v ) sodium deoxycholate , 150 mM NaCl , 1 mM EDTA , 1% Nonidet P40 , 1 mM sodium vanadate , 1 mM AEBSF , 10 ug/ml Aprotinin , 10 ug/ml Leupeptin , 10 ug/ml Pepstatin , and 1 mM sodium fluoride ) . Total protein concentration in cleared extracts was estimated using Pierce’s BCA ( bicinchoninic acid ) Protein Assay Kit ( Thermo Fisher Scientific , Carlsbad , CA ) using bovine serum albumin as a standard . Western blots were performed as previously described ( Ghiani and Gallo , 2001; Ghiani et al . , 2010 ) . Each extract was analysed for relative protein levels of phosphoCREB by using an antibody directed against the phosphorylated form at Ser133 ( rabbit polyclonal , Millipore ) and then for tCREB ( rabbit polyclonal , Millipore ) . Equal protein loading was verified by Ponceau S solution ( Sigma ) reversible staining of the blots and each extract was also analysed for relative protein levels of β-actin ( Sigma ) . Relative intensities of the protein bands were quantified by scanning densitometry using the NIH Image Software ( Image J , http://rsb . info . nih . gov/ij/ ) , and each value background-corrected . Data are shown as arbitrary units and are the average ± SEM of 4–5 animals/group . We confirmed by two way ANOVA comparisons that levels of β-actin do not vary with time of day in either group ( p = 0 . 9 ) and is no different between aligned and misaligned mice ( p = 0 . 9; Figure 5—figure supplement 1 ) . Mice were perfused intracardially with 4% paraformaldehyde ( PFA ) . Brains were dissected out , post-fixed in 4% PFA at 4°C overnight , cryoprotected in 15% sucrose . Immunolabelling of frozen sections ( 30 μm ) was performed as previously described ( Ghiani et al . , 2011 ) . Briefly , sections were blocked in carrier solution ( 1% BSA and 0 . 3% Triton X-100 ) containing 20% normal goat serum for 1 hr and incubated for 48h at 4°C with primary antibodies against tCREB ( 1:500 rabbit polyclonal , Millipore ) diluted in carrier solution containing 5% normal goat serum . Sections were incubated with a goat anti-rabbit secondary antibodies conjugated to Cy3 ( Jackson ImmunoResearch Laboratories , West Grove , PA ) and mounted with Vectashield medium with DAPI ( 4′ , 6-diamidino-2-phenylindole; Vector Laboratories , Burlingame , CA ) . Immunostained sections were visualized using a Zeiss Axio Imager 2 with an AxioCam MRm . Images were acquired through Axiovision ( Zeiss , Thornwood NY ) , using a 5X objective in order to visualize the entire hippocampus . Hippocampal slice preparation and electrophysiological recordings were performed as previously described ( Carlisle et al . , 2008 ) . Briefly , slices ( 400 µm ) were maintained in oxygenated ( 95% O2 / 5% CO2 ) , warmed ( 30°C ) artificial cerebrospinal fluid ( ACSF ) containing 124 mM NaCl , 4 . 4 M KCl , 25 mM NaHCO3 , 1 . 0 mM NaH2PO4 , 2 . 0 mM CaCl2 , 1 . 2 mM MgSO4 , and 10 mM glucose , and allowed to recover for at least 2 hr prior to the start of an experiment . A bipolar , nichrome wire stimulating electrode was placed in stratum radiatum of the hippocampal CA1 region to activate Schaffer collateral–commissural fibre synapses and an extracellular glass microelectrode filled with ACSF ( resistance = 5–10 MΩ ) was used to record evoked field excitatory postsynaptic potentials ( fEPSPs ) . Extracellular recordings were done under interface conditions . The intensity of presynaptic fibre stimulation was adjusted to evoke fEPSPs with amplitude approximately 50% of the maximal fEPSP amplitude that could be elicited in each slice . fEPSPs were then elicited at 0 . 02 Hz throughout the experiment . LTP was induced by high-frequency stimulation ( HFS , 2 trains of 100 Hz stimulation , 1 sec duration , inter-train interval = 10 sec ) . The average slope of fEPSPs normalized to baseline measured between 55 and 60 min post-HFS was used for statistical comparisons ( two tailed t tests ) . The paired pulse stimulation pulses were delivered with interpulse intervals of 25 , 50 , 100 and 200 ms . Contextual fear conditioning was performed as previously described ( Chaudhury and Colwell , 2002; Wang et al . , 2009; Loh et al . , 2010 ) with minor modifications . In brief , aligned and misaligned mice were tested and trained either in the day ( ZT 2 ) or the night ( ZT 14 ) . Mice were trained and tested only once to avoid effects of prior manipulation . Results reported for daytime and night-time tests are from distinct cohorts of animals . Mice were habituated to the testing room for 30 min under the relevant lighting conditions ( dim red light , 2 lux , at night ) . The animals were individually introduced to the novel environment ( shock chamber ) and allowed to familiarize to context for 3 min , after which the mice were trained to associate the context with a fearful unconditioned stimulus ( US ) : foot-shock ( 0 . 2 mA ) . The training protocol consisted of 2 US with an inter-trial interval of 64 sec . At the end of the last US , mice were left in the chamber for a further 64 sec , after which they were returned to their home cages . 24 hr later , mice were placed individually into the same conditioning chamber for 6 min . Video recordings of the acquisition and recall tests were done with CCTV ( Gadspot ) cameras with supplemental infrared lighting during both times of day . Freezing behaviour was scored as previously described ( Chaudhury and Colwell , 2002 ) . Cued fear conditioning was performed as previously described ( Wang et al . , 2009 ) . Using the same stimulus protocol as contextual FC , 30 sec of white noise ( 80 dB; conditioned stimulus , CS ) preceded the presentation of US1 and US2 . 24 hr later , mice were placed individually into a novel testing chamber . Following 2 min of baseline , the cued tone was presented for 2 min; mice were left in the chamber for a further 2 min . Freezing behaviour was scored as previously described . Habituation to the testing arena , object familiarization , and testing for NOR over 5 consecutive days were performed in the active phase at ZT 21 ( n = 8 per feeding condition ) for night-time tests under dim red light ( 2 lux ) , with day tests at ZT 9 ( n = 8 per feeding condition ) . Mice were habituated to the testing arena ( 60 × 48 × 30 cm ) in 10 min trials on 2 consecutive days . During familiarization trials , two identical objects were placed equidistant from each other and the walls of the arena , and mice were allowed to explore the arena for 10 min on 2 consecutive days . Mice that failed to travel 20 m or interact with objects for more than 20 sec in the second familiarization trial were eliminated from further analysis . On the day of testing for NOR , one familiar object was replaced with a novel object with a different shape and made of a different material ( plastic or glass ) . The testing trial was 5 min in duration . Arenas and objects were wiped between animals with 10% Windex and dried using paper towels . Video feeds of each arena from an overhead CCTV camera supplemented with infrared lighting were fed to the ANY-maze software . The arena and object maps were defined in ANY-maze , and allowed for automated tracking of the animal’s area , with the ability to distinguish between the head and tail . Entry of the animal’s head to the defined object’s area was scored for time interacting with the object . An experimenter watched the recorded videos and overlaid tracking in real-time to verify that tracking of the animal’s area and head/tail orientation was consistent and that scoring of object interaction was accurate . Performance on the NOR test is indicated by a discrimination index calculated of time spent with the novel object ( Tnovel ) divided by the sum of time spent with both objects ( Tnovel+Tfamiliar ) . One animal from the misaligned group ( out of 8 ) tested during the day failed to sufficiently explore both objects ( < 10 sec during the second familiarization trial ) and was eliminated from further analysis . For all novel object recognition experiments , mice were trained and tested only once to avoid effects of prior manipulation . Results reported for daytime and night-time tests are from distinct cohorts of animals . All statistical analysis was performed using Sigma Stat ( ver . 12 ) . Effects of misaligned feeding on activity , sleep , PER2-driven bioluminescence , and LTP were determined using unpaired t-tests in comparison to the aligned group . In cases of unequal variance , the Mann-Whitney rank sum test was applied . To determine the effects and interaction of time of day and scheduled feeding condition on protein expression , and hippocampal-dependent learning and memory , we applied two factor analysis of variance ( ANOVA ) tests . The Holm-Sidak post hoc test was used to distinguish differences due to misaligned feeding . Differences with p < 0 . 05 were deemed significant in all analyses . Where appropriate ( Figure 3 , 4 , 6 ) , box plots indicate the 25th and 75th percentiles , with error whiskers indicating the 10th and 90th percentiles . Values in the text , line graphs , and bar graphs are reported as mean ± standard error mean ( SEM ) . | Many processes within the body follow an approximately 24-hour cycle . In addition to patterns of sleep and wakefulness , such circadian rhythms help to regulate body temperature , blood pressure and hormone levels . They also affect when we feel hungry , when our muscles work most efficiently , and when we are mentally at our sharpest . A region of the brain called the suprachiasmatic nucleus ( SCN ) generates and maintains circadian rhythms , and thus acts as the body’s master clock . Daily exposure to light keeps the SCN synchronized with the 24-hour day/night cycle . However , most organs , from the heart to the pancreas , also possess their own clocks , which help to regulate organ-specific processes . These secondary clocks normally operate in synchrony with the SCN . Exposure to light has long been known to influence circadian rhythms . However , more recent evidence suggests that the timing of meals may also affect circadian clocks , particularly those within the digestive system . Loh et al . therefore decided to investigate whether eating outside normal waking hours would also affect other key physiological processes , specifically the cognitive processes of learning and memory . Mice normally consume most of their food after sunset . Loh et al . showed that rodents that were instead fed during the day performed less well on cognitive tests than other mice who received the same food at night . The daytime-fed mice showed changes in a region of the brain called the hippocampus , which supports learning and memory . In particular , daytime feeding changed the timing of the secondary circadian clock within the hippocampus , although it had no effect on the master clock in the SCN . Loh et al . therefore suggest that the misalignment of these circadian clocks impairs cognition . Further experiments are needed to determine whether a similar relationship exists between the timing of meals and cognitive performance in humans . If so , these findings will have implications for the many individuals whose mealtimes , for work or social reasons , are out of synchrony with their body clocks . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Misaligned feeding impairs memories |
Chromosome instability ( CIN ) is deleterious to normal cells because of the burden of aneuploidy . However , most human solid tumors have an abnormal karyotype implying that gain and loss of chromosomes by cancer cells confers a selective advantage . CIN can be induced in the mouse by inactivating the spindle assembly checkpoint . This is lethal in the germline but we show here that adult T cells and hepatocytes can survive conditional inactivation of the Mad2l1 SAC gene and resulting CIN . This causes rapid onset of acute lymphoblastic leukemia ( T-ALL ) and progressive development of hepatocellular carcinoma ( HCC ) , both lethal diseases . The resulting DNA copy number variation and patterns of chromosome loss and gain are tumor-type specific , suggesting differential selective pressures on the two tumor cell types .
Aneuploidy , the presence of an abnormal number of chromosomes in a cell , is a hallmark of solid tumors and in human cancers is frequently an indicator of poor prognosis . Genomic instability has the potential to promote loss of tumor suppressors and increases in oncogene copy number , thereby driving tumorigenesis . However , experiments in non-transformed cells show that chromosome imbalance imposes a physiological burden that reduces cell fitness ( Torres et al . , 2007; Williams et al . , 2008; Kops et al . , 2004; Mao et al . , 2003 ) . Primary murine embryonic fibroblasts ( MEFs ) engineered to carry an extra chromosome grow more slowly than wild-type cells and exhibit significant changes in metabolism . The same is true of cells from Down syndrome patients , which carry a supernumerary chromosome 21 ( Williams et al . , 2008; Jones et al . , 2010 ) . It remains poorly understood how the oncogenic effects of genomic instability as a driver of gene gain and loss , and the burden of aneuploidy in reducing fitness play out in real tumors . It has been suggested that tumors experience a burst of chromosome instability ( CIN ) leading to the emergence of clones with greater oncogenic potential but that CIN is then suppressed so that cancer cells maintain a relatively stable karyotype ( Lengauer et al . , 1997; Wang et al . , 2014 ) . This model of ‘genome restabilization’ is supported by statistical analysis of tumor karyotype across large numbers of human cancer genomes including identification of a common cancer karyotype , overlaid by tissue-specific differences , in genomic data from The Cancer Genome Atlas ( Davoli et al . , 2013 ) . In addition , aneuploidy appears most common in areas of the genome having more oncogenes and tumor suppressors , implying positive selection . Conversely , however , many solid tumors are genetically heterogeneous , suggesting that CIN is not fully suppressed in growing tumors ( Nicholson and Cimini , 2013 ) . Mouse models of CIN provide an excellent tool for studying the role of aneuploidy in tumorigenesis . One method to induce CIN is to disrupt the spindle assembly checkpoint ( SAC ) . The SAC senses the presence of maloriented or detached kinetochores and blocks exit from mitosis until pairs of sister chromatids achieve the bipolar geometry that is uniquely compatible with normal chromosome disjunction ( Jallepalli and Lengauer , 2001; Taylor et al . , 2004; Rieder et al . , 1995 ) . Studying the SAC in mice is complicated by the fact that germline deletion of murine SAC genes is lethal by ~E10 . 5 ( Li et al . , 2009; Dobles et al . , 2000; Babu et al . , 2003; Baker et al . , 2004; García-Higuera et al . , 2008; Iwanaga et al . , 2007; Perera et al . , 2007; Putkey et al . , 2002; Wang et al . , 2004 ) . As a result , most studies of SAC knockouts to date have employed heterozygous animals or hypomorphic alleles , resulting in weak and sporadic tumor development at long latencies ( Iwanaga et al . , 2007; Burds et al . , 2005; Dai et al . , 2004; Michel et al . , 2001 ) . We have previously shown that deletion of Mad2l1 ( HUGO MD2L1; UniProt Q9Z1B5 ) , an essential component of the SAC , is tolerated by murine interfollicular epidermal cells , which terminally differentiate to form the outer layers of the skin , but not by hair follicle bulge stem cells , a specialized self-renewing cell type required for hair follicle maintenance ( Foijer et al . , 2013 ) . These findings support the idea that a functional SAC is required in cells undergoing repeated division , but not necessarily in differentiated cells with limited proliferative potential . The implications for cancer are unclear , since cancers grow from one or a small number of cells which must divide many times to create a macroscopic tumor . In this paper we describe an analysis of tumorigenesis in mice carrying a conditional Mad2l1 deletion in a highly proliferative cell type ( T-cells ) and in a second cell type in which proliferation is induced by injury ( hepatocytes ) . To tolerize cells to checkpoint loss we also introduced conditional deletions or mutations in Trp53 . We found that T-cells and hepatocytes survive checkpoint loss in both the presence and absence of Trp53 mutation but that Trp53 mutations promote oncogenesis ( Jacks et al . , 1994; Purdie et al . , 1994; Donehower et al . , 1992 ) . In T-cells , loss of Mad2l1 and Trp53 causes rapidly growing acute lymphoblastic leukemia ( T-ALL ) and in hepatocytes it causes progressive disease that ends in lethal hepatocellular carcinoma ( HCC ) . Single-cell sequencing shows that Mad2l1-null T-ALLs experienced an elevated rate of chromosome mis-segregation relative to normal T cells and murine T-cells in which the SAC is partially inactivated by truncation of Mps1 ( Mps1 is another SAC component; ( Bakker et al . , 2016; Foijer et al . , 2014 ) . In contrast , when Mad2l1-null T-ALLs and HCCs were assayed at a population level using array-based comparative genomic hybridization ( aCGH ) recurrent and tissue-specific patterns of chromosome loss and gain were observed . The differences between single-cell and population-level measures of aneuploidy are most parsimoniously explained by postulating that Mad2l1-null tumors experience ongoing CIN but that specific aneuploid genomes predominate in a tumor as a result of tissue-specific selection .
To cause CIN in a tissue-restricted fashion , we engineered a conditional flanked-by-LOX allele of Mad2l1 ( Mad2l1f ) . LoxP sites were inserted upstream of exon 2 and downstream of exon 5 so that Cre expression would result in deletion of ~90% of the Mad2l1 ORF ( Mad2l1Δ; Figure 1A and Figure 1—figure supplement 1a ) . Correct targeting of the construct in ES cells was confirmed by Southern blotting ( Figure 1—figure supplement 1b ) . By crossing Mad2l1f and Lck-Cre transgenic animals we induced recombination of Mad2l1f in CD4– CD8– T cells ( Molina et al . , 1992 ) and by crossing with Alb-Cre carrying mice we induced Mad2l1f recombination in developing hepatocytes ( Weisend et al . , 2009 ) , a post-mitotic cell type that normally exhibits polyloidy ( Duncan et al . , 2010 ) . In both tissues , genotyping showed that Mad2l1 excision was efficient ( Figure 1B , C ) . We generated a tumor-sensitized background by crossing Cre transgenic Mad2l1f/fand FLOX-Trp53 ( Trp53f ) mice ( Jonkers et al . , 2001 ) ; Trp53 loss has been shown to promote survival of Mad2l1-deficient murine cells ( Burds et al . , 2005 ) . 10 . 7554/eLife . 20873 . 003Figure 1 . Tissue specific loss of Mad2l1 leads to T-cell acute lymphoblastic lymphoma in T cells in a permissive Trp53null background . ( A ) Schematic overview of the Mad2l1 conditional allele before and after Cre-mediated recombination . The red triangles refer to the loxP sites that surround exon 2 to exon 5; roman numerals refer to exons . ( B , C ) PCR for Mad2l1 genotypes and recombination of the Mad2l1f allele in ( B ) thymocytes and ( C ) liver tissue ( L ) or tail tissue ( T ) . ( D ) Kaplan Meier plots showing survival of the indicated genotypes for Lck-Cre::Mad2l1f/f::Trp53f/f compared to control mice . Statistical tests for compared Mad2l1f/f and Mad2l1+/+ having same Trp53 genotype , **p<0 . 01 ( Mantel-Cox test ) . Control curves ( Lck-Cre::Trp53f/fand Lck-Cre ) were same animals as used in Foijer et al . ( 2014 ) . ( E ) Images showing enlarged thymus and spleen in a Lck-Cre::Mad2l1f/f::Trp53f/f mouse compared to a healthy control . ( F ) Average thymus and spleen weights for tumor-bearing Lck-Cre::Mad2l1f/f::Trp53f/f mice compared to unaffected control mice . ( G ) Representative H&E staining of control thymus ( upper panel ) and Lck-Cre::Mad2l1f/f::Trp53f/f acute T acute lymphoblastic lymphoma sample with staining indicating an undifferentiated cell state ( lower panel ) . Scale bar 10 microns . ( H ) Forward and side scatter ( FSC , SCC ) plots for normal ( appearing ) thymuses and a T-ALL showing the emergence of a larger blasting population , before thymus size increased ( upper panels ) . FSC and SCC plots for thymus , blood and spleen of a tumor-bearing mouse , showing blasting cells in thymus and spleen , but not blood ( lower panels ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 00310 . 7554/eLife . 20873 . 004Figure 1—figure supplement 1 . Complete Targeting Vector and Generation of Mad2l1f/f Mice . ( A ) Targeting vector and locus with restriction site changes highlighted . Recombination of floxed allele results in excision of exons 2–5 . Arrows indicate PCR primer sites . ( B ) Restriction digest and outhern blot showing correct recombination at the murine Mad2l1f/f locus . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 00410 . 7554/eLife . 20873 . 005Figure 1—figure supplement 2 . Representative array CGH profiles for 3 Lck-Cre::Mad2l1f/f::Trp53f/f tumors showing clonal loss at the TCR loci on chromosomes 6 and 14 indicating tumor clonality . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 005 Lck-Cre::Mad2l1f/f::Trp53+/+ mice did not experience malignancies within the first year of life ( Figure 1D ) and adult T cells from these animals developed normally , suggesting that T cells are tolerant of Mad2l1 loss . On a Trp53-heterozygous background ( a Lck-Cre::Mad2l1f/f::Trp53f/+ genotype ) loss of Mad2l1 resulted in death of ~50% of animals by ~8 mo . ( from T-ALL , see below ) whereas control animals heterozygous for a Trp53 deletion but carrying wild type Mad2l1 had the same lifespan as wild-type littermate controls ( Figure 1D; blue lines , p<0 . 01 ) . On a Trp53-homozygous deletion background , loss of Mad2l1 ( i . e . an Lck-Cre::Mad2l1f/f::Trp53f/f genotype ) resulted in rapid disease progression with half of double mutant animals dead by ~4 mo . ( Figure 1D; red line ) . Double mutant mice experienced a statistically significant acceleration in cancer development relative to mice homozygous for Trp53 deletion , which itself is known to be highly tumorigenic in thymocytes ( Figure 1D; compare green and red lines , p<0 . 01 ) . Dyspnea ( labored breathing ) was observed shortly before the death of Mad2l1-mutant animals , consistent with thymic hypertrophy . Post-mortem analysis of tissues revealed a ~10–15 fold increase in the average mass of the thymus and 70% increase in the mass of the spleen ( Figure 1E , F ) . Histological analysis of thymi demonstrated the presence of rapidly dividing blasts with irregular nuclei and abnormal DNA , consistent with lymphoma ( Figure 1G , compare top and bottom panels ) . We conclude that Mad2l1 and Trp53 loss cooperate in oncogenic transformation of T-cells and that the combination is rapidly lethal . To characterize cellular defects in double knockout mice , ~20 Lck-Cre::Mad2l1f/f::Trp53f/f animals were euthanized prior to the appearance of dyspnea and thymocytes then analyzed . FACS showed that thymi from these animals contained numerous dividing and undifferentiated ( blasting ) cells in comparison to thymi from control animals ( Figure 1H; blasts , red arrow , normal cells black arrow ) . In ~10% of Lck-Cre::Mad2l1f/f::Trp53f/f animals thymi were macroscopically normal but they also contained an abnormal number of dividing and undifferentiated cells showing that this phenotype was fully penetrant . In most animals , the spleen also contained blasting cells and was enlarged , suggesting metastasis of T-cells to this organ ( Figure 1H compare blasting population highlighted by red arrow in lower right panel with blasting cells in thymus ) . However , blasts were not observed in the peripheral blood ( Figure 1H bottom center panel ) . These and related data show that the majority of Lck-Cre::Mad2l1f/f::Trp53f/f animals suffered from poorly differentiated CD4+ and CD8+ T-acute lymphoblastic lymphoma ( T-ALL ) while a subset suffer from more differentiated CD4+ or CD8+ T-ALL as described previously for Trp53null thymic lymphoma ( Donehower et al . , 1995 ) . Array-based comparative genomic hybridization ( CGH ) analysis of TCR α and β loci on chromosomes 14 and 6 revealed a single dominant rearranged TCR in each animal implying that T-ALLs were clonal ( see Figure 1—figure supplement 2 , array CGH data was deposited at GSE63686 in NCBI GEO ) . In sum , these data demonstrate synergy between loss of Mad2l1 and Trp53 in the transformation of T-cells to malignant T-ALL and show that tumors grow large enough to kill animals while remaining clonal at TCR loci . Mad2l1-null cells must therefore proliferate extensively subsequent to a tumor-initiating genetic event ( a common characteristic of cancer ) . The lifespan of animals carrying a conditional knockout of Mad2l1 in hepatocytes ( Alb-Cre::Mad2l1f/f::Trp53+/+ mice ) was unchanged relative to wild-type littermates ( Figure 2A ) but double mutant animals deleted for Mad2l1 and one or both Trp53 alleles ( Alb-Cre::Mad2l1f/f::Trp53f/+ and Alb-Cre::Mad2l1f/f::Trp53f/f mice ) died significantly younger than littermate Mad2l1+/+::Trp53+/+ controls ( p<0 . 01 and p<0 . 001 respectively ) . In contrast , liver-specific deletion of one or both Trp53 alleles in Mad2l1-wild type animals had no detectable impact on lifespan ( Figure 2A; Trp53f/+ blue lines , p<0 . 01; Trp53f/f red and green lines , p<0 . 001 ) consistent with previous data showing that Trp53 loss is only mildly oncogenic in hepatocytes ( Harvey et al . , 1993 ) . Post-mortem analysis of Alb-Cre::Mad2l1f/f::Trp53f/f and Alb-Cre::Mad2l1f/f::Trp53f/+ animals revealed the presence of one or more liver tumors per mouse . In many cases these tumors were so large and invasive that the tri-lobular structure of the liver was unrecognizable ( Figure 2B ) . 10 . 7554/eLife . 20873 . 006Figure 2 . Loss of Mad2l1 in hepatocytes results in multifocal hepatocellular carcinoma . ( A ) Kaplan Meier plots showing survival of the indicated genotypes for Alb-Cre::Mad2l1f/f::Trp53f/f compared to control mice . Statistical tests compared Mad2l1f/f::Trp53f/f and Mad2l1f/f::Trp53f/+ to Mad2l1+/+::Trp53+/+ mice , **p<0 . 01 , ***p<0 . 001 ( Mantel-Cox test ) . ( B ) Images showing multifocal disease in a Alb-Cre::Mad2l1f/f::Trp53f/f mouse compared to a healthy control . ( C ) Histological sections from Alb-Cre::Mad2l1f/f::Trp53+/+ mice . Left panels show regenerative nodules ( RN ) marked by black arrows in low magnification field and cells in one nodule at high magnification . Centre panels show an HCA marked by the asterisk . Right panels show an HCC marked by an asterisk ( top ) or the entire field ( bottom ) . Scale bars in top fields 1 mm , bottom fields 0 . 1 mm . ( D ) Incidence of HCA and HCC in the livers of Control ( Mad2l1+/+::Trp53+/+ or Alb-Cre– ) , Alb-Cre::Mad2l1f/f::Trp53+/+ , and Alb-Cre::Mad2l1f/f::Trp53f/f . ( E ) Kaplan Meier plot showing survival of Alb-Cre::Mad2l1f/f::Trp53R246S mice compared to control . ***p<0 . 001 ( Mantel-Cox test ) . ( F ) Representative MRI images of an Alb-Cre::Mad2l1f/f::Trp53f/f with a tumor ( white arrow ) over time in weeks . EOVIST is used as a contrast agent and tumors exclude the reagent and are dark . ( G ) Volumetric measurements of tumors over time . Each symbol represents a different mouse . ( H ) Doubling time as determined by semi-log regression of data in ( G ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 00610 . 7554/eLife . 20873 . 007Figure 2—figure supplement 1 . Characterizing aneuploid T-ALLs and HCCs . Workflow from MRI imaging to histological examination of liver malignancies . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 007 In the case of single-mutant Alb-Cre::Mad2l1f/f animals , widespread liver damage and formation of regenerative nodules was evident by ~4 mo . of age . Regenerative nodules are non-neoplastic sites of liver proliferation and repair commonly present following liver damage . By 8–12 mo . of age benign hepatocellular adenoma ( HCA ) was evident in >50% of animals ( ~5% of wild type animals also had HCA , which is normal for this genetic background [Leenders et al . , 2008] ) and by 12–16 mo . half of mice had hepatocellular carcinoma ( HCC; Figure 2C ) . Thus Alb-Cre::Mad2l1f/f mice experienced benign and malignant liver cancer at high penetrance . Deletion of Trp53 on this background ( Alb-Cre::Mad2l1f/f::Trp53f/f animals; Figure 2D ) dramatically accelerated the onset of cancer , with 75% of 8–12 mo . old animals exhibiting HCC , occasionally in combination with HCA or cholangiocarcinoma ( Table 1 ) . Note that the reduction in the proportion of HCC Alb-Cre::Mad2l1f/fmice in the cohort older than 16 mo . arises simply because tumor-bearing animals die at an accelerated rate leaving behind disease-free animals . This is less obvious for Alb-Cre::Mad2l1f/f::Trp53f/f animals because nearly all of them eventually get HCC ( so increasing death is matched by increasing disease prevalence ) . 10 . 7554/eLife . 20873 . 008Table 1 . Incidence of HCA , HCC , and CC ( cholangiocarcinoma ) in Alb-Cre::Mad2l1::Trp53 mice . The HCA/CC column describes the number of mice with HCC that also have HCA or CC . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 0084–8 ( Mo ) 8–12 ( Mo ) 12–16 ( Mo ) 16–20 ( Mo ) 20–24 ( Mo ) HCAHCCHCACCHCAHCCHCAHCAHCCHCAHCAHCCHCAHCAHCCHCACCCCCCCCCCCCMad2l1f/f: :Trp53+/+2/170/17/0/178/141/14/1/42/4/2/136/135/619/345/343/50/34///Mad2l1f/f: :Trp53f/f2/145/142/51/141/2722/278/220/99/94/90/1513/156/130/44/41/40/42/5/1/92/4Mad2l1f/f: :Trp53f/+2/122/121/21/122/173/171/30/85/83/511/2512/256/123/105/103/50/10/2/32/12Mad2l1f/+: :Trp53f/f0/100/10/0/100/100/10/0/10/1/1/112/11/2/134/131/41/13///1/22/4Mad2l1f/+: :Trp53f/+0/130/13/0/130/70/7/0/30/3/2/80/8/0/82/8/0/8/////Mad2l1f/+: :Trp53+/+0/150/15/0/150/60/6/0/30/3/1/80/8/2/132/131/20/13////Mad2l1+/+: :Trp53f/f0/20/2/0/20/40/4/0/21/21/23/102/101/20/11/1/0/1//1/2//Mad2l1+/+: :Trp53f/+NANANANANA0/20/2/0/40/4/0/10/1/0/1Mad2l1+/+: :Trp53+/+0/30/3/0/30/20/2/0/30/3/0/30/3/0/40/4/0/4Alb-cre–0/60/6/0/60/90/9/1/70/7/NANANANANA People living in areas of the globe in which hepatotoxic agents are endemic often suffer from HCC that involves a dominant-negative Trp53-R249S mutation ( Yin et al . , 1998; Hsu et al . , 1991; Lee and Sabapathy , 2008 ) . When we crossed the murine analog of this mutation ( Trp53-R246S ) into Mad2l1-mutant animals ( Alb-Cre::Mad2l1f/f::Trp53R246S mice; [Yin et al . , 1998] ) we observed early death and extensive HCC ( Alb-Cre::Mad2l1f/f::Trp53R246S vs . Alb-Cre:Trp53R246S littermates; p<0 . 001; compare Figure 2A and E , red lines ) . Such animals therefore recapitulate a known feature of human disease . We conclude that Mad2l1 deletion in hepatocytes is sufficient to cause HCC but that tumor formation is significantly accelerated by deletion of Trp53 or introduction of a mouse analog of a Trp53 mutation commonly observed in human liver cancer . To analyze disease progression in Alb-Cre::Mad2l1f/f::Trp53f/f animals we used magnetic resonance imaging with EOVIST as a contrast reagent . EOVIST-excluding regions were confirmed to be tumors by fixing livers immediately after MRI followed by serial sectioning and H&E histology of the same region of the tissue ( Figure 2—figure supplement 1 ) . In nine animals examined ( a total of 32 tumors ) we observed that tumor volume increased approximately exponentially ( Figure 2F , G ) with an average doubling time of ~28 days ( Figure 2H ) . Moreover , imaging revealed that the number of tumors in each animal increased with age so that by 12 mo . an average of 3 morphologically distinct tumors were present in each animal ( range 1–7; n = 9 ) . We conclude that liver cancer induced by loss of Mad2l1 and Trp53 results in progressive multi-focal cancer and that tumors grow exponentially once established . This resembles human HCC , which is also multi-focal and progressive . Although we have previously shown that partially inactivating the SAC via truncation of Mps1 in T-cells accelerates Trp53-induced lymphomagenesis ( Foijer et al . , 2014 ) , the observation that Mad2l1-null T-ALLs and HCCs can grow rapidly is surprising . Germline deletion of murine Mad2l1 is lethal and RNAi-mediated depletion of Mad2l1 in cancer cells results in mitotic catastrophe and cell death within approximately six cell doublings ( Kops et al . , 2004; Dobles et al . , 2000 ) . To establish that the Mad2l1 locus was in fact lost in tumor cells and that no RNA or protein was expressed , we analyzed DNA structure by genomic PCR ( Figure 3A ) and aCGH ( Figure 3B ) , mRNA levels using qPCR ( Figure 3C ) , and protein levels by Western blotting of tumor samples ( Figure 3D ) . In all but one T-ALL tumor from Lck-Cre::Mad2l1f/f::Trp53f/f animals ( tumor 33 , in which switching was incomplete and protein present ) Mad2l1 DNA , mRNA , and protein were below the level of detection . In the case of HCCs from Alb-Cre::Mad2l1f/f::Trp53f/f animals , PCR revealed bands corresponding to both recombined and unrecombined Mad2l1 ( Mad2l1Δ and Mad2l1f ) . Probe values by aCGH were generally intermediate between probe values for wild-type cells and Mad2l1-null T-ALLs ( compare Figure 3B to Figure 3E ) . The presence in HCC DNA from recombined and unrecombined Mad2l1f loci is expected since liver comprises multiple cell types and tumors contain high levels of infiltrating Mad2l1-proficient immune cells . Mad2l1-null HCCs were invasive so resected tumors were invariably contaminated with surrounding normal tissue ( including tissue in which recombination might have been incomplete ) . It is difficult to estimate the contribution of such cells to PCR signals , but in a subset of HCCs , Mad2l1Δ was the dominant PCR product ( tumors in lanes 6 and 11 in Figure 3F ) . It is also possible that cells heterozygous for Mad2l1 deletion can contribute to tumorigenesis . However , germline Mad2l1 heterozygosity did not cause HCC either alone or in combination with Trp53 deletion and we therefore find this explanation less likely ( Table 1 ) . Overall we conclude that Mad2l1 protein and mRNA are present in amounts below the level of detection in virtually all T-ALLs and that in some HCCs , the extent of recombination is sufficiently high that some , and perhaps all transformed hepatocytes lack a functional Mad2l1 gene . 10 . 7554/eLife . 20873 . 009Figure 3 . Excision of Mad2l1 DNA and loss of SAC function in normal and tumor cells . ( A ) Recombination efficiency of Mad2l1f and Trp53f alleles in T-ALL samples as measured by genomic PCR . Numbers refer to tumor ID . ( B ) Recombination efficiency at Mad2l1 and Trp53 loci in T-ALL or samples as measured by array CGH . Each rectangle represents a single aCGH probe value , three probes values are shown per conditional gene: one probe recognizing the Mad2l1 or Trp53 deleted fragment ( middle ) and two probes flanking the 5’and 3’ sides of the deleted region . ( C ) Quantitative PCR showing complete deletion of Mad2l1 ( probe A ) and Trp53 ( probes A and B ) in T-ALL samples . Error bars show SEM for six ( Mad2l1 , Trp53 ) and three ( Trp53 ) tumors ( biological replicates ) . ( D ) Western blots showing loss of Mad2l1 expression in T-ALL samples . ( E ) Recombination efficiency at Mad2l1 and Trp53 loci in HCC samples as measured by array CGH . ( F ) Genomic PCR of tumor tissue for WT , FLOX , and recombined alleles of Mad2l1 and Trp53 . Black vertical line shows where an empty lane was removed . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 009 To demonstrate that recombination in Mad2l1f/f cells abrogates checkpoint control we established Mad2l1f/f mouse embryonic fibroblasts ( MEFs ) from E12 . 5-E13 . 5 embryos and infected cells with a retrovirus that expressed doxycycline- ( Dox ) inducible Cre ( Dox-Cre; [Foijer et al . , 2014] ) . Addition of Dox to Mad2l1f/f MEFs that express Dox-Cre resulted in efficient excision of Mad2l1 as judged by PCR genotyping and Western blotting for Mad2l1 protein ( Figure 4A ) . The resulting Mad2l1-null MEFs could be passaged multiple times in low oxygen conditions . Exposure of wild-type MEFs to nocodazole for 6 hr . increased mitotic index ~20 fold , reflecting checkpoint-dependent detection of microtubule depolymerization and imposition of cell cycle arrest . In contrast , when Mad2l1Δ/Δ MEFS were exposed to nocodazole , only a modest increase in mitotic index was observed , consistent with a loss of SAC–mediated mitotic delay ( Figure 4B and Video 1 ) . 10 . 7554/eLife . 20873 . 010Figure 4 . Mad2l1 inactivation in mouse embryonic fibroblasts fully alleviates the spindle assembly checkpoint . ( A ) Recombination efficiency measured by genomic PCR ( top ) and Western blot ( bottom ) showing partial to complete Mad2l1 deletion in MEFs following retroviral doxycycline-inducible Cre . Actin serves as loading control . ( B ) Average phospho-Histone H3 staining of dox-inducible Cre-transduced MEFs following 6 hr of nocodazole treatment . Error bars show the SEM of at least two biological replicates . ( C ) Average mitotic index of thymocytes isolated from Paclitaxel or control-injected mice 4–6 hr post-treatment . Error bars show SEM of at least four biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 01010 . 7554/eLife . 20873 . 011Video 1 . SAC loss in dox-inducible Cre-transduced Mad2l1f/f MEFs . Video shows the instant mitotic exit of dox-inducible Cre-transduced Mad2l1f/f MEFs in the presence of nocodazole , which indicates loss of SAC function . Chromatin is labeled with H2B-GFP . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 011 When wild type and Lck-Cre::Mad2l1f/f::Trp53+/+ mice were injected with the microtubule stabilizing drug Paclitaxel , the fraction of phospho-H3 positive CD4+CD8+ thymocytes ( the most rapidly dividing thymic population ) increased 6-fold in Mad2l1-sufficient animals but only 2 . 5-fold in Mad2l1-null animals ( Figure 4C ) . Thus , the mitotic checkpoint is functionally impaired in Mad2l1Δ/Δ thymocytes in vivo , consistent with loss of expression of a protein required for the spindle assembly checkpoint protein . We conclude that loss of SAC function is compatible with rapid growth of both liquid and solid tumors in the mouse . The partial mitotic arrest observed in CD4+CD8+ thymocytes from Lck-Cre::Mad2l1f/f::Trp53+/+ animals might reflect the presence of Mad2l1-proficient T-cells in circulation combined with the inability of paclitaxel to impose a strong arrest on wild-type murine thymocytes ( so the positive signal is not very high ) . Alternatively , it is possible that Mad2l1Δ/Δ thymocytes are partially responsive to microtubule depolymerization and retain some SAC function . We are as-yet unable to distinguish between these possibilities . Despite attempts to raise suitable anti-mouse Mad2l1 antibodies or purchase them from commercial suppliers , we have been unable to perform sufficiently good immunofluorescence microscopy against murine Mad2l1 to determine whether the subset of thymocytes that arrest in the presence of paclitaxel express Mad2l1 . Further development of single-cell methods will be needed to resolve this issue . To study changes in the genome accompanying Mad2l1 deletion we performed genome-wide aCGH and microarray transcriptome analysis of T-ALL and HCCs ( see Materials and methods , NCBI GEO , GSE63689 ) . T-ALLs had a median of 3 whole chromosome loss or gain events per tumor , whereas HCCs had a median of 2 ( Figure 5A , p>0 . 05 ) . Loss of Chr13 was frequent in both tumor types ( Figure 5B , C; green point in Figure 5C ) as was gain of Chr15 ( Figure 5B , C; red point in Figure 5C ) which was also the most common whole-chromosome aneuploidy overall . Chromosomes also exhibited tissue-specific patterns of loss and gain: T-ALLs frequently gained Chr4 and Chr12 while HCCs lost these chromosomes ( Figure 5B , C; Chr4: p<0 . 001 , Chr12: p<0 . 01 , blue points in 4E ) . In contrast , Chr18 exhibited preferential gain in HCCs but loss in T-ALLs ( Figure 5B , C; p<0 . 001 , blue points in Figure 5C; aCGH of individual tumors is shown in Figure 5—figure supplement 1A for T-ALLs and Figure 5—figure supplement 1B for HCCs ) . One known effect of aneuploidy is to alter mRNA expression through gene dosage . On a per-chromosome basis we observed statistically significant correlation between chromosome ploidy and levels of mRNA expression from genes expressed on that chromosome ( Figure 5D , E ) . However , the genes that were differentially expressed in HCC and T-ALL were not the same . This was true even for genes expressed on Chr15 , which was gained in both HCC and T-ALL ( Figure 5F , G ) . When we used Webgestalt ( Zhang et al . , 2005 ) to determine whether the same pathways were affected in T-ALLs and HCC , the only common denominators were metabolic pathways , in agreement with earlier findings showing that aneuploidy disrupts metabolism in multiple organisms ( NCBI GEO GSE63689 ) ( Williams et al . , 2008; Torres et al . , 2007; Foijer et al . , 2014 ) . We conclude that deletion of Mad2l1 results in loss and gain of whole chromosomes in both HCC and T-ALL but that the pattern of gain and loss is tissue specific , most likely due to changes in gene expression that are themselves tissue-specific . Moreover , even when the same chromosome is gained in HCC and T-ALL , the genes that exhibit differential expression are not the same . 10 . 7554/eLife . 20873 . 012Figure 5 . Aneuploidy events are a recurrent genetic lesion in T-ALLs and HCCs . ( A ) Box and whiskers plot of whole chromosome gain-loss events in tumors , scored by averaging the log2 ratio of aCGH fluorescence from tumor over normal ( aCGH ratio ) for each chromosome with a ±0 . 3 cut-off . Line – median , Box – interquartile range , Whisker – range . ( B ) Box and whiskers plot of the aCGH ratio for each chromosome . Two-way ANOVA , Bonferroni post-test **p<0 . 01 , ***p<0 . 001 ( C ) Scatter plot showing average chromosome aCGH ratio of HCC and thymus tumors . Gain of Chr15 ( red ) , loss of Chr13 ( green ) , and tissue specific gain or loss of Chr4 , Chr12 , and Chr18 ( blue ) . ( D , E ) Scatter plots of average chromosome aCGH ratio plotted against average mRNA ratio for ( D ) HCC and ( E ) T-ALL . r is the Pearson Correlation with indicated P value . ( F , G ) Expression analysis of genes on chromosome 15 ( F ) and chromosome 18 ( G ) comparing HCC and T-ALL samples . ( H ) Scatter plot showing chromosome normalized aCGH ratio for every gene in HCC and T-ALL . The three listed genes are likely hybridization artifacts due to a mixed 129/C57BL6 background . ( I , J ) Number of focally amplified or deleted regions per tumor ( I ) and the number of genes amplified or deleted per tumor ( J ) scored with ±0 . 3 cut-off for the chromosome normalized aCGH ratio . *p<0 . 05 , **p<0 . 01 , Mann-Whitney Test comparing T-ALL and HCC . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 01210 . 7554/eLife . 20873 . 013Figure 5—figure supplement 1 . Copy number changes in T-ALL and HCC as assessed by aCGH . ( A , B ) Individual aCGH plots for T-ALLs ( A ) and HCCs ( B ) showing chromosome copy number alterations for individual tumors . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 01310 . 7554/eLife . 20873 . 014Figure 5—figure supplement 2 . T-ALL and HCC specific CNVs . ( A ) Graph of the chromosome normalized aCGH ratio for the genomic region around Pten for T-ALL samples . Colored lines have deletions in Pten . ( B ) Graph of the normalized aCGH ratio for Chr6 ( left ) focused on the genomic region around Met ( right ) . Blue dots are individual aCGH probes . Red line is a moving average of 3 adjacent probes . ( C , D ) Met is over expressed by ( C ) mRNA as measured by quantitative PCR , mouse number indicates normal while mouse number + LT indicates tumor , and ( D ) protein as measured by Western Blot . NT is normal tissue , T is tumor tissue . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 014 For both T-ALL and HCC aCGH revealed recurrent copy number variants ( CNVs ) , focal changes in chromosome structure , some of which included known oncogenes and tumor suppressors . We found that ~30% of T-ALLs carried a 1–8 gene deletion spanning Pten ( Figure 5—figure supplement 2A ) , a negative regulator of the PI3 kinase and 5–10% of HCCs carried an amplification in Met , a known oncogene in human HCC ( Zender et al . , 2006 ) . In the latter case , we confirmed that Met RNA and protein were over-expressed ( Figure 5—figure supplement 2B–D ) . We detected no consistently amplified or deleted genes common to both thymus and liver tumors: three possible candidates uncovered by aCGH proved to be artifacts arising from our use of a mixed 129 x C57BL/6 background ( Boyden and Dietrich , 2006 ) ( Figure 5H ) . Moreover , the overall structure and pattern of gene amplifications and deletions appeared to differ in the two tumor types . Focal CNVs were more frequent in T-ALL than HCC: T-ALLs carried a median of 29 CNVs per tumor with an average size of ~40 genes per CNV whereas HCCs carried ~20 CNVs per tumor with an average size of ~2–3 genes ( both differences were statistically significant , ~p< 0 . 05 and p<0 . 01 respectively; Figure 5I , J ) . This implies either differential selection for CNVs in T cells and hepatocytes or differences in chromosome breakage and rejoining as a consequence of SAC loss . We note a potential complication in the interpretation of this data . Whereas T-ALLs appear to be clonal based on TCR sequence HCCs are likely to be polyclonal even after physical dissociation of tumor masses . The degree of clonality might affect our assessment of CNV characteristics by aCGH ( which averages across all cells in the sample ) . Analysis of additional tumors using single-cell methods will therefore be required to confirm the observation that Mad2l1 deletion generates CNVs with different sizes in different tissues . Mice lacking Mad2l1 in hepatocytes exhibited a characteristic progression from regeneration nodules to HCA and then HCC . To determine if HCA actually gives rise to HCC in Alb-Cre::Mad2l1f/f::Trp53f/f and Alb-Cre::Mad2l1f/f::Trp53f/+ animals , gain and loss of chromosomes was profiled by aCGH in HCA , HCC , and unaffected liver ( see Materials and methods , NCBI GEO GSE63689 and GSE63100 ) . Chromosomes 15 , 16 , and 18 were frequently gained in HCC ( as described above ) and also in HCA , although signals were weaker in HCAs implying that only a subset of cells had undergone chromosome gain events ( Figure 6A ) . Overall , HCA appeared more aneuploid than normal tissue and HCC more aneuploid than HCA ( Figure 6B ) . We assayed the relatedness of HCA and HCC by sequencing tumors from eight animals carrying a macroscopic HCA and one HCC and an additional mouse carrying one HCA and two HCCs ( see Materials and methods , data deposited at Sequence Read Archive accession number SRA191233 ) . CNVs in the benign and malignant tumors were compared pairwise using the Jaccard Index , which scores similarity while accounting for differences in the total number of alterations in each tumor ( Lohr et al . , 2014 ) . To assess statistical significance , Jaccard Indexes were transformed into Z-scores . A pair of tumors was considered strongly related ( at 95% confidence ) if the Z-score was >1 . 96 in a two-way test ( comparing the HCA to the HCC and the HCC to the HCA ) , while a pair was considered weakly related if only one side of the Z-score was >1 . 96 . We identified two pairs of HCA and HCC tumors that were strongly related and found that the two HCCs that arose in a single mouse were also strongly related ( relative to tumors from different animals; Table 2 ) . In three animals HCA and HCCs were weakly related and in four animals we did not detect significant similarity ( Table 2 ) . In aggregate , tumor pairs from the same mouse were significantly more similar to each other than to tumors from other mice , but only 10–20% of CNVs overlapped among benign and malignant tumors from the same animal . We conclude that HCC can indeed evolve from HCA in our mouse model but that HCA may not be a necessary precursor to HCC . The frequency of progression is difficult to judge: when HCA and HCC appear unrelated , it is possible that a related HCA was present at an earlier time . Nonetheless , our data clearly demonstrate the possibility of evolution from HCA to HCC in a single animal and of a single HCC into a macroscopically distinct HCC in the same organ . Both progression and clonal evolution are accompanied by ongoing gain and loss of whole chromosomes and CNVs . 10 . 7554/eLife . 20873 . 015Figure 6 . Mad2l1 deficiency results in clonal abnormalities despite ongoing chromosomal instability in murine T-ALL . ( A ) Box and whiskers plot for each chromosome of unaffected liver ( blue n = 12 ) , HCA ( green n = 10 ) , and HCC ( red n = 18 ) from Alb-Cre::Mad2l1f/f::Alb-Cre mice with mixed Trp53 genotypes . Statistical significance assessed by Two-way ANOVA and Tukey multiple comparison test with comparison between each group shown in the table below , *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . ( B ) Sum of the absolute value of the aCGH ratio for each chromosome . Statistical significance assessed by One-way ANOVA and Tukey multiple comparison test , *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . ( C–F ) AneuFinder plots revealing perfect euploidy in control thymus ( 45 freshly isolated T-cells ( C ) ) and recurrent chromosomal abnormalities as well as intratumor karyotype heterogeneity in three Lck-Cre::Mad2l1f/f::Trp53f/f T-ALLs for which 46 ( D ) , 44 ( E ) and 43 ( F ) primary tumor cells were analyzed by single cell sequencing , respectively . Colors refer to chromosome copy number state . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 01510 . 7554/eLife . 20873 . 016Figure 6—figure supplement 1 . Heterogeneity and aneuploidy scores for control thymus and individual T-ALLs analyzed by single cell sequencing . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 01610 . 7554/eLife . 20873 . 017Table 2 . Evolutionary relationship of tumors within the same animal . Focal copy number variants were compared between tumors using the Jaccard Index and tested for significance by transforming the Jaccard Indices into Z-scores ( 1 . 96 cutoff for significance ) . Z-scores were calculated for every tumor individually . Thus , each tumor pair has two Z-scores and was considered weakly related if one comparison was significant ( * ) and strongly related if both comparisons were significant ( ** ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 017MouseTumorZ scoreTumorZ scoreRelated7122HCA11 . 08HCC11 . 362985HCA11 . 1HCC11 . 156717HCA11 . 55HCC11 . 376705HCA10 . 13HCC10 . 566718HCA11 . 88HCC10 . 466546HCA1 . 62HCC11 . 476891HCA11 . 21HCC11 . 216891HCC12 . 81HCC21 . 37*6755HCA12 . 49HCC12 . 69**6228HCA12 . 5HCC12 . 03** To measure intratumor karyotype heterogeneity in Mad2l1-null tumors directly we performed single-cell sequencing from three animals with T-ALL and wild-type control . Single cell suspensions of T-ALL and control thymocytes ( ~45 cells per animal ) were separated by flow cytometry into 96 well plates followed by barcoding and low-coverage next-generation sequencing ( van den Bos et al . , 2016 ) . As expected , no chromosomal abnormalities were observed in T-cells from wild-type thymi ( Figure 6C ) but extensive aneuploidy was evident in 3 T-ALL samples ( Figure 6D–F ) . Most cells from all three T-ALLs subjected to single-cell analysis had gained Chr . 14 and 15 ( Figure 6D–F ) and in one of two tumors examined , a preponderance of cells had gained Chr . 1 , 2 , 4 , 5 , and 17 ( Figure 6F ) . In contrast , changes in the ploidy of other chromosomes appeared random: > 70% of the cells in a single T-ALL had unique karyotypes . Intratumor differences in karyotype were further evidenced by high heterogeneity scores for all three T-ALLs ( Figure 6—figure supplement 1 ) . These scores were also higher than those of T-ALLs expressing truncated Mps1 ( Foijer et al . , 2014; Bakker et al . , 2016 ) . Because T-ALLs were clonal based on TCR loci , the most likely explanation for intratumor karyotypic heterogeneity is ongoing chromosome loss and gain , presumably due to loss of SAC function . Human cancers in the Cancer Genome Atlas ( TCGA ) have previously been reported to have an abnormal pan-cancer karyotype involving Chrs . 7 , 20 ( which are frequently gained ) and Chrs . 10 , 13 , 22 ( which are frequently lost ) ( Davoli et al . , 2013; Nicholson and Cimini , 2013 ) as well as tissue-specific changes in chromosome ploidy . To compare these findings with our results , we reanalyzed TGCA data for epithelial cancers such as HCC and breast cancer and also non-epithelial cancers such as glioma and glioblastoma multiforrme ( Figure 7 ) . In human HCCs , chromosomes Chr1q and Chr8q are commonly gained and Chr8p lost ( Figure 7; top panel ) , an event that has been associated with worse clinical outcomes ( Emi et al . , 1993 ) . Human 8q is related to murine Chr . 15 , which we find frequently gained in Mad2l1-null HCCs ( Hertz et al . , 2008; Guan et al . , 2000 ) . Similar karyotypic abnormalities are found in HCC and breast cancer ( Figure 7 , top panel ) , consistent with the existence of a pan-cancer karyotype . However , non-epithelial cancers exhibit a very different karyotypic pattern than carcinomas ( compare upper and lower panels Figure 7 ) and see also ( Zack et al . , 2013 ) . Thus , epithelial and non-epithelial human cancers are similar to Mad2l1-null HCC and T-ALL in having distinct karyotypes , perhaps because they experience different selective pressures . 10 . 7554/eLife . 20873 . 018Figure 7 . Overlaid TCGA copy number data at kilobase resolution for epithelial ( Hepatocellular carcinoma and Breast Cancer , upper panel ) and non-epithelial ( Glioma and Glioblastoma multiforme , lower panel ) represented as the number of tumors above a log2 threshold of 0 . 3 for gains and below −0 . 3 for losses . The Y-axis is scaled to the total number of tumors analyzed . Analyzed data was sequenced at the Broad institute ( Boston , MA ) on SNP6 . 0 chips . DOI: http://dx . doi . org/10 . 7554/eLife . 20873 . 018
The rapid growth of Mad2l1-deficient liquid and solid tumors stands in contrast to previous data showing that Mad2l1 is essential for the survival of cancer cell lines and for development of early mouse embryos ( Kops et al . , 2004; Dobles et al . , 2000 ) . Mad2l1 loss substantially accelerates tumor development in Trp53-positive and/or mutant backgrounds ( depending on cell type ) emphasizing that checkpoint inactivation and CIN are not just tolerated by cells , they accelerate disease onset in animals already mutant for a potent tumor suppresser . In cultured MEFs and thymocytes assayed in vivo , the microtubule poisons nocodazole and taxol provoke a significantly weaker cell cycle arrest than in wild-type cells . The near-complete absence of response to nocodazole in MEFs is consistent with checkpoint inactivation but in vivo experiments with thymocytes reveal partial responsiveness to taxol . It is unclear whether this reflects the presence of a Mad2l1-independent pathway for detecting and responding to spindle damage , or the presence of unrecombined Mad2l1-positive cells that have a normal checkpoint response; additional single-cell experiments will be required to resolve this matter . Parallel experiments in our labs have shown that checkpoint deficiency is also tolerated by basal epidermal skin cells but is lethal to hair follicle stem cells ( Foijer et al . , 2013 ) . These data established that SAC inactivation is compatible with proliferation of some cell types but not others . However , our previous hypothesis that the SAC would be required in highly proliferating cells appears not be true . Since most T-ALLs are clonal and show no evidence of Mad2l1 expression , they must arise from a single checkpoint-null initiating cell following many rounds of cell division . Further experiments with Mad2l1f/f animals and Cre drivers should reveal which tissues tolerate checkpoint loss and which do not . The consequences of Mad2l1 inactivation for chromosome structure also appear to differ with cell type: in Mad2l1-deficient HCCs we observe fewer and smaller lesions that in T-ALL . The reasons for this are unknown and the result is subject to the caveat that T-ALLs are clonal and HCC polyclonal . Nonetheless , the data suggest that the consequences of SAC inactivation can vary , providing a rationale for expanding the study of chromosome segregation in human cells from a few transformed lines to multiple primary cell types . Loss of Mad2l1 creates a mutator phenotype in the absence of an overt oncogenic driver . In the current study we detect frequent loss of known tumor suppressor genes , Pten in T-ALL for example , and amplification of known oncogenes , such as MET in HCC . CNVs are also found in many other genes whose function in cancer remains unknown . Large-scale genomic analysis of Mad2l1-deficient tumors may be a generally useful means for identifying new oncogenes and tumor suppressors as well as genes ( such as Trp53 ) whose mutation tolerizes cells to checkpoint loss and aneuploidy ( Torres et al . , 2010; Fujiwara et al . , 2005 ) . Mad2l1f/f animals may also be useful in studying the role of genomic instability in drug resistance and tumor recurrence ( Burrell and Swanton , 2014b , 2014a ) and in the clonal evolution of tumors ( Sotillo et al . , 2010 ) . Single-cell sequencing of T-ALLs from Mad2l1-deficient mice reveals extensive aneuploidy and intratumor karyotypic heterogeneity . In some animals with liver cancer , we found evidence of clonal progression from HCA to HCC as well as genetic progression among physically distinct HCCs in a single animal . This is precisely what we would expect of cells experiencing ongoing CIN . Our data are also consistent with findings from ultra-deep whole genome sequencing of lung cancers demonstrating extensive intra-tumor heterogeneity at the level of point mutations and structural chromosome abnormalities ( de Bruin et al . , 2014; Gerlinger et al . , 2012 ) . In these studies , different physical regions from the same tumor were repeatedly sequenced , revealing the presence of shared mutations as well as mutations unique to each region , a pattern consistent with genomic instability . In Mad2l1-deficient murine T-ALLs , HCAs and HCCs , the average karyotype of cell populations is shown by aCGH to involve pan-cancer and tissue-specific patterns of whole chromosome loss and gain . For example , Chr13 is lost and Chr15 gained in both T-ALL and HCC whereas Chr4 and Chr12 are gained in T-ALL and lost in HCC . In humans , recurrent cancer karyotypes have also been observed and these differ between carcinomas , sarcomas and liquid tumors ( Zack et al . , 2013 ) . Single cell and aCGH data are most easily reconciled by postulating that tumor cells experience ongoing CIN but that specific karyotypes have a selective advantage within the environment of a tumor . Karyotypic variation presumably results in differential loss and gain of oncogenes and tumor suppressors as discussed above , but it also causes large-scale changes in gene expression . In general , genes that are located on chromosomes with abnormal ploidy also change in levels of expression , but even when T-ALL and HCC experience the same change in ploidy for a particular chromosome , GSEA shows that differentially expressed genes do not overlap . We therefore speculate that karyotypic selection is imposed both at the level of structural rearrangements in specific genes and broad changes in gene expression . Selection in the face of ongoing CIN contrasts with a model of a genome restabilization in which CIN is a transient phenomenon and tumor karyotype maintained because aneuploid or structurally abnormal chromosomes are subsequently transmitted with good fidelity . However , further analysis of tumor cells passaged in culture or in syngeneic animals will be required to distinguish unambiguously among these two possibilities .
The conditional targeting vector ( shown in Figure 1a and Extended Data Figure 1 ) was constructed to delete a genomic fragment containing exons 2 and part of exon 5 of the Mad2l1 gene by homologous recombination . One loxP site was introduced into intron 1 and the other loxP site together with Frt-Neo-Frt cassette was inserted into exon 5 , such that Mad2l1 exon 2 and part of exon 5 were flanked by the loxP sites . Cre-mediated deletion will remove entire exons 2 , 3 , 4 and part of exon 5 including stop codon , producing a Mad2l1Δ allele . Embryonic stem cells derived from 129Sv mice were transfected and selected by genomic southern blot . Homologous recombinant clones were isolated and the loxP-flanked PGKneo cassette was excised by transient expression of FLP recombinase . Chimeric mice were created by injecting Mad2l1-targeted ES cell line ( from 129 background ) into C57BL/6 blastocysts generated by superovulation . Chimeras were crossed to C57BL/6 wild-type animals to generate founder lines . Mad2l1f/f mice were crossed to Lck-Cre or Alb-Cre transgenic mice to generate T cell specific and parenchymal liver cell-specific knockouts of Mad2l1 respectively . Lck-Cre and Alb-Cre::Mad2l1f/fmice were then inter-crossed with Trp53f/f mice ( Jonkers et al . , 2001 ) . All animals were kept in pathogen-free housing under guidelines approved by the Center for Animal Resources and Comparative Medicine at Harvard Medical School or at the Wellcome Trust Sanger Institute . Animal protocols were approved by the Massachusetts Institute of Technology , Harvard Medical School Committees on Animal Care ( IACUC numbers I04272 and IS00000178 ) , UK Home Office , and UMCG animal facility ( DEC 6369 ) . Tail DNA was isolated using NucleoSpin Tissue Kit from Macherey-Nagel according to the manufacturer’s protocol . The following primers were used for Mad2l1+ ASOL233 , GCAGACCAAACGAACCTAAGTT . ASOL 238 , GCAAGAGGTGGTTCAATAGTGAG , Mad2l1f MOL232 , AGGCTGAGCCGGGCCTTAGGAC; MOL233 , GTAACCGTGTAATAACGTTTAAGTCTC , Mad2l1Δ MOL231 , GTCTGCGGTGAGGTTGG; ASOL233 , GCAGACCAAACGAACCTAAGTT . Alb-Cre AlbF , GTTAATGATCTACAGTTATTGG and AlbR , CGCATAACCAGTGAAACAGCATTGC . Lck-Cre LckF , CCTTGGTGGAGGAGGGTGGAATGAA , LckR , TAGAGCCCTGTTCTGGAAGTTACAA , and CreT2R , CGCATAACCAGTGAAACAGCATTGC . MEFs were isolated as described previously ( Foijer et al . , 2005 ) and cultured in DMEM containing 10% FCS , pyruvate , non-essential amino acids and penicillin/ streptomycin ( Invitrogen ) . Cells were genotyped using the above-mentioned primers to confirm genotypes and routinely tested for mycoplasma contamination . For spindle checkpoint integrity measurements , cells were exposed to 250 ng/ml nocodazole ( Sigma ) for 4–6 hr , fixed in 70% ethanol and labeled with Alexa Fluor-488-conjugated pHistoneH3 antibodies ( Cell Signaling , RRID:AB_10694488 ) as described previously ( Foijer et al . , 2014 ) . For time-lapse imaging , cells were transduced with H2B-GFP ( Foijer et al . , 2014 ) as described previously ( Foijer et al . , 2005 ) and seeded on 4-well imaging slides ( LabTek , Thermo Fisher ) in the presence of 250 ng/ml nocodazole ( Sigma ) . Cells were imaged on a DeltaVision Elite imaging station ( GE Healthcare ) . Animals were euthanized and their thymus or liver were removed and rinsed in PBS . Tissues collected were fixed overnight in formalin . Fixed tissues were then stored in 70% ethanol until they were embedded in paraffin . Section slides were prepared and standard H&E staining were done at Rodent Histopathology Core facility at Dana-Farber/Harvard Cancer Center . Protein from tumors was isolated using protein lysis buffer ( Millipore ) in the presence of protease inhibitors ( Millipore ) . Protein concentration was quantified using the Bradford assay ( Biorad ) . 20 µg of total protein was run on a 4–12% gradient gel ( Invitrogen ) per sample and blotted on PVDF membrane ( Millipore ) . Antibodies used were mouse monoclonal Mad2l1 ( BD Biosciences , RRID:AB_398005 ) , mouse monoclonal Actin ( Cell Signaling , RRID: AB_2223172 ) and HRP-labeled goat-anti mouse ( New England Biolabs ) . Mouse thymus and liver genomic DNA was extracted with NucleoSpin Tissue Kit ( Macherey-Nagel ) . Sex-mismatched wild type liver DNA was used as control . Mouse Genome CGH Microarrays 44K or 244K from Agilent were used . Array hybridization and data analysis were performed at the Wellcome Trust Sanger Institute , the Partners HealthCare Center for Personalized Genetic Medicine at Harvard Medical School , and the BioMicro Center at the Massachusetts Institute of Technology . Low coverage Next-Gen sequencing of liver tumor DNA isolated as described above was performed at the BioMicro Center at the Massachusetts Institute of Technology . For single cell sequencing , T-ALL samples or primary thymus were dissected and homogenized through a tissue strainer . Single cells in G1 were sorted into 96 wells plate by flow cytometry using a Hoechst/Propidium iodide double staining . Cells were then lysed , DNA sheared and DNA was barcode labeled followed by library preparation as described previously ( van den Bos et al . , 2016 ) in an automated fashion ( Agilent Bravo robot ) . Single cell libraries were pooled an analyzed on an Illumina Hiseq2500 sequencer . Single cell sequencing data was analyzed using AneuFinder software as described previously ( Bakker et al . , 2016 ) . RNA was isolated using the RNeasy kit ( Qiagen ) . For qPCR reactions , 1 µg of total RNA was used for a reverse transcriptase reaction ( Superscript II , Invitrogen ) . The resulting cDNA was used as a template for qPCR ( ABI PRISM 7700 Sequence Detector ) in the presence of SYBR-green ( Invitrogen ) to label the product . The relative amounts of cDNA were compared to Actin to correct for the amount of total cDNA . Average values and standard deviations were calculated as indicated in Figure legends and compared to the expression values in control mice ( normalized to the value of 1 ) . We used the following primers: Trp53 A Fw TGTTATGTGCACGTACTCTCC , Trp53 A Rv GTCATGTGCTGTGACTTCTTG Trp53 B Fw TCCGAAGACTGGATGACTG , Trp53 B Rv AGATCGTCCATGCAGTGAG , Mad2l1 A Fw AAACTGGTGGTGGTCATCTC , Mad2l1 A Rv TTCTCTACGAACACCTTCCTC , Actin A Fw CTAGGCACCAGGGTGTGATG , and Actin A Rv GGCCTCGTCACCCACATAG . Illumina expression microarrays were performed as described previously ( Foijer et al . , 2014 ) . Array data is publically available via GEO accession numbers GSE63689 and GSE63100 . Sequencing data of liver tumors is available at the Sequence Read Archive accession number SRA191233 . Single cell sequencing data of lymphomas is available at the Sequence Read Archive accession numbers under GSE63689 . Graphing plots and statistical testing was performed using GraphPad Prism ( GraphPad Software ) or MATLab ( Mathworks ) . | An estimated 350 billion of the cells in the human body are dividing at any given moment . Every cell division requires the 46 chromosomes in the cell , which store the genetic information that the cell needs to survive , to be copied and distributed evenly between the two new cells . Sometimes mistakes in cell division can result in cells that have the wrong number of chromosomes – a state called aneuploidy . Aneuploidy is rare in healthy cells but occurs in over 75% of cancers . It is the result of a process called chromosomal instability that often leads to the death of healthy cells . However , it is not well understood how aneuploidy affects how cancer cells develop or behave . Mice are commonly used to investigate cancer because they have many genetic similarities with humans . To better understand the relationship between aneuploidy and cancer , Foijer , Albacker et al . engineered mice in which they could induce aneuploidy in liver cells and immune cells called T-cells . This modification accelerated the formation of liver cancer and lymphoma – a cancer of the immune system . The number of chromosomes in the cells of these cancers varied greatly , demonstrating that these cells experience constant chromosomal instability . Overall , this suggests that aneuploidy increases the likelihood of cancer developing . The mouse cancer cells closely resemble their human counterparts , and so could potentially be used to test new cancer drugs . In the future , developing new therapies that selectively target aneuploid cells could result in cancer treatments that have fewer side effects than existing treatments . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"cancer",
"biology"
] | 2017 | Deletion of the MAD2L1 spindle assembly checkpoint gene is tolerated in mouse models of acute T-cell lymphoma and hepatocellular carcinoma |
Determining the structures , kinetics , thermodynamics and mechanisms that underlie conformational exchange processes in proteins remains extremely difficult . Only in favourable cases is it possible to provide atomic-level descriptions of sparsely populated and transiently formed alternative conformations . Here we benchmark the ability of enhanced-sampling molecular dynamics simulations to determine the free energy landscape of the L99A cavity mutant of T4 lysozyme . We find that the simulations capture key properties previously measured by NMR relaxation dispersion methods including the structure of a minor conformation , the kinetics and thermodynamics of conformational exchange , and the effect of mutations . We discover a new tunnel that involves the transient exposure towards the solvent of an internal cavity , and show it to be relevant for ligand escape . Together , our results provide a comprehensive view of the structural landscape of a protein , and point forward to studies of conformational exchange in systems that are less characterized experimentally .
Proteins are dynamical entities whose ability to change shape often plays essential roles in their function . From an experimental point of view , intra-basin dynamics is often described via conformational ensembles whereas larger scale ( and often slower ) motions are characterized as a conformational exchange between distinct conformational states . The latter are often simplified as a two-site exchange process , G⇌E , between a highly populated ground ( G ) state , and a transiently populated minor ( or ‘excited’ , E ) state . While the structure of the ground state may often be determined by conventional structural biology tools , it is very difficult to obtain atomic-level insight into minor conformations due to their transient nature and low populations . As these minor conformations may , however , be critical to protein functions , including protein folding , ligand binding , enzyme catalysis , and signal transduction ( Mulder et al . , 2001; Tang et al . , 2007; Baldwin and Kay , 2009 ) it is important to be able to characterize them in detail . While it may in certain cases be possible to capture sparsely populated conformations in crystals under perturbed experimental conditions , or to examine their structures by analysis of electron density maps ( Fraser et al . , 2009 ) , NMR spectroscopy provides unique opportunities to study the dynamical equilibrium between major and minor conformations ( Baldwin and Kay , 2009; Sekhar and Kay , 2013 ) via e . g . chemical-exchange saturation transfer ( Vallurupalli et al . , 2012 ) , Carr-Purcell-Meiboom-Gill ( CPMG ) relaxation dispersion ( Hansen et al . , 2008 ) , or indirectly via paramagnetic relaxation enhancement ( Tang et al . , 2007 ) or residual dipolar coupling ( Lukin et al . , 2003 ) experiments . In favourable cases such experiments can provide not only thermodynamic and kinetic information ( i . e . the population of G and E states and the rate of exchange between them ) , but also structural information in the form of chemical shifts ( CS ) , that can be used to determine the structure of the transiently populated state ( Sekhar and Kay , 2013 ) . Despite the important developments in NMR described above , it remains very difficult to obtain structural models of minor conformations , and a substantial amount of experiments are required . Further , it is generally not possible to use such experiments to infer the mechanisms of interconversion , and to provide a more global description of the multi-state free energy landscape ( Zhuravlev and Papoian , 2010; Wang et al . , 2012 ) . In the language of energy landscape theory ( Onuchic et al . , 1997 ) , free energy basins and their depths control the population and stability of functionally distinct states , while the relative positions of basins and the inter-basin barrier heights determine the kinetics and mechanism of conformational exchange . As a complement to experiments , such functional landscapes can be explored by in silico techniques , such as molecular dynamics ( MD ) simulations , that may both be used to help interpret experimental data and provide new hypotheses for testing ( Karplus and Lavery , 2014; Eaton and Muñoz , 2014 ) . Nevertheless , the general applicability of simulation methods may be limited by both the accuracy of the physical models ( i . e . force fields ) used to describe the free energy landscape and our ability to sample these efficiently by computation . We therefore set out to benchmark the ability of simulations to determine conformational free energy landscapes . The L99A variant of lysozyme from the T4 bacteriophage ( T4L ) has proven an excellent model system to understand protein structure and dynamics . Originally designed a ‘cavity creating’ variant to probe protein stability ( Eriksson et al . , 1992b ) it was also demonstrated that the large ( 150 Å3 ) internal cavity can bind hydrophobic ligands such as benzene ( Eriksson et al . , 1992a; Liu et al . , 2009 ) . It was early established that the cavity is inaccessible to solvent in the ground state , but that ligand binding is rapid ( Feher et al . , 1996 ) , suggesting protein dynamics to play a potential role in the binding process . This posts a long-standing question of how the ligands gain access to the buried cavity ( Mulder et al . , 2000; López et al . , 2013; Merski et al . , 2015; Miao et al . , 2015 ) . NMR relaxation dispersion measurements of L99A T4L demonstrated that this variant , but not the wild type protein , displayed conformational exchange on the millisecond timescale between the ground state and a minor state populated at around 3% ( at room temperature ) ( Mulder et al . , 2001 ) . Such small populations generally lead only to minimal perturbations of ensemble-averaged experimental quantities making structural studies difficult , and hence it was difficult to probe whether the exchange process indeed allowed for ligand access to the cavity . A series of additional relaxation dispersion experiments , however , made it possible to obtain backbone and side chain CSs of the minor E state of L99A ( Mulder et al . , 2002; Bouvignies et al . , 2011 ) . The backbone CS data were subsequently used as input to a CS-based structure refinement protocol ( CS-ROSETTA ) to produce a structural model of the E state ( EROSETTA; Figure 1 ) of the L99A mutant ( Bouvignies et al . , 2011 ) . This model was based in part on the crystal structure of the ground state of L99A ( referred to in what follows as GXray ) , but perturbing the structure in regions that the experiments demonstrated to undergo conformational change in a way so that the final model ( EROSETTA ) agrees with experiments . The structure was further validated by creating and solving the structure of a triple mutant variant that inverts the populations of the G and E states . The EROSETTA structure revealed substantial local rearrangements in T4L L99A , in particular near the cavity which gets filled by the side chain of a phenylalanine at position 114 ( F114 ) . Because the cavity is filled and solvent inaccessible in the E-state , the structure did , however , not reveal how ligands might access the cavity . 10 . 7554/eLife . 17505 . 003Figure 1 . Structures of the major G and minor E states of L99A T4L and the hidden state hypothesis . The X-ray structure of the G state ( GXray; PDB ID code 3DMV ) has a large internal cavity within the core of the C-terminal domain that is able to bind hydrophobic ligands . The structure of the E state ( EROSETTA; PDB ID code 2LC9 ) was previously determined by CS-ROSETTA using chemical shifts . The G and E states are overall similar , apart from the region surrounding the internal cavity . Comparison of the two structures revealed two remarkable conformational changes from G to E: helix F ( denoted as HF ) rotates and fuses with helix G ( HG ) into a longer helix , and the side chain of phenylalanine at position 114 ( F114 ) rotates so as to occupy part of the cavity . As the cavity is inaccessible in both the GXray and EROSETTA structures it has been hypothesized that ligand entry occurs via a third ‘cavity open’ state ( Merski et al . , 2015 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 003 In an attempt to benchmark the ability of simulations to map conformational free energy landscapes , we have here employed a series of in silico experiments designed to probe the structure and dynamics of L99A T4L and have compared the results to NMR measurements . We used enhanced-sampling MD simulations in explicit solvent and with state-of-the-art force fields to map the free-energy landscape including the exchange between the major and minor conformations of the protein . We used a series of recently developed metadynamics methods ( Laio and Parrinello , 2002 ) to sample the conformational exchange process and associated structure and thermodynamics , as well as to determine the kinetics and mechanisms of exchange . We obtained additional insight into the structural dynamics of the E state using simulations that employed the experimental CSs as replica-averaged restraints . Our results provide a coherent picture of the conformational dynamics in L99A and extend the insights obtained from recent simulations of a triple mutant of T4L ( Vallurupalli et al . , 2016 ) , by providing new information about the mechanisms of exchange and the transient exposure of the internal cavity . Together with previous results for Cyclophilin A ( Papaleo et al . , 2014 ) the results described here reiterate how simulation methods have now reached a stage where they can be used to study slow , conformational exchange processes such as those probed by NMR relaxation dispersion even in cases where less information is available from experiments .
As the average lifetime of the G and E states are on the order of 20–50 ms and 1 ms , respectively ( Mulder et al . , 2001 , 2002; Bouvignies et al . , 2011 ) , direct and reversible sampling of the G-E transition at equilibrium would be extremely demanding computationally . Indeed , a recent set of simulations of a triple mutant of T4L , which has a substantially faster kinetics , was able only to observe spontaneous transitions in one direction ( Vallurupalli et al . , 2016 ) . We therefore resorted to a set of flexible and efficient enhanced sampling methods , collectively known as ‘metadynamics’ ( Laio and Parrinello , 2002 ) , that have previously been used in a wide range of applications . In metadynamics simulations , a time-dependent bias is continuously added to the energy surface along a small number of user-defined collective variables ( CVs ) . In this way , sampling is enhanced to reach new regions of conformational space and at the same time allows one to reconstruct the ( Boltzmann ) free-energy surface . The success of the approach hinges on the ability to find a set of CVs that together describe the slowly varying degrees of freedom and map the important regions of the conformational landscape . We first performed a set of metadynamics simulations in the well-tempered ensemble ( Barducci et al . , 2008 ) using so-called path CVs ( Spath and Zpath ) ( Branduardi et al . , 2007; Saladino et al . , 2012 ) with the aid of recently developed adaptive hills to aid in the convergence of the sampling ( Branduardi et al . , 2012; Dama et al . , 2014 ) ( see details in Appendix and Appendix 1—table 1 ) . In short , the Spath variable describes the progress of the conformational transition between the GXray and EROSETTA structures with additional ‘interpolation’ using an optimal ‘reference’ path in a simplified model ( see details in Appendix and Figure 2—figure supplement 1 ) , while Zpath measures the distance to this reference path . In this way , the two-dimensional free energy landscape along Spath and Zpath provides a useful description on conformational exchange between ground and excited states that does not assume that the initial reference path describes perfectly the actual path ( s ) taken . Projecting the sampled free energy landscape along Spath ( upper panel of Figure 2 ) reveals a deep , narrow free energy basin around Spath=0 . 2 ( labeled by red sphere and corresponding to the G state ) , and a broader , shallow free energy basin with Spath ranging from 0 . 6 to 0 . 8 ( labeled by blue sphere and corresponding to the E state ) . Additional information is obtained from the two-dimensional landscape ( shown as a negative free energy landscape , -F ( Spath , Zpath ) , in the lower panel of Figure 2 ) which reveals a complex and rough landscape with multiple free energy minima ( corresponding to mountains in the negative free energy landscape ) . Subsequently , structural inspection of these minima identified that the conformations in the basins around Spath=0 . 2 and Spath=0 . 75 correspond to the structures of GXray and EROSETTA , respectively . 10 . 7554/eLife . 17505 . 004Figure 2 . Free energy landscape of the L99A variant of T4L . In the upper panel , we show the projection of the free energy along Spath , representing the Boltzmann distribution of the force field employed along the reference path . Differently colored lines represent the free energy profiles obtained at different stages of the simulation , whose total length was 667ns . As the simulation progressed , we rapidly found two distinct free energy basins , and the free energy profile was essentially constant during the last 100 ns of the simulation . Free energy basins around Spath=0 . 2 and Spath=0 . 75 correspond to the previously determined structures of the G- and E-state , respectively ( labelled by red and blue dots , respectively ) . As discussed further below , the E-state is relatively broad and is here indicated by the thick , dark line with Spath ranging from 0 . 55 to 0 . 83 . In the lower panel , we show the three-dimensional negative free energy landscape , -F ( Spath , Zpath ) , that reveals a more complex and rough landscape with multiple free energy minima , corresponding to mountains in the negative free energy landscape . An intermediate-state basin around Spath=0 . 36 and Zpath=0 . 05 nm2 , which we denote I0 . 36 , is labeled by a yellow dot . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 00410 . 7554/eLife . 17505 . 005Figure 2—figure supplement 1 . Approximately equidistant frames along the reference path . The plot reveals a ‘gullwing’ shape of the matrix of pairwise RMSDs of the frames of the reference path , indicating that frames along the reference path are approximately equidistant . We used 31 structures to discretize the path . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 00510 . 7554/eLife . 17505 . 006Figure 2—figure supplement 2 . One and two dimensional free energy landscape of L99A and the triple mutant . ( A ) The two-dimensional free energy surface F ( Spath , Zpath ) of L99A sampled by a 667 ns PathMetaD simulation . ( B ) The two-dimensional free energy surface F ( Spath , Zpath ) of the triple mutant sampled by a 961 ns PathMetaD simulation . ( C ) The free energy profiles as a function of Spath of both L99A ( black ) and the triple mutant ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 006 The broad nature of the free energy landscape in the region of the minor state is consistent with the observation that our MD simulations initiated from EROSETTA display significant conformational fluctuations ( RUN20 and RUN22 in Appendix 1—table 1 ) . Furthermore , our metadynamics simulations revealed multiple local free energy minima adjacent to the EROSETTA basin , together composing a wider basin ( highlighted by the black curve in Figure 2 ) . Thus , these simulations suggest that the E state displays substantial conformational dynamics , a result corroborated by simulations that have been biased by the experimental data ( see section ‘Simulations of the minor state using chemical shift restraints’ ) . In addition to free-energy minima corresponding to the G and E states , we also found a free energy minimum around Spath=0 . 36 and Zpath=0 . 05 nm2 ( denoted as I0 . 36 and labeled by a yellow sphere in Figure 2 ) that is located between the G and E states on the one-dimensional free-energy surface . We note , however , that it is difficult to infer dominant reaction pathways from such free energy surfaces , and so from this data alone , we cannot determine whether I0 . 36 occurs as an intermediate in G-E conformational transitions . Indeed , it appears from the two-dimensional surface that there exist multiple possible pathways between G and E , as illustrated by white lines along the mountain ridges of the negative free energy landscape in the lower panel of Figure 2 . ( We also explored the mechanism of exchange by reconnaissance metadynamics simulations ( Tribello et al . , 2011 ) , the results of which are described and discussed further below . ) Based on the encouraging results above for L99A T4L , we examined whether simulations could also capture the effect of mutations on the free energy landscape . Using Rosetta energy calculations on the GXray and EROSETTA structures it was previously demonstrated that two additional mutations , G113A and R119P , when introduced into the L99A background , cause an inversion in the populations of the two states ( Bouvignies et al . , 2011; Vallurupalli et al . , 2016 ) . Indeed , NMR data demonstrated that the triple mutant roughly inverts the populations of the two states so that the minor state structure ( of L99A ) now dominates ( with a 96% population ) the triple mutant . We repeated the calculations described above for L99A also for the triple mutant . Remarkably , the free energy profile of the triple mutant obtained using metadynamics simulations reveals a free energy landscape with a dominant minimum around Spath=0 . 7 and a higher energy conformation around Spath=0 . 15 ( Figure 2—figure supplement 2 ) . Thus , like our previous observations for a ‘state-inverting mutation’ in Cyclophilin A ( Papaleo et al . , 2014 ) , we find here that the force field and the sampling method are sufficiently accurate to capture the effect of point mutations on the free energy landscape . Further , we note that the barrier height for the conformational exchange in the triple mutant is very similar to the value recently estimated using a completely orthogonal approach ( Vallurupalli et al . , 2016 ) . Finally , we attempted to determine the free energy landscape of the L99A , G113A double mutant , which has roughly equal populations of the two states ( Bouvignies et al . , 2011 ) , but this simulation did not converge on the simulation timescales at which the two other variants converged . With a free-energy surface in hand and a method to distinguish G- and E-state conformations , we calculated the free energy difference , ΔG , between the two conformational states , and compared with the experimental values . We divided the global conformational space into two coarse-grained states by defining the separatrix at Spath=0 . 46 which corresponds to a saddle point on the free energy surface , on the basis of the observations above that the E state is relatively broad . Although a stricter definition of how to divide the reaction coordinate certainly helps the precise calculation , here we just used this simple definition to make an approximate estimation of the free energy difference . Further , since the barrier region is sparsely populated , the exact point of division has only a modest effect on the results . By summing the populations on the two sides of the barrier we calculated ΔG as a function of the simulation time ( Figure 3 ) . Initially during the simulations the free energy profile varies substantially ( Figure 2 ) and the free energy difference equally fluctuates . As the simulations converge , however , the free energy difference between the two states stabilize to a value at approximately ΔG=3 . 5 kcal mol−1 ( Figure 3 , black line ) . This value can be compared to the value of 2 . 1 kcal mol−1 obtained from NMR relaxation dispersion experiments ( Mulder et al . , 2001 ) , revealing reasonably good , albeit not exact , agreement with the experiments . 10 . 7554/eLife . 17505 . 007Figure 3 . Estimation of free energy differences and comparison with experimental measurements . We divided the global conformational space into two coarse-grained states by defining the separatrix at Spath=0 . 46 ( 0 . 48 for the triple mutant ) in the free energy profile ( Figure 2—figure supplement 2 ) which corresponds to a saddle point of the free energy surface , and then estimated the free energy differences between the two states ( ΔG ) from their populations . The time evolution of ΔG of L99A ( upper time axis ) and the triple mutant ( lower axis ) are shown as black and blue curves , respectively . The experimentally determined values ( 2 . 1 kcal mol−1 for L99A and −1 . 9 kcal mol−1 for the triple mutant ) are shown as yellow lines . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 007 Similar calculations using the simulations of the triple mutant also converge , in this case to about −1 . 6 kcal mol−1 ( Figure 3 , blue line ) , in excellent agreement with the experimental measurement ( −1 . 9 kcal mol−1 ) ( Bouvignies et al . , 2011 ) . Combining these two free energy differences we find that the G113A , R119P mutations cause a shift in the G-E free energy of 5 . 1 kcal mol−1 in simulations compared to 4 . 0 kcal mol−1 obtained by experiments . Thus , we find that the simulations with reasonably high accuracy are able to capture the thermodynamics of the conformational exchange between the two states . While the generality of such observations will need to be established by additional studies , we note here that comparably good agreement was obtained when estimating the effect of the S99T mutations in Cyclophilin A ( Papaleo et al . , 2014 ) . In our previous work on Cyclophilin A ( Papaleo et al . , 2014 ) we sampled the conformational exchange using parallel-tempering metadynamics simulations ( Bonomi and Parrinello , 2010 ) using four CVs that we chose to describe the structural differences between the G and E states in that protein . We note here that we also tried a similar approach here but unfortunately failed to observe a complete G-to-E transition , even in a relatively long trajectory of about 1 μs per replica ( CVs summarized in Appendix 1—table 2 , parameters shown in Appendix 1—table 1 ) . This negative result is likely due to the CVs chosen did not fully capture the relevant , slowly changing degrees of freedom , thus giving rise to insufficient sampling even with the use of a parallel tempering scheme . Enhanced-sampling simulations such as those described above provide an effective means of mapping the free-energy landscape and hence the structural and thermodynamic aspects of conformational exchange . While the same free-energy landscape also determines the kinetics and mechanisms of exchange it may be more difficult to extract this information from e . g . path-CV-based metadynamics ( PathMetaD ) simulations . To examine how well simulations can also be used to determine the rates of the G-to-E transitions , quantities that can also be measured by NMR , we used the recently developed ‘infrequent metadynamics’ method ( InMetaD , see details in Appendix ) ( Tiwary and Parrinello , 2013; Salvalaglio et al . , 2014; Tiwary et al . , 2015a; 2015b ) . Briefly described , the approach calculates first-passage times for the conformational change in the presence of a slowly-added bias along a few CVs , here chosen as the path CVs also used to map the landscape . By adding the bias slowly ( and with lower amplitude ) we aim to avoid biasing the transition-state region and hence to increase the rate only by lowering the barrier height; in this way it is possible to correct the first-passage times for the bias introduced . Using this approach on L99A T4L we collected 42 and 36 independent trajectories with state-to-state transition starting from either the G state or E state , respectively ( Appendix 1—figure 1 and Appendix 1—figure 2 ) . The ( unbiased ) rates that we calculated ( Table 1 and Appendix 1—figure 3 ) are in good agreement with the experimental rates ( Mulder et al . , 2001; Bouvignies et al . , 2011 ) ( within a factor of 10 ) , corresponding to an average error of the barrier height of ∼1 kcal mol−1 . We also performed similar calculations for the ‘population-inverting’ triple mutant , where we collected 30 transitions ( 15 for each direction ) using InMetaD simulations . As for L99A , we also here find similarly good agreement with experimental measurements ( Vallurupalli et al . , 2016 ) ( Table 1 and Appendix 1—figure 4 ) . We estimated the reliability of this computational approach using a Kolmogorov-Smirnov test to examine whether the first-passage times conform to the expected Poisson distribution ( Salvalaglio et al . , 2014 ) , and indeed the results of this analysis suggest good agreement ( Table 1 , Appendix 1—figure 5 and Appendix 1—figure 6 ) . 10 . 7554/eLife . 17505 . 008Table 1 . Free energy differences and rates of conformational exchange . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 008τG→E ( ms ) τE→G ( ms ) ΔG ( kcal mol−1 ) L99ANMR200 . 72 . 1InMetaD175 ± 561 . 4 ± 0 . 62 . 9 ± 0 . 5PathMetaD3 . 5L99A/G113A/R119PNMR0 . 24-1 . 9InMetaD2 . 0 ± 1 . 714 . 3 ± 8 . 3-1 . 2 ± 1 . 1PathMetaD-1 . 6 The ability to calculate forward and backward rates between G and E provided us with an alternative and independent means to estimate the free energy difference between the two states ( Table 1 ) , and to test the two-state assumption used in the analysis of the experimental NMR data . We therefore calculated the free energy difference from the ratio of the forward and backward reaction rates . The values obtained ( 2 . 9 ± 0 . 5 kcal mol−1 and −1 . 2 ± 1 . 1 kcal mol−1 for L99A and the triple mutant , respectively ) are close both to the values obtained above from the equilibrium free energy landscape ( 3 . 5 kcal mol−1 and −1 . 6 kcal mol−1 ) and experiment ( 2 . 1 kcal mol−1 and −1 . 9 kcal mol−1 ) . In particular , the relatively close agreement between the two independent computational estimates lends credibility both to the free energy landscape and the approach used to estimate the kinetics . The observation that both values for L99A are slightly larger than the experimental number suggests that this discrepancy ( ca . 1 kcal mol−1 ) can likely be explained by remaining force field deficiencies rather than lack of convergence or the computational approach used . While the simulations described above used available structural information of G and E states to guide and enhance conformational sampling , the resulting free energy surfaces represent the Boltzmann distributions of the force field and are not otherwise biased by experimental data . To further refine the structural model of the E state we used the relaxation-dispersion derived CSs that were used to determine of EROSETTA [BMRB ( Ulrich et al . , 2008 ) entry 17604] as input to restrained MD simulations . In these simulations , we used the experimental data as a system-specific force field correction to derive an ensemble of conformations that is compatible both with the force field and the CSs . Such replica-averaged simulations use the experimental data in a minimally-biased way that is consistent with the Principle of Maximum Entropy ( Pitera and Chodera , 2012; Roux and Weare , 2013; Cavalli et al . , 2013; Boomsma et al . , 2014 ) . We performed CS-restrained MD simulations of the E state of L99A averaging the CSs over four replicas . Although the number of replicas is a free parameter , which should in principle be chosen as large as possible , it has been demonstrated that four replicas are sufficient to reproduce the structural heterogeneity accurately ( Camilloni et al . , 2013 ) without excessive computational requirements . The agreement between calculated and experimental CSs was quantified by the root-mean-square deviation between the two ( Figure 4—figure supplement 1 ) . In particular , it is important not to bias the agreement beyond what can be expected based on the inherent accuracy of the CS prediction methods ( we assumed that the error in the experimental CS measurement , even for the E state , is negligible in comparison ) . Thus , we compared the experimental CS values of the minor state with the values calculated using the EROSETTA structure as input to CamShift ( Kohlhoff et al . , 2009 ) , Sparta+ ( Shen and Bax , 2010 ) and ShiftX ( Neal et al . , 2003 ) ( Figure 4—figure supplement 2 ) . The average RMSDs for five measured nuclei ( Hα , HN , N , C′ and Cα ) are 0 . 2 , 0 . 4 , 2 . 0 , 0 . 8 and 1 . 1ppm , respectively ( Appendix 1—table 1 ) , which are close to the inherent uncertainty of the CS back-calculation , indicating that the level of agreement enforced is reasonable . To compare the results of these experimentally-biased simulations with the experimentally-unbiased simulations described above , we projected the CS-restrained MD trajectories onto either one ( Figure 4 ) or both ( Figure 4—figure supplement 3 ) of the Spath and Zpath variables used in the path-variable-driven simulations ( PathMetaD ) . The distribution of conformations obtained using the E-state CSs as restraints is in good agreement with the broad free energy profile of the E-state obtained in the metadynamics simulations that did not include any experimental restraints . To ensure that this observation is not merely an artifact of both simulations using the same force field ( CHARMM22* ) , we repeated the biased simulations using the Amber ff99SB*-ILDN force field and obtained comparable results . We also verified that the conclusions obtained are reasonably robust to other variables such as the number of replicas and the strength of restraints ( Figure 4—figure supplement 4 ) . 10 . 7554/eLife . 17505 . 009Figure 4 . Conformational ensemble of the minor state as determined by CS biased , replica-averaged simulations . We determined an ensemble of conformations corresponding to the E-state of L99A T4L using replica-averaged CSs as a bias term in our simulations . The distribution of conformations was projected onto the Spath variable ( orange ) and is compared to the free energy profile obtained above from the metadynamics simulations without experimental biases ( black line ) . To ensure that the similar distribution of conformations is not an artifact of using the same force field ( CHARMM22* ) in both simulations , we repeated the CS-biased simulations using also the Amber ff99SB*-ILDN force field ( magenta ) and obtained similar results . Finally , we used the ground state CSs of a triple mutant of T4L , which was designed to sample the minor conformation ( of L99A ) as its major conformation , and also obtained a similar distribution along the Spath variable ( cyan ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 00910 . 7554/eLife . 17505 . 010Figure 4—figure supplement 1 . Equilibrium sampling of conformational regions of the E state of L99A by CS-restrained replica-averaged simulation . We calculated the RMSD between the experimental CSs and the values back-calculated by CamShift ( Kohlhoff et al . , 2009 ) as implemented in ALMOST ( Fu et al . , 2014 ) . We projected a 250 ns MD trajectory sampled using the CHARMM22* force field ( RUN3 in Appendix 1—table 1 ) onto the RMSDs . The average RMSDs for the five measured nuclei ( Hα , HN , N , C′ and Cα ) are 0 . 23 ppm , 0 . 38 ppm , 1 . 97 ppm , 0 . 83 ppm and 1 . 06 ppm , respectively ( Appendix 1—table 2 ) , which are close to the inherent uncertainty of the chemical shift calculation ( Figure 4—figure supplement 2 ) . This indicates the simulation yielded an ensemble in good agreement with experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 01010 . 7554/eLife . 17505 . 011Figure 4—figure supplement 2 . Estimation of the inherent uncertainty of the chemical shift calculation by different algorithms: CamShift ( Kohlhoff et al . , 2009 ) , ShiftX ( Neal et al . , 2003 ) and Sparta+ ( Shen and Bax , 2010 ) . Using EROSETTA as the reference structure , we calculated the chemical shifts using different algorithms and compared the correlation coefficients and RMSD between them . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 01110 . 7554/eLife . 17505 . 012Figure 4—figure supplement 3 . Force field dependency of the replica averaged MD simulations of L99A with chemical shift restraints . The chemical shifts of the E state of L99A ( BMRB 17604 ) were used . ( A ) The simulation with CHARMM22* force field . ( B ) The simulation with Amber ff99SB*-ILDN force field . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 01210 . 7554/eLife . 17505 . 013Figure 4—figure supplement 4 . Effect of changing the force constant and number of replicas in CS-restrained simulation of L99A . ( A ) N = 4 , ϵCS=24 KJ · mol−1 . ( B ) N = 2 , ϵCS=24 KJ · mol−1 . ( C ) N = 2 , ϵCS=12 KJ · mol−1 . N refers to the number of replicas that the CS values are averaged over . The CHARMM22* force field was used in these simulations . The results also support the conclusion that the conformational space of the minor ( E ) state covers a relatively wide set of structures including the EROSETTA structure . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 01310 . 7554/eLife . 17505 . 014Figure 4—figure supplement 5 . Replica-averaged CS-restrained MD simulation of a T4L triple mutant ( L99A/G113A/R119P ) . Chemical shift restraints were from BMRB 17 , 603 and CHARMM22* force field was used . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 014 As a final and independent test of the structural ensemble of the minor conformation of L99A we used the ground state CSs of the triple mutant ( BMRB entry 17603 ) , which corresponds structurally to the E state of L99A , as restraints in replica-averaged CS-biased simulations ( Figure 4—figure supplement 5 ) . Although not fully converged , these simulations also cover roughly the same region of conformational space when projected along Spath ( Figure 4 ) . Thus , together our different simulations , which employ different force fields , are either unbiased or biased by experimental data , and use either dispersion-derived ( L99A ) or directly obtained ( triple mutant ) CS all provide a consistent view of the minor E-state conformation of L99A . We also note that the CS-derived ensembles of the E-state support the way we divided the G- and E-states when calculating conformational free energy differences between the two states . Having validated that our simulations can provide a relatively accurate description of the structure , thermodynamics and kinetics of conformational exchange we proceeded to explore the molecular mechanism of the G-to-E transitions . We used the recently developed reconnaissance metadynamics approach ( Tribello et al . , 2010 ) , that was specifically designed to enhance sampling of complicated conformational transitions and has been employed to explore the conformational dynamics of complex systems ( Tribello et al . , 2011; Söderhjelm et al . , 2012 ) . We performed three independent reconnaissance metadynamics simulations of L99A starting from the G state ( summarized in Appendix 1—table 1 ) using the same geometry-based CVs that we also used in the parallel-tempering simulations described above . We observed complete conformational transitions from the G to E state in the reconnaissance simulations in as little as tens of nanoseconds of simulations ( Figure 5—figure supplement 1 ) — at least 1–2 orders of magnitude faster than standard metadynamics . These G-to-E and E-to-G transitions , although biased by the CVs , provide insight into the potential mechanisms of exchange . To ease comparison with the equilibrium sampling of the free energy landscape we projected these transitions onto the free energy surface F ( Spath , Zpath ) ( Figure 5 ) . The results reveal multiple possible routes connecting the G and E states , consistent with the multiple gullies found on the free energy surface ( Figure 2 ) . The trajectories also suggested that the G-to-E interconversion can either take place directly without passing the I0 . 36 state or indirectly via it . 10 . 7554/eLife . 17505 . 015Figure 5 . Mechanisms of the G-E conformational exchanges explored by reconnaissance metadynamics . Trajectories labeled as Trj1 ( magenta ) , Trj2 ( blue ) and Trj3 ( green and orange ) are from the simulations RUN10 , RUN11 and RUN12 ( Appendix 1—table 1 ) , respectively . There are multiple routes connecting the G and E states , whose interconversions can take place directly without passing the I0 . 36 state or indirectly via it . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 01510 . 7554/eLife . 17505 . 016Figure 5—figure supplement 1 . Complete G-to-E transitions of L99A obtained by reconnaissance metadynamics simulations . The state-specific fraction of contacts ( Wang et al . , 2012 ) , and , were employed to monitor the conformational transitions to G and E state , respectively . Trajectories Trj1 , Trj2 and Trj3 are from the simulations RUN10 , RUN11 and RUN12 ( Appendix 1—table 1 ) , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 01610 . 7554/eLife . 17505 . 017Figure 5—figure supplement 2 . Conformational transitions between the G and E states monitored by other order parameters . Trajectories Trj1 ( magenta ) , Trj2 ( blue ) and Trj3 ( green and orange ) are from the simulations RUN10 , RUN11 and RUN12 ( Appendix 1—table 2 ) , respectively . The steepest descent path ( SDP , black ) used to define the initial path in PathMetaD is also shown as a reference . To measure the distance between helix F and helix I , and between F144 and helix D , we employed order parameters RHF-HI and RF114-HD . is defined as the Cα distance between E108 and R137 , while RF114-HD is defined as the distance between the Cδ4 atom of F114 and the Cα atom of Y88 . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 01710 . 7554/eLife . 17505 . 018Figure 5—figure supplement 3 . Solvent accessible surface area ( SASA ) calculation of the side chain of F114 . The figure suggests in the direct G⇋E transitions ( Trj1 and first half of Trj3 ) F114 can rotate its side chain inside the protein core . In contrast , in the G⇋I0 . 36⇋E route ( Trj2 and second half of Trj3 ) the side chain of F114 , which occupies the cavity in the E state , gets transiently exposed to solvent during the transition . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 018 In the context of coarse-grained kinetic models the results above would suggest at least two possible mechanisms operate in parallel: G⇌E or G⇌I0 . 36⇌E . Further inspection of the structures along these different kinetics routes ( see the trajectories of other order parameters in Figure 5—figure supplement 2 and Videos 1–4 ) suggested an interesting distinction between the two . In the G⇌I0 . 36⇌E route the side chain of F114 , which occupies the cavity in the E state , gets transiently exposed to solvent during the transition , whereas in the direct G⇌E transitions F114 can rotate its side chain inside the protein core ( see also the solvent accessible surface area calculation of F114 in Figure 5—figure supplement 3 ) . 10 . 7554/eLife . 17505 . 019Video 1 . Trajectory of the G-to-E conformational transition observed in Trj1 , corresponding to the red trajectory in Figure 5 . The backbone of L99A is represented by white ribbons , Helices E , F and G are highlighted in blue , while F114 is represented by red spheres . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 01910 . 7554/eLife . 17505 . 020Video 2 . Trajectory of the G-to-E conformational transition observed in Trj2 , corresponding to the blue trajectory in Figure 5 . The backbone of L99A is represented by white ribbons , Helices E , F and G are highlighted in blue , while F114 is represented by red spheres . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 02010 . 7554/eLife . 17505 . 021Video 3 . Trajectory of the G-to-E conformational transition observed in Trj3 , corresponding to the green trajectory in Figure 5 . The backbone of L99A is represented by white ribbons , Helices E , F and G are highlighted in blue , while F114 is represented by red spheres . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 02110 . 7554/eLife . 17505 . 022Video 4 . Trajectory of the E-to-G conformational transition observed in Trj3 , corresponding to the yellow trajectory in Figure 5 . The backbone of L99A is represented by white ribbons , Helices E , F and G are highlighted in blue , while F114 is represented by red spheres . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 022 As the internal cavity in L99A T4L remains buried in both the G and E states ( and indeed occupied by F114 in the E state ) it remains unclear how ligands access this internal cavity and how rapid binding and release is achieved . Visual inspection of our trajectories and solvent-accessible surface area analysis revealed structures with transient exposure of the internal cavity towards the solvent . The structures were mostly found in a region of conformational space that mapped onto the I0 . 36 basin ( Figure 2 ) , and the events of that basin mostly took place between 430 ns and 447 ns ( see Video 5 ) . Thus , we mapped these structures to the free energy surface ( Figure 6—figure supplement 1 ) and analysed them . Overall , the structure is more similar to the G- than E-state , though is more loosely packed . The similarity to the G-state is compatible with rapid binding and position of F114 in this state . 10 . 7554/eLife . 17505 . 023Video 5 . Movie of the calculated two-dimensional free energy landscape of L99A as a function of simulation time . The figure shows the time evolution of the free energy surface as a function of Spath and Zpath sampled in a 667 ns PathMetaD simulation of L99A . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 023 We used CAVER3 ( Chovancova et al . , 2012 ) ( see parameters in Appendix 1—table 4 ) to analyse the structures and found multiple tunnels connecting the cavity with protein surface ( Figure 6—figure supplement 1 and 2 ) . The tunnels are relatively narrow with the typical radius of the bottleneck ( defined as the narrowest part of a given tunnel ) between ∼1 Å − ∼2 Å . We used CAVER Analyst1 . 0 ( Kozlikova et al . , 2014 ) ( see details in Appendix and parameters in Appendix 1—table 4 ) to separate the tunnels into different clusters ( Figure 6—figure supplement 3 and Appendix 1—table 5 ) with the dominant cluster ( denoted tunnel#1 ) having a entrance located at the groove between HF and HI . A typical representative structure of I0 . 36 is shown in Figure 6A . The radii along the structures in cluster #1 vary , but share an overall shape ( Figure 6—figure supplement 1 ) , and we find that the maximal bottleneck radius is ∼2 . 5 Å , the average bottleneck radius is ∼1 . 3 Å , and the average length ∼11 . 2 Å . 10 . 7554/eLife . 17505 . 024Figure 6 . A transiently formed tunnel from the solvent to the cavity is a potential ligand binding pathway . ( A ) We here highlight the most populated tunnel structure ( tunnel#1 ) , that has an entrance located at the groove between helix F ( HF ) and helix I ( HI ) . Helices E , F and G ( blue ) and F114 ( red ) are highlighted . ( B ) The panel shows a typical path of benzene ( magenta ) escaping from the cavity of L99A , as seen in ABMD simulations , via a tunnel formed in the same region as tunnel #1 ( see also Video 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 02410 . 7554/eLife . 17505 . 025Figure 6—figure supplement 1 . A transiently formed tunnel from the solvent to the cavity forms in the I0 . 36 state . ( A ) Typical structures from the I0 . 36 state sampled in the simulation ( between 430 ns and 447 ns ) are mapped onto the free energy surface , and represented by yellow points . ( B ) The cavity-related regions ( helix E , F and G ) are coloured in blue , while F114 is coloured in red . F114 tends to be partially solvent exposed in I0 . 36 , resulting in the internal cavity to be open . The tunnel#1 connecting the internal cavity and protein surface is coloured in yellow , and has a peanut-shell like shape . ( C ) shows the radius along the tunnel of structures belong to the cluster of tunnel#1 . Lines in different colours represent different structures . Grey dotted line represents the average tunnel radius . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 02510 . 7554/eLife . 17505 . 026Figure 6—figure supplement 2 . Representative structures of the cavity region in the I0 . 36 state . The figure shows six representative structures of the cavity region revealing multiple tunnels that connect the cavity with the protein surface . The different colours correspond to different tunnels observed , and a structure can have different tunnels with different widths present at the same time . The colours represent the relative size with yellow , purple and green corresponding to tunnels of decreasing width . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 02610 . 7554/eLife . 17505 . 027Figure 6—figure supplement 3 . Tunnel clustering analysis on I0 . 36 state . The clustering of tunnels was performed using the CAVER Analyst software ( Kozlikova et al . , 2014 ) and the average-link hierarchical algorithm based on the calculated matrix of pairwise tunnel distances . We found that the most weighted tunnel ( denoted as tunnel#1 ) populates 27% of the I0 . 36 basin . The second and third tunnels populate 20% and 15% , respectively , but their maximal bottleneck radii are 1 . 4 and 1 . 3 Å , in contrast to the maximal bottleneck radius of tunnel#1 of 2 . 5 Å . ( A ) Heat map visualization of the tunnel profile of tunnel#1 . The colour map represents the radius of the tunnel#1 along the tunnel length . ( B ) Average tunnel radius and minimal tunnel radius of individual structures belonging to tunnel#1 cluster . Note that the gaps indicate these snapshots do not have tunnels . ( C ) The tunnel radius as a function of the tunnel index which is ranked by the average radius ( R ) . The second widest tunnel ( tunnel#1 ) has the highest population and is highlighted in yellow . ( D ) A typical structure of I0 . 36 with an open tunnel#1 . HE , HF and HG are coloured in blue , F114 is coloured in red , and tunnel#1 is coloured in yellow . ( E ) The figure shows the location of an alkylbenzene ( magenta ) in a crystal structure of L99A T4L ( PDB ID: 4W59 ) . The figure further shows ( in yellow ) the tunnel induced in the structure by the alkyl chain , as revealed by CAVER3 when applied to the structure after removing the ligand . Because the tunnel overlaps with the alkyl chain of the ligand , only the phenyl moiety of the ligand is visible . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 02710 . 7554/eLife . 17505 . 028Figure 6—figure supplement 4 . Ligand unbinding pathways revealed by ABMD simulations . The figure shows how ABMD simulations allow us to observe the ligand benzene escape from the internal binding site . We performed two sets of 20 simulations using two different force constants for the ABMD ( upper: 50 KJ · mol−1 · nm−2; lower: 20 KJ · mol−1 · nm−2 ) ; note also the different time scales on the two plots . The simulations used the RMSD of the ligand to the bound state as the reaction coordinate , but are here shown projected onto the distance between the benzene molecule and the side chain of F114 . The three structures in the bottom panel provide representative structures . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 028 Interestingly , a series of structures of L99A were recently described , in which the internal cavity where filled with eight congeneric ligands of increasing size to eventually open the structure size ( Merski et al . , 2015 ) . We performed a comparable tunnel analysis on those eight ligand-bound structures ( PDB ID codes: 4W52 – 4W59 ) , revealing the maximal bottleneck radius of 1 . 8 Å ( bound with n-hexylbenzene , 4W59 ) . Although the size of the tunnel in these X-ray structures is slightly smaller than that in I0 . 36 structures , the location of the tunnel exit is consistent with the dominant tunnel#1 in I0 . 36 ( Figure 6—figure supplement 3 ) . We note , however , that the tunnels observed in our simulation and in the ligand-induced cavity-open X-ray structure ( 4W59 ) , are too narrow to allow for unhindered passage of e . g . benzene with its a van der Waals’ width of 3 . 5 Å ( Eriksson et al . , 1992a ) . Thus , we speculate that the transient exposure in I0 . 36 might serve as a possible starting point for ligand ( un ) binding , which would induce ( Koshland , 1958; López et al . , 2013; Wang et al . , 2012 ) further the opening of the tunnel . As an initial step towards characterizing the mechanism of ligand binding and escape we used adiabatic biased molecular dynamics ( ABMD ) simulations ( Marchi and Ballone , 1999; Paci and Karplus , 1999 ) to study the mechanism of how benzene escapes the internal cavity ( see Appendix for details ) . In ABMD the system is perturbed by a ‘ratcheting potential’ , which acts to ‘select’ spontaneous fluctuations towards the ligand-free state . In particular , the biasing potential is zero when the reaction coordinate ( here chosen to be the RMSD of the ligand to the cavity-bound state ) increases , but provides a penalty for fluctuations that brings the ligand closer to the cavity . In this way , we were able to observe multiple unbinding events in simulations despite the long lifetime ( 1 . 2 ms ) of the ligand in the cavity . Most of trajectories ( 15 of the 20 events observed ) reveal that benzene escapes from the cavity by following tunnel #1 ( Figure 6—figure supplement 4 and Appendix 1—table 6 ) . A typical unbinding path is shown in the right panel of Figure 6 ( see also Video 6 ) . Because the ABMD introduces a bias to speed up ligand escape , we ensured that the observed pathway was the same at two different values of the biasing force constants ( Figure 6—figure supplement 4 and Appendix 1—table 6 ) . Future work will be aimed to perform a more quantitative analysis of the ligand binding and unbinding kinetics . 10 . 7554/eLife . 17505 . 029Video 6 . A typical trajectory of the benzene escaping from the buried cavity of L99A via tunnel #1 revealed by ABMD simulations . The backbone of L99A is represented by white ribbons , Helices E , F and G are highlighted in blue , while F114 and benzene are represented by spheres in red and magenta , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 17505 . 029 The ability to change shape is an essential part of the function of many proteins , but it remains difficult to characterize alternative conformations that are only transiently and sparsely populated . We have studied the L99A variant of T4L as a model system that displays a complicated set of dynamical processes which have been characterized in substantial detail . Our results show that modern simulation methods are able to provide insight into such processes , paving the way for future studies for systems that are more difficult to study experimentally . Using a novel method for defining an initial reference path between two conformations , we were able to sample the free energy landscape described by an accurate molecular force field . In accordance with experiments , the simulations revealed two distinct free energy basins that correspond to the major and minor states found by NMR . Quantification of the free energy difference between the two states demonstrated that the force field is able to describe conformational free energies to an accuracy of about 1 kcal mol−1 . This high accuracy is corroborated by previous studies of a different protein , Cyclophilin A , where we also calculated conformational free energies and compared to relaxation dispersion experiments and found very good agreement . For both proteins we were also able to capture and quantify the effect that point mutations have on the equilibrium between the two states , and also here found good agreement with experiments . We note , however , that comparable simulations of the L99A/G113A mutant did not reach convergence . Moving a step further , we here also calculated the kinetics of conformational exchange using a recently developed metadynamics method . For both the L99A variant and a population-inverting triple mutant we find that the calculated reaction rates are in remarkably good agreement with experiments . The ability to calculate both forward and backward rates provided us with the opportunity to obtain an independent estimate the calculated free energy difference . The finding that the free energy differences estimated in this way ( for both L99A and the triple mutant ) are close to those estimated from the free energy landscape provides an important validation of both approaches , and we suggest that , when possible , such calculations could be used to supplement conventional free energy estimates . The free-energy landscape suggested that the E state is relatively broad and contains a wider range of conformations . To validate this observation , we used the same chemical shift information as was used as input to Rosetta and performed replica-averaged CS-restrained simulations . The resulting ensemble demonstrates that the experiments and force field , when used jointly , indeed are compatible with a broader E state . Thus , we suggest that the EROSETTA structure and CS-restrained ensemble jointly describe the structure and dynamics of the E state . While NMR experiments , in favourable cases , can be used to determine the structure , thermodynamics and kinetics of conformational exchange , a detailed description mechanism of interconversion remains very difficult to probe by experiments . We explored potential mechanisms of conformational exchange between the two states , finding at least two distinct routes . One route involved a direct transition with the central F114 entering the cavity within the protein , whereas a different possible mechanism involves transient partial-loosening of the protein . In both cases , the mechanism differs from the reference path that we used as a guide to map the free energy landscape , demonstrating that high accuracy of the initial guess for a pathway is not absolutely required in the metadynamics simulations , suggesting also the more general applicability of the approach . Finally , we observed a set of conformations with a transiently opened tunnel that leads from the exterior of the protein to the internal cavity , that is similar to a recently discovered path that is exposed when the cavity is filled by ligands of increasing size . The fact that such a tunnel can be explored even in the absence of ligands suggests that intrinsic protein motions may play an important role in ligand binding , and indeed we observed this path to be dominant in simulations of ligand unbinding . In total , we present a global view of the many , sometimes coupled , dynamical processes present in a protein . Comparison with a range of experimental observations suggests that the simulations provide a relatively accurate description of the protein , demonstrating how NMR experiments can be used to benchmark quantitatively the ability of simulations to study conformational exchange . We envisage that future studies of this kind , also when less is known about the structure of the alternative states , will help pave the way for using simulations to study the structural dynamics of proteins and how this relates to function .
Our simulations were initiated in the experimentally determined structures of the ground state of L99A ( GXray; PDB ID code 3DMV ) or minor , E state ( EROSETTA; 2LCB ) . The structure of the ground state of the L99A , G113A , R119P triple mutant , corresponding to the E state of L99A was taken from PDB entry 2LC9 ( GROSETTATriple ) . Details can be found in the Appendix . Taking GXray and EROSETTA as the models of the initial and final structures , we calculated an initial reaction path between them with the MOIL software ( Elber et al . , 1995 ) , which has been used to explore the mechanism of conformational change of proteins ( Wang et al . , 2011 ) . Further details can be found in the Appendix and in refs . ( Majek et al . , 2008; Wang et al . , 2011 ) . The adaptive-hill version of metadynamics updates the Gaussian width on the fly according to the local properties of the underlying free-energy surface on the basis of local diffusivity of the CVs or the local geometrical properties . Here , we used the former strategy . Simulation were performed using Gromacs4 . 6 ( Pronk et al . , 2013 ) with the PLUMED2 . 1 plugin ( Tribello et al . , 2014 ) . See parameter details in Appendix 1—table 1 . We performed replica-averaged CS restrained MD simulations by using GPU version of Gromacs5 with the PLUMED2 . 1 and ALMOST2 . 1 ( Fu et al . , 2014 ) plugins . Equilibrated structures of EROSETTA and GROSETTATriple were used as the starting conformations . CS data of EROSETTA and GROSETTATriple were obtained from the BMRB database ( Ulrich et al . , 2008 ) as entries 17604 and 17603 , respectively . Reconnaissance metadynamics ( Tribello et al . , 2010 ) uses a combination of a machine learning technique to automatically identify the locations of free energy minima by periodically clustering the trajectory and a dimensional reduction technique that can reduce the landscape complexity . We performed several reconnaissance metadynamics simulations with different combinations of CVs starting from GXray using Gromacs4 . 5 . 5 with PLUMED1 . 3 plugin . See parameter details in Appendix 1—table 1 . The key idea of infrequent metadynamics is to bias the system with a frequency slower than the barrier crossing time but faster than the slow intra-basin relaxation time , so that the transition state region has a low risk of being substantially biased . As the first transition times should obey Poisson statistics , the reliability of the kinetics estimated from InMetaD can be assessed by a statistical analysis based on the Kolmogorov-Smirnov ( KS ) test ( Salvalaglio et al . , 2014 ) . See parameter details on Appendix and Appendix 1—table 1 . | Proteins are the workhorses of cells , where they perform a wide range of roles . To do so , they adopt specific three-dimensional structures that enable them to interact with other molecules as necessary . Often a protein needs to be able to shift between different states with distinct structures as it goes about its job . To fully understand how a protein works , it is important to be able to characterize these different structures and how the protein changes between them . Many of the experimental techniques used to study protein structure rely on isolating the individual structural forms of a protein . Since many structures only exist briefly , this can be very difficult . To complement experimental results , computer simulations allow researchers to model how atoms behave within a molecule . However , a number of factors limit how well these models represent what happens experimentally , such as the accuracy of the physical description used for the modeling . Wang et al . set out to test and benchmark how well computer simulations could model changes in structure for a protein called T4 lysozyme , which has been studied extensively using experimental techniques . T4 lysozyme exists in two different states that have distinct structures . By comparing existing detailed experimental measurements with the results of their simulations , Wang et al . found that the simulations could capture key aspects of how T4 lysozyme changes its shape . The simulations described the structure of the protein in both states and accurately determined the relative proportion of molecules that are found in each state . They could also determine how long it takes for a molecule to change its shape from one state to the other . The findings allowed Wang et al . to describe in fine detail – down to the level of individual atoms – how the protein changes its shape and how mutations in the protein affect its ability to do so . A key question for future studies is whether these insights can be extended to other proteins that are less well characterized experimentally than T4 lysozyme . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2016 | Mapping transiently formed and sparsely populated conformations on a complex energy landscape |
Increases in ocean temperature are associated with changes in the distribution of fish stocks , and the foraging regimes and maternal attendance patterns of marine mammals . However , it is not well understood how these changes affect offspring health and survival . The maternal attendance patterns and immunity of South American fur seals were assessed in a rookery where hookworm disease is the main cause of pup mortality . Pups receiving higher levels of maternal attendance had a positive energy balance and a more reactive immune system . These pups were able to expel hookworms through a specific immune mediated mechanism and survived the infection . Maternal attendance was higher in years with low sea surface temperature , therefore , the mean hookworm burden and mortality increased with sea surface temperature over a 10-year period . We provide a mechanistic explanation regarding how changes in ocean temperature and maternal care affect infectious diseases dynamics in a marine mammal .
Marine mammals are a diverse group of top predators highly sensitive to changes in aquatic ecosystems ( Constable et al . , 2014 ) . Within this group , fur seals and sea lions ( otariids ) breed and give birth on land but forage at sea , alternating periods of foraging in the ocean with periods of offspring attendance and nursing on land ( income breeders ) ( Stephens et al . , 2009 ) . Therefore , otariids , like other marine mammals , are highly sensitive to local changes in prey distribution and abundance ( Trillmich et al . , 1991 , Constable et al . , 2014 , Elorriaga-Verplancken et al . , 2016 ) . One of the most important indexes of the abundance of marine mammal prey is sea surface temperature ( SST ) ( Soto et al . , 2006 , Elorriaga-Verplancken et al . , 2016 ) . Warmer SST indicates reduced nutrient upwelling , which is associated with reduced primary productivity and abundance of mesopelagic marine organisms ( Lewandowska et al . , 2014 ) . This decrease in food resources forces otariid females to change their foraging strategies by increasing their foraging trip lengths , resulting in decreased time spent on land with their pup ( maternal attendance ) ( Trillmich et al . , 1991 , Costa , 2008 ) . These changes in patterns of maternal attendance have been associated with decreased pup growth and increased mortality ( Soto et al . , 2006 , Jeanniard-du-Dot et al . , 2017 ) . Regardless , the mechanisms that drive decreased survival during years with low ocean productivity have not been intensely explored beyond assuming that this results from direct mortality because of starvation . However , in some otariid populations , in years with abnormal SST , the immune competence of pups decreases ( Banuet-Martínez et al . , 2017 ) , suggesting that environmental variables can affect the health of marine mammals by impairing their immune function . If these immunological changes impact offspring survival , there could be additional negative consequences between a warmer ocean , health , and survival of marine vertebrates . In some marine mammal populations , infectious diseases are one of the most significant causes of mortality among young individuals ( Gulland and Hall , 2007; Spraker et al . , 2007; Seguel et al . , 2013 ) . In otariids , hookworms ( Uncinaria sp . ) have been described in nearly all species , and while some populations suffer few adverse effects , others experience up to 70% of hookworm-related mortality being one of the most significant infectious diseases of young fur seals and sea lions ( Spraker et al . , 2007; Lyons et al . , 2011a; Seguel et al . , 2013; Seguel and Gottdenker , 2017 ) . Fur seals are infected with hookworms ( Uncinaria sp . ) during their first 1–4 days of life through their mother’s colostrum ( Lyons et al . , 2011b; Seguel et al . , 2018 ) . These nematodes live in the small intestine where they bite the mucosa to feed on blood , causing substantial tissue damage , anemia , and death ( Marcus et al . , 2015 , Seguel et al . , 2017 , Seguel et al . , 2018 ) ; however , it is unclear how the host responds to this infection . Long term studies in fur seal populations show that hookworm prevalence and mortality varies over time , but the mechanisms driving these patterns are unknown ( Lyons et al . , 2011a; Seguel et al . , 2013 ) . In this paper , we describe how oceanographic environmental variables , via the modification of maternal care , are associated with immune-mediated parasite clearance , and survival of a marine mammal , the South American fur seal ( SAFS , Arctocephalus australis ) .
The hookworm ( Uncinaria sp . ) prepatent period varied from 14 to 18 days and based on the coprological analyses and necropsies of recaptured pups , the number of days a pup released hookworm eggs ( infectious period ) ranged from 5 to 55 days ( 2014–15 and 2017 , mean = 25 . 7 ± 10 . 9 , n = 146 ) . Seven to 15 days before having a negative coprological test , fur seal pups experienced a decline of more than 50% in the number of eggs shed in previous exams . At this stage , pups were considered to be in a hookworm clearance state . When presenting the first negative coprological exam , they were considered to have cleared hookworm infection ( Figure 1A ) . Between 81% to 100% of pups examined through necropsy between 2005–08 ( n = 124 ) and 2012–17 ( n = 154 ) had evidence of hookworm infection , and hookworm-related mortality corresponded to 13–50% of all pups found dead ( n = 56 , Figure 1B ) . Total hookworm mortality could be calculated in a subset of marked pups in 2014 ( n = 38 ) , 2015 ( n = 53 ) , and 2017 ( n = 54 ) ( Figure 1—source data 1 ) . Hookworms killed 42 . 1% of pups born in 2014 , 20 . 7% of pups born in 2015 , and 24% of pups born in 2017 at Guafo Island ( GLM , 2014 = 1 . 02 ± 0 . 47 , Z = 2 . 16 , p = 0 . 0304 ) . Based on multimodel inference using generalized linear mixed models , pups that had higher hookworm burden , delayed hookworm clearance , and lower plasma concentration of parasite specific IgG , blood urea nitrogen , and glucose were more likely to die from hookworm disease ( Figure 1C–F ) ( Supplementary file 1 and 2 ) . Therefore , the most important host-related factors affecting hookworm mortality were energy balance and immune response against the parasite . The parasite-related factors affecting mortality suggested that hookworm clearance , by reducing infectious period and hookworm burden , enhanced host survival . To determine the mechanisms that drive hookworm clearance and affect host mortality , the immune response to hookworms was investigated during 2017 at different infection stages in 54 fur seal pups , and compared to 24 hookworm-free ( ivermectin-treated ) age-matched controls ( Figures 2 , 4A and B , Figure 2—source data 1 ) . The number of peripheral blood leukocytes ( lymphocytes , macrophages , neutrophils , eosinophils , and basophils ) was obtained as a basic tool to indirectly measure the level of proliferation of these different immune cell types in infected and control animals . During the patent and clearance period , fur seal pups that survived infection ( n = 41 ) experienced a significant increase in the number of peripheral blood lymphocytes ( GLMM , 0 . 9 ± 0 . 003 , Z = 231 , p = 2 . 0×10−16 ) and basophils ( GLMM , 4 . 8 ± 0 . 08 , Z = 56 . 7 , p = 2 . 0×10−16 ) , and had higher numbers of these cells when compared to age-matched controls and to the pups that died from hookworm infection ( Figure 2A–B ) . The number of neutrophils in peripheral blood was similar between controls and pups that survived but slightly lower in pups that died ( n = 13 ) from hookworm disease ( GLMM , died = −0 . 53 ± 0 . 06 , Z = −8 . 04 , p = 9 . 1×10−16 ) . During the patent period , lower numbers of monocytes were found in animals that died from hookworm disease compared to controls ( GLMM , died = −0 . 88 ± 0 . 11 , Z = −7 . 64 , p = 1 . 54×10−14 ) , and eosinophils were higher in animals that survived when compared to controls and animals that died ( GLMM , survived = 0 . 86 ± 0 . 16 , Z = 5 . 19 , p = 2 . 0×10−7 ) ; however , during the clearance and post-clearance periods , eosinophils ( GLMM , survived = 0 . 26 ± 0 . 14 , Z = 1 . 88 , p = 0 . 07 ) and macrophages ( GLMM , survived = −0 . 03 ± 0 . 09 , Z = 0 . 36 , p = 0 . 71 ) were in similar numbers in pups that survived infection and controls . Pups that cleared the infection developed medium to high levels of parasite-specific IgG , whereas the level of these antibodies was significantly lower in pups that died from hookworm infection and almost non-existent in the control group ( Figure 2J–M ) . There was moderate to marked immunolabelling of the hookworm intestinal brush border using serum from six pups with moderate to high levels of parasite-specific IgG ( 23–100 arbitrary units ) ( Figure 2I ) , suggesting that anti-hookworm antibodies bind proteins located in the hookworm intestine . To determine the morphological and immune cell population changes in the anatomical site of hookworm infection , sections of small intestine and mesenteric lymph nodes were collected from pups that died from hookworm disease ( n = 21 ) , pups that were undergoing clearance ( n = 18 ) , and pups that were never infected with hookworms ( controls , n = 6 ) ( Figure 3—source data 1 ) . The small intestine mucosa , submucosa , and the mesenteric lymph nodes of pups undergoing hookworm clearance contained larger numbers of T-lymphocytes when compared to pups that died from hookworm infection or pups never infected with adult Uncinaria sp ( Generalized linear models with negative binomial distribution ( GLM . NB ) , mucosa clearance = 0 . 86 ± 0 . 11 , Z = 8 . 312 , p = 2 . 0×10−16 , submucosa clearance = 1 . 08 ± 0 . 16 , Z = 6 . 86 , p = 7 . 0×10−12 , mesenteric lymph node clearance = 0 . 78 ± 0 . 06 , Z = 14 . 21 , p = 2 . 0×10−16 ) ( Figure 3 ) . B-lymphocytes and plasma cells were more numerous in the mesenteric lymph node of pups clearing hookworm infection versus controls and pups dead from hookworm infection ( GLM . NB , B-lymphocytes clearance = 0 . 29 ± 0 . 07 , Z = 4 . 1 , p = 4 . 1×10−5 , plasma cell clearance = 0 . 59 ± 0 . 05 , Z = 10 . 2 , p = 2 . 0×10−16 ) . Similarly , there were higher numbers of mast cells ( GLM . NB , clearance = 1 . 14 ± 0 . 25 , Z = 4 . 6 , p = 4 . 2×10−6 ) and more mucus ( GLM , clearance = 0 . 03 ± 0 . 003 , Z = 7 . 84 , p = 9 . 4×10−10 ) in the mucosa , and more leukocytes expressing IL-4 in the intestine ( GLM . NB , clearance = 1 . 97 ± 0 . 15 , Z = 12 . 72 , p = 2 . 0×10−16 ) and mesenteric lymph node ( GLM . NB , clearance = 1 . 57 ± 0 . 11 , Z = 14 . 63 , p = 2 . 0×10−16 ) of pups that cleared hookworm infection when compared to controls and pups with hookworm enteritis and bacteremia . Pups that died from hookworms , however , had larger numbers of macrophages in the intestinal submucosa ( GLM . NB , mortality = 0 . 52 ± 0 . 09 , Z = 5 . 79 , p = 6 . 94×10−9 ) and mesenteric lymph nodes ( GLM . NB , mortality = 0 . 63 ± 0 . 05 , Z = 12 . 68 , p = 2 . 0×10−16 ) compared to pups never infected with hookworms and pups clearing hookworm infection ( Figure 3 ) . Maternal attendance patterns and pup-related health parameters were assessed in the 2017 reproductive season ( n = 78 ) ( Figure 4A and B , Figure 4—source data 1 ) . Among measured serum chemistry variables , the average level of blood glucose was the best predictor of growth rate ( GLM . NB , 0 . 18 ± 0 . 03 , Z = 5 . 9 , p = 2×10−16 ) . Among the considered external factors that could affect growth , the number of nursing events observed in a pup was the most significant predictor of growth rate , and although hookworm burden and hookworm infectious period were included in some top ranked models , their effect was not significant ( Figure 4A , supplementary file 4 and 5 ) . Additionally , there were no significant differences in growth rates between pups treated with ivermectin ( n = 24 ) and non-treated ( n = 54 ) ( GLM . NB , 0 . 17 ± 0 . 15 , Z = 1 . 16 , p = 0 . 26 ) . Nevertheless , when pups that died from hookworm disease were considered ( n = 13 ) , they had significantly slower growth rates ( GLM . NB , −1 . 05 ± 0 . 16 , Z = −6 . 6 , p = 2 . 7×10−11 ) compared to pups that survived ( n = 41 ) and pups treated with ivermectin; however , the animals that died also had the lowest levels of maternal attendance ( GLM . NB , −0 . 78 ± 0 . 23 , Z = −3 . 45 , p = 5 . 5×10−4 ) ( Figure 4B ) . Regarding the factors that affected overall immune reactivity ( Figure 4C , Figure 4—source data 1 ) , pups with more nursing events , faster growth rate , and higher hookworm burden were more likely to recruit higher numbers of T-cells ( CD3+ lymphocytes ) in the skin in response to ( Phytohemagglutinin ) PHA challenge ( Figure 4C , supplementary file 5 and 6 ) . Pups with lower parasite-specific IgG concentrations ( GLMM . NB , coeff = −0 . 017 ± 0 . 002 , Z = 6 . 54 , p = 2×10−16 , n = 146 ) and higher hookworm burden ( GLMM . NB , coeff = 0 . 06 ± 0 . 022 , Z = 2 . 78 , p = 0 . 005 , n = 146 ) had longer infectious periods ( Supplementary file 7 and 8 ) , suggesting that among measured immune parameters , parasite-specific IgG was the most significant factor affecting the permanence of hookworms in the intestine . Based on the PHA immune challenge performed when pups were 1-mo-old ( Figure 4D , Figure 4—source data 1 ) , animals with high T-cell response had higher levels of IgG , maternal attendance , glucose , growth rate , and shorter infectious periods at the end of the study when compared to the average levels in pups with low T-cell response ( Figure 4D ) . However , hookworm burden was similar between the two groups ( Figure 4D ) , suggesting , in conjunction with the previous analyses , that maternal attendance and growth rate accounted for most of the difference in T-cell reactivity between these groups . SAFS females were observed more frequently arriving to the rookery from foraging trips early in the morning ( 2007 = 78/115 , 67 . 8% returning events in the morning , 2017 = 87/135 , 64% returning events in the morning ) . Foraging trip length was correlated with the number of nursing events , indicating that the more time females spend at sea makes it less likely to observe them nursing their pup ( Figure 5A ) . In 2017 , a year with SST above Guafo Island average , SAFS females ( n = 21 ) spend more time foraging at sea compared to 2007 ( n = 23 ) , a year with SST below Guafo Island average , therefore in 2017 ( n = 79 ) the level of maternal attendance and pup growth rate were lower than in 2007 ( n = 128 ) ( Figure 5B ) ( Figure 5A and B , Figure 5—source data 1 ) . Between 2012 and 2017 ( Figure 5C , Figure 5—source data 1 ) , in years with high SST ( e . g . 2014 ) , the average concentrations of glucose , cholesterol , parasite-specific IgG , and peripheral blood lymphocytes and basophils were lower than in years with low SST ( Figure 5C ) . Similarly , the average hookworm infectious period was shorter in years with low SST ( GLM , X2 = 6 . 95 , df = 1 , p = 0 . 00036 ) . Over a 10-y period ( 2005–08 , 2012–17 ) ( Figure 6 , Figure 6—source data 1 and 2 ) , there was a significant positive correlation between mean hookworm burdens of necropsied pups and SST ( Linear regression , Ad-R2 = 0 . 86 , p < 0 . 001 ) , and between hookworm mortality and SST ( Ad-R2 = 0 . 56 , p = 0 . 016 ) ; however , in the case of hookworm prevalence at necropsy the correlation with SST was not significant ( Figure 6 , supplementary file 9–11 ) ( Ad-R2 = 0 . 29 , p = 0 . 064 ) . A similar but negative correlation existed between the same hookworm epidemiological parameters and average chlorophyll-a concentrations ( Figure 6 ) .
In Chilean Patagonia , during years with high SST , ocean productivity decreases , forcing adult female fur seals to increase their foraging trip length and decrease their levels of maternal attendance . Pups receiving less maternal care had reduced growth rates and decreased energy budget , impairing the ability of their immune system to mount an effective response against hookworms to expel the parasite from the intestine . These pups , with longer hookworm infection periods , usually die as a consequence of hookworm disease establishing a pattern in which hookworm disease severity and mortality are correlated to indexes of oceanographic environmental conditions such as sea surface temperature . The sensitivity of otariid hookworm disease to ocean temperature and marine productivity presents a scenario where global climate change could increase the extent and severity of a disease present in most fur seal and sea lion populations .
From 2012 through 2017 South American fur seal pups were captured by hand every 7–15 d between December 15th and March 10th . At the first capture , pups were marked with a number in the fur using commercial hair decoloring solution . During each capture procedure , standard length , weight , sex , and body condition were recorded . The pup age was calculated based on the peak of parturition for Guafo Island rookery ( December 15 , Pavés et al . , 2016 ) and based on the assessment of rest of placenta ( 1–3-d-old ) and umbilical cord ( 2–7-d-old ) during the first capture . For a subset of pups during 2014 ( n = 10 ) , 2015 ( n = 20 ) , and 2017 ( n = 40 ) , age was exactly known because their parturition was observed and they were marked 24 h later . Blood was drawn from the caudal gluteal vein of pups into EDTA , heparin , and plain ( serum ) vacutainer tubes . Plain blood tubes were centrifuged within 1–3 h post-collection in the field laboratory to obtain serum , which was preserved at −20°C until later long-term storage ( −80°C ) or analyses in the mainland laboratory . Plasma was obtained and stored following similar procedures with heparin non-coagulated blood . During each capture procedure , a rectal swab was collected and stored in Sheather’s sucrose for later semi-quantitative determination of hookworm egg burden according to standardized methods for this fur seal population ( Seguel et al . , 2018 ) . Hookworm burden of pups found dead was determined by collection , sexing , and counting of all nematodes present in the small intestine and correlated with egg burden through a fecal swab collected during necropsy ( Seguel et al . , 2017; Seguel et al . , 2018 ) . Using non-coagulated blood , hematocrit , hemoglobin concentration , total red blood cell count ( RBC ) , total white blood cell count ( WBC ) , and differential leukocyte counting were determined for each pup as previously described ( Seguel et al . , 2016 ) . The total serum concentrations of albumin , globulins , cholesterol , glucose , triglycerides , blood urea nitrogen , and creatinine were determined in the mainland laboratory using previously described methods for this population ( Seguel et al . , 2016 ) . All blood and serum ( or plasma ) measured parameters were obtained for every capture procedure . Because previous studies indicated a hookworm prevalence close to 100% in this population ( Seguel et al . , 2018 ) , a ‘hookworm-free’ control group was created in 2017 by treating 60 pups with a subcutaneous injection of ivermectin ( 300 µg/kg ) when they were between 1 and 7 d old . These pups were subjected to the same capture , handling , health assessments , and data acquisition procedures as indicated for the non-treated pups . These pups never presented hookworm eggs in their feces during the duration of the study . In 2014 ( n = 38 ) , 2015 ( n = 53 ) , and 2017 ( n = 54 ) marked pups were observed at least once a week during the study period . Pups observed dead were retrieved from the rookery to perform complete necropsies , collect tissue samples for histopathology , and determine the cause of death according to previously described diagnostic criteria ( Seguel et al . , 2011; Seguel and Gottdenker , 2017 ) . The minimal number of pups to capture every year was calculated at the beginning of the reproductive season based on the known recapture rates at Guafo Island ( 60–80% ) and sample size simulations to reach a power of at least 80% ( R packages ‘pwr’ and ‘SIMR’ ) . Therefore , hookworm disease outcome ( dead vs . survived ) was known in these pups and registered in the final data sheet to calculate total hookworm mortality and to fit models to identify the most significant health-related parameters that predicted hookworm mortality . In 2017 , a PHA immune challenge experiment was performed in a group of pups when they were approximately 8 wk old ( n = 75 ) . For these animals there were enough recaptures ( at least four ) to measure the average of all health-related parameters , hookworm infection history , and outcome ( survival vs . death ) at the end of the study period ( February–March ) . The challenge consisted of injection of 0 . 1 ml of a 1 . 0 mg * ml−1 solution of phytohemagglutinin ( PHA ) into the interdigital skin of the right posterior flipper ( Vera-Massieu et al . , 2015 ) . The same volume of a saline solution was injected in the same location of the left flipper ( control ) . Swelling was measured in both injection sites 12 h after challenge and a 4-mm punch biopsy was collected from each site ( PHA and control ) following anesthesia with 5% isoflurane . Biopsy samples were stored in 10% buffered formalin and routinely processed for histopathology and immunohistochemistry for CD3 . The number of CD3+ lymphocytes in control and treatment biopsies were counted and the difference between these two recorded and used in statistical analyses . In 2017 , another subset of pups of known age ( n = 55 ) was immune-challenged when they were approximately 30-d-old ( during the acute hookworm infection phase ) . Sample size was calculated based on a minimum power of 80% using data collected in a preliminary study in 2016 ( differences in skin swelling and T-cell recruitment ) . Pups were divided into two groups based on level of skin swelling and number of CD3+ lymphocytes detected during examination of biopsy samples . Animals with more than 20 CD3+ lymphocytes per section were considered high responders , whereas pups with less than 16 CD3+ lymphocytes were categorized as low responders . Only one pup was in the middle range ( 17 cells ) and was not included in comparison analyses . At the end of the study period , data on growth rate , maternal attendance , and hookworm infection status were available for these pups . All data are available in the manuscript or the supplementary materials . | Every year off the coasts of Chile , Guafo Island becomes a nursery for South American fur seals pups . Mother fur seals leave their young on the beaches , going out at sea to hunt for fish before returning to the shore to nurse . These first few months are dangerous for young seals , with many dying because of hookworms , parasites that latch to the wall of the bowels to suck blood . However , the immune system of the pups is usually able to mount a response and fight off these parasites . Even though the pups stay on land , their lives depend on the health of the ocean that feeds their nursing mothers . In recent years , sea temperature has been rising rapidly , which modifies winds and water currents . This can set off a chain of events that results in fewer fish being available for seals and other marine mammals to eat . Researchers know that years with warmer waters are associated with changes in the pattern of the mothers’ hunting trips , more pups’ deaths , and a weaker immune system in young fur seals . However , the mechanisms that connect these different factors are still unclear . To explore this , Seguel et al . followed South American fur seals colonies on Guafo Island for several years , tracking the mothers’ trips and monitoring the health of the pups by looking at their levels of blood sugar , whether they carry hookworms , and certain elements of their immune system . Results showed that in years when the sea is warmer , fur seal mothers are gone hunting for longer: they spend less time nursing their young , which then grow slower . These young seals also have lower levels of blood sugar , and so they have less energy to create the immune response necessary to clear off parasitic worms . In fact , in years with warmer seas , almost half of the pups die from hookworm infections . The work by Seguel et al . shows that warmer oceans directly weaken the immune defenses of certain marine mammals . If temperatures keep rising , infectious diseases may kill more of these animals . Further work is now needed to explore if strategies could be developed to help seal populations , for example by treating the pups with drugs that eliminate the parasites . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"microbiology",
"and",
"infectious",
"disease"
] | 2018 | Immune-mediated hookworm clearance and survival of a marine mammal decrease with warmer ocean temperatures |
Cdc48 is a AAA+ ATPase that plays an essential role for many cellular processes in eukaryotic cells . An archaeal homologue of this highly conserved enzyme was shown to directly interact with the 20S proteasome . Here , we analyze the occurrence and phylogeny of a Cdc48 homologue in Actinobacteria and assess its cellular function and possible interaction with the bacterial proteasome . Our data demonstrate that Cdc48-like protein of actinobacteria ( Cpa ) forms hexameric rings and that the oligomeric state correlates directly with the ATPase activity . Furthermore , we show that the assembled Cpa rings can physically interact with the 20S core particle . Comparison of the Mycobacterium smegmatis wild-type with a cpa knockout strain under carbon starvation uncovers significant changes in the levels of around 500 proteins . Pathway mapping of the observed pattern of changes identifies ribosomal proteins as a particular hotspot , pointing amongst others toward a role of Cpa in ribosome adaptation during starvation .
Energy-dependent chaperones and chaperone-protease complexes comprise important cellular components guarding protein homeostasis in all kingdoms of life . Chaperones in the context of protein degradation form complexes with compartmentalizing protease cylinders and act by unfolding their protein substrates , translocating them into the proteolytic compartment for degradation . Mycobacteria and many other actinobacteria possess several independent chaperone-protease complexes: the canonical bacterial Clp protease system and membrane-bound FtsH protease as well as a eukaryotic-like proteasome system ( Imkamp et al . , 2015 ) . The existence of proteasomes in bacteria is an unusual feature restricted to actinobacteria , a diverse phylum with different members showing various examples of eukaryotic-like complexes or activities ( Cavalier-Smith , 2002 ) . Like in eukaryotes , the bacterial 20S proteasome core particle must associate with an activator to form the fully active proteolytic complex . The first activator described was ARC ( ATPase forming ring-shaped complexes ) , also called Mpa ( mycobacterial proteasomal ATPase ) in mycobacteria , a divergent AAA+ family protein that catalyzes the ATP-dependent unfolding and translocation into the proteasome chamber of proteins post-translationally modified with prokaryotic ubiquitin-like protein Pup ( Burns et al . , 2009; Pearce et al . , 2008; Striebel et al . , 2014 ) . Proteomic studies on whole-cell purified ‘pupylomes’ ( pupylated proteomes ) identified hundreds of cellular proteins as pupylation targets ( Compton et al . , 2015; Festa et al . , 2010; Watrous et al . , 2010 ) . One of the best characterized pupylation targets and favorite in vitro substrates is ketopantoate hydroxymethyltransferase ( PanB ) , the enzyme that catalyzes the first committed step of pantothenate biosynthesis ( Chaudhuri et al . , 2003; Pearce et al . , 2006 ) . More recently , a second proteasomal activator was discovered , the bacterial proteasome activator Bpa ( also referred to as PafE ) ( Delley et al . , 2014; Jastrab et al . , 2015 ) . Bpa acts in an ATP-independent manner and is thought to aid in proteasomal degradation of proteins damaged under stress conditions , since it supports degradation of the unstructured model substrate casein and is involved in degradation of the conformationally unstable transcriptional heat shock protein repressor HspR in vivo ( Jastrab et al . , 2017; Jastrab et al . , 2015 ) . The two non-homologous activators Mpa/ARC and Bpa both form ring-shaped complexes with a central pore and share as one additional important feature with eukaryotic proteasome interactors a C-terminal proteasome interaction motif featuring a penultimate tyrosine ( Delley et al . , 2014; Imkamp et al . , 2015 ) . In eukaryotes , this motif is called HbYX motif , since the tyrosine is preceded by a hydrophobic residue . Mpa/ARC and Bpa on the other hand both carry the sequence ‘GQYL’ , where Q can be mutated freely without changing association behavior of the activator with the proteasome ( Jastrab et al . , 2015 ) . In eukaryotes , 20S proteasomes are found in the cytoplasm as well as in the nucleus , where they degrade proteins upon association with the 19S regulatory particle . The regulatory particle contains six ATPase subunits that form the hexameric basal ring stacking onto the 20S cylinder . However , the proteasome is also responsible for degradation of ER-resident proteins via the so-called ER-associated degradation ( ERAD ) pathway that requires retrotranslocation of substrate proteins out of the ER into the cytoplasm . In this context , the 20S proteasome additionally cooperates with another AAA+ protein , Cdc48 ( also known as p97 or VCP ) ( Baek et al . , 2013; DeLaBarre et al . , 2006; Wolf and Stolz , 2012 ) . Cdc48 has also been implicated in a multitude of other cellular processes like membrane fusion , autophagy and gene expression ( Baek et al . , 2013; Yamanaka et al . , 2012 ) . The different Cdc48 functions are largely mediated by various adaptor proteins ( >40 ) binding to the N-terminal domain or C-terminal region of Cdc48 and aiding in recognition of different substrates ( Baek et al . , 2013; Buchberger et al . , 2015 ) . Although the exact role played by Cdc48 during ERAD is not fully understood , it is involved in pulling proteins from the membrane or from ribosomes during co-translational degradation , and recent evidence suggests even a direct association with the 20S particle ( Barthelme and Sauer , 2013 ) . Such a direct role as 20S proteasome activator has been shown for the Cdc48 homolog from archaea in vitro ( Barthelme et al . , 2014 ) . Archaeal Cdc48 forms a complex with the 20S proteasome that is capable of degrading ssrA-tagged model substrates , likely due to the disordered , extended nature of the ssrA-tag at their C-terminus ( Barthelme and Sauer , 2013 ) . Eukaryotic Cdc48 on the other hand is not able to recognize ssrA-tagged substrates in vitro , which was proposed to be due to the absence of two hydrophobic residues in the entry pore loop . These residues must be involved in recognition of the ssrA tag , since their introduction into the eukaryotic p97 enabled it to unfold ssrA-tagged model substrates ( Barthelme and Sauer , 2013; Rothballer et al . , 2007 ) . Although the ssrA-tag is not a bona fide degradation signal in eukaryots and archaea , the D1 pore-1 loops and the nature of the residues within these loops appear to play a general role in substrate recruitment . A recent study demonstrated that mutating residues in these entry loops in vivo deregulated cellular protein turnover ( Esaki et al . , 2017 ) . In the mammalian homolog , the same residues are required for proper function of the ERAD pathway ( DeLaBarre et al . , 2006 ) . In this study , we characterize the Cdc48-like protein of actinobacteria ( Cpa ) by a combination of bioinformatical analysis , genetic modification and biochemical characterization to explore ring formation of Cpa , to probe its ability to interact with the 20S proteasome as well as to map proteomic changes resulting from its deletion in Mycobacterium smegmatis ( Msm ) . We show that Cpa forms hexameric rings , their assembly strongly dependent on pH and ionic strength . Once formed , the rings are capable of interacting with the proteasomal core particle , identifying Cpa as a novel proteasomal interactor in actinobacteria . Moreover , we report the comparative proteomic profile of an Msm parent and cdc48-deficient strain , providing a basis for assessing the function of this AAA+ ATPase in mycobacteria .
Actinobacteria encode a number of AAA+ proteins that were shown to be involved in protein degradation pathways , amongst them the chaperone complexes of the Clp protease system , ClpX and ClpC , as well as the energy-dependent proteasome activator Mpa in mycobacteria ( Laederach et al . , 2014 ) . As Cdc48 in eukaryotes and archaea is amongst other functions also tied to protein degradation , we set out to investigate if its homolog might have an analogous function in actinobacteria . The Mycobacterium tuberculosis ( Mtb ) Cpa protein Rv0435c is annotated in the UniProt database as Cdc48 based on sequence homology to the eukaryotic Cdc48 ( p97 or VCP ) ( The UniProt Consortium , 2017 ) . In order to gain a better understanding of the position of the actinobacterial homolog within the family of Cdc48-like proteins we performed a multiple sequence alignment of 1167 sequences including members of five previously identified eukaryotic Cdc48 families ( PEX1/6 , NSF , Spaf , Spaf-like , p97 ) as well as archaeal and actinobacterial Cdc48-like proteins using the ClustalO algorithm . The alignment was used for construction of a phylogenetic tree using the approximately-maximum-likelihood method of the FastTree software ( Price et al . , 2010 ) . As expected , all the sequences cluster into subgroups according to their previously assigned classification ( Figure 1A ) . Interestingly , mycobacterial Rv0435c , as well as its actinobacterial orthologs , form an independent cluster in the Cdc48-family tree ( Figure 1A , blue branch ) . In contrast , sporadically occurring Cdc48 homologs of bacterial species outside of the phylum Actinobacteria do not cluster with their actinobacterial counterparts , but rather with archaeal Cdc48 family members ( Figure 1A , grey branches within the green archaeal branch and lime-green group in Figure 1—figure supplement 1B ) . This , in combination with their infrequent occurrence in bacterial genomes would indicate that these bacteria acquired the cdc48 gene by means of horizontal gene transfer from Archaea while the actinobacterial homolog Cpa is a descendant from a common Cdc48 ancestor . To assess the possibility of Rv0435c acting in the context of the proteasome , we analyzed the occurrence of the cpa gene with respect to the proteasomal subunit genes prcA and prcB across the actinobacterial phylum ( Figure 1B ) . Interestingly , none of the actinobacteria lacking proteasomal subunit genes encode a Cdc48 ortholog . In other words , where Cpa is present in actinobacteria , it co-occurs with the 20S proteasome . This observation supports the possibility that Cpa could be another proteasomal interactor , as it was shown for the eukaryotic and archaeal homologs ( Barthelme and Sauer , 2013 ) . Analysis of the sequence alignment of 76 representative actinobacterial orthologs together with the recently solved crystal structure of the mycobacterial Cpa monomer show that Cpa features the canonical domain structure of tandem AAA-module proteins ( Unciuleac et al . , 2016 ) . A small N-terminal domain ( referred to as N-domain ) , usually responsible for binding adaptors and/or substrates , is followed by two consecutive canonical AAA modules ( D1 and D2 , respectively ) . To visualize the location of the domains and residues within a predicted hexameric ring , we built a structural homology model for the rhodococcal Cpa sequence using human p97 as a template , for which a structure of the assembled complex was available ( Hänzelmann and Schindelin , 2016 ) ( Figure 1—figure supplement 1D ) . The AAA modules each contain Walker A and B motifs required for ATP binding and hydrolysis; however , only the D2 module carries the so-called sensor asparagine ( assisting the Walker B motif in coordinating water molecules ) and arginine finger ( stimulating the ATPase activity ) ( Figure 1C and Figure 1—figure supplement 1D ) ( Unciuleac et al . , 2016 ) . The alignment around the Walker A and B motifs shows strict conservation of the motifs in the D2 module , while the D1 module exhibits mild degeneration of the motifs ( sequence logos in Figure 1C ) . To assess the sequence conservation more quantitatively and in context of the aforementioned Cdc48-like families , we employed the method described by Wang and Kennedy ( 2014 ) using principal component analysis to estimate the variance between the sequences being analyzed ( Figure 1—figure supplement 1A–C ) . This analysis confirms that actinobacterial Cdc48-like proteins form a separate , tight cluster rather than clustering together with any specific other group ( Figure 1—figure supplement 1B ) . Analysis of the second component loadings , which showcase the separation between the different families , revealed that the N-domains are the main distinguishing feature between the members of the different families , while the ATPase modules are highly conserved across the families ( Figure 1—figure supplement 1D ) . Proteasomal activator complexes usually feature a ring-shaped architecture and associate with the 20S cylinder faces by a ring-stacking interaction aligning the activator ring pore with the 20S entry pore to form a conduit for substrate translocation . A crystal structure was recently determined for the Mycobacterium smegmatis ( Msm ) Cpa ( Msm0858 ) in its monomeric form and the study concludes that Msm0858 does not form the canonical hexameric ring structures but rather acts as a monomer in solution ( Unciuleac et al . , 2016 ) . As interaction with the 20S cylinder would require a ring-shaped complex , we hypothesized that actinobacterial Cdc48 homologs also assemble into hexamers like their eukaryotic and archaeal homologs . To perform size-exclusion experiments to test this hypothesis , we used recombinantly produced Cpa from Rhodococcus erythropolis due to its higher solubility compared to mycobacterial Cpa . In the absence of nucleotide , RerCpa eluted in one main peak at an elution volume equivalent to a monomer/dimer assembly state ( Figure 2A , black profile ) . Incubation of RerCpa with the ATP-hydrolysis transition state analog ADP-AlFx prior to size exclusion resulted in a shift of the monomer/dimer elution peak to a position equivalent to just above 500 kDa ( as indicated by the molecular weight standards ) , which is in good agreement with a hexameric assembly state ( expected molecular weight: 450 kDa ) . An additional , smaller fraction elutes even earlier , at a position equivalent to larger than 700 kDa . In order to overcome the limitation of column calibration and obtain an accurate assessment of complex size , we also performed size exclusion chromatography with multi-angle light scattering detection ( SEC-MALS ) . The light scattering signal revealed the mass of the assembled species in the main peak to correspond to 430 ± 39 kDa . This is in good agreement with the molecular weight of six monomers ( 75 kDa ) in a hexameric assembly ( Figure 2B ) . A fraction in Figure 2A eluted at an earlier position which corresponds to the molecular weight of a dodecameric assembly ( 943 kDa ) and is likely due to the high protein concentration which was used for this experiment . To assess the shape of the assembled complex , we recorded electron micrographs of negatively stained Cpa particles . A double Walker B variant of the rhodococcal enzyme ( D312N , E566Q ) capable of nucleotide binding but not hydrolysis was used in this experiment to allow observation of the complex in the presence of ATP rather than ADP-AlFx . The evenly distributed particles display as spherical shapes , most of which appear to be top views ( Figure 2C and Figure 2—figure supplement 1 ) . The 2D classification of the imaged particles revealed that most of the classes represent hexameric assemblies ( Figure 2—figure supplement 1 ) . Some of the particles , however , look slightly distorted , suggesting either high flexibility ( presumably of the N-domains ) or ring instability leading to ring opening in some of the 2D classes . Nevertheless , the diameter of the particles is rather uniform with roughly 150 Å and the appearance and dimensions of the particles agree well with hexameric ring assemblies that were observed previously for archaeal or eukaryotic Cdc48 . To test whether the mycobacterial Cpa can also form hexameric rings like the rhodococcal protein , we repeated the size-exclusion experiment described in the previous paragraph using MsmCpa incubated in the presence of ADP-AlFx ( Figure 2—figure supplement 1 ) . The elution profile clearly shows that mycobacterial Cpa also forms hexamers in the presence of nucleotide . Additionally , we tested the ATPase activity of the mycobacterial enzyme at two different temperatures ( 28°C – temperature at which activity for the rhodococcal enzyme was determined , and 37°C – optimum growth temperature of Msm ) ( Figure 2—figure supplement 2B ) . Although mycobacterial Cpa exhibits low ATPase activity at 28°C , we observed a five-fold increase in activity when the temperature was increased to 37°C ( change from 7 . 9 ± 2 . 9 to 42 . 5 ± 3 . 2 min−1 hexamer−1 ) . Since the rhodococcal enzyme showed an overall better behaviour in vitro , subsequent in vitro experiments were performed using this homologue . As demonstrated by the size-exclusion analysis , Cpa , like many of the AAA+ proteins that feature tandem AAA-modules , assembles into the hexamer only upon nucleotide binding . ATPase activity for these complexes is usually dependent on the formation of the assembled , hexameric ring ( Kress et al . , 2007 ) . We therefore explored the dependence of ATPase activity on buffer conditions by varying pH and salt concentration , both factors known to affect assembly of AAA proteins . For example , ClpB/Hsp104 requires low-salt concentrations for assembly of the ring in vitro ( Wendler et al . , 2012 ) . The rhodococcal Cpa exhibited a strong dependence of its ATPase activity on both , pH and KCl concentration , where increasing pH and decreasing salt concentration led to an increase in the measured enzyme activity by up to sevenfold ( Figure 3A ) . We hypothesized that this strong dependence of the ATPase activity on salt was the result of enhanced ring formation under those conditions rather than a direct effect on the ATPase activity . To test this , we recorded assembly time courses at different salt concentrations ( 0–400 mM KCl ) while keeping the pH constant at 7 . 8 . Assembly was triggered by addition of Cpa into the assembly buffer at the given salt concentration and containing ADP-AlFx as nucleotide to avoid complications from ATP turnover . Aliquots were taken from the assembly reaction and subjected to size exclusion to determine the elution peak area of assembled Cpa . Interestingly , increasing the salt concentration from 0 to 400 mM led to a strong deceleration of ring formation ( Figure 3B ) . When no salt was present in the sample the oligomerization was complete within a few minutes , while at the highest KCl concentration only 40% of Cpa was assembled into rings after 5 hr of incubation . This result indicates that during the time frame in which the ATPase assay is performed the assembly is still slowly ongoing with its velocity dependent on the initial KCl concentration . Therefore , the measured activity reports mainly on the protein assembly state and indicates that ring formation is indeed strongly dependent on ionic strength . Having established that Cpa is capable of hexameric ring formation , we were interested in testing whether the protein could interact with the 20S proteasome . To test for interaction between Cpa and the proteasome core particle , we employed the bacterial adenylate cyclase two-hybrid assay previously used for identifying interaction between the actinobacterial proteasome and another activator , Bpa ( Battesti and Bouveret , 2012; Delley et al . , 2014 ) . The assay is based on the reconstitution of adenylate cyclase activity from the individually inactive T25 and T18 adenylate cyclase subdomains , which are tethered to the two potential interaction partners . In this case , Cpa from Mtb ( Rv0435c ) was fused to the T25 subdomain , while the T18 subdomain was fused to the α subunit of the 20S proteasome . Plasmids encoding the two respective fusion constructs were co-transformed into an E . coli strain deficient in adenylate cyclase activity . In case the two subdomain-carrying proteins , MtbCpa and Mtb20S proteasome , physically interact , the active adenylate cyclase complex can be reconstituted from the two subdomains , leading to production of cyclic AMP ( cAMP ) that in turn activates the lac operon ( Figure 4A ) . To test for cAMP production , we therefore performed a β-galactosidase assay on cells permeabilized with SDS/chloroform . In addition to the wild type MtbCpa we also tested a variant where five C-terminal amino acids were removed . Although Cpa orthologs do not carry the classical HbYX proteasome interaction motif at their C-terminus , the C-terminus itself might nevertheless contribute to this interaction . Figure 4B shows the results of the β-galactosidase assay , including the negative ( T25 against T18-CP ) control . The activity of around 60 pmol 4-methylumbelliferone produced per min per OD unit indicates that the full-length mycobacterial Cpa is capable of directly interacting with the proteasomal core particle in the context of a bacterial cell . Additionally , we observe that removal of the C-terminal amino acids does not change the interaction strength between Cpa and the 20S core particle , indicating that those residues do not significantly mediate this interaction . A milder contribution cannot be excluded on the basis of this assay due to the inherent positive feedback circuit that precludes precise quantification . To probe for the interaction between Cpa and the proteasome in vitro , negatively stained EM micrographs of RerCpa rings pre-formed in the presence of ADP-AlFx and purified by size exclusion chromatography were recorded in presence of rhodococcal closed-gate proteasomes . We observed a small fraction of coaxial Cpa-proteasome complexes in side view ( Figure 4C and Figure 4—figure supplement 1 , white arrows ) next to uncapped 20S cylinders ( Figure 4C , yellow arrows ) and free Cpa rings ( Figure 4C , blue arrows ) . Additionally , we tested the ability of RerCpa to compete with MtbMpa for complex formation with the proteasome . In this assay , degradation of a pupylated substrate by the MtbMpa-20S complex is observed in the absence or presence of an excess of RerCpa over MtbMpa . Competitive inhibition of the degradation reaction indicates complex formation between Cpa and the proteasome . In the absence of RerCpa , pupylated ketopanthoate hydroxymethyltransferase ( MtbPanB-Pup ) , a well-characterized degradation substrate of the MtbMpa-20S complex , is completely degraded within 20 min under the assay conditions , while in the presence of RerCpa no significant degradation takes place within the same time frame ( Figure 5A ) . Using as a spectroscopically tracable model substrate a linear fusion of Pup to GFP ( Pup-GFP ) , we also conducted the competition experiment with an all rhodococcal system ( Figure 5B ) . Like in the heterologous setup , RerCpa could inhibit degradation of Pup-GFP by the RerARC-20S complex in a concentration-dependent manner , indicating interaction between Cpa and the 20S proteasome . Finally , to quantify the interaction strength , we employed microscale thermophoresis ( MST ) to measure the dissociation constant between RerCpa and Mtb20S CP . For this purpose , we fused the small orange fluorescent protein mKO2 ( monomeric Kusabira-Orange 2 ) to the N-terminus of RerCpa and performed titration experiments with closed- and open-gate mycobacterial proteasomes in presence and absence of ATP ( Figure 6 ) . In presence of ATP , we measured a dissociation constant of 0 . 36 ± 0 . 02 μM for the closed-gate proteasome and 0 . 56 ± 0 . 05 μM for the open-gate proteasome , indicating that the interaction between Cpa and closed-gate core particle is marginally stronger than with the open-gate variant . Importantly , we could not observe a signal change in absence of ATP , demonstrating that the measured interaction is specific to the Cpa hexamer that is formed only when nucleotide is present . In order to gain insight into the potential cellular function of mycobacterial Cpa , we generated an Msm cpa knockout strain by means of homologous recombination ( Msm Δcpa ) and confirmed the absence of the protein using antibodies raised against MsmCpa produced recombinantly in E . coli . ( Figure 7—figure supplement 1 ) . The knockout strain exhibited no growth defect compared to the parent strain under standard culture conditions ( Middlebrook 7H9 medium with glycerol and Tween 80 at 37°C ) ( Figure 7A ) . However , transfer of the parent and knockout strain cultures grown in minimal medium supplemented with glycerol ( adapted from ( Elharar et al . , 2014 ) ) to a medium devoid of glycerol as the main carbon source reveals a growth defect in the knockout strain ( Figure 7B ) . The wild-type Msm cells undergo two to three divisions during the first 72 hr after transfer , while the Msm ∆cpa cells stop dividing already after 24 hr . Once cell division has ceased , both strains can persist for at least 2 weeks and can be recovered when plated on 7H10 solid medium ( data not shown ) . This result might point to a role of Cpa in adaptation of Msm to nutrient starvation stress . Since bacterial operonic arrangement can give indications about the functional context in which a gene is required , we analyzed the genomic cpa locus across actinobacteria . As is evident from the phylogenetic tree shown in Figure 1A , the actinobacterial cdc48 homolog cluster consists of three distinct branches based on their sequence alignment . Careful investigation of the genomic context in the three branches shows that in branches 1 and 2 ( Figure 1A and B ) cpa is encoded in a putative operon together with two other genes: psd ( phosphatidylserine decarboxylase proenzyme ) and pssA ( phosphatidylserine synthase ) , two consecutively acting enzymes responsible for the synthesis of phosphatidylethanolamine from CDP-diacylglycerol and serine . In the genomes of those two branches , the proteasomal genes are usually located within a relatively fixed distance from the cpa locus . In the actinobacteria belonging to the third branch , the psd/pssA genes exist , but are not located in the same operon as or in close proximity of the cpa gene . To probe if psd , pssA and cpa are co-transcribed , we designed six pairs of primers spanning different regions of the polycistronic RNA expected in case of co-transcription ( Figure 8A ) . We used total RNA extracted from wild-type Msm to generate a single cDNA using a primer specific to the end of the cpa gene . In case the genes are transcribed from a single promoter placed upstream of the psd gene , we should generate all amplification products from the six primer pairs ( labeled A-F in Figure 8A ) . Consistent with a single mRNA generated from the three genes , all six amplification products were present: two probes spanning joints between psd/pssA and pssA/cpa ( probes A and B ) , a cpa-specific probe ( probe C ) as well as three longer probes , spanning two or three genes at the same time ( probes D , E and F ) . Evidently , the psd , pssA and cpa genes are expressed together from a polycistronic operon . Considering that phosphatidylethanolamine has been implicated in membrane curvature determination as well as cell division in some organisms ( Mileykovskaya et al . , 1998 ) and that the genes responsible for its synthesis are co-expressed with the cpa gene , we were interested to see whether removal of Cpa from the cell would influence the shape of starved cells . To test this , we grew Msm cultures of wt and cpa-knockout strains to the end of exponential phase ( using medium with and without glycerol , as described above ) and subsequently analyzed cell morphology using light microscopy ( Figure 8—figure supplement 1 ) . Since mycobacterial cells typically tend to form bigger cell aggregates in standard culture and individual cells are difficult to discern , we only inspected the cells visually without measuring individual cell lengths – no differences were apparent . Under carbon starvation , the cells did not form as many clumps and they were generally shorter compared to those grown in the presence of glycerol . To test whether the cells differ in length between the two strains we estimated the average cell length using light microscopy images . The knockout cells were longer by approximately 13% compared to the wild-type cells . The difference , albeit statistically significant , seems too small to conclude that Cpa is involved in the regulation of cell division , although it cannot be excluded that Cpa could influence cell shape in a different way . In order to assess the effects of Cpa activity on the Msm proteome , we carried out differential proteomic analysis of the wild-type and ∆cpa Msm strains . To this end both strains were grown under standard culture conditions as well as under carbon starvation to the early stationary phase , when the cells were harvested and the soluble proteomes ( enriched with membrane proteins; see Materials and methods section ) of both strains determined using label-free quantification mass spectrometry ( LFQ-MS ) . The relative protein changes together with their statistical significance are depicted in ‘volcano’ plots in Figure 7A . We observed 45 proteins during normal growth and 251 proteins during carbon starvation that were at least 1 . 5 times more abundant in the wild type strain compared to Msm ∆cpa . Likewise , 49 proteins were found to accumulate in the knockout cells during normal growth and 254 proteins during carbon starvation compared to Msm wt . Consequently , while during standard growth the differences are rather mild , exposing the cells to starvation stress led to significantly higher accumulation or depletion of a larger number of proteins in the cpa knockout strain . This observation strengthens the notion of Cpa playing a role during adaptation to nutrient-limited conditions . In order to gain a general understanding of the process in which Cpa may be involved directly , we classified the significantly changed proteins into functional classes using the Clusters of Orthologous Groups ( COG ) classification system ( Figure 7B ) . We observed that in both cases ( cells starved and non-starved ) , the class with most of the changed proteins was ‘transcription’ . Due to the low number of changes during standard growth , many COG classes under this condition were represented by only one or two proteins . This changed , however , when the cells were starved for carbon . We found that many changes occurred in classes like ‘transcription’ ( most represented , similarly to standard growth ) , ‘energy production and conversion’ , ‘amino acid transport and metabolism’ and ‘lipid transport and metabolism’ . From the perspective of Cpa acting in a protein degradation context , those proteins accumulating in the knockout strain could under the most simplistic interpretation be degradation substrate candidates . However , the fact that equal number of proteins accumulated as were depleted in the knockout strain suggests that this view is too simplistic . For example , degradation of a transcription factor that is present in the cell only with low abundance , could lead to significant proteomic changes . Indeed , as mentioned above , the most represented class in both our datasets is ‘transcription’ , which could explain the observed changes in both directions . To identify functional connections within the accumulating proteome , we performed a pathway enrichment analysis using STRING and DAVID tools ( Huang et al . , 2009; Szklarczyk et al . , 2017 ) . The result of this analysis showed an unusually high number of structural components of the ribosome ( Figure 9—figure supplements 1 and 2 ) accumulating in the cpa-knockout cells . In fact , COG class J , encompassing proteins connected to translation and ribosome , was one of the very few classes that contained mostly proteins accumulating in the knockout cells and few with decreased levels ( Figure 7B ) . Finally , we were interested in testing whether some of the identified proteins could be Cpa binders in vivo under carbon starvation stress . For this purpose , we performed a co-immunoprecipitation ( co-IP ) experiment using an anti-MsmCpa antibody and wild-type M . smegmatis cells ( with Msm ∆cpa strain as a control for nonspecific binding to the beads/antibody ) . Using shotgun LC-MS/MS , we could identify a few dozen proteins in at least two out of three co-IP replicates that were specific to the presence of Cpa ( i . e . did not bind to the antibody in the cpa-knockout lysate ) . We also submitted the results of the pull-down experiment to pathway enrichment analysis by STRING and DAVID tools and observed that again ribosomal proteins ( indicated by bold rectangles in Figure 9—figure supplement 2 ) comprise a significantly enriched fraction of the immunoprecipitated protein mixture ( see Figure 9—source data 2 for the full list of identified binders ) .
Despite the vast amount of structural and functional information available for eukaryotic members of the Cdc48 family ( p97 and NSF in particular ) , comparatively little is known about their prokaryotic homologs ( Bodnar and Rapoport , 2017; Kienle et al . , 2016 ) . Only recently has it been established that archaeal equivalents of eukaryotic p97 are capable of directly interacting with the 20S proteasomal core and that they can support protein degradation in vitro ( Barthelme and Sauer , 2012 ) . The only previous study of an actinobacterial Cdc48-like ATPase is for the ortholog from Msm , and it reports that the protein exists as a monomer in solution as well as in the crystal structure ( Unciuleac et al . , 2016 ) . Even though the authors suggest that the protein would need to undergo an oligomerization step in order to gain ATPase activity ( due to ATP hydrolysis usually occuring at the interface between adjacent protomers ) , no evidence for assembly was presented . In our study , we investigate different structural and functional aspects of Cdc48-like protein from actinobacteria ( Cpa ) using both rhodococcal and mycobacterial Cpa . In particular , the solubility and stability of the rhodococcal ATPase across a range of buffer conditions allowed us to study its oligomerization behavior by time-resolved size exclusion analysis and size exlusion chromatorgaphy coupled to a light scattering detection system . Our results clearly demonstrate that Cpa forms hexameric rings in the presence of nucleotide and that ATPase activity correlates with this assembly state . The most complete assembly into hexamers was observed in the presence of ADP-AlFx , an analog of ATP that mimics the transition state of ATP hydrolysis , suggesting particular stabilization of the ring during ATP turnover . At least in vitro , the assembly into hexamers occurred most readily at low ionic strength , indicating that electrostatic interactions are crucial for rapid oligomerization . Another tandem AAA+ protein , the chaperone ClpB , can even assemble in absence of nucleotide in vitro if the ionic strength is kept low enough ( Schirmer et al . , 2001; Schlee et al . , 2001 ) . It is possible that like its eukaryotic counterparts , Cpa in vivo associates with a range of adaptor proteins that could further stabilize the ring . Such adaptor-dependent assembly behavior was for example shown for ClpC from Bacillus subtilis ( Kirstein et al . , 2006; Schlothauer et al . , 2003 ) . The assembly of Cpa into a hexameric , ATPase-active ring complex , which is clearly demonstrated by our in vitro analysis , opens up the possibility that the Cpa ring can stack to the 20S proteasome α-rings , thereby forming a complex with centrally aligned ring pores . Importantly , the occurrence pattern of the cpa gene in actinobacterial genomes lends support to this argument , as cpa is found only in those members of actinobacteria carrying the two proteasomal subunit genes ( Figure 1B ) . Genomic co-occurrence can indicate a functional connection between the co-occurring gene products , as is for example the case for components involved in the two known proteasomal degradation pathways in mycobacteria ( Barandun et al . , 2012; Delley et al . , 2014 ) . Furthermore , using a bacterial two-hybrid assay and electron microscopy we could demonstrate physical association between MtbCpa and the Mtb proteasome . This interaction occurs at the α-ring face of the proteasome , as can be seen in the negatively stained electron micrographs , and which is also supported by RerCpa competing with Mpa for binding to the proteasome in vitro , thereby inhibiting degradation of pupylated substrates . However , the molecular elements of this interaction differ from those employed by the other two actinobacterial proteasomal interactors , Mpa/Arc and Bpa , that both carry the canonical penultimate tyrosine motif , while Cpa does not . Interestingly , it has been shown for the archaeal and eukaryotic Cdc48 homologs , that an additional proteasome-interaction element exists that considerably contributes to the binding affinity ( Barthelme and Sauer , 2013 ) . Our attempt to test whether the removal of this molecular feature , the so-called D2 pore-2 loop , from Cpa would abolish its interaction with the 20S CP was unsuccessful due to the fact that removal or replacement of this element resulted in assembly defects of the hexameric ring . Nevertheless , it remains unclear , why the actinobacterial Cdc48 homolog does not carry both interaction motifs like its archaeal/eukaryotic cousins . One possible reason might be that formation of the Cpa-proteasome should be transient in nature . Alternatively , the different molecular arrangement around the actinobacterial α-ring of the proteasome might not be optimal for interaction with the Cpa C-terminus and therefore its contribution to the interaction minor or non-existent , so that the penultimate tyrosine was eventually lost . In fact , it would not be the only known case in the family of proteasomal regulators lacking the penultimate aromatic residue: the eukaryotic 11S ( PA28 ) activator also does not carry the HbYX motif: instead , the C-terminal residue of each subunit provides proteasome binding energy and the so-called ‘activation loops’ trigger gate opening allowing for substrate entry ( Stadtmueller and Hill , 2011; Whitby et al . , 2000 ) . However , our two-hybrid analysis showed that the last five C-terminal residues do not play a major role in proteasome binding , suggesting that instead other elements of Cpa must mediate the interaction . Our in vivo analysis of a cpa-disrupted Msm strain suggests that Cpa plays a role during the adaptation to conditions where nutrients ( carbon in particular ) become limiting . In an attempt to better understand which aspects of cellular function Cpa might influence , we compared the proteomes of the parent Msm strain with the cpa-knockout strain grown in presence and absence of glycerol . Under both conditions , we identified proteins accumulating in the knockout strain as well as proteins with decreasing levels , indicating that the observed changes include not only potential degradation substrates but also proteins affected either through secondary effects like for example degradation of a regulatory protein or through Cpa functions independent of the proteasome . The proteins affected by the cpa-gene disruption fall into multiple functional categories , those involved in transcription , energy production/conversion , amino acid metabolism/transport and lipid metabolism represented the most . Given the hampered growth of the knockout and the increased protein level changes under starvation conditions , these results suggest that Cpa could be involved in adaptation of the cell to a lack of nutrients . The proteins accumulating in the cpa-knockout strain potentially could comprise substrates that are usually removed with Cpa-involvement , either by direct degradation or by secondary effects . Our functional analysis of the proteome changes upon starvation ( Figure 9 and Figure 9—figure supplement 1 ) combined with identification of Cpa binders in the wild-type cells under the same conditions ( Figure 9—figure supplement 2 ) could indicate that removal of Cpa from the cell more directly influences the level of ribosomal proteins compared to the other detected changes . In other bacteria it has been observed that nutrient stress ( carbon starvation among others ) results in ribosome disassembly and rRNA degradation ( Zundel et al . , 2009 ) . Additionally , another recent study reported that an excess of ribosomal proteins is removed from yeast cells by the ubiquitin-proteasome system ( Sung et al . , 2016 ) . It is tempting to hypothesise that actinobacterial Cpa is involved in a similar process; however , more effort will be required to investigate whether the role played by Cpa in disassembly and/or removal of ribosomal proteins under nutrient limitation requires cooperation with the proteasome core or is mainly mediated by its AAA pore-threading activity . At the same time , it is difficult to speculate about the substrate clientele based on the homology of Cpa to eukaryotic Cdc48 . For example , eukaryotic Cdc48 is implicated in processing of ubiquitinated proteins ( Richly et al . , 2005; Rumpf and Jentsch , 2006 ) , a post-translational modification that does not occur in actinobacteria . Actinobacteria possess , however , the functionally analogous covalent modifier Pup that renders proteins as substrates for degradation by the Mpa-proteasome complex ( Striebel et al . , 2010; Striebel et al . , 2014 ) . The well-characterized pupylation substrate PanB in fact accumulates in the cpa-knockout strain . However , in vitro degradation tests with recombinantly produced and in vitro pupylated PanB showed no degradation in presence of Cpa and the proteasome ( data not shown ) . Similarly , although the archaeal Cdc48-CP assembly is capable of degrading an ssrA-tagged model substrate in vitro , archaea do not harbor the tmRNA , a crucial element of trans-translation encoding the ssrA-tag ( Hayes and Keiler , 2010; Karzai et al . , 2000 ) . Actinobacteria feature a tmRNA and carry out trans-translation . However , no Cpa-mediated degradation of ssrA-tagged model substrates could be detected in vitro ( data not shown ) . As the Clp protease system is responsible for ssrA-tagged substrate degradation in mycobacteria and related bacteria , this finding was not unexpected ( Laederach et al . , 2014; Raju et al . , 2012 ) . The search for substrates is likely further hampered by the fact that Cdc48 family members almost always act in the context of adaptor proteins . In the absence of these adaptors , the activity toward the adaptor-mediated set of substrates is simply not detectable . The co-expression of Cpa from a polycistronic mRNA together with Psd and PssA , two consecutively acting enzymes responsible for the synthesis of phosphatidylethanolamine , might give an indication about the functional context of Cpa activity . Several studies indicate a functional link between phosphatidylethanolamine and cytokinesis , both in eukaryotes and bacteria ( Luo et al . , 2009; Mileykovskaya et al . , 1998 ) . This would suggest that Cpa might play a role in context of cell division . Removal of Cpa from the cells did not , however , show a significant effect on cell division as gauged by their cell size and morphology . Further investigations are clearly needed in order to gain a deeper understanding of the physiological role of Cpa in the biology of actinobacteria , including the search for adaptor proteins and analysis of their substrate recruitment clientele .
All sequences were taken from the NCBI database ( see Supplementary file 1 for a full list of accession numbers ) and aligned using the ClustalO multiple sequence alignment algorithm ( Sievers et al . , 2011 ) . The resulting alignment was used to construct a phylogenetic tree with FastTree 2 . 1 . 1 ( Price et al . , 2010 ) applying default parameters . The results were visualized with FigTree 1 . 4 . 3 . Gene co-occurrence was investigated using the STRING tool ( Szklarczyk et al . , 2017 ) and verified by searching selected actinobacterial genomes for the presence of bpa , pup-prcAB , arc , psd and pssA genes with the help of BLAST ( Altschul et al . , 1990 ) . Walker A/B sequence logos were generated from 76 Cpa ( Cdc48-like protein of actinobacteria ) sequences using WebLogo 2 . 8 . 2 ( Crooks et al . , 2004 ) . The multiple sequence alignment ( described above ) was used as an input for the principal component analysis as described by ( Wang and Kennedy , 2014 ) . Briefly , all the amino acid positions were ranked according to their occurrence at a given position and amino acids with the same occurrence were ranked according to their alphabetical order . The original residues were replaced with calculated ranks and the resulting table was used to perform PCA . All the calculations were carried out using a Python script ( available at https://github . com/misialq/pca-protein-analysis; copy archived at https://github . com/elifesciences-publications/pca-protein-analysis ) . A structural model of the rhodococcal Cpa was obtained using the SWISS-MODEL homology-modeling server where structure of the human p97 ( PDB-ID: 5C1A ) served as a template ( Biasini et al . , 2014 ) . Sequence conservation was plotted according to the PCA loadings of the second principal component: positions whose PCA loadings exceeded 30% of the maximal PC2 loading were colored in red ( variable ) and those below 10% were colored in blue ( conserved ) . Structure drawing was done in PyMOL 1 . 5 ( Schrödinger , LLC ) . Mycobacterium smegmatis mc ( 2 ) 155 ( Msm ) was grown in liquid Middlebrook 7H9 medium ( Difco ) supplemented with 0 . 025% ( w/v ) Tween 80 and 0 . 2% ( w/v ) glycerol , unless otherwise stated . The cells were grown at 37°C with shaking at 180 rpm . For carbon starvation , the cells were first cultured small-scale in minimal medium ( adapted from ( Elharar et al . , 2014 ) containing: 40 mM K2HPO4 , 22 mM KH2PO4 , 15 mM ( NH4 ) 2SO4 , 1 . 7 mM sodium citrate , 0 . 4 mM MgSO4 , 0 . 05% ( w/v ) Tween 80 and 0 . 4% ( w/v ) glycerol . Washed cells from the small-scale culture were transferred into a large-scale culture in the same medium but lacking glycerol . The genes encoding Rv0435c ( MtbCpa ) , MSMEG_0858 ( MsmCpa ) and RER_15150 ( RerCpa ) were amplified from genomic DNA ( M . tuberculosis – Mtb , M . smegmatis – Msm and R . erythropolis – Rer , respectively ) by PCR using Q5 DNA polymerase ( New England Biolabs; NEB ) and cloned into a pET28a expression vector using Gibson assembly ( NEBuilder HiFi DNA Assembly Kit; NEB ) . The double Walker B ( D312N , E566Q ) mutant of RerCpa was cloned into the same backbone by Gibson assembly using gene pieces containing mutations introduced into PCR primers and amplified from rhodococcal genomic DNA . A His6-tag followed by a Tobacco Etch Virus ( TEV ) protease cleavage site preceded the N-terminus of either protein . Constructs were transformed into E . coli Tuner ( DE3 ) cells for heterologous expression which was carried out in ZYP-5052 autoinduction medium as described by Studier ( 2005 ) with minor modifications: the expression culture was inoculated to an OD600 = 0 . 025 using an overnight preculture and incubated at 30°C/180 rpm for 6 hr . After that time , the temperature was reduced to 18°C and the cells were harvested after an additional 14 hr . Cells were lysed in buffer A ( 50 mM HEPES-NaOH pH 7 . 8 at 8°C , 300 mM NaCl , 40 mM imidazole ) using a high-pressure homogeniser ( Microfluidics ) and the cleared lysate was loaded onto Ni2+-IMAC Sepharose 6 FF resin . After washing the resin with 15 column volumes of buffer A , the protein was eluted using buffer B containing 250 mM imidazole . Protein-containing fractions were pooled and dialysed for 30 min at 4°C against buffer C ( 50 mM HEPES-NaOH pH 7 . 8 at 8°C , 150 mM NaCl , 5 mM β-mercaptoethanol ) . The N-terminal hexahistidine tag was cleaved by addition of TEV protease and further dialysis for 14 hr . TEV protease was removed via Ni2+ affinity chromatography , and Cpa was further purified by size exclusion chromatography using a Superdex 200 column in buffer C without β-mercaptoethanol . PrcAB , Arc/Mpa , PanB-Pup and Pup-GFP were purified as described previously ( Striebel et al . , 2010; Striebel et al . , 2009 ) . To test for interaction between Cpa and the 20S proteasome using a bacterial adenylate cyclase two-hybrid assay ( BACTH ) , the Mtb cpa full-length gene , as well as its variant with the last five C-terminal residues removed , was fused to the C-terminus of the T25 adenylate cyclase subdomain . The cpa gene ( Rv0435c ) was amplified from Mtb genomic DNA using Q5 DNA polymerase ( NEB ) and cloned into the pKT25 vector using Gibson assembly ( NEB ) . The prcAB genes were cloned into the pUT18 vector where prcA was fused to the N-terminus of the T18 subdomain of adenylate cyclase . The lac promoter region was duplicated in this vector downstream of the prcA-T18 fusion followed by the prcB∆pro gene ( propeptide residues 1–57 were removed ) . Empty pKT25 and pUT18 backbones , as well as T18-zip and T25-zip leucine zipper gene fusions ( used as positive control ) , were obtained from Euromedex . Analytical SEC was performed on a Superose 6 Increase 10/300 GL column ( GE Healthcare ) connected to an Agilent 1260 Infinity HPLC system . All runs were performed at 23°C at a flowrate of 1 ml min−1 in 50 mM HEPES-KOH pH7 . 8 , 25 mM KCl , 20 mM MgCl2 . RerCpa was injected at a concentration of 30 µM ( protomer ) in a volume of 15 µl and was detected by following absorption at 280 nm . Molecular weight standards were run under the same conditions . The sample with ADP-AlFx contained additionally 2 mM ADP , 2 mM Al ( NO3 ) 3 and 12 mM NaF . For the assembly time trace , each sample was placed in the HPLC autosampler set to 18°C immediately after mixing and the injections onto the column were carried out at the indicated time points . Fraction of assembled protein was calculated by integrating all peaks and dividing the sum of the peak areas of hexamer and dodecamer by the sum of all peaks ( hexamer , dodecamer and monomer ) . To determine the molecular weight of the rhodococcal Cpa , 10 µl of 130 µM ( protomer ) protein solution were loaded onto a Superdex 200 Increase 10/300 GL column ( GE Healthcare ) . The sample was analyzed by refractive index using an Optilab T-rEX differential refractometer ( Wyatt Technology ) and by MALS using a miniDAWN TREOS light scattering detector ( Wyatt Technology ) . RerCpa-D312N , E566Q sample ( 300 nM ) was incubated for 10 min at room temperature in the presence of ATP ( 2 mM ) in 50 mM HEPES-KOH pH7 . 8 , 25 mM KCl , 20 mM MgCl2 . After incubation , sample was diluted to final concentration of 75 nM and 8 μL were applied to Quantifoil grids R2/2 freshly coated with a thin layer of carbon and incubated for 2 min . Excess liquid was blotted away and the grids were stained with 2% uranyl acetate for 2 min . Images were collected on FEI F20 electron microscope operated at 200 keV equipped with Faclon II direct electron detector ( FEI Company ) at x 82 , 350 magnification with a dose of 30 electron per Å2 , defocus value of −2 μm , and a pixel size of 1 . 7 Å per pixel . Particles were selected semi-automatically with BOXER implemented in the EMAN package ( Ludtke et al . , 1999 ) . Particles ( 17 , 375 ) were extracted using RELION ( Scheres , 2012 ) and binned three-fold resulting in a pixel size of 5 . 1 Å per pixel . Two-dimensional ( 2D ) maximum-likelihood classification was performed with the RELION package ( Scheres , 2012 ) using K = 30 and was run for 15 iterations . To image Cpa-proteasome interaction , 75 nM RerCpa rings in the presence of ADP-AlFx and 30 nM 20S closed-gate rhodococcal proteasome were incubated at room temperature in 50 µl reactions and then applied to Quantifoil grids R2/2 freshly coated with a thin layer of carbon followed by staining with 2% uranyl acetate similarly as described above . The samples were then imaged using a FEI Morgagni 268 transmission electron microscope operating at 100 keV . ATPase activity was determined using an ATPase assay coupled to pyruvate kinase and lactate dehydrogenase ( Nørby , 1988 ) . Reactions were performed in 96-well plates at 28°C ( or 37°C for MsmCpa ) , and absorption at 340 nm was monitored using a Synergy 2 plate reader ( BioTek ) . The assay was performed in 50 mM HEPES-KOH , 2 . 5 mM phosphoenolpyruvate , 1 mM NADH , 40 U ml−1 of both pyruvate kinase and lactate dehydrogenase , 20 mM MgCl2 and 2 mM DTT , where pH was varied in the range of pH 7 to pH 8 and KCl was varied from 0 to 200 mM . The total reaction volume was 100 µl , and the assay was carried out with 1 mM ATP and 1 . 5 µM RerCpa ( or 3 µM MsmCpa ) enzyme . The rate of ATP hydrolysis was calculated using an extinction coefficient of 6 . 22 mM−1 cm−1 at 340 nm for NADH . To test the interaction between MtbPrcAB and MtbCpa , a bacterial adenylate cyclase two-hybrid ( BACTH ) assay was performed as described by Battesti and Bouveret with some modifications ( Battesti and Bouveret , 2012 ) . Briefly , E . coli BTH101 ∆cya cells were co-transformed with the pKT25 and pUT18 fusion constructs described above . As a negative control , bacteria were co-transformed with empty pKT25 vector . Co-transformants were plated on LB agar plates supplemented with 100 µg ml−1 ampicillin and 50 µg ml−1 kanamycin and incubated for 72 hr at 30°C . Several clones were grown for 20 hr at 30°C in liquid LB medium containing both antibiotics and 0 . 5 mM IPTG . 100 µl of the overnight cultures were transferred into 400 µl of buffer Z ( 44 . 7 mM Na2HPO4 , 45 . 3 NaH2PO4 , 10 mM KCl , 1 mM MgSO4 , 38 . 5 mM β-mercaptoethanol; pH 7 ) , followed by addition of 25 µl 0 . 01% SDS and 50 µl chloroform and vigorous mixing . 20 µl of each cell extract were transferred to a 96-well nonbinding plate ( Corning ) , and the β-galactosidase reaction was started by addition of 30 µl 0 . 83 mM 4-methylumbelliferyl β-D-galactopyranoside ( Sigma-Aldrich ) . After 20 min of incubation at room temperature , the reactions were stopped by addition of 50 µl 1 M Na2CO3 and the fluorescence was measured in a Synergy 2 plate reader ( BioTek ) using a 360/40 excitation filter and a 460/40 emission filter . The amount of released 4-methylumbelliferone was determined using a standard curve prepared the same day ( Sigma-Aldrich ) . The β-galactosidase activity was normalized to OD600 of the overnight cultures used for the assay . The final activity is expressed as pmol of 4-methylumbelliferone min-1 OD unit-1 . The gel-based competition assay with Mpa was performed as described previously ( Delley et al . , 2014 ) with minor modifications . MtbPanB-Pup ( 4 µM protomer ) was pre-incubated for 15 min at 30°C with Mtb∆7PrcAB ( 0 . 1 µM complex ) , 5 mM ATP , 40 mM phosphocreatine , 0 . 5 U ml−1 creatine phosphokinase and 1 mM DTT in buffer M2 ( 50 mM HEPES-KOH pH 7 . 5 at 23°C , 100 mM KCl , 20 mM MgCl2 , 10% glycerol ) in presence or absence of RerCpa ( 12 µM protomer ) . The degradation reaction was started by addition of 0 . 2 µM Mpa ( hexamer ) . Aliquots were withdrawn at the indicated time points , and the reactions were stopped by addition of Laemmli sample buffer . The aliquots were analyzed by SDS-PAGE followed by staining with Coomassie Brilliant Blue . For the assay with a fluorescent GFP readout , RerPup-GFP ( 0 . 5 µM ) was pre-incubated for 15 min at 30°C with Rer∆7PrcAB ( 25 nM complex ) , 5 mM ATP , 40 mM phosphocreatine , 0 . 5 U ml−1 creatine phosphokinase and 1 mM DTT in buffer M2 in presence or absence of RerCpa ( 0 . 6 , 1 . 5 or 3 µM protomer ) . The degradation reaction was started by addition of 50 nM RerArc ( hexamer ) . The decrase in fluorescence ( Ex485/20 , Em528/20 ) was monitored using a Synergy 2 plate reader ( BioTek ) at 28°C . For MST experiments , rhodococcal Cpa was N-terminally fused to the orange fluorescent protein mKO2 ( monomeric Kusabira-Orange 2 ) , expressed and purified as described for the wild-type Cpa . 20S mycobacterial proteasomes ( closed- and open-gate variant ) were titrated to 50 nM mKO2-RerCpa using a 1 . 5x dilution series between 0 . 09 and 3 . 5 µM in 50 mM HEPES-KOH , 50 mM NaCl , 0 . 005% Tween 20 ( and 5 mM ATP , 20 mM MgCl2 in the samples containing ATP ) in a total of 16 µl . After 5 min of incubation at room temperature , the samples were transferred to Monolith NT . 115 premium coated capillaries ( NanoTemper Technologies ) and MST traces were recorded at 25°C in a Monolith NT . 115 instrument ( NanoTemper Technologies; LED power 60% , MST power 20% ) . All measurements were performed in three independent replicates . The MST data were fitted using the T-Jump section of the traces using the Hill equation: y = U + ( B – U ) ( Sh/ ( Kdh + Sh ) ) , where U is the signal from the unbound state , B is the signal from the bound state , S is the concentration of the titrated species , Kd is the dissociation constant and h is the Hill coefficient . Fitting was performed using GraphPad Prism 7 . A targeted allelic exchange of the MSMEG_0858 gene was performed by application of a two-step selection with the p2NIL/pGOAL system ( Gopinath et al . , 2015 ) . First , the competent Msm WT cells were transformed with the suicide knockout plasmid p∆cpa and plated on 7H10 medium supplemented with 20 µg ml−1 kanamycin and 50 µg ml−1 hygromycin B to select for single-crossover mutants . An ‘X-Gal underlay’ was performed to visualize the clones that were successful recombinants: 200 µl of 0 . 4% X-Gal solution were pipetted under the agar , and the plate was incubated for an additional 1–2 days . Blue clones were used to inoculate fresh Middlebrook 7H9 medium containing kanamycin and hygromycin and incubated for several days at 37°C . To select for double-crossover mutants , the cells were plated onto Middlebrook 7H10 agar medium supplemented with 2% sucrose and control medium without sucrose and incubated at 37°C for several days . The ‘X-gal underlay’ was repeated and several sucrose-resistant white clones were picked and re-plated onto fresh Middlebrook 7H10 agar plates with and without kanamycin . Several kanamycin-sensitive clones were picked , and absence of the cpa gene was confirmed by colony PCR using OneTaq polymerase ( NEB ) and primers flanking the disruption site . The resulting mutant strain ( referred to as Msm ∆cpa ) was additionally confirmed by PCR amplification of the whole psd-pssA-∆cpa locus using Q5 polymerase ( NEB ) followed by sequencing of the resulting fragment , as well as western blot detection of the protein product using a rabbit polyclonal antibody ( Biogenes , Germany ) raised against recombinant MsmCpa expressed and purified from E . coli . Total RNA isolated from Msm SMR5 was used as a template for synthesis of cDNA with a primer specific to the 3’ end of the cpa mRNA . cDNA synthesis was performed with 200 U of Maxima Reverse Transcriptase ( Thermo Fisher Scientific ) in the presence of 20 U of the RiboLock RNase Inhibitor ( Thermo Fisher Scientific ) using a step-wise temperature gradient: 50°C for 30 min , 55°C for 15 min and 65°C for 30 min . The resulting cDNA was used directly in a PCR reaction with OneTaq polymerase ( NEB ) using six primer pairs for amplification of different regions of the transcript: two pairs for joints between psd/pssA and pssA/cpa genes , one cpa-specific pair , two pairs spanning more than one gene and one pair spanning all three genes of interest . To control for genomic DNA contamination , a set of samples was prepared where the reverse transcriptase was omitted . Additionally , to check the expected correct size of all the fragments , PCR reactions were also performed directly with genomic DNA where no RNA was present . PCR products were later analyzed using a 1% agarose gel and visualized in the presence of ethidium bromide . To obtain information about the relative proteome changes in the wild-type compared to the ∆cpa Msm strain we used the label-free quantification mass spectrometry approach . For this purpose WT and knockout cells were grown in pentaplicates in minimal medium ( described above ) in a similar way as during growth curve determination , to an OD600 = 2 for cultures with glycerol and OD600 = 0 . 12–0 . 25 for cultures without glycerol . Cells were harvested by centrifugation and washed three times with 1x PBS . Cell pellets were resuspended in 1 ml of 1x PBS/2 mM EDTA ( containing 1x Halt Protease Inhibitor Cocktail; Thermo Fisher Scientific ) and transferred to 2 ml screw-cap tubes containing ~ 0 . 5 g of 0 . 15 mm zirconium oxide beads ( Next Advance ) . The cells were lysed by bead-beating two times for 30 s at maximum speed with a 1 min cooling pause in a Minilys homogeniser ( Bertin Instruments ) . Intact cells were removed by centrifuging twice at 3000 g at 4°C for 15 min . Resulting supernatants were transferred to fresh tubes and membrane fractions were recovered by centrifugation at 100000 g at 4°C for 4 hr . The supernatant ( soluble fraction ) was transferred to a new tube and the pellet was gently washed with 200 µl of 1x PBS/2 mM EDTA and spun down again at 100000 g at 4°C for 50 min . Subsequently , washed pellets were resuspended in 200 µl of 50 mM NH4HCO3 by incubation in an ultrasonic bath for 1 hr ( SONOREX 10 P , 70% power , 4°C ) . Membrane proteins were solubilised by addition of 250 µl of 2% SDS ( final concentration of 1 . 1% ) and an overnight incubation at room temperature . Insoluble material was spun down at 21000 g for 30 min at room temperature and the pellet was resuspended in 2% SDS . After an additional 1 hr incubation , the sample was centrifuged again and the supernatant was combined with the supernatant from the previous step ( solubilised membrane proteins ) . The proteins were precipitated by addition of 5 volumes of ice-cold acetone , followed by an overnight incubation at −20°C . Precipitated proteins were collected by centrifugation at 21000 g at 4°C for 30 min , the pellets were air-dried and resuspended in 100 µl of 3 M urea . Protein concentration was determined using the BCA method with bovine serum albumin ( BSA ) as a standard . The soluble fraction was mixed with the membrane fraction in a ratio 2:1 for a total concentration of 0 . 75 mg ml−1 . Equal amounts of protein were analysed by SDS-PAGE followed by immunoblot with an antibody against MsmCpa ( see above ) or EcoliRpoB as a loading control . Cell lysates were submitted for LFQ-MS analysis to the Functional Genomics Center Zurich ( FGCZ ) . For the α-Cpa co-IP experiment , 100 ml cultures of Msm wild-type and ∆cpa cells were grown in absence of glycerol as described in the previous section . Following harvesting , cell pellets were resuspended in 1 ml of 1x PBS/2 mM EDTA and lysed as described above . The lysates were then spun down at 20000 g at 4°C for 10 min and the supernatants were kept on ice . Anti-MsmCpa antibody was immobilized on 50 µl ( per sample ) DynabeadsTM Protein A ( Thermo Fisher Scientific ) and crosslinked using BS3 crosslinker , following manufacturer instructions . The beads were then mixed with 600 µl of the lysate , supplied with 5 mM ATP and 15 mM MgCl2 and incubated at 30°C for 15 min with gentle agitation . Next , the beads were washed three times with 200 µl of 1x PBS/5 mM ATP/15 mM MgCl2 and the immobilized proteins were eluted with 50 µl of 0 . 2 M glycine pH 2 . 1 and immediately neutralized by addition of 10 µl of 1 M NaOH . 25 µl of each elution were analyzed using shotgun LC-MS/MS at the Functional Genomics Center Zurich ( FGCZ ) . Co-immunoprecipitations were performed in triplicate . | Cells use proteins to carry out the biological processes necessary for life . If a protein becomes damaged or is no longer needed , cells must dispose of it , just as we might take out the trash . The cell’s main ‘garbage disposal unit’ is the proteasome , a barrel-shaped molecular machine that breaks down unwanted proteins . The proteasome binds to other molecules called regulators , which select the proteins to be dismantled . The proteasomes of mycobacteria – a group that includes the bacteria that cause tuberculosis – help them to survive hostile or rapidly changing environments . Mycobacteria contain a molecule called Cpa whose structure is like a regulator that is found in many non-bacterial cells . Ziemski et al . therefore set out to investigate whether Cpa performs a similar role in bacteria . The results of biochemical experiments performed in test tubes revealed that Cpa forms rings made up of six copies of itself . These rings can bind to proteasomes . Ziemski et al . also created genetically modified mycobacteria that could not produce Cpa and studied how they coped with starvation . These modified bacteria stopped growing sooner than their similarly starved genetically normal counterparts . The two groups of bacteria also produced different amounts of some proteins . Ziemski et al . used a technique that pulled Cpa out of the starving genetically normal cells to analyse the proteins that Cpa physically interacts with . These proteins included building blocks of the ribosome , the cellular machinery that produces new proteins . It therefore appears that Cpa helps mycobacteria to cope with starvation by reducing the amount of protein made by the cell . Cpa may also help mycobacteria to survive in other stressful conditions , such as those that the bacteria experience when they infect the human body . Developing drugs that prevent Cpa from working could therefore potentially lead to new treatments for a number of diseases caused by mycobacteria , such as tuberculosis . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology"
] | 2018 | Cdc48-like protein of actinobacteria (Cpa) is a novel proteasome interactor in mycobacteria and related organisms |
Metagenomics has revealed hundreds of species in almost all microbiota . In a few well-studied cases , microbial communities have been observed to coordinate their metabolic fluxes . In principle , microbes can divide tasks to reap the benefits of specialization , as in human economies . However , the benefits and stability of an economy of microbial specialists are far from obvious . Here , we physically model the population dynamics of microbes that compete for steadily supplied resources . Importantly , we explicitly model the metabolic fluxes yielding cellular biomass production under the constraint of a limited enzyme budget . We find that population dynamics generally leads to the coexistence of different metabolic types . We establish that these microbial consortia act as cartels , whereby population dynamics pins down resource concentrations at values for which no other strategy can invade . Finally , we propose that at steady supply , cartels of competing strategies automatically yield maximum biomass , thereby achieving a collective optimum .
In this section , we present a model for the population dynamics of cell types metabolically competing for external resources ( see Figure 1 ) . Importantly , biomass production is governed by a physical model that respects flux conservation . 10 . 7554/eLife . 22644 . 003Figure 1 . Model for metabolically competing cell types . ( A ) The rate of biomass production g ( c1 , … , cp ) is a function of the internal building-block concentrations . ( B ) Biologically relevant growth-rate functions g ( c1 , … , cp ) are increasing with respect to ci with diminishing returns . ( C ) Different cell types , i . e . metabolic strategies , are defined as specific distributions of enzymes for import αi and conversion κji , subject to a finite budget . ( D ) Cell types ( e . g . σ1 and σ2 ) compete for external building blocks that are steadily and homogeneously supplied in volume V . DOI: http://dx . doi . org/10 . 7554/eLife . 22644 . 003
In this section , we demonstrate the possibility of stable coexistence at steady supply rates by simulating competitive population dynamics subject to continual invasion by new metabolic variants . We consider that coexistence is stable when a population of distinct cell types can resist invasion by any other metabolic variants . In our simulations , cell types have distinct metabolic strategies defined by randomly chosen enzyme distributions {αi , κji} satisfying the budget constraint Equation ( 2 ) , with the universal growth-rate function Equation ( 1 ) and uniform biomass stoichiometry ( bi=1 ) . First , we show that competitive population dynamics with the continual introduction of new cell types leads to a stationary state with fixed building-block availability and with fixed populations of distinct cell types . Second , we show that these final cell types achieve optimal growth given the fixed building-block supply rates . Third , we show that final , optimal populations generally consist of consortia of distinct cell types and that a consortium of identical cell types can emerge for different building-block supplies . In this section , we mathematically elucidate the emergence of microbial consortia at steady state and characterize the benefit of the division of labor in these consortia . Our analysis exploits the demonstrated convergence of competitive population dynamics toward a stationary state , which allows us to analytically derive the metabolic strategies of optimal cell types . The benefit of division of labor among these optimal cell types follows geometrical considerations in the space of stationary states . First , we exploit arguments from transport-network theory to systematically identify the metabolic classes of optimal cell types at steady state . Second , we elucidate the structure of microbial consortia by establishing which metabolic classes can be jointly optimal within a consortium . Third , we characterize the benefit of division of labor showing that consortia can act as cartels , whereby population dynamics pins down resource concentrations at values for which no other strategy can invade .
For simplicity , we assumed linear metabolic fluxes and uniform enzymatic rates , production costs , and building-block stoichiometries . However , the emergence of optimal cartels does not rely on these assumptions . Even allowing for fluxes that are nonlinear ( e . g . Michaelis-Menten ) with respect to building-block concentrations , microbes must utilize their enzymes in the linear regime to be metabolically optimal: Because resources are depleted by competitive growth between metabolic classes , fluxes mediated by saturated enzymes do not limit growth . Cells can improve their growth rate by reallocating their enzyme budget from saturated enzymes to the unsaturated enzymes mediating growth-limiting linear fluxes . Moreover , independent of rates , production costs , and stoichiometries , optimal metabolic types must consist of non-overlapping trees of conversions . Indeed , the optimality of such metabolic networks , obtained from transport-network theory , only requires the linearity of metabolic fluxes with respect to enzyme concentrations . As a result , optimal metabolic types , as well as cartels , can still be enumerated . Interestingly , we discovered that distinct cartels can arise for very similar external building-block availabilities , and cartels can even merge under special conditions . In an extended model that includes fluctuations , e . g . in enzyme expression ( Wang and Zhang , 2011; Kiviet et al . , 2014 ) , we expect ‘ghosts’ of these neighboring cartels associated with similar resource availabilities to persist against the background of the dominant cartel . As our primary concern is the emergence of a division of labor , we consider only relatively large populations of cells for which we can neglect stochastic population fluctuations . What relevance might our results have for real metabolic networks ? Microbes regulate metabolic processes via complex networks with , e . g . , multistep reaction chains and metabolic branch points ( Almaas et al . , 2004 ) . However , there is evidence of optimal partitioning of enzymes in these real networks: microbes produce components of multiprotein complexes in precise proportion to their stoichiometry , whereas they produce components of functional modules differentially according to their hierarchical role ( Li et al . , 2014 ) . Recent experimental studies have revealed that optimal metabolic flux partitioning is an operating principle for resource allocation in the proteome economy of the cell ( Hui et al . , 2015; Hermsen et al . , 2015 ) . Provided optimality considerations apply to real metabolic networks , the approach we have taken can provide insight into flux partitioning and division of labor in microbial communities . For instance , we expect that for a group of interconvertible resources that are collectively growth limiting , the expressed metabolic network should have the topological properties discussed above — no ‘futile cycles’ and no ‘convergent pathways’ . Such predictions are not at odds with the existence of well-known metabolic cycles such as the TCA cycle and the GOGAT cycle because these cycles are not futile but rather are energy yielding or assimilatory , respectively . Our predictions apply directly to irreversible conversion processes , e . g via chains of reactions with committed steps , as well as to reversible chains of reactions , for which the only cycles in optimal metabolic networks are two-cycles due to reversibility . The overall acyclic nature of anabolic fluxes can be tested experimentally by measuring reaction fluxes in metabolic networks , e . g . using isotope tracers and mass spectrometry . Abiotic and biotic processes controlling resource turnover in nutrient reservoirs , such as the ocean or soil sediments , operate on many different temporal and spatial scales ( Braswell et al . , 1997; Whitman et al . , 1998 ) . In our framework , steady but spatially inhomogeneous supply of diffusive building blocks should lead to the tiling of space by locally dominant cartels . Because of our model cells’ ability to shape their environment , we expect sharp transitions between neighboring tiles , consisting of cartels that differ by a single metabolic class . We expect spatial tiling to emerge in real microbial communities growing in inhomogeneous conditions , e . g . in a gradostat with spatially structured nutrient supply ( Lovitt and Wimpenny , 1981 ) . In such spatial communities , the detection of well-delimited patches of resource availabilities , with specific nutrient ratios , would be evidence of spatial tiling by microbial cartels . The spatial structure of microbial communities may also reflect the extracellular division of labor . Extracellular division of labor involves metabolic pathways with obligatory external reactions , i . e . with enzymes that are public goods . In a homogeneous environment , ‘cheating’ cell types that do not produce the public good are always at an advantage and their introduction causes the collapse of the entire population . In our framework , we expect producer cartels to spatially segregate from neighboring non-producer cartels ( Drescher et al . , 2014 ) . Temporally varying supply can also be addressed within our framework . For supply fluctuations on long timescales ≫1/δ ( the lifetime of a cell ) , the population dynamics within cartels keeps resource levels fixed , whereas fluctuations on short timescales ≪1/δ are self-averaging . In practice , slow supply fluctuations can arise due to seasonal biogeochemical cycles ( Schoener , 2011 ) , while fast supply fluctuations can arise from the transient biomass release upon cell death ( Yoshida et al . , 2003 ) . The effect of supply fluctuations occurring on timescales ∼1/δ , which includes day-night cycles , is more complex . Transport-network theory predicts that fluctuating resource conditions select for networks with metabolic cycles , whose structures depend on the statistics of the driving fluctuations ( Katifori et al . , 2010; Corson , 2010 ) . Characterizing the benefit of cycles in such networks may well reveal new optimization principles that underlie the microbial metabolic diversity . Microbes also adjust to fluctuating conditions by switching their metabolic type via gene regulation instead of relying on population dynamics . Within our framework , to consistently implement the optimal mix of metabolic strategies , the role of sensing and regulation is then primarily to determine the relevant ‘supply sector’ by assessing the relative abundance of various resources . Thus , in principle , division of labor within a single species can lead to cartels with distinct metabolic strategies associated with distinct phenotypic states . However , the persistence of cartels requires the coexistence of all cartel strategies , which within a single species could be facilitated by cell-to-cell communication ( quorum sensing ) . We therefore anticipate that extension of our analysis to fluctuating supply conditions may provide insight into the design principles underlying regulation and signaling in microbial communities . | Microbes are found in virtually every environment on Earth . Like other organisms , microbes grow by using enzymes to convert nutrients into proteins , DNA and other molecules that make up their cells . Together , these chemical transformations define the “metabolism” of a microbe . In any given environment , there is almost always a diverse variety of microbes living together . Different microbes in these communities will use different combinations of enzymes to exploit the available nutrients , and members of well-studied communities have been found to work together to make the most of the nutrient source . This is remarkable because one might expect competition between microbes to select for a single “best” microbe , rather than diverse communities . The economic concept of “division of labor” suggests that if microbes divide chemical tasks between each other , they will use the available resources more efficiently . The concept provides a possible explanation for metabolic diversity amongst microbes , yet it remains to be shown whether microbial communities actually benefit from a division of labor . Here , Taillefumier et al . used mathematical models to reveal that even in a uniform environment , metabolic competition generally leads to the steady coexistence of distinct microbes , collectively called a “consortium” . In a consortium , distinct microbes organize themselves to create a community-level metabolism that best exploits the nutrients present . The models showed that while growing , a consortium depletes the available pool of nutrients to such low levels that only members of the consortium can survive . The findings suggest that the benefit of metabolic diversity stems from the ability of a consortium to automatically deplete nutrients to levels at which no other microbes can invade . Taillefumier et al . propose that consortia that arise naturally under conditions where there is a steady supply of nutrients produce the maximum mass of microbes . Future experiments that analyze the impact of fluctuating nutrient supply may help us to understand the benefit of metabolic diversity in real-world microbial communities . | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"ecology",
"physics",
"of",
"living",
"systems"
] | 2017 | Microbial consortia at steady supply |
Mechanotransducer channels at the tips of sensory stereocilia of inner ear hair cells are gated by the tension of 'tip links' interconnecting stereocilia . To ensure maximal sensitivity , tip links are tensioned at rest , resulting in a continuous influx of Ca2+ into the cell . Here , we show that this constitutive Ca2+ influx , usually considered as potentially deleterious for hair cells , is in fact essential for stereocilia stability . In the auditory hair cells of young postnatal mice and rats , a reduction in mechanotransducer current , via pharmacological channel blockers or disruption of tip links , leads to stereocilia shape changes and shortening . These effects occur only in stereocilia that harbor mechanotransducer channels , recover upon blocker washout or tip link regeneration and can be replicated by manipulations of extracellular Ca2+ or intracellular Ca2+ buffering . Thus , our data provide the first experimental evidence for the dynamic control of stereocilia morphology by the mechanotransduction current .
The sense of hearing depends on stereocilia , the microvilli-like mechanosensory projections at the apical surface of inner ear hair cells . A hair cell bundle consists of stereocilia rows with precisely graded heights according to the cell’s location along the cochlea , suggesting that the exact shape of the bundle is crucial for the normal function of the hair cell ( Engström and Engström , 1978 ) . Mammalian auditory hair cells do not regenerate and , therefore , have to maintain their precisely arranged stereocilia throughout the lifespan of the organism . Indeed , recent data show that the actin core is remarkably stable along the length of stereocilia , except for a small region at their tips ( Zhang et al . , 2012; Drummond et al . , 2015; Narayanan et al . , 2015 ) . Hair cell stereocilia are interconnected by extracellular ‘tip links’ ( Pickles et al . , 1984 ) . Sound-induced deflections of a hair bundle modulate the tension of the tip links , which controls the opening of mechano-electrical transduction ( MET ) channels ( Assad et al . , 1991 ) . These channels are located at the tips of shorter but not the tallest row stereocilia ( Beurg et al . , 2009 ) . In the resting bundle , the tip links are under a certain degree of tension , ensuring responses to the smallest sound-induced deflections ( Howard and Hudspeth , 1987; Hacohen et al . , 1989; Assad et al . , 1991 ) . This resting tension is thought to be responsible for the wedge-shaped stereocilia tips at the lower end of the tip links in shorter row stereocilia ( Furness and Hackney , 1985; Kachar et al . , 2000; Rzadzinska et al . , 2004 ) . In addition , the resting tip link tension increases the open probability of MET channels , resulting in a continuous influx of Ca2+ into the cell through these non-selective cation channels ( Corey and Hudspeth , 1979 ) . This constitutive Ca2+ influx is perceived as a potentially deleterious consequence of the extreme sensitivity of the MET apparatus in the auditory hair cells ( Beurg et al . , 2010 ) . However , we show here that the resting MET current also controls the structural stability of the transducing stereocilia in auditory hair cells .
In our previous study ( Indzhykulian et al . , 2013 ) , we noticed relatively slow changes ( within ~20 min ) of the stereocilia tip shape after breaking the tip links . We observed these changes in scanning electron microscopy ( SEM ) images obtained from samples fixed at different time points after tip link breakage . Therefore , we decided to explore whether these changes of stereocilia tip shape could be initiated by the loss of the resting MET current . We blocked the MET channels with extracellular amiloride or benzamil at concentrations of 100 µM and 30 µM , respectively . At these concentrations , the blockers are expected to inhibit ~75% ( amiloride ) or ~90% ( benzamil ) of the MET current induced by hair bundle deflections ( Rüsch et al . , 1994 ) . Indeed , both these blockers substantially reduced the entry of FM1-43 dye through the partially open at rest MET channels in both inner ( IHCs ) and outer ( OHCs ) hair cells of young postnatal mouse organ of Corti explants ( Figure 1A ) . We also cultured organ of Corti explants at postnatal day four ( P4 ) in the presence of benzamil for 24 hr and confirmed that , after a long-term MET blockage , the MET channels remain inhibited but are still functional ( Figure 1B ) . 10 . 7554/eLife . 24661 . 003Figure 1 . Long-term blockage of the MET channels causes selective shortening of the second and third , but not the first ( tallest ) , rows of stereocilia in mouse outer hair cell ( OHC ) bundles . ( A and B ) Assessment of MET blockage with MET channel-permeable dye , FM1-43 . ( A ) Left panels show maximal projection images of FM1-43 fluorescence in mouse organ of Corti explants immediately after the tissue dissection , in control conditions ( top ) and in the presence of non-saturating concentrations of MET blockers: amiloride ( 100 μM , middle ) or benzamil ( 30 μM , bottom ) . Right panels show reference bright-field images of the same cochlear explants at the focal plane of the hair cell bodies . Data are representative of two independent series . ( B ) Similar maximal projection FM1-43 ( left ) and bright-field ( right ) images at the end of 24 hr incubation at 37°C with 30 μM of benzamil ( top ) and after washout of this drug ( bottom ) . ( C ) Representative scanning electron microscopy ( SEM ) images of OHC stereocilia bundles ( false-colored ) in mouse organ of Corti explants cultured for 24 hr at 37°C in vehicle control conditions ( top ) , 100 μM of amiloride ( middle ) , or 30 μM of benzamil ( bottom ) . Right panels show higher magnification images of OHC stereocilia . Arrows point to examples of retracted stereocilia . ( D ) Heights of individual stereocilia in different rows of OHC bundles in mouse organ of Corti explants cultured for 24 hr in control conditions ( black; n = 103–120 stereocilia ) or in the presence of the MET blockers ( gray ) , amiloride ( 100 μM , n = 99–108 ) or benzamil ( 30 μM , n = 75–80 ) . Error bars indicate mean ± SD . The data are from a single series of experiments ( 8–17 cells per treatment ) with control and drug-treated explants processed in parallel , representative of one ( amiloride ) and three ( benzamil ) independent series . ( E and F ) Representative false-colored SEM images of OHC bundles ( E ) and quantification of stereocilia heights ( F ) in the first ( blue ) , second ( yellow ) and third ( red ) rows of the bundle ( n = 40–130 ) , indicating the dose-dependent effect of a 32 hr incubation in the presence of 0 , 5 . 5 , 10 and 30 μM of benzamil . Data ( 4–12 cells per treatment ) are shown as mean ± SD . For D and F: *p<0 . 05; ****p<0 . 0001; n . s . , non-significant ( Welch’s t tests ) . ( G ) Percentage of shorter ( second and third ) row OHC stereocilia having tip links after 24 hr culturing in control conditions ( black , n = 632 ) or with MET blockers ( gray , n = 262–527 ) . Combined data from two independent series ( 7–17 cells per treatment ) are shown as mean ± SE . n . s . , non-significant ( Student’s t tests ) . Age of explants A: P5; B–G: P4 +24–32 hr incubation . Original SEM images can be found in Vélez-Ortega et al . ( 2017 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 00310 . 7554/eLife . 24661 . 004Figure 1—figure supplement 1 . Blockage of the MET current causes selective shortening of the second- and third-row ( transducing ) stereocilia in rat auditory hair cells . Representative false-colored SEM images of OHC ( left and middle ) and IHC ( right ) stereocilia bundles from rat organ of Corti explants cultured for 24 hr at 37°C in vehicle control conditions ( top ) or in the presence of 100 μM of amiloride ( middle ) or 30 μM of benzamil ( bottom ) . Middle panels show higher magnification views of stereocilia from the OHC bundles on the left . The arrows indicate examples of shortened stereocilia . Age of explants: P6 +24 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 00410 . 7554/eLife . 24661 . 005Figure 1—figure supplement 2 . Organ of Corti examinations were limited to the middle cochlear region . ( A ) Montage of a representative SEM image of an organ of Corti explant ( harvested at P4 and cultured for 30 hr ) and two drawings illustrating the flexible glass fibers used to hold the tissue in place during culturing . The percentage distance from the apex is indicated in four locations , and the arrow shows the region where we performed the high-resolution imaging of the hair cell bundles . ( B ) Representative false-colored SEM images of stereocilia bundles , from the second row of OHCs , after 24 hr incubation in culture medium alone ( top ) or in the presence of 0 . 05% of the drug vehicle DMSO ( second panel ) , 100 μM of amiloride ( third panel ) , or 30 μM benzamil ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 00510 . 7554/eLife . 24661 . 006Figure 1—figure supplement 3 . Quantification of stereocilia heights from SEM images . ( A ) Schematic diagram illustrating the heights of stereocilia to be determined in the first ( h1 ) , second ( h2 ) and third ( h3 ) rows of the bundle . ( B ) Quantification of the height of a tallest row stereocilium from two projections ( p1 and p2 ) measured in two SEM images of the lateral ( ‘back’ ) side of the same hair bundle at different tilt angles separated by τ degrees . The actual height of a stereocilium ( h ) is calculated by solving the system of two equations shown in the box with two unknown variables ( h and α , the initial tilt angle ) . ( C ) Alternatively , we acquired several images of the same bundle at different tilt angles to determine the angle at which the tallest row stereocilia are parallel to the EM beam ( 0° in the illustrated example ) . ( D ) Next , we calculated the height of the tallest row stereocilia ( h1 ) using the equation in the box , where p1 is the measured projection and τ1 is the ‘angle of view’ , that is , the difference between the tilt angle of the acquired image and the angle where stereocilia are parallel to the beam . The technique illustrated in panel D is less time-consuming than the technique in panel B and , therefore , it was used throughout the study . However , the key measurements of the heights of the tallest row stereocilia in Figure 1D and Figure 2B were reproduced with both measurement techniques that gave identical results . ( E ) Heights of the second- ( h2 ) and third-row ( h3 ) stereocilia were determined from an image of the medial ( ‘front’ ) side of the hair bundle following the equations in the box , where p2 and p3 are the measured projections of height differences between first - second and first - third row stereocilia , τ2 is the known angle of view determined as in panel C , h1 is the height of the tallest stereocilium determined as in panel D , and d is the distance between stereocilia rows at the base . The average value of parameter d was determined from images taken at an angle where stereocilia were parallel to the EM beam . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 006 Next , we used scanning electron microscopy ( SEM ) to examine the effect of this long-lasting inhibition of MET channels on the morphology of the hair bundle . Although a small re-shaping of the stereocilia tips was expected based on previous observations after breakage of the tip links ( Kachar et al . , 2000; Rzadzinska et al . , 2004 ) , we were surprised to discover that both amiloride and benzamil caused dramatic changes to the staircase morphology of OHC stereocilia bundles both in mice ( Figure 1C ) and in rats ( Figure 1—figure supplement 1 ) . To quantify these changes , we acquired SEM images of the same hair cell bundles from different viewing angles ( between 0° and ±52° ) in the mid-cochlear region ( Figure 1—figure supplement 2 ) . Typically , we determined an angle of view to the top of the bundle and obtained at least two views of each side of the bundle ( medial and lateral ) , which ensured accurate measurements of stereocilia heights ( Figure 1—figure supplement 3 ) . Our analysis showed that neither of the MET channel blockers affected the height of the tallest ( first row ) stereocilia in OHC bundles ( Figure 1D , left ) . Of note , similar heights of the tallest row stereocilia were reported in hamster OHCs at the same mid-cochlear location and the same age ( P5 ) ( Kaltenbach et al . , 1994 ) . Furthermore , throughout the 24 hr culturing of the mouse cochlear explants harvested at P4 , we did not detect any height changes in the tallest row stereocilia in the mid-cochlear OHCs ( data not shown ) , consistent with previous reports showing that OHC ( but not IHC ) stereocilia stop growing and reach a plateau at the ages of P2-P4 throughout most of the cochlea except at the very apex ( Roth and Bruns , 1992; Kaltenbach et al . , 1994 ) . In contrast to the non-transducing tallest row stereocilia , transducing stereocilia of the second and third rows in the OHC bundles exhibited dramatic changes of their morphology after incubation with MET channel blockers ( Figure 1C ) . Especially in response to benzamil , many stereocilia shortened to heights that we never observed in control samples processed in parallel . These results indicate the retraction or disassembly of the stereocilium actin core ( Figure 1C , bottom and Figure 1D , right columns ) . The effect of amiloride was similar , but it produced larger variability of stereocilia heights ( Figure 1C , middle and Figure 1D , middle columns ) . Overall , we detected a decrease in the average height of the second- and third-row stereocilia after incubation with either of the MET channel blockers ( Figure 1D ) . The effect of MET current blockage on the height of transducing stereocilia was dose-dependent ( Figure 1E , F ) . Among the different rows of OHCs in the organ of Corti , the largest effects were observed in the third ( the outermost ) row of OHCs ( data not shown ) . Notably , the number of visible tip links per stereocilium in OHCs did not change after long-term MET blockage ( Figure 1G ) . It is not clear whether these tip links are newly formed links or the same links that slid down . However , their presence suggests that stereocilia shortening after the blocking of the MET channels is likely to occur even in the presence of tip link-generated mechanical tension and other tip link-associated signaling events . The effects of MET blockers on the staircase morphology of IHC bundles were qualitatively similar to those observed in OHCs ( Figure 2 ) . However , IHCs exhibited a smaller decrease in the heights of transducing stereocilia as compared to OHCs from the same explants and location along the cochlea ( Figure 2A–B ) . As in OHCs , long-term blockage of MET channels in IHCs did not result in the loss of tip links ( Figure 2C ) . We also noticed the accelerated ‘pruning’ of the supernumerary ( fourth and fifth rows and unranked ) stereocilia after MET channel blockers in both IHCs ( control = 25 . 1 ± 0 . 9 vs . benzamil = 20 . 4 ± 1 . 3 supernumerary stereocilia per cell , n = 19–31 cells , p<0 . 04 ) and OHCs ( control = 10 . 9 ± 0 . 8 vs . benzamil = 6 . 8 ± 0 . 7 , n = 32–65 , p<0 . 002 ) . It may indicate that these supernumerary stereocilia express functional MET channels or that the developmentally regulated program of their retraction depends on the intracellular Ca2+ concentration , which is expected to decrease after MET channel blockage . 10 . 7554/eLife . 24661 . 007Figure 2 . Blockage of the MET channels causes selective shortening of transducing second- and third-row stereocilia but not the tallest first-row stereocilia in mouse inner hair cells ( IHCs ) . ( A ) Representative scanning electron microscopy ( SEM ) images of IHC stereocilia bundles in mouse organ of Corti explants cultured for 24 hr at 37°C in vehicle control conditions ( top ) , 100 μM of amiloride ( middle ) , or 30 μM of benzamil ( bottom ) . The insets show higher magnification images of the tips of second-row stereocilia . ( B ) Heights of individual stereocilia in different rows of IHC bundles in mouse organ of Corti explants cultured for 24 hr in control conditions ( black; n = 80–91 stereocilia ) or in the presence of the MET blockers ( gray ) , amiloride ( n = 38–48 ) or benzamil ( n = 55–62 ) . The heights of the tallest stereocilia are similar to those reported for the hamster IHCs at the same age and mid-cochlear location ( Kaltenbach et al . , 1994 ) . Error bars indicate Mean ± SD . *p<0 . 05; ****p<0 . 0001; n . s . , non-significant ( Welch’s t tests ) . The data are from a single series of experiments ( 8–18 cells per treatment ) with control and drug-treated explants processed in parallel , representative of one ( amiloride ) and three ( benzamil ) independent series . ( C ) Percentage of the second-row stereocilia with tip links in IHC bundles after 24 hr incubation in control conditions ( black , n = 459 ) or with MET blockers ( gray , n = 163–325 ) . The data from two independent series are shown ( 12–31 cells per treatment ) as mean ± SE . n . s . , non-significant ( Student’s t tests ) . Age of explants in A-C: P4 +24 hr incubation . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 007 Besides amiloride and benzamil , a similar retraction of shorter ( but not the tallest ) row stereocilia was observed in mouse OHCs and IHCs after 24 hr incubation with 30 µM of tubocurarine ( data not shown ) , a larger molecule that also blocks MET channels but is unlikely to permeate through them ( Farris et al . , 2004 ) . Altogether , the selective shortening of the transducing second- and third-row stereocilia , but not of those in the tallest row , argues against a non-specific action of the MET blockers on the actin core of the stereocilium . Next , we explored whether the observed stereocilia shortening was permanent . We cultured several organ of Corti explants with or without benzamil for 24 hr and , as expected , found retraction of the transducing stereocilia ( Figure 3A , C ) . Then , we removed the benzamil and allowed the explants to recover at 37°C for an additional 24 hr . After the recovery period , the second-row stereocilia and the majority of third-row stereocilia re-grew and reached the same stereocilia heights as in the control explants that were processed in parallel ( Figure 3B , D ) . This data suggest that the effects of the MET blockers on the transducing stereocilia are reversible and indicate that the MET current is likely to dynamically regulate the height of these stereocilia . 10 . 7554/eLife . 24661 . 008Figure 3 . Transducing stereocilia that have retracted due to MET blockage are able to regrow after drug washout . ( A and B ) Representative false-colored SEM images of mouse OHC stereocilia after ( A ) 24 hr incubation either in control conditions ( left ) or with 30 μM of benzamil ( right ) , and after ( B ) 24 additional hours of recovery after washout . Arrows down point to retracted stereocilia , while arrows up indicate re-growth . ( C and D ) Heights of second- and third-row stereocilia after ( C ) 24 hr incubation in control conditions ( black , n = 83 ) or with benzamil ( gray , n = 95 ) , and ( D ) 24 hr after washout ( n = 30 , control; n = 38 , benzamil ) . Stereocilia heights are shown as a percent relative to the size of tallest ( first ) row . Data are from 7 to 16 cells per treatment . Error bars indicate mean ± SD . ****p<0 . 0001; n . s . , non-significant ( Welch’s t tests ) . Age of the explants: P4 +24–48 hr . All incubations were performed at 37°C . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 008 In the next experiment , we compared the effects of three chemically unrelated MET blockers . We used benzamil , ruthenium red , and tubocurarine , at concentrations of 30 , 10 and 30 μM , respectively , previously shown to block 80–90% of the MET current ( Rüsch et al . , 1994; Farris et al . , 2004 ) . We used a shorter incubation period , but , as before , we processed all samples in parallel and imaged the hair cell bundles at the same mid-cochlear location . As expected , the incubation with benzamil , ruthenium red or tubocurarine for 5 hr led to a smaller , but still statistically significant , retraction of shorter row stereocilia in the mouse OHCs ( Figure 4A , B ) . Although all these chemically unrelated drugs may have different side effects , the fact that they all produced qualitatively similar shortening of transducing stereocilia in the OHCs indicates that such shortening is likely to arise from their common action—the blockage of the MET channels . 10 . 7554/eLife . 24661 . 009Figure 4 . MET-dependent shortening of transducing stereocilia in OHCs initiates with thinning at the tips . ( A ) SEM images of mouse OHC stereocilia bundles after incubation at 37°C for 5 hr in vehicle control conditions ( top ) or in the presence of the MET blockers benzamil ( 30 μM , second row ) , ruthenium red ( 10 μM , third row ) or tubocurarine ( 30 μM , bottom ) . Right panels show stereocilia at higher magnification; arrows point to retracted stereocilia and arrowheads indicate some examples of thin stereocilia tips . ( B ) Height differences between first- and second-row stereocilia ( Steps 1–2 , left ) and between first- and third-row stereocilia ( Step 1–3 , right ) in OHC bundles after 5-hr incubation in control conditions ( black; n = 76 ) or with MET blockers ( gray , n = 58–117 ) . The inserts show the measurement procedure , which traced each stereocilium to its highest point and did not account for the shape changes at the tips . Data are from 7 to 13 cells per treatment and representative of one ( ruthenium red , tubocurarine ) and two ( benzamil ) independent series . Error bars indicate mean ± SD . Note that the staircase ‘step’ measurement procedure requires fewer calculations than the estimation of the absolute height of the stereocilium ( as in Figures 1D and 2B ) and , therefore , it is more accurate for quantifying smaller changes in the staircase morphology of the bundles ( see Figure 1—figure supplement 3C , E ) . ( C ) Diameter of second-row stereocilia at the shaft and at several positions near the stereocilia tip in OHCs after a 5-hr incubation in control conditions ( black , n = 40 ) or with MET blockers: benzamil ( dark grey , n = 44 ) , ruthenium red ( light grey , n = 42 ) or tubocurarine ( white , n = 36 ) . Data are from 3 to 5 cells per treatment and are shown as mean ± SE . For B and C: ***p<0 . 001; ****p<0 . 0001; n . s . , non-significant ( Welch’s t tests ) . ( D ) Representative transmission electron microscopy ( TEM ) images of the upper part of first- and second-row stereocilia from mouse OHCs after incubation for 5 hr in control conditions ( left ) or in the presence of 30 μM benzamil ( right ) . Notice the actin filaments within the abnormally thin tips of second-row stereocilia after treatment with benzamil . Age of the explants in A–D: P4-5 . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 00910 . 7554/eLife . 24661 . 010Figure 4—figure supplement 1 . MET current regulates the shape of the tips of transducing stereocilia in OHCs . ( A and B ) Superimposed contours of OHC stereocilia tips ( n = 14–18 individual stereocilia per each cartoon , each represented by a thin grey line ) from organ of Corti explants cultured for 5 hr in control conditions ( A ) or in the presence of 30 μM of benzamil ( B ) . Notice that the blockage of MET channels by benzamil leads to changes in the tip shape of second- and third-row stereocilia ( i . e transducing stereocilia ) , while the tip shape of the non-transducing first-row stereocilia remains unchanged . The contours were derived from SEM images of OHCs ( at the same mid-cochlear location and the same angle of view ) and were aligned to the point at the very tip to reveal the differences in tip shapes . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 010 The 5-hr incubation with MET blockers also allowed us to explore the initial steps of stereocilia shortening . After this incubation , we observed an increased number of thin and ‘pointed’ tips in the second- and third-row stereocilia of OHC bundles but not in the non-transducing tallest row stereocilia ( Figure 4A−arrowheads , and Figure 4—figure supplement 1 ) . We quantified these shape changes in the second row of stereocilia , where the effect was most prominent . We found that all MET channel blockers tested produce highly significant thinning of the tips of these stereocilia in OHCs ( Figure 4C ) . We examined the actin core of the abnormally thin stereocilia tips with transmission electron microscopy ( TEM ) in plunge-frozen freeze-substituted preparations . In all these stereocilia ( n = 30 ) , actin filaments filled the tips completely , without any signs of ‘over-tented’ membranes ( Figure 4D ) . We concluded that the MET current-dependent thinning of the tips of transducing stereocilia is caused by the remodeling of the stereocilia actin core . In fact , these data also suggest that the actin filaments located at the circumference of the stereocilia core , farther away from the transducer channel , are more susceptible to the blockage of the MET current . The 5-hr incubation with MET channel blockers affected transducing stereocilia in IHCs to a lesser degree than in OHCs ( Figure 5A ) . These differences between IHCs and OHCs were expected based on the results observed after the 24 hr incubations ( Figures 1 and 2 ) . However , we still detected statistically significant changes in the staircase ‘steps’ of IHC bundles after the 5-hr incubation with some ( but not all ) MET channel blockers ( Figure 5B ) . Despite the larger variability of tip shapes in the second-row stereocilia in IHCs as compared to OHCs , we were also able to detect statistically significant changes in the shape of IHC second-row stereocilia tips after incubation with any of these three MET channel blockers ( Figure 5C ) . Similar to OHCs , we observed the remodeling of the underlying actin core after MET current blockage ( Figure 5D ) in all second-row IHC stereocilia with abnormally thin tips that we examined by TEM ( n = 10 ) . We concluded that the effects of MET current blockage on the shapes of IHC stereocilia tips are less prominent but qualitatively similar to those in OHCs . 10 . 7554/eLife . 24661 . 011Figure 5 . MET current regulates the shape of transducing stereocilia in IHCs . ( A , B ) Representative SEM images ( A ) and quantification of height differences between stereocilia rows ( B ) of IHC stereocilia bundles after incubation at 37°C for 5 hr in vehicle control conditions or in the presence of MET blockers: benzamil ( 30 μM ) , ruthenium red ( 10 μM ) or tubocurarine ( 30 μM ) . Panel layouts are identical to Figure 4A , B . ( B ) Staircase ‘step’ measurements: control ( n = 57 ) , benzamil ( n = 62 ) , ruthenium red ( n = 52 ) , and tubocurarine ( n = 51 ) . Data are from 8 to 12 cells per treatment . ( C ) Heights of the wedged tips ( left cartoon ) from individual IHC stereocilia in the second row of the bundle after incubation for 5 hr in vehicle control conditions ( black , n = 54 ) or in the presence of MET blockers ( gray , n = 59–66 ) . Data are from 3 to 5 cells per treatment . In B and C: Error bars indicate mean ± SD . **p<0 . 01; ***p<0 . 001; ****p<0 . 0001; n . s . , non-significant ( Welch’s t tests ) . Age of explants in all panels: P4-5 . ( D ) Representative TEM images of the tips of second-row stereocilia in IHCs after incubation for 5 hr in vehicle control conditions ( left ) or in the presence of 30 μM of benzamil ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 011 To determine whether Ca2+ influx is the component of the MET current responsible for maintaining the stability of transducing stereocilia , we loaded organ of Corti explants with the membrane-permeable acetoxymethyl ( AM ) ester derivative of BAPTA ( BAPTA-AM ) , which is cleaved inside the cell by endogenous esterases . This treatment results in the accumulation of BAPTA in the cell , increasing intracellular Ca2+ buffering and limiting any Ca2+-dependent effects to the vicinity of the sites of Ca2+ entry into the cytosol . It does not affect the integrity of stereocilia links ( Figure 6A , B , insets ) , because micromolar concentration of BAPTA-AM outside of the cell is not sufficient to significantly decrease the millimolar concentration of extracellular Ca2+ . In both OHCs and IHCs , intracellular BAPTA resulted in the appearance of the abnormally thin tips in the second but not the tallest row stereocilia ( Figure 6A–D ) as well as the shortening of many transducing second- and third-row stereocilia ( Figure 6E , F ) . Similar to our results with extracellular MET blockers , the effects of intracellular BAPTA on the average height of transducing stereocilia were more prominent in OHCs than in IHCs ( Figure 6E , F ) . This stereocilia remodeling after BAPTA-AM cannot be attributed to the loss of MET current because intracellular BAPTA does not block this current or change the resting open probability of MET channels in either OHCs or IHCs , even at a very large concentration of 10 mM ( Peng et al . , 2013 ) . However , the intracellular BAPTA should decrease the size of the Ca2+‘hotspot’ at the point of Ca2+ influx through the MET channels ( Figure 6G ) . A steeper Ca2+ gradient inside the stereocilium would decrease the Ca2+ concentration experienced by the peripheral actin filaments , causing their preferential retraction or disassembly ( Figure 6G , arrows ) and resulting in a ‘pointed’ shape at the tips of second-row stereocilia in both OHCs ( Figure 6C ) and IHCs ( Figure 6D ) . A similar mechanism may shape the tips of the second-row stereocilia in OHCs and IHCs in the presence of nearly saturating ( but not completely blocking ) concentrations of MET blockers ( Figures 4D and 5D ) . In contrast to the second-row stereocilia , tip shape changes were barely noticeable in the significantly shorter third-row OHC stereocilia after intracellular BAPTA ( Figure 6A , B ) or after MET channel blockage ( Figure 4—figure supplement 1 ) . However , even in these stereocilia , we observed similar shortening after both treatments ( Figures 4B and 6E , right panels ) . We concluded that , similar to the MET blockers , the increase of intracellular Ca2+ buffering can initiate the remodeling and shortening of transducing stereocilia in auditory hair cells . Therefore , the stability of the transducing shorter row stereocilia in the auditory hair cells may be controlled by the Ca2+ influx through the MET channels that are partially open at rest . 10 . 7554/eLife . 24661 . 012Figure 6 . Increased intracellular Ca2+ buffering leads to thinning and shortening of transducing stereocilia in mouse auditory hair cells . ( A and B ) SEM images of mouse OHC ( left and middle ) and IHC ( right ) stereocilia bundles after incubation for 1 hr ( A ) or 5 hr ( B ) at 37°C in vehicle-control conditions ( top ) or in the presence of 20 μM of the membrane-permeable BAPTA-AM ( bottom ) . Middle panels and insets show higher magnification images of OHC and IHC stereocilia , respectively . Arrowheads indicate examples of stereocilia with abnormally thin tips . ( C and D ) Stereocilia diameter at the shaft and at several regions near the tip in second-row stereocilia from OHCs ( C ) and IHCs ( D ) cultured in vehicle-control conditions ( black ) or with BAPTA-AM ( gray ) for 1 hr or 5 hr ( n = 41–47 ) . Data are from 3 to 4 cells per treatment and are shown as mean ± SE . ( E and F ) Height differences between the first and second ( left ) , and first and third ( right ) rows of stereocilia in OHCs ( E ) or IHCs ( F ) cultured in vehicle-control conditions ( black , n = 54–60 , OHCs; n = 40–52 , IHCs ) or with BAPTA-AM for 1 hr or 5 hr ( gray; n = 78–85 , OHCs; n = 64–79 , IHCs ) . Insert cartoons in E clarify that we measured the length of a stereocilium to its highest point and did not account for the shape changes at the tips . Data are from 8 to 10 cells per treatment . Error bars indicate mean ± SD . For C–F: Data shown are from a single series of experiments , representative of three independent series . **p<0 . 01; ***p<0 . 001; ****p<0 . 0001; n . s . , non-significant ( Welch’s t tests ) . Age of explants in A–F: P4 plus 1 or 5 hr of culturing . ( G ) Schematic diagram illustrating the intracellular Ca2+ gradient at the tip of a transducing stereocilium . Higher Ca2+ concentrations near the MET channel might prevent actin remodeling , while actin filaments further away from the channel might be more susceptible to depolymerization or other types of remodeling due to low Ca2+ concentration . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 01210 . 7554/eLife . 24661 . 013Figure 6—figure supplement 1 . Remodeling of transducing stereocilia initiated by changes in extracellular Ca2+ concentration . ( A ) SEM images of OHC stereocilia bundles after 1 hr incubations in control DMEM with 1 . 8 mM CaCl2 ( top ) or in BAPTA-buffered DMEM to lower the Ca2+ concentration to ~ 100 μM ( bottom ) . Right panels show higher magnification images of OHC stereocilia . Arrows point to shortened stereocilia and arrowheads to stereocilia with abnormally thin tips . ( B ) Height differences between stereocilia rows ( as indicated in the cartoons ) from OHCs incubated for 1 hr in normal ( black , n = 55 ) or low ( gray , n = 44 ) Ca2+ conditions . Data are from 3 to 7 cells per treatment . Error bars indicate mean ± SD . ( C ) SEM images of IHC stereocilia bundles after incubations for 1 hr in normal cell culture medium ( top ) or supplemented with 5 mM ( middle ) and 10 mM ( bottom ) CaCl2 . ( D ) Diameter of second-row stereocilia at the shaft and at several regions near the tip from IHCs cultured in control conditions ( black , n = 17 ) or in high extracellular Ca2+ ( light and dark gray , n = 19–25 ) . Data are from 2 to 3 cells per treatment and shown as mean ± SEM . For B and D: *p<0 . 05; **p<0 . 01; ***p<0 . 001; ****p<0 . 0001 ( Welch’s t tests ) . Age of explants: P4 + 1 hr in culture at 37°C . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 013 To study whether the remodeling of transducing stereocilia could be initiated by changes in the concentration of extracellular Ca2+ , we cultured organ of Corti explants at 37°C for 1 hr in extracellular media with different concentrations of free Ca2+ . Low extracellular Ca2+ decreases the concentration of free Ca2+ inside the auditory hair cell stereocilia , while high extracellular Ca2+ increases it ( Beurg et al . , 2010 ) , despite inhibition of MET channels in these cells by Ca2+ ions ( Kennedy et al . , 2003 ) . We observed shortening of transducing stereocilia in the OHCs cultured in low extracellular Ca2+ ( Figure 6—figure supplement 1 ) and thickening of the tips of second-row stereocilia in the IHCs cultured in high extracellular Ca2+ ( Figure 6—figure supplement 1 ) . With these very short incubations , we were not able to detect statistically significant effects of high extracellular Ca2+ in OHCs and low extracellular Ca2+ in IHCs . Longer incubations were damaging to the hair cells ( data not shown ) . We concluded that stereocilia remodeling may be initiated not only by changes in the intracellular Ca2+ buffering but also by changes in the extracellular Ca2+ concentration around the hair cell bundle . Although it has been previously reported that the loss of a tip link results in remodeling of the wedge-shaped tip of a stereocilium into a round-shaped tip ( Kachar et al . , 2000 ) , this effect has generally been associated with the loss of tip link tension and not the loss of the resting MET current ( see for example: [Rzadzinska et al . , 2004] ) . In addition , all these studies focused on the tips of stereocilia , leaving potential stereocilia shortening uninvestigated . Therefore , we re-examined the effects of tip link breakage on stereocilia shape . We disrupted the tip links with extracellular Ca2+-free medium supplemented with BAPTA , as previously described ( Indzhykulian et al . , 2013 ) . After BAPTA treatment for 15 min , we started to observe shortening of the second and third row stereocilia in OHCs , which became obvious 1 hr later ( Figure 7A ) . At several time points of recovery after BAPTA treatment , quantitative measurements revealed a significant decrease in the height of shorter row stereocilia ( Figure 7C ) . Similar to the effects of MET channel blockers ( Figure 1D ) , the absolute heights of the tallest OHC stereocilia were not affected after BAPTA and throughout recovery ( Figure 7B ) , indicating the selective shortening of only transducing stereocilia . However , we did not observe abnormally thin stereocilia tips in the transducing stereocilia of OHCs after the treatment with extracellular BAPTA ( Figure 7A , insets ) . This is expected because tip link disruption with BAPTA eliminates the MET current completely , in contrast to the experiments with non-saturating concentrations of MET blockers that may result in a Ca2+ gradient across the stereocilium diameter and the preferential remodeling of peripheral actin filaments ( Figure 6G ) . The regeneration of tip links led to the regrowth of transducing shorter row stereocilia by 6 hr of recovery ( Figure 7A , C ) , when ~ 70% of the MET current has reappeared ( Indzhykulian et al . , 2013 ) . Thus , there is only a relatively short time window after BAPTA treatment ( less than 6 hr ) when the shortening of transducing stereocilia in the OHCs could be detected . A similar but less prominent shortening of transducing stereocilia with their subsequent recovery was observed in the IHCs after treatment with extracellular BAPTA ( data not shown ) . As expected , the MET blocker benzamil inhibited stereocilia re-growth after extracellular BAPTA treatment in both OHCs and IHCs ( Figure 7—figure supplement 1 ) , confirming that this re-growth is likely to be driven by the recovery of the MET current . 10 . 7554/eLife . 24661 . 014Figure 7 . Disruption of tip links leads to remodeling of the transducing shorter row stereocilia . ( A ) Representative false-colored SEM images of OHC stereocilia bundles in mouse organ of Corti explants incubated in control conditions for 1 hr ( top ) , immediately after treatment with BAPTA-buffered Ca2+-free solution for 15 min ( second panel ) , and after 1 hr ( third panel ) and 6 hr ( bottom ) recovery in Ca2+-containing culture medium . Panels on the right show OHC stereocilia at higher magnification . The arrows point to shortened or regrown stereocilia . ( B and C ) Absolute heights of individual stereocilia in the first row ( B ) , and relative heights of the second ( left ) and third ( right ) rows ( C ) , in mouse OHC bundles at different times of recovery after BAPTA ( gray ) or in control conditions ( black ) . Error bars indicate mean ± SD . ****p<0 . 0001; n . s . , non-significant ( Welch’s t tests ) . Data are from 6 to 11 cells per time point and representative of two independent series . ( D ) SEM images of representative mouse IHC bundles before ( top left ) , immediately after treatment with BAPTA-buffered Ca2+-free medium for 15 min ( top right ) , and after recovery periods of 15 min ( bottom left ) and 7 hr ( bottom right ) . Insets show higher magnification views of the tips of second-row stereocilia . ( E ) The height of the wedged tips ( left cartoon ) of second-row stereocilia ( green ) and the percentage of tip links ( magenta ) before ( Control ) and at 0 min , 10 min and 7 hr of recovery after treatment with BAPTA . Quantifications of the wedged tip size and number of tip links were performed in the same IHCs . n = 33–105 stereocilia ( from 4 to 12 cells ) per time point . Pooled data from seven independent series , shown as mean ± SE . *p<0 . 05; ****p<0 . 0001 ( Student’s t tests ) . Age of explants: P4 . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 01410 . 7554/eLife . 24661 . 015Figure 7—figure supplement 1 . Regrowth of stereocilia after tip link regeneration is inhibited by MET current blockage . ( A ) Representative SEM images of mouse OHC ( left and middle ) and IHC ( right ) bundles that were treated with BAPTA-buffered Ca2+-free extracellular solution for 15 min and allowed to recover for 6 hr in Ca2+-containing medium in the absence ( top ) or presence ( bottom ) of the MET channel blocker benzamil ( 30 μM ) . Middle panels show higher magnification views of OHC stereocilia from the bundles on the left . The arrow indicates a stereocilium that was unable to re-grow , and the arrowheads point to two stereocilia that exhibit abnormally thin ‘pointed’ tips . ( B ) Quantification of height differences between first- and second-row stereocilia ( Steps 1–2 ) and between first- and third-row stereocilia ( Steps 1–3 ) in OHC ( left ) and IHC ( right ) bundles in the absence ( black , n = 32–70 ) or presence ( grey , n = 41–93 ) of the MET channel blocker benzamil ( 30 μM ) . The cartoons on the left show the measurement procedure . Data are from 5 to 6 cells per treatment . Age of explants: P4 . DOI: http://dx . doi . org/10 . 7554/eLife . 24661 . 015 We also re-analyzed our previously published experiments ( Indzhykulian et al . , 2013 ) to examine the changes in the wedge-shaped tips of IHC stereocilia after tip link breakage . Immediately after treatment with BAPTA for 15 min at room temperature , the conical wedge at the tips of the second-row stereocilia was still present ( Figure 7D , E ) . However , within ~10–20 min of recovery in normal extracellular medium and 37°C , the tips of these stereocilia became round ( Figure 7D , E ) . The wedge-shaped tips of the second-row stereocilia reappeared with the regeneration of the tip links at 7 hr of recovery ( Figure 7D ) . We concluded that the previously reported ‘rounding’ of the stereocilia tips after tip link breakage ( Kachar et al . , 2000; Rzadzinska et al . , 2004 ) is consistent with our current results on the MET current-dependent remodeling of transducing stereocilia .
Our study provides the first experimental evidence for the role of the MET current in the maintenance of hair bundle structural stability in mammalian auditory hair cells . Somewhat similar mechanisms were hypothesized to explain the development of the remarkable staircase morphology of the hair cell stereocilia bundle ( Tilney et al . 1992 ) or to explain particular mouse mutant phenotypes ( Alagramam et al . , 2011; Caberlotto et al . , 2011a , 2011b ) , but they have never been experimentally proved . In this study , remodeling at the tips of stereocilia and their subsequent retraction were observed ( i ) after treatment with various chemically unrelated blockers of the MET channels , ( ii ) after breakage of the tip links and associated loss of the MET current , and ( iii ) only in mechanotransducing shorter row stereocilia but not in the tallest row stereocilia . This evidence strongly suggests that stereocilia remodeling in our experiments is initiated by the reduction ( or complete loss ) of the resting MET current . Reproduction of these phenomena with the changes in intracellular Ca2+ buffering or extracellular Ca2+ concentration suggests that Ca2+ influx might represent an essential component of the MET current that controls the stability of the transducing stereocilium . However , further experimentation is needed to determine the exact role of Ca2+ in stereocilia remodeling . An important limitation of our study is that it was performed in young postnatal hair cells with stereocilia bundles that are not entirely mature . Unfortunately , it is hard to determine whether the observed MET-dependent stereocilia remodeling is present in the adult auditory hair cells , because the adult hair cells do not survive in culture . However , several lines of evidence indicate that the auditory hair cells in the middle of the cochlea at P4-P5 may demonstrate phenomena that are present in older hair cells . First , these hair cells exhibit well-developed MET current ( Waguespack et al . , 2007; Lelli et al . , 2009 ) . Second , the actin core of stereocilia in these cells seems to be already stable after a period of initial growth , while actin remodeling at the stereocilia tips is qualitatively similar to that in adult cells ( Zhang et al . , 2012; Drummond et al . , 2015; Narayanan et al . , 2015 ) . Third , this developmental age seems to coincide with the switch of myosin 15a isoforms that underlies the transition from predominant growth to the maintenance of the hair bundle ( Fang et al . , 2015 ) . Fourth , we observed the MET-dependent remodeling of transducing stereocilia also in rat and mouse OHCs and IHCs at P6 ( Figure 1—figure supplement 1 and data not shown ) . Last but not least , the developmental growth of stereocilia bundles in the mid-cochlear OHCs is already at the plateau at P4 ( Kaltenbach et al . , 1994 ) , while the growth of IHC stereocilia bundles continues until ~P18 ( Kaltenbach et al . , 1994; Peng et al . , 2009 ) . Therefore , the MET-dependent stereocilia remodeling observed in our study may reflect basic mechanisms controlling the height and the shape of transducing stereocilia after their initial growth . Ca2+-dependent mechanisms of actin core remodeling are likely to operate at the tips of stereocilia , where MET channels are in close proximity to the proteins regulating stereocilia growth ( Belyantseva et al . , 2005; Beurg et al . , 2009 ) . It is generally believed that most of the Ca2+ entering through the MET channels is extruded at the stereocilium level by the plasma membrane Ca2+ ATPase type 2 ( PMCA2 ) , which effectively ‘shields’ a stereocilium from the rest of the cell in both non-mammalian ( Lumpkin and Hudspeth , 1998; Yamoah et al . , 1998; Dumont et al . , 2001 ) and mammalian ( Dumont et al . , 2001; Beurg et al . , 2010 ) hair cells . The exact gradient of free Ca2+ concentration in the stereocilium and , hence , the concentration at the very tip , also depend on mobile and fixed intracellular Ca2+ buffers ( Lumpkin and Hudspeth , 1998; Beurg et al . , 2010 ) . In this study , all experiments aimed to reduce intrastereociliar Ca2+ resulted in shortening of the transducing stereocilia or their thinning at the tips , while all experiments designed to recover or increase intrastereociliar Ca2+ caused re-growth of transducing stereocilia or their thickening at the tips . Observed quantitative ( but not qualitative ) differences in the MET-dependent remodeling of stereocilia in the IHCs and OHCs can be easily explained by available data . The expected larger resting Ca2+ influx through the MET channels in OHCs ( Beurg et al . , 2010 ) should be extruded more effectively by PMCA2 that is expressed in OHC stereocilia at a higher density than in IHCs ( Chen et al . , 2012 ) . Therefore , the blockage of the MET channels in OHCs is expected to produce a larger drop in the resting Ca2+ concentration at the very tips of transducing stereocilia , causing a more prominent remodeling of these stereocilia in OHCs . In addition , the larger diameter of the second-row stereocilia in IHCs may promote a more prominent Ca2+ gradient across the diameter of a stereocilium ( see Figure 6G ) and , hence , produce larger MET- and Ca2+-dependent changes to the shape of stereocilia tips in IHCs than in OHCs . It is harder to explain why , after MET channel blockage , the third-row stereocilia in OHCs shorten but do not thin at the tips as prominently as do the second-row stereocilia ( see Figure 4—figure supplement 1 ) . Unfortunately , currently available techniques for the imaging of Ca2+ transients in cochlear hair cell stereocilia ( Beurg et al . , 2009 , 2010; Delling et al . , 2016 ) are limited to non-ratiometric imaging ( which cannot determine the actual concentration of free Ca2+ ) and are too crude to resolve Ca2+ gradients across a stereocilium diameter . There are essential molecules that are expressed differently in the auditory hair cell bundle between the first- and second-row stereocilia ( Furness et al . , 2013; Fang et al . , 2015; Ebrahim et al . , 2016 ) but , perhaps , also between the third and second rows . The latter differences have never been quantitatively investigated in the cochlear hair cells . It is worth mentioning that , despite the remarkable stability of stereocilia , the processes underlying this stability are physiologically vulnerable . Culturing organ of Corti explants at a room temperature of 25°C results in a less prominent MET-dependent stereocilia remodeling and initiates disruption of the hair bundle morphology ( data not shown ) . Therefore , we avoided commonly used techniques for the manipulation of intracellular Ca2+ , such as the application of Ca2+ ionophores or the inhibition of PMCA2 , because they could be deleterious to the hair cells incubated for several hours at 37°C . Our very limited understanding of stereocilia development and maintenance does not yet allow the proposal of a particular molecular mechanism for the MET-dependent stereocilia remodeling observed in this study . First , the Ca2+ influx through the MET channels may have a variable effect on different actin isoforms . The stereocilia core contains both β- and γ- isoforms of actin ( Furness et al . , 2005; Perrin et al . , 2010 ) . Although both these isoforms are distributed along the entire length of the stereocilium , variations in the ratio between isoforms are observed ( Perrin et al . , 2010 ) and , in fact , exposure to noise leads to visible changes in the ratio between these two isoforms at the stereocilia tips ( Belyantseva et al . 2009 ) . When bound to Ca2+ , γ-actin exhibits slower polymerization and depolymerization kinetics than β-actin ( Bergeron et al . , 2010 ) . Therefore , Ca2+ influx through MET channels may stabilize the actin core and even shift the equilibrium toward filament growth in the areas with an increased ratio of γ- to β- actin . Second , at least some proteins controlling actin dynamics in the stereocilia are Ca2+ sensitive . For example , the actin-bundling protein plastin 1 ( a homologue of fimbrin ) is expressed in hair cell stereocilia ( Flock et al . , 1982; Taylor et al . , 2015 ) and contains two EF hand Ca2+-binding sites ( Lin et al . , 1994 ) . Plastin 1 knock-out mice exhibit progressive hearing loss and stereocilia width abnormalities ( Taylor et al . , 2015 ) . Members of the gelsolin family enhance actin dynamics upon an increase of intracellular Ca2+ ( Kinosian et al . , 1998; Revenu et al . , 2007 ) and are also present in hair cell stereocilia ( Mburu et al . , 2010; Olt et al . , 2014 ) . The exact function of these and other potential Ca2+-sensitive regulators of actin in the hair cell stereocilia are yet to be determined . Third , the Ca2+ influx may affect various myosin motors expressed in the stereocilium , similarly to the proposed effects of Ca2+ on the myosin-based adaptation motor ( Gillespie and Cyr 2004 ) . Particularly interesting are myosin 15a and myosin 3 . Both these myosins are involved in stereocilia length regulation by delivering their cargoes , whirlin ( Belyantseva et al . , 2005 ) , Eps8 ( Manor et al . , 2011; Zampini et al . , 2011 ) , Eps8L2 ( Furness et al . , 2013 ) , espin-1 ( Salles et al . , 2009 ) and ESPNL ( Ebrahim et al . , 2016 ) to the tips of stereocilia . Some of these proteins—the long isoform of myosin 15a ( Fang et al . , 2015 ) , Eps8L2 ( Furness et al . , 2013 ) and ESPNL ( Ebrahim et al . , 2016 ) —are expressed predominantly at the tips of transducing stereocilia in mammalian auditory hair cells . Knockout or mutant mice with functional deficiencies in these proteins exhibit selective disassembly of stereocilia in the shortest but not tallest rows of the auditory hair cell bundles ( Furness et al . , 2013; Fang et al . , 2015; Ebrahim et al . , 2016 ) . Thus , these molecules are also good candidates for the molecular machinery involved in the MET-dependent stereocilia remodeling observed in our study . Finally , the actin-severing proteins AIP1 and ADF are thought to be responsible for actin disassembly at the stereocilia tips and for balancing of the continuous incorporation of new actin monomers to the tips ( Narayanan et al . , 2015 ) . Mice lacking ADF or expressing a mutant AIP1 exhibit defects in the hair bundle morphology ( Narayanan et al . 2015 ) that are very similar to the ones observed after the blockage of MET channels in our experiments . However , it is yet unknown whether AIP1 and ADF deficiencies influence the stereocilia actin core directly or secondarily to the loss of MET current . Independent of the molecules involved , our data demonstrate a functional link between the resting MET current and stereocilia remodeling . After the initial report on the incorporation of exogenous β-actin into the stereocilia of young postnatal rats ( Schneider et al . , 2002 ) , it was hypothesized that the stereocilia actin core is maintained through the continuous treadmill of actin ( Rzadzinska et al . , 2004 ) . A similar relatively fast actin remodeling was demonstrated in zebrafish stereocilia but without evidence of treadmilling ( Hwang et al . , 2015 ) . On the other hand , several independent groups have now established that in adult and young mammalian and non-mammalian hair cells , the active actin remodeling occurs only in a small ( ~0 . 5 µm ) region at the tips of stereocilia but not along their shafts ( Zhang et al . , 2012; Drummond et al . , 2015; Narayanan et al . , 2015 ) . Our data reconcile these different points of view on the stability of stereocilia actin core , at least in the transducing stereocilia of mammalian hair cells . Apparently , in the presence of a normal resting current through the MET channels , actin remodeling is limited to the tips of stereocilia . However , when the Ca2+ concentration inside the stereocilia changes significantly after blocking or unblocking the MET channels , the equilibrium is shifted and stereocilia start to retract or re-grow respectively , expanding the area of active actin remodeling . Interestingly , the areas of incorporation of fluorescently labelled actin to the tips of stereocilia in the second row of IHCs are larger than that in the tallest row stereocilia and vary significantly between individual stereocilia ( Narayanan et al . , 2015 ) , similarly to the variations of MET-dependent Ca2+ influx into these stereocilia ( Beurg et al . , 2009 ) . Furthermore , deletions in the genes encoding currently known components of the MET machinery—TMC1/TMC2 ( see Supplemental Figure 4B in [Kawashima et al . , 2011] ) , TMHS ( see Fig . 1D in [Xiong et al . , 2012] ) , TMIE ( see Fig . 5A-B in [Zhao et al . , 2014] ) —result in the loss of MET current and changes of the hair bundle morphology that seem to be limited to the shortest but not tallest row stereocilia in the auditory hair cells . These abnormalities are very similar to the changes that we observed in this study after blocking MET channels . Thus , we believe that the changes of stereocilia bundle morphology in these mouse mutants are likely initiated by the loss of the MET current . What is the physiological significance of the MET-dependent stereocilia remodeling ? It is unlikely to drive the initial formation of the hair bundle due to the relatively low amplitude of the MET current in the first postnatal days ( Waguespack et al . , 2007; Lelli et al . , 2009 ) and the rather normal stereocilia bundle formation in mutant mice lacking MET currents ( Kawashima et al . , 2011; Xiong et al . , 2012; Zhao et al . , 2014 ) . However , this mechanism may be essential for maintenance and/or fine tuning of the staircase shape of the mature hair bundle after substantial MET current has developed . Additionally , our data may represent an exaggerated manifestation of stereocilia remodeling at the stereocilia tips that operates at a faster time scale as compared to the overall changes of stereocilia height . If this is the case , then the MET-dependent remodeling of stereocilia tips may also contribute to processes such as the dynamic regulation of tip link tension . The dynamic control of stereocilia remodeling by the Ca2+ influx through the MET channels works in parallel with any other mechanisms that are responsible for tensioning the tip links , such as the operation of myosin-based molecular motors ( reviewed in [Gillespie and Cyr , 2004] ) . Our data indicate that , even after the 24 hr incubation with MET channel blockers and the significant shortening of the transducing stereocilia , the resting tension in the transduction machinery is still present , which allows FM1-43 accumulation inside the OHCs immediately after blocker washout ( Figure 1 ) . Furthermore , this tension is likely to be essential for the recovery of the MET current and the stereocilia re-growth after washing out the MET blockers ( Figure 3 ) . In fact , it is tempting to speculate that the upward force of myosin motors may eventually determine the exact staircase architecture of a stereocilia bundle , which represents one of the most enigmatic problems in hair cell biology . The height of a transducing stereocilium may be set by a delicate equilibrium between the modulatory influence of the Ca2+ influx on actin assembly/disassembly and the rate of delivery of essential molecules to the tip , which is likely to be inversely proportional to the height of a stereocilium . By setting a certain tension of the tip link and MET current at rest , the myosin motors may set this equilibrium at a precise stereocilium height and determine , for example , the final heights of the second- and third-row stereocilia when they re-grow after the washout of MET blockers ( Figure 3 ) . Independent of whether these speculations are true or not , the MET-dependent stereocilia remodeling demonstrated in our study is likely to represent an important mechanism for maintenance and repair of the hair bundles in non-regenerating mammalian auditory hair cells .
Organ of Corti explants were isolated from C57BL/6 ( RRID:IMSR_JAX:000664 , Jackson Laboratories , Bar Harbor , ME ) or CD1 ( Charles River Laboratories , Wilmington , MA ) wild-type mice ( both male and female ) at postnatal days 4 ( P4 ) through P6 , or from Sprague-Dawley rats ( RRID:RGD_734476 , Charles River Laboratories ) at P6 . The explants were held by two flexible glass fibers ( ~1–2 cm in length ) glued to the bottom of plastic Petri dishes ( Electron Microscopy Sciences ) using the silicone elastomer Sylgard ( World Precision Instruments ) . Explants were cultured at 37°C and 5% CO2 in DMEM ( Invitrogen , Carlsbad , CA ) either alone or supplemented with 7% fetal bovine serum ( FBS , Atlanta Biologicals , Flowery Branch , GA ) and 10 μg/mL ampicillin ( Calbiochem , San Diego , CA ) . Each experiment was typically performed for at least two to three independent series . Each series consisted of comparisons between tissue samples ( organs of Corti ) from littermates , including comparisons between the two ears of the same animal ( i . e . control vs . treated ) . All animal procedures were approved by the Institutional Animal Care and Use Committee ( IACUC ) at the University of Kentucky ( protocol 00903M2005 ) . Freshly isolated organ of Corti explants were incubated for 30 s in ice-cold Ca2+-containing standard Hank's balanced salt solution ( HBSS , catalog number 14025 , Invitrogen ) supplemented with 6 μM of FM1-43FX in the absence or presence of various blockers of MET channels ( see below ) . Then , the explants were rinsed thoroughly with cold Ca2+-containing HBSS and fixed in 4% paraformaldehyde ( PFA ) solution for 30 min . Immediately after fixation , the samples were rinsed with HBSS and imaged . For the recovery experiments ( Figure 1B ) , the organ of Corti explants were cultured for 24 hr in the presence of benzamil ( 30 μM ) and exposed to FM1-43 for 30 s either before or after drug washout . Samples were rinsed thoroughly with HBSS and imaged immediately after . Imaging was performed using an upright Olympus BX51WI microscope equipped with a 40X ( 0 . 8 NA ) LUMPlanFL water-immersion objective and spinning disc confocal attachment ( DSU ) . In all experiments , the osmolarity of HBSS was adjusted to 310 mOsm with ~20 mM of D-glucose . Freshly isolated organ of Corti explants were incubated ( cultured ) in FBS/ampicillin-supplemented DMEM at 37°C and 5% CO2 for 5–32 hr with or without the following MET channel blockers: benzamil ( 5 . 5 , 10 or 30 μM ) , amiloride ( 100 μM ) , ruthenium red ( 10 μM ) and tubocurarine ( 30 μM ) ( Sigma-Aldrich , St . Louis , MO ) . The maximum concentrations used were chosen based on previously reported dose-response curves in order to block 75–90% of the MET current ( Rüsch et al . , 1994; Farris et al . , 2004 ) and to minimize potential deleterious effects of these drugs during long-term incubations . The control explants were incubated in parallel in the same medium but without drugs , except for the vehicle control for benzamil that included 0 . 05% of DMSO ( Molecular Probes , Eugene , OR ) . The samples treated for 5 hr were briefly rinsed with HBSS and fixed immediately after incubation . After long incubations for 24–32 hr , the samples were first placed in HBSS ( with or without the same MET blocker that was used in the experiment ) and observed with an upright microscope ( E600FN , Nikon ) . The fibrous material from tectorial membrane outgrowth was gently removed with a ~2–4 µm suction pipette mounted on a micromanipulator ( MHW-3 , Narishige , Tokyo , Japan ) . Then , the explants were fixed for electron microscopy . Freshly isolated mouse organ of Corti explants were first rinsed with standard HBSS . Next , the explants were incubated for 15 min at room temperature in Ca2+-free HBSS ( catalog number 14175 , Invitrogen ) supplemented with 5 mM of Ca2+ chelator , 1 , 2-bis ( o-aminophenoxy ) ethane-N , N , N' , N'-tetraacetic acid , BAPTA ( Sigma-Aldrich ) and 0 . 5 mM or 0 . 9 mM of Mg2+ . After incubation , the explants were rinsed with the standard Ca2+-containing HBSS and allowed to recover in FBS/ampicillin-supplemented DMEM at 37°C and 5% CO2 for different periods of time up to 6 hr . At the end of the recovery period or immediately after incubation with BAPTA , the explants were fixed for electron microscopy ( see below ) . The membrane-permeable Ca2+ chelator BAPTA-AM ( Molecular Probes ) was pre-mixed with a 20% Pluronic F-127 solution in DMSO ( Molecular Probes ) . Freshly isolated P4 mouse organ of Corti explants were incubated in FBS/ampicillin-supplemented DMEM at 37°C and 5% CO2 for 1 to 5 hr in the presence of 20 μM BAPTA-AM ( in Pluronic/DMSO ) or in vehicle control conditions ( 0 . 1% of the Pluronic/DMSO solution ) . At the end of the incubation , the explants were rinsed with standard HBSS and placed in cold fixative . Freshly isolated P4 mouse organ of Corti explants were fixed after 1 hr incubation at 37°C and 5% CO2 . Longer incubations , especially with high extracellular Ca2+ , were found to be deleterious to the hair cells . Some explants were incubated in DMEM alone ( ~1 . 8 mM Ca2+ ) or in DMEM supplemented with 1 . 7 mM BAPTA to lower the free Ca2+ concentration ( ~100 μM Ca2+ ) . Other explants were incubated in FBS/ampicillin-supplemented DMEM , alone ( ~1 . 85 mM Ca2+ ) or supplemented with 5 mM CaCl2 ( ~6 . 85 mM Ca2+ ) or 10 mM CaCl2 ( ~11 . 85 mM Ca2+ ) . Organ of Corti explants were fixed for at least 2 hr in a mixture of 3% PFA and 3% glutaraldehyde in 0 . 1 M sodium cacodylate buffer , pH 7 . 4 ( Electron Microscopy Sciences , Hatfield , PA ) supplemented with 2 mM of CaCl2 ( Sigma-Aldrich ) . The samples were rinsed with distilled water , dehydrated through a graded series of ethanol , critical point dried from liquid CO2 ( EMS 850 , Electron Microscopy Sciences ) , sputter-coated with 5-nm platinum ( Q150T , Quorum Technologies , Guelph , Canada ) , and imaged with a field-emission scanning electron microscope ( Helios Nanolab 660 , FEI , Hillsboro , OR ) . To avoid damage of the sample with the electron beam , imaging was performed at a small working distance ( ~4 mm ) , which improved signal-to-noise ratio and allowed the use of smaller apertures . To accurately quantify the height of individual stereocilia in the different rows of a hair bundle , we obtained images of the same bundle from different angles , including views from the lateral ( ‘back’ ) and medial ( ‘front’ ) sides of the bundle . Figure 1—figure supplement 3 describes the methods of quantification . Samples with any signs of failed SEM preparation were discarded . These signs include: ( i ) fused or curved stereocilia due to encountering surface tension; ( ii ) lack of tip links in the untreated control samples; and ( iii ) mounting errors that would not allow imaging of the same bundle from the ‘front’ and ‘back’ sides . Measurements were performed by an examiner blind to the experimental conditions using ImageJ ( RRID:SCR_003070 ) . A ‘tip’ link was defined as a link that extends obliquely from the top of a lower row stereocilium to the side of a taller stereocilium in the direction of mechanosensitivity of the bundle . Any other link originating at the hemisphere of the tip of a shorter stereocilium was not considered as a tip link . See ( Indzhykulian et al . , 2013 ) for more details . Organ of Corti explants were fixed overnight in 3% glutaraldehyde in 0 . 15 M sodium cacodylate buffer ( pH 7 . 4 ) at 4°C , rinsed thoroughly with sodium cacodylate buffer , and post-fixed for 1 hr in 1% tannic acid . Then , the samples were rinsed with distilled water and cryoprotected by overnight incubations in 5 , 10 and 30% glycerol solutions . The explants were then placed on 3 mm copper grids and plunge frozen in liquid Freon before being transferred to 1% uranyl acetate in methanol at −90°C in a Leica EM AFS2 freeze substitution machine . Methanol was exchanged for Lowicryl HM-20 resin and polymerized by long-wave UV radiation . All reagents were obtained from Electron Microscopy Sciences . The resin blocks were trimmed on an ultramicrotome ( UC6 , Leica , Wetzlar , Germany ) and then milled with a focused ion beam and imaged in ‘Slice and View’ mode with a backscattered electron detector using the FEI Helios 660 Nanolab system . | Our sense of hearing depends on cells known as hair cells that line the inner ear . Each hair cell has tiny projections called stereocilia , which are arranged in a bundle with rows of increasing height like a staircase and are connected to each other by tiny filaments called tip-links . When sound waves hit the stereocilia , the tension on the tip-links increases , which opens “mechanotransduction” channels on the shorter stereocilia that allow calcium ions to flow into the cells . To ensure that the ears can detect even the softest sounds , the tip-links always have a small amount of tension which allows a small , but continuous flow of calcium ions into the cell . Scientists generally consider this continuous flow of calcium ions as a potentially harmful byproduct of sensitive hearing . Vélez-Ortega et al . isolated inner ear tissues from young mice and rats and exposed them to drugs that either block the flow of calcium ions through the mechanotransduction channels or break the tip-links on stereocilia . Surprisingly , these drugs made profound changes in the shape of individual stereocilia and the staircase architecture of the stereocilia bundle . When the drugs were rinsed out of the hair cells , the stereocilia went back to their normal shape . Sequestering of free calcium ions inside the hair cells had a similar effect on the shape of stereocilia . These findings show that the flow of calcium ions into the sterocilia via mechanotransduction channels controls the exquisite staircase-like architecture of the stereocilia bundle . More research is needed to identify which structural proteins cause the stereocilia shape changes and to work out exactly how calcium ions are involved . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2017 | Mechanotransduction current is essential for stability of the transducing stereocilia in mammalian auditory hair cells |
Schnyder corneal dystrophy ( SCD ) is an autosomal dominant disorder in humans characterized by abnormal accumulation of cholesterol in the cornea . SCD-associated mutations have been identified in the gene encoding UBIAD1 , a prenyltransferase that synthesizes vitamin K2 . Here , we show that sterols stimulate binding of UBIAD1 to the cholesterol biosynthetic enzyme HMG CoA reductase , which is subject to sterol-accelerated , endoplasmic reticulum ( ER ) -associated degradation augmented by the nonsterol isoprenoid geranylgeraniol through an unknown mechanism . Geranylgeraniol inhibits binding of UBIAD1 to reductase , allowing its degradation and promoting transport of UBIAD1 from the ER to the Golgi . CRISPR-CAS9-mediated knockout of UBIAD1 relieves the geranylgeraniol requirement for reductase degradation . SCD-associated mutations in UBIAD1 block its displacement from reductase in the presence of geranylgeraniol , thereby preventing degradation of reductase . The current results identify UBIAD1 as the elusive target of geranylgeraniol in reductase degradation , the inhibition of which may contribute to accumulation of cholesterol in SCD .
SCD ( Schnyder corneal dystrophy ) is a rare autosomal dominant eye disease that is characterized by bilateral opacification of the cornea ( Klintworth , 2009; Weiss , 2009 ) . The clinical manifestation of SCD can be apparent early in life , usually within the first decade . Thereafter , opacification of the cornea progresses slowly and ultimately leads to reduced visual acuity , which is postulated to be caused by light scattering ( Weiss , 2009 ) . The significance of this visual impairment is highlighted by the frequency in which corneal transplant surgery is utilized in treatment of SCD; approximately 50% of SCD patients ≥50 years of age undergo corneal transplant surgery for restoration of normal vision acuity ( Weiss , 2007 ) . The frequency of the procedure increases to greater than 70% for SCD patients 70 years and older . Analyses of corneas removed from SCD patients during transplantation surgery revealed a marked accumulation of unesterified cholesterol and lesser accumulation of esterified cholesterol and phospholipids ( McCarthy et al . , 1994; Gaynor et al . , 1996; Yamada et al . , 1998 ) , suggesting dysregulation of cholesterol metabolism may underlie pathogenesis of the disease . Systemic dyslipidemia has been reported to be associated with some , but not all cases of SCD ( Thiel et al . , 1977; Brownstein et al . , 1991; Crispin , 2002 ) . In 2007 , two groups independently identified SCD-associated mutations in the gene encoding UBIAD1 ( UbiA prenyltransferase domain-containing protein-1 ) ( Orr et al . , 2007; Weiss et al . , 2007 ) . UBIAD1 ( also known as TERE1 ) was first described as being absent or markedly diminished in bladder and prostate tumors ( McGarvey et al . , 2001 , 2003 ) and belongs to the UbiA superfamily of prenyltransferases ( Heide , 2009 ) . These enzymes , which are found in a wide variety of species , contain 8–10 transmembrane helices and catalyze transfer of isoprenyl groups to aromatic acceptors , producing a diverse range of molecules including ubiquinone , chlorophylls , hemes , vitamin E , and vitamin K ( Forsgren et al . , 2004; Nakagawa et al . , 2010; Bonitz et al . , 2011 ) . Indeed , UBIAD1 catalyzes transfer of the 20-carbon geranylgeranyl group from geranylgeranyl pyrophosphate to menadione ( vitamin K3 ) derived from plant-derived phylloquinone ( vitamin K1 ) , generating MK-4 ( menaquinone-4 , vitamin K2 ) ( Figure 1 ) ( Nakagawa et al . , 2010; Hirota et al . , 2013 ) . It has also been proposed that UBIAD1 mediates polyprenylation of 4-hydroxybenzoate to produce 3-polyprenyl-4-hydroxybenzoate ( PPHB ) , an intermediate in the synthesis of CoQ10 ( coenzyme Q10 or ubiquinone-10 ) ( Mugoni et al . , 2013 ) . 10 . 7554/eLife . 05560 . 003Figure 1 . Biosynthesis of cholesterol and menaquinone-4 ( MK-4 , vitamin K2 ) in mammalian cells . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 003 To date , 24 UBIAD1 mutations have been identified in ∼50 SCD families ( Nickerson et al . , 2013; Nowinska et al . , 2014 ) . Several of these mutations alter amino acid residues that localize to the active site of the UBIAD1 prenyltransferase domain as determined by molecular modeling using structural models of bacterial and archaeal UbiA prenyltransferases ( Bräuer et al . , 2004 , 2008; Nickerson et al . , 2010; Cheng and Li , 2014; Huang et al . , 2014 ) . Cells from patients harboring four SCD-associated mutations ( N102S , D112N , and G177E/R ) exhibited reduced MK-4 biosynthetic activity ( Nickerson et al . , 2013 ) . A possible link between UBIAD1 and cholesterol metabolism was provided by co-immunoprecipitation studies that showed an association of UBIAD1 with the cholesterol biosynthetic enzyme HMG CoA ( 3-hydroxy-3-methylglutaryl coenzyme A ) reductase ( Nickerson et al . , 2013 ) . However , the relevance of this association to regulation of cholesterol metabolism and pathogenesis of SCD is not clear . In mammalian cells , the ER ( endoplasmic reticulum ) -localized reductase catalyzes reduction of HMG CoA to mevalonate ( Figure 1 ) . This reaction constitutes a rate-limiting step in synthesis of cholesterol as well as nonsterol isoprenoids such as the farnesyl and geranylgeranyl groups that are transferred to many cellular proteins and utilized in synthesis of ubiquinone , heme , and dolichol ( Goldstein and Brown , 1990; Casey and Seabra , 1996 ) . Multiple feedback mechanisms converge on reductase to ensure a constant supply of nonsterol isoprenoids , while avoiding overaccumulation of cholesterol ( Brown and Goldstein , 1980 ) . This feedback system is exploited therapeutically by a group of competitive inhibitors of reductase called statins , which trigger regulatory responses that result in the lowering of plasma levels of LDL ( low-density lipoprotein ) -cholesterol and thereby reduce the incidence of cardiovascular disease ( Stossel , 2008; Marais et al . , 2014 ) . Statins also block production of sterol and nonsterol isoprenoids that govern feedback regulation of reductase ( Goldstein and Brown , 1990 ) . Thus , a marked accumulation of reductase occurs in livers of statin-treated animals and humans ( Brown and Goldstein , 1980; Kita et al . , 1980; Reihner et al . , 1990 ) , which can blunt the cholesterol-lowering effects of these drugs . This accumulation results in part , from slowed ERAD ( ER-associated degradation ) of reductase ( Nakanishi et al . , 1988; Inoue et al . , 1991; Ravid et al . , 2000 ) . The ERAD of reductase is initiated by intracellular accumulation of sterols , which causes the enzyme to bind to ER membrane proteins called Insig-1 and Insig-2 ( Sever et al . , 2003a , 2003b ) ( Figure 2 ) . Insig binding occurs through the membrane domain of reductase , which contains eight membrane-spanning helices and precedes a large cytosolic domain with enzymatic activity ( Liscum et al . , 1985; Roitelman et al . , 1992 ) . Insig-associated ubiquitin ligases gp78 and Trc8 facilitate ubiquitination of cytosolically exposed lysine residues in the membrane domain of reductase ( Sever et al . , 2003a; Song et al . , 2005; Jo et al . , 2011; Liu et al . , 2012 ) . This ubiquitination marks reductase for recognition by the AAA ( ATPases associated with diverse cellular activities ) -ATPase VCP/p97 , which mediates extraction of reductase across ER membranes ( Morris et al . , 2014 ) . Once extracted , ubiquitinated reductase is then released into the cytosol and delivered into the proteolytic chamber of the 20S proteasome through reactions mediated by the proteasome 19S regulatory particle , which contains six AAA-ATPases ( Ehlinger and Walters , 2013 ) . Geranylgeraniol , the alcohol derivative of geranylgeranyl pyrophosphate , augments sterol-accelerated ERAD of reductase but does not appreciably affect sterol-induced ubiquitination ( Sever et al . , 2003a ) . This observation led to the notion that geranylgeraniol augments post-ubiquitination steps in reductase ERAD ( Figure 2 ) . Indeed , geranylgeraniol enhances sterol-induced membrane extraction and cytosolic dislocation of reductase as judged by assays carried out in vitro and in intact cells ( Song and DeBose-Boyd , 2004; Elsabrouty et al . , 2013; Morris et al . , 2014 ) . 10 . 7554/eLife . 05560 . 004Figure 2 . Insig-mediated , sterol-accelerated degradation of HMG CoA reductase in mammalian cells . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 004 In the current studies , we use proximity-dependent biotinylation ( Roux et al . , 2012 ) to identify UBIAD1 as an associated protein of the membrane domain of reductase , which is necessary and sufficient for sterol-accelerated ERAD ( Gil et al . , 1985; Skalnik et al . , 1988 ) . Endogenous UBIAD1 binds to reductase in sterol-treated cells as determined by co-immunoprecipitation . This sterol-induced binding is inhibited by geranylgeraniol , releasing UBIAD1 for translocation from membranes of the ER to the Golgi . CRISPR/Cas9-mediated knockout and RNA interference ( RNAi ) -mediated knockdown of UBIAD1 relieves the geranylgeraniol requirement for reductase ERAD . UBIAD1 harboring the SCD-associated N102S mutation resists geranylgeraniol-mediated displacement from reductase and remains associated with the protein in the ER , thereby blocking sterol-accelerated degradation of reductase . These studies identify UBIAD1 as the elusive target of geranylgeraniol in the ERAD of reductase and provide insight into mechanisms through which the nonsterol isoprenoid modulates the reaction . Moreover , our results suggest that slowed ERAD of reductase may contribute to the accumulation of cholesterol observed in SCD .
To identify proteins that associate with reductase , we employed a novel chimeric protein composed of the membrane domain of reductase ( amino acids 1–346 ) fused to the promiscuous biotin ligase BirA* , which features proximity-dependent biotinylation of the chimera's near-neighbor proteins ( Roux et al . , 2012 ) . Cells stably expressing high levels of the reductase-BirA* chimera were subjected to treatment with the oxysterol 25-HC ( 25-hydroxycholesterol ) in the presence of biotin . Following treatments , cells were harvested and lysed; the lysates were then incubated with streptavidin-conjugated agarose beads to capture biotinylated proteins that were subsequently identified by tandem mass spectrometry . Figure 3—figure supplement 1B shows an experiment in which we used co-immunoprecipitation to measure the interaction of endogenous reductase with several proteins that were identified as associated with the reductase-BirA* chimera ( Figure 3—figure supplement 1A ) . SV-589 cells , a line of transformed human fibroblasts ( Yamamoto et al . , 1984 ) , were depleted of sterols and subsequently treated in the absence or presence of 25-HC . Following treatments , the cells were harvested for lysis in detergent-containing buffer; the resulting lysates were then immunoprecipitated with polyclonal antibodies against reductase . Immunoblot analysis revealed that treatment of cells with 25-HC triggered co-immunoprecipitation of endogenous reductase with Insig-1 ( Figure 3—figure supplement 1B , lanes 3 and 4 ) ; the control ER membrane protein calnexin failed to interact with reductase , regardless of the absence or presence of 25-HC ( lanes 5 and 6 ) . ERGIC-53 similarly failed to appear in the anti-reductase immunoprecipitates ( lanes 9 and 10 ) . In contrast , lamina-associated peptide-2 ( lanes 7 and 8 ) , peroxiredoxin-4 ( lanes 11 and 12 ) , PGRMC2 ( lanes 13 and 14 ) , and annexin-A1 ( lanes 15 and 16 ) co-precipitated with reductase in both the absence and presence of 25-HC; the significance and specificity of these interactions are unclear . In the absence of 25-HC , a small amount of UBIAD1 ( UbiA prenyltransferase domain-containing protein-1 ) appeared in the pellet fraction of the reductase immunoprecipitation ( lane 17 ) ; this appearance was markedly enhanced when the cells were subjected to treatment with 25-HC ( lane 18 ) . UBIAD1 belongs to the UbiA superfamily of prenyltransferases ( Heide , 2009; Bonitz et al . , 2011 ) . Similar to other UbiA prenyltransferases , UBIAD1 is very hydrophobic; the human enzyme is comprised of 338 amino acids and is predicted to contain at least eight transmembrane helices ( Figure 3A ) . Besides its regulated association with reductase ( this study and Nickerson et al . , 2013 ) , we chose UBIAD1 for further examination owing to the observation that mutations in the UBIAD1 gene are associated with SCD , which may result from dysregulated cholesterol metabolism ( McCarthy et al . , 1994; Gaynor et al . , 1996; Yamada et al . , 1998 ) . The experiment of Figure 3B shows a time course of 25-HC-mediated association of UBIAD1 with reductase as determined by co-immunoprecipitation . Immunoblot analysis of anti-reductase immunoprecipitates revealed co-precipitation of endogenous UBIAD1 with reductase following treatment of cells for 30 min with 25-HC ( Figure 3B , top panel , lane 6 ) . This sterol-induced co-precipitation continued after prolonged incubation ( top panel , lanes 7–14 ) , even though the total amount of reductase was reduced by accelerated degradation ( second and fourth panels , lanes 12 and 14 ) . The structurally related 1 , 1-bisphosphonate esters SR-12813 and Apomine mimic 25-HC in accelerating reductase ubiquitination and degradation ( Roitelman et al . , 2004; Sever et al . , 2004; Nguyen et al . , 2009 ) . As shown in Figure 3C , Apomine triggered co-precipitation between reductase and UBIAD1 in a manner similar to that of 25-HC ( second panel , compare lane 2 with lanes 3–5 ) . 10 . 7554/eLife . 05560 . 005Figure 3 . Identification of UBIAD1 as an associated protein of HMG CoA reductase . ( A ) Amino acid sequence and predicted topology of human UBIAD1 . Amino acid residues mutated in Schnyder corneal dystrophy ( SCD ) families are shaded in red; asparagine-102 ( N102 ) and glycine-177 ( G177 ) , the two most frequently altered UBIAD1 residues in SCD , are indicated by arrows . The predicted active site in the prenyltransferase domain of UBIAD1 is indicated by the green dotted line . ( B and C ) SV-589 cells were set up for experiments on day 0 at a density of 2 × 105 cells per 100-mm dish in medium A supplemented with 10% FCS . On day 3 , cells were switched to medium A containing 10% NC-LPPS , 10 µM sodium compactin , and 50 µM sodium mevalonate to deplete sterols . ( B ) After 16 hr at 37°C , the cells received sterol-depleting medium in the absence or presence of 1 µg/ml 25-HC and were further incubated for the indicated period of time 37°C . Cells were then harvested , lysed in PBS containing 1% digitonin , and the resulting lysates were subjected to immunoprecipitation with polyclonal anti-reductase antibodies as described in ‘Materials and methods’ . Aliquots of precipitated material ( IP Pellet ) and lysates were subjected to SDS-PAGE and immunoblot analysis was carried out with IgG-H8 ( against UBIAD1 ) , IgG-A9 ( against reductase ) , IgG-17H1 ( against Insig-1 ) , and anti-calnexin IgG . ( C ) Following sterol depletion , cells were incubated for 45 min at 37°C in sterol-depleting medium containing the indicated concentration of Apomine . Cells were then harvested , lysed , and immunoprecipitated with polyclonal anti-reductase; aliquots of precipitated material and lysates were analyzed by SDS-PAGE , followed by immunoblot as described in ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 00510 . 7554/eLife . 05560 . 006Figure 3—figure supplement 1 . Identification of proteins associated with HMG-Red ( TM1-8 ) -BirA* . ( A ) HEK-293S/pHMG-Red ( TM1-8 ) -BirA* cells were set up on day 0 , depleted of sterols on day 3 , harvested on day 4 for lysis and affinity purification using streptavidin-coupled beads as described in ‘Materials and methods’ . Precipitated proteins were subjected to SDS-PAGE and the gel was subjected to staining with colloidal blue . Three segments of the gel ( indicated by boxes ) were excised and the identities of the proteins were determined by tandem mass spectrometry . The spectral count for the most abundant proteins identified in each segment is indicated in parentheses . ( B ) SV-589 cells were set up on day 0 , depleted of sterols on day 3 , and subjected to treatment in the absence or presence of 25-HC ( 1 µg/ml ) for 45 min at 37°C on day 4 as described in the legend to Figure 3 . Cells were then harvested , lysed in PBS containing 1% digitonin , and the resulting lysates were subjected to immunoprecipitation with polyclonal anti-reductase antibodies as described in ‘Materials and methods’ . Aliquots of precipitated material ( IP Pellet ) and lysates were subjected to SDS-PAGE and immunoblot analysis was carried out with antibodies against the indicated protein . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 006 We next used RNA interference ( RNAi ) to examine a role for Insigs in the 25-HC-induced binding of reductase to UBIAD1 . SV-589 cells were transfected with duplexes of small interfering RNAs ( siRNAs ) against mRNAs encoding green fluorescent protein ( GFP ) , which is not expressed in the cells , or Insig-1 and Insig-2 . Following transfection , the cells were depleted of sterols , incubated in the absence or presence of 25-HC , and lysed for subsequent immunoprecipitation with polyclonal anti-reductase . Immunoblotting of precipitated proteins revealed that 25-HC stimulated co-precipitation of both UBIAD1 and Insig-1 with reductase in SV-589 cells transfected with control GFP siRNA ( Figure 4A , second and third panels , lane 2 ) . The sterol-induced co-precipitation of UBIAD1 with reductase was markedly reduced in cells that received siRNAs against Insig-1 and Insig-2 ( second panel , lanes 3 and 4 ) . The experiment of Figure 4B shows that knockdown of UBIAD1 reduced , but did not eliminate the sterol-induced binding of reductase to Insig-1 ( third panel , compare lanes 2 and 4 ) . Finally , we examined the requirement of reductase for the association of UBIAD1 with Insig-1 . Immunoblot analysis of anti-UBIAD1 immunoprecipitates revealed that as expected , reductase and Insig-1 co-precipitated with UBIAD1 when control siRNA-transfected cells were treated with 25-HC ( Figure 4C , top and bottom panels , lane 2 ) . However , RNAi-mediated knockdown of reductase abolished co-precipitation of Insig-1 with UBIAD1 ( bottom panel , compare lane 4 with lane 2 ) . 10 . 7554/eLife . 05560 . 007Figure 4 . Specificity of sterol-dependent UBIAD1-HMG CoA reductase association . SV-589 cells were set up on day 0 at 1 × 105 cells per 100-mm dish in medium A containing 10% FCS . On day 3 , cells were transfected with siRNAs targeting mRNAs encoding GFP , Insig-1 and Insig-2 , UBIAD1 , or reductase as indicated and described in ‘Materials and methods’ . Cells transfected with siRNA duplexes against reductase received 200 mM mevalonate to provide essential nonsterol isoprenoids . On day 4 , cells were depleted of sterols through incubation for 16 hr at 37°C in medium A containing 10% NC-LPPS , 10 µM compactin , and 50 µM mevalonate . The cells then received identical medium in the absence or presence of 1 µg/ml 25-HC . After 45 min at 37°C , cells were harvested , lysed , and immunoprecipitated with polyclonal antibodies against either reductase ( A and B ) or UBIAD1 ( C ) . The resulting precipitated material and lysates were subjected to SDS-PAGE and immunoblot analysis with IgG-A9 ( against reductase ) , IgG-H8 ( against UBIAD1 ) , IgG-17H1 ( against Insig-1 ) , and anti-calnexin IgG . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 007 Results of Figure 3B show that UBIAD1 continued to appear in anti-reductase immunoprecipitates after prolonged treatment with 25-HC , even though total reductase was reduced by accelerated degradation . We reasoned that the proteasome inhibitor MG-132 would block this degradation , leading to increased co-immunoprecipitation of UBIAD1 with stabilized reductase . Immunoblot analysis of anti-reductase immunoprecipitates from lysates of cells treated in the absence of MG-132 shows that 25-HC stimulated reductase degradation ( Figure 5—figure supplement 1A , top panel , compare lanes 1 and 2 ) , which was blocked by MG-132 ( lanes 3 and 4 ) . Insig-1 co-precipitated with reductase in the presence of 25-HC ( second panel , lane 2 ) and this co-precipitation was enhanced by MG-132 ( lane 4 ) . UBIAD1 also co-precipitated with reductase in 25-HC-treated cells ( third panel , lane 2 ) ; however , the interaction was reduced , rather than enhanced , in the presence of MG-132 ( lane 4 ) . Taking into consideration results of Figure 5—figure supplement 1A together with the previous observation that geranylgeraniol augments post-ubiquitination steps in reductase ERAD ( Sever et al . , 2003a; Elsabrouty et al . , 2013 ) , we next sought to determine whether geranylgeraniol modulates UBIAD1-reductase binding . As expected , 25-HC dose-dependently stimulated co-precipitation of UBIAD1 with reductase ( Figure 5A , top panel , lanes a–c ) ; however , this sterol-regulated association was inhibited in the presence of 10 or 20 µM geranylgeraniol ( lanes d–i ) . Geranylgeraniol-mediated inhibition of reductase-UBIAD1 binding was also observed in the repeat experiment shown in Figure 5—figure supplement 1B . In contrast , sterol-induced binding of UBIAD1 to reductase continued in the presence of 20 µM farnesol ( Figure 5B , top panel , compare lanes a–d with lanes e–h ) , a 15-carbon nonsterol isoprenoid that does not augment sterol-accelerated ERAD of reductase ( Sever et al . , 2003a ) . 10 . 7554/eLife . 05560 . 008Figure 5 . The nonsterol isoprenoid geranylgeraniol inhibits sterol-induced binding of UBIAD1 to HMG CoA reductase and promotes its translocation to the Golgi . ( A and B ) SV-589 cells were set up for experiments on day 0 and depleted of sterols on day 3 as described in the legend to Figure 3 . Following sterol-depletion , cells received medium A containing 10% NC-LPPS , 10 µM compactin , 50 µM mevalonate with the indicated concentration of 25-HC in the absence or presence of 10 or 20 µM geranylgeraniol ( A ) or 20 µM farnesol ( B ) . Following incubation for 45 min at 37°C , cells were harvested , lysed , and immunoprecipitated with polyclonal antibodies against reductase . Aliquots of the resulting immunoprecipitates and lysates were subjected to immunoblot analysis with IgG-A9 ( against reductase ) , IgG-H8 ( against UBIAD1 ) , IgG-17H1 ( against Insig-1 ) , and anti-calnexin IgG . ( C ) SV-589 cells were set up on day 0 at 7 . 5 × 104 cells/well of six-well plates with glass coverslips in medium A containing 10% FCS . On day 1 , the cells were switched to identical medium or medium A containing 10% NC-LPPS , 10 µM compactin , and 50 µM mevalonate as indicated . Following incubation for 16 hr at 37°C , the cells were treated in the absence or presence of 30 µM geranylgeraniol ( GGOH ) or 1 µg/ml 25-HC for an additional 4 hr at 37°C . The cells were subsequently fixed for microscopy as described in ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 00810 . 7554/eLife . 05560 . 009Figure 5—figure supplement 1 . The proteasome inhibitor MG-132 and geranylgeraniol inhibit sterol-induced binding of UBIAD1 to HMG CoA reductase . SV-589 cells were set up for experiments on day 0 and depleted of sterols on day 3 as described in the legend to Figure 3 . Following sterol-depletion , cells received medium A containing 10% NC-LPPS , 10 µM compactin , 50 µM mevalonate in the absence or presence of 10 µM MG-132 for 1 hr at 37°C , followed by treatment in the absence or presence of 1 µg/ml 25-HC ( A ) or the indicated concentration of 25-HC in the absence or presence of 15 µM geranylgeraniol ( B ) . Following incubation for 45 min at 37°C , cells were harvested , lysed , and immunoprecipitated with polyclonal antibodies against reductase . Aliquots of the resulting immunoprecipitates and lysates were subjected to immunoblot analysis with IgG-A9 ( against reductase ) , IgG-H8 ( against UBIAD1 ) , IgG-17H1 ( against Insig-1 ) , and anti-calnexin IgG . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 00910 . 7554/eLife . 05560 . 010Figure 5—figure supplement 2 . Geranylgeraniol , but not 25-HC , FOH , or cholesterol , stimulates translocation of endogenous UBIAD1 to the Golgi in cells deprived of sterol and nonsterol isoprenoids . ( A ) SV-589 cells were set up on day 0 at 7 . 5 × 104 cells/well of six-well plates with glass coverslips in medium A containing 10% FCS . On day 1 , the cells were switched to identical medium or medium A containing 10% NC-LPPS , 10 µM compactin , and 50 µM mevalonate as indicated . Following incubation for 16 hr at 37°C , the cells were treated in the absence or presence of 30 µM geranylgeraniol ( GGOH ) , 30 µM farnesol ( FOH ) , 1 µg/ml 25-HC , or 10 µg/ml cholesterol for an additional 4 hr at 37°C . The cells were subsequently fixed and subjected to immunostaining as described in the legend to Figure 5 . Coverslips were mounted using Fluoromount G ( Electron Microscopy Sciences , Hatfield , PA ) . Images were obtained with a Zeiss Axio Observer Epifluorescence microscope using a 63× oil Plan-Apochromat objective and Ziess Axiocam color digital camera in black and white mode . For the purpose of presentation , brightness levels were adjusted across the entire image using ImageJ software ( National Institution of Health , USA ) . ( B ) SV-589 cells were set up on day 0 at 3 × 104 cells/well of a twelve-well plate with glass coverslips . On day 1 , the cells were transfected using FuGENE6 with 200 ng DsRed-Golgi ( Clontech Laboratories , Inc . , Mountain View , CA ) ; the total amount of DNA/well was adjusted to 500 ng by the addition of empty pcDNA3 . 1 vector . 4 hr after transfection , cells received a direct addition of medium A containing 10% FCS or 10% NC-LPPS supplemented with 10 µM compactin and 50 µM mevalonate ( final concentrations ) as indicated . Following incubation for 16 hr at 37°C , cells were treated for 4 hr in the absence or presence of 30 µM geranylgeraniol ( GGOH ) , 30 µM farnesol ( FOH ) , 1 µg/ml 25-HC , or 10 µg/ml cholesterol . Cells were subsequently fixed and subjected to immunostaining and imaging as described in ( A ) . Shown are cropped images representing a 64 × 64 micron-portion of the original images 219 × 174 microns in size . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 01010 . 7554/eLife . 05560 . 011Figure 5—figure supplement 3 . Geranylgeraniol , but not 25-HC or farnesol , stimulates translocation of transfected UBIAD1 to the Golgi in cells deprived of sterol and nonsterol isoprenoids . SV-589/pMyc-UBIAD1 cells , a line of SV-589 cells that stably express Myc-UBIAD1 , were generated as follows . SV-589 cells were set up on day 0 at a density of 7 × 105 cells per 100-mm dish in medium A supplemented with 10% FCS . On day 1 , cells were transfected with 2 µg/dish of pCMV-Myc-UBIAD1 using FuGENE6 transfection reagent as described in ‘Materials and methods’ . Following incubation for 16 hr at 37°C , cells were switched to medium A supplemented with 10% FCS and 700 µg/ml G418 . Fresh medium as added every 2–3 days until colonies formed after 2 weeks . Individual colonies were isolated using cloning cylinders , and expression of Myc-UBIAD1 was determined by immunoblot analysis . Select colonies were expanded and then further purified by serial dilution in 96-well plates . Individual clones were screened by immunofluorescense using IgG-9E10 against the Myc epitope . For experiments , SV-589/pMyc-UBIAD1 cells were set up on day 0 at 7 . 5 × 104 cells/well of six-well plates with glass coverslips in medium A containing 10% FCS . On day 1 , the cells were switched to identical medium or medium A containing 10% NC-LPPS , 10 µM compactin , and 50 µM mevalonate as indicated . Following incubation for 16 hr at 37°C , the cells were treated in the absence or presence of 30 µM geranylgeraniol , 30 µM farnesol , or 1 µg/ml 25-HC for an additional 4 hr at 37°C . The cells were subsequently fixed and subjected to immunostaining and analysis as described in the legend to Figure 5—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 011 Consistent with a role in the ERAD of reductase , UBIAD1 has been localized to membranes of the ER ( Nakagawa et al . , 2010; Nickerson et al . , 2013 ) . However , it should be noted that the prenyltransferase has also been localized to membranes of the mitochondria ( Nickerson et al . , 2010 ) as well as the Golgi ( Mugoni et al . , 2013; Wang et al . , 2013 ) . In the experiment shown in Figure 5C , we examined the subcellular localization of endogenous UBIAD1 in SV-589 cells grown in medium containing FCS ( fetal calf serum ) . The results show that endogenous UBIAD1 localized to juxtanuclear structures that resembled the Golgi and also stained with antibodies against the Golgi protein GM130 ( Nakamura et al . , 1995 ) ( Figure 5C , panels 1 and 2 ) . The Golgi localization of UBIAD1 was markedly diminished when the cells were depleted of sterol and nonsterol isoprenoids through incubation in medium containing LPPS ( lipoprotein-poor serum ) and the reductase inhibitor compactin ( Figure 5C , panels 3 and 4 ) . Notably , localization of Golgi-localized GM130 was unaffected by depletion of sterol and nonsterol isoprenoids ( compare panels 2 and 4 ) . The Golgi localization of UBIAD1 was restored in sterol- and nonsterol isoprenoid-depleted cells by the addition of geranylgeraniol ( Figure 5C , panel 5 ) , but not by the addition of 25-HC ( panel 7 ) . A repeat experiment confirms that sterol and nonsterol isoprenoid depletion disrupted the Golgi localization of UBIAD1 ( Figure 5—figure supplement 2 ) . Importantly , this localization was restored by geranylgeraniol , but not by farnesol , 25-HC , or cholesterol . We obtained similar results with SV-589 cells stably transfected with pCMV-Myc-UBIAD1 encoding Myc-tagged human UBIAD1 . Depletion of sterol and nonsterol isoprenoids disrupted Golgi localization of transfected Myc-UBIAD1 ( Figure 5—figure supplement 3 ) . This localization was restored by geranylgeraniol , but not by farnesol or 25-HC . Notably , endogenous UBIAD1 exhibited relatively more non-Golgi staining compared to Myc-UBIAD1 in FCS-cultured cells , suggesting anti-Myc may have a greater degree of specificity compared to that of anti-UBIAD1 . The role of UBIAD1 in sterol-accelerated reductase ERAD was next examined using RNAi-mediated knockdown and CRISPR/Cas9-mediated knockout ( Cong et al . , 2013; Mali et al . , 2013 ) . For RNAi experiments , SV-589 cells were transfected with siRNA duplexes targeting the GFP or UBIAD1 mRNAs and subsequently treated in the absence or presence of 25-HC and geranylgeraniol prior to subcellular fractionation . Figure 6A shows that 25-HC stimulated reductase degradation from membranes of control siRNA-transfected cells ( top panel , lane 2 ) ; this degradation was enhanced by geranylgeraniol ( lanes 3 and 4 ) . In UBIAD1 knockdown cells , 25-HC stimulated complete degradation of reductase , even in the absence of geranylgeraniol ( lanes 5–8 ) . Nearly identical results were obtained using an siRNA duplex targeting a different region of the UBIAD1 mRNA ( Figure 6—figure supplement 1 ) . Geranylgeraniol also augmented the ERAD of reductase that was stimulated by Apomine ( Figure 6B , top panel , lanes 1–4 ) ; however , the 1 , 1-bisphosphonate ester caused reductase to become completely degraded in UBIAD1 knockdown cells ( lanes 5–8 ) . Results consistent with those observed using RNAi were obtained in cells subjected to CRISPR/Cas9-mediated knockout of UBIAD1 . In wild type SV-589 cells , Apomine stimulated reductase ERAD through a mechanism augmented by geranylgeraniol ( Figure 6C , top panel , lanes 1–4 ) , whereas the compound caused complete degradation of reductase in cells deficient in UBIAD1 ( designated UBIAD1− ) ( lanes 5–8 ) . Figure 6D shows that in a time-dependent fashion , geranylgeraniol augmented Apomine-induced degradation of reductase in SV-589 cells ( compare lanes 1–6 with lanes 7–12 ) . However , Apomine alone caused complete degradation of reductase in UBIAD1− cells ( Figure 6D , lanes 13–18 ) and augmentation by geranylgeraniol was markedly diminished ( lanes 19–24 ) . Importantly , transfection of UBIAD1− cells with pCMV-Myc-UBIAD1 restored the requirement of geranylgeraniol for maximal degradation of reductase ( Figure 6E , top panel , compare lanes 1–3 with lanes 4–6 ) . 10 . 7554/eLife . 05560 . 012Figure 6 . RNA interference-mediated knockdown or CRISPR/Cas9-mediated knockout of UBIAD1 alleviates geranylgeraniol requirement in sterol-accelerated degradation of HMG CoA reductase . ( A and B ) SV-589 cells were set up for experiments on day 0 , transfected with the indicated siRNA duplexes on day 3 , and depleted of sterols as described in the legend to Figure 4 . The sterol-depleted cells were then treated with medium A containing 10% NC-LPPS , 10 µM compactin , and 50 µM mevalonate in the absence or presence of 1 µg/ml 25-HC and geranylgeraniol ( GGOH ) as indicated . Following incubation for 4 hr at 37°C , cells were harvested for subcellular fractionation . Aliquots of resulting membrane fractions ( 20 µg protein/lane ) were subjected to SDS-PAGE and immunoblot analysis was carried out with IgG-A9 ( against reductase ) , IgG-H8 ( against UBIAD1 ) , and anti-calnexin IgG . ( C–E ) SV-589 , UBIAD1− , UBIAD1−/pCDNA3 . 1 , and UBIAD1−/pMyc-UBIAD1 cells were set up for experiments on day 0 at 3 . 5 × 105 cells per 60-mm dish in medium A containing 10% FCS . On day 1 , cells were switched to medium A supplemented with 10% NC-LPPS , 10 µM compactin , and 50 µM mevalonate . Following incubation for 16 hr at 37°C , cells received the identical medium in the absence or presence of 10 µM Apomine and the indicated concentration ( C ) or 10 µM ( D and E ) of geranylgeraniol ( GGOH ) . After the indicated period of time ( D ) or 5 hr ( C and E ) at 37°C , cells were subjected to subcellular fractionation and membrane fractions ( 12 µg protein/lane ) were analyzed by immunoblot as described in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 01210 . 7554/eLife . 05560 . 013Figure 6—figure supplement 1 . RNA interference-mediated knockdown of UBAD1 alleviates requirement for geranylgeraniol in sterol-accelerated reductase degradation . SV-589 cells were set up for experiments on day 0 , transfected with the indicated siRNA duplexes on day 3 , and depleted of sterols as described in the legend to Figure 3 . Notably , the siRNA duplex targeting UBIAD1 ( 5′-UCUUGGAGCCGCAGGAUGUUU-3′ , Dharmacon/ThermoScientfic ) was distinct from that used in Figures 3 , 6 . The sterol-depleted cells were then treated with medium A containing 10% NC-LPPS , 10 µM compactin , and 50 µM mevalonate in the absence or presence of 1 µg/ml 25-HC and 20 µM geranylgeraniol ( GGOH ) . Following incubation for 4 hr at 37°C , cells were harvested for subcellular fractionation . Aliquots of resulting membrane fractions ( 20 µg protein/lane ) were subjected to SDS-PAGE and immunoblot analysis was carried out with IgG-A9 ( against reductase ) and IgG-H8 ( against UBIAD1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 013 To date , 24 mutations in the UBIAD1 gene have been identified in SCD families ( Nickerson et al . , 2013; Nowinska et al . , 2014 ) ; these mutations result in the alteration of 20 amino acid residues in UBAID1 ( Figure 3A ) . Asparagine-102 ( N102 ) , which localizes to the prenyltransferase domain of UBIAD1 ( Cheng and Li , 2014; Huang et al . , 2014 ) , is the most frequently altered residue ( ∼30% ) in SCD families . Considering that SCD is an autosomal dominant disorder , we reasoned that overexpression of UBIAD1 harboring the SCD-associated N102S mutation would block reductase ERAD . Figure 7A shows that when overexpressed in SV-589 cells , the T7-tagged membrane domain of reductase became degraded in the presence of 25-HC and geranylgeraniol ( top panel , lanes 1 and 2 ) . Co-expression of a plasmid encoding wild type UBIAD1 stabilized reductase in a dose-dependent fashion and 25-HC plus geranylgeraniol continued to accelerate its ERAD ( lanes 3–10 ) . Notably , the highest level of wild type UBIAD1 co-expressed with reductase blunted the enzyme's accelerated degradation ( lanes 11 and 12 ) . Co-expression of UBIAD1 ( N102S ) also stabilized the membrane domain of reductase ( Figure 7B , top panel , compare lane 1 with lanes 3 , 5 , 7 , 9 , and 11 ) , suggesting the protein continued to bind to the enzyme . However , UBIAD1 ( N102S ) blunted sterol-accelerated ERAD of reductase , even at low levels of expression ( Figure 7B , top panel , compare lane 2 with lanes 4 , 8 , 10 , and 12 ) . Similar results were obtained with T7-tagged , full-length reductase ( Figure 7—figure supplement 1 ) . The protein was subjected to sterol-accelerated degradation when transfected alone or together with up to 100 ng of plasmid encoding wild type UBIAD1 ( Figure 7—figure supplement 1A , top panel , lanes 1–8 ) ; co-transfection of higher levels ( 300 and 1000 ng ) of the UBIAD1-encoding plasmid inhibited reductase degradation ( lanes 9–12 ) . Consistent with results obtained with the reductase membrane domain , inhibition of full-length reductase degradation was observed upon co-transfection with a significantly lower amount ( 30 ng ) of plasmid encoding UBIAD1 ( N102S ) ( Figure 7—figure supplement 1B , top panel , compare lanes 1–4 with lanes 5–12 ) . Glycine-177 ( G177 , see Figure 3A ) is the second-most frequently altered UBIAD1 residue in SCD families ( Nickerson et al . , 2013 ) . The results of Figure 7C show that UBAID1 harboring the SCD-associated G177R mutation blocked reductase ERAD to a degree similar to that observed with UBIAD1 ( N102S ) ( top panel , compare lanes 1 and 2 with lanes 3–6 ) . 10 . 7554/eLife . 05560 . 014Figure 7 . The Schnyder corneal dystrophy ( SCD ) -associated N102S and G177R mutants of UBIAD1 block sterol-accelerated ERAD of HMG CoA reductase . SV-589 cells were set up for experiments on day 0 at 4 × 105 cells per 60-mm dish in medium A containing 10% FCS . On day 1 , cells were transfected with 3 µg/dish of pCMV-HMG-Red ( TM1-8 ) -T7 in the absence or presence of the indicated concentration of pCMV-Myc-UBIAD1 ( WT or N102S ) ( A and B ) or 3 µg/dish of pCMV-HMG-Red ( TM1-8 ) -T7 together with 30 ng of pCMV-Myc-UBIAD1 ( WT , N102S , or G177R ) ( C ) as described in ‘Materials and methods’ . 4 hr after transfection , cells received a direct addition of medium A containing 10% NC-LPPS , 10 µM compactin , and 50 µM mevalonate ( final concentrations ) . Following incubation for 16 hr at 37°C , cells were treated with identical medium in the absence or presence of 1 µg/ml 25-HC plus 20 µM geranylgeraniol ( GGOH ) as indicated . After 4 hr at 37°C , cells were harvested and subjected to subcellular fractionation . Aliquots of resulting membrane fractions were then subjected to SDS-PAGE and immunoblot analysis was carried out with anti-T7 IgG ( against reductase ) , IgG-9E10 ( against UBIAD1 and Insgi-1 ) , and anti-calnexin IgG . Proteins corresponding to reductase in ( A and B ) were quantified using ImageJ software . The intensities of these signals in the absence of 25-HC plus geranylgeraniol were arbitrarily set as 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 01410 . 7554/eLife . 05560 . 015Figure 7—figure supplement 1 . Schnyder corneal dystrophy ( SCD ) -associated N102S mutant of UBIAD1 blocks sterol-accelerated ERAD of full-length HMG CoA reductase . SV-589 cells were set up for experiments on day 0 at 4 × 105 cells per 60-mm dish in medium A containing 10% FCS . On day 1 , cells were transfected with 3 µg/dish of pCMV-HMG-Red-T7 in the absence or presence of the indicated concentration of wild type ( A ) or N102S ( B ) versions of pCMV-Myc-UBIAD1 as described in ‘Materials and methods’ . 4 hr after transfection , cells received a direct addition of medium A containing 10% NC-LPPS , 10 µM compactin , and 50 µM mevalonate ( final concentrations ) . Following incubation for 16 hr at 37°C , cells were treated with identical medium in the absence or presence of 1 µg/ml 25-HC plus 20 µM geranylgeraniol ( GGOH ) as indicated . After 4 hr at 37°C , cells were harvested and subjected to subcellular fractionation . Aliquots of resulting membrane fractions were then subjected to SDS-PAGE and immunoblot analysis was carried out with anti-T7 IgG ( against reductase ) , IgG-9E10 ( against UBIAD1 and Insgi-1 ) , and anti-calnexin IgG . Asterisks denote a non-specific cross-reactive band . Proteins corresponding to reductase in ( A and B ) were quantified using ImageJ software . The intensities of these signals in the absence of 25-HC plus geranylgeraniol were arbitrarily set as 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 015 We next compared the geranylgeraniol-mediated displacement of wild type and N102S UBIAD1 from reductase using co-immunoprecipitation . UBIAD1− cells stably transfected with empty mock vector , pCMV-Myc-UBIAD1 , or pCMV-Myc-UBIAD1 ( N102S ) were briefly incubated with 25-HC in the absence or presence of geranylgeraniol . The cells were then harvested , lysed , and subjected to anti-reductase immunoprecipitation . The results show that wild type UBIAD1 co-precipitated with reductase in the 25-HC-treated cells ( Figure 8A , second panel , lane e ) . This co-precipitation was inhibited by geranylgeraniol ( lanes g and h ) . UBIAD1 ( N102S ) similarly co-precipitated with reductase ( lane i ) ; however , the mutant UBAID1 resisted geranylgeraniol-mediated displacement and remained associated with reductase ( lanes j–l ) . UBIAD1 ( N102S ) exhibited a similar resistance to geranylgeraniol-mediated displacement from reductase in two independent experiments shown in Figure 8—figure supplement 1 . 10 . 7554/eLife . 05560 . 016Figure 8 . SCD-associated UBIAD1 mutant resists geranylgeraniol-mediated displacement from HMG CoA reductase and remains sequestered in ER membranes . ( A ) UBIAD1−/pCDNA3 . 1 , UBIAD1−/pMyc-UBIAD1 ( WT ) , and UBIAD1−/pMyc-UBIAD1 ( N102S ) cells were set up for experiments on day 0 at a density of 4 × 105 cells per 60-mm dish in medium A containing 10% FCS . On day 3 , cells were depleted of sterols as described in the legend to Figure 4 . After 16 hr at 37°C , cells received the identical medium containing 1 µg/ml 25-HC in the absence or presence of the indicated concentration of geranylgeraniol . After 45 min at 37°C , cells were harvested , lysed , and immunoprecipitated with polyclonal anti-reductase antibodies . Aliquots of the precipitated material and the lysates were subjected to SDS-PAGE and immunoblot analysis was carried out with IgG-A9 ( against reductase ) , IgG-H8 ( against UBIAD1 ) , and anti-calnexin IgG . Proteins corresponding to immunoprecipitated UBIAD1 were quantified using ImageJ software . The intensities of these signals in the absence of geranylgeraniol were arbitrarily set as 1 . ( B ) SV-589 cells were set up on day 0 at 3 × 104 cells/well of a twelve-well plate with a glass coverslip in medium A containing 10% FCS . On day 1 , the cells were transfected using FuGENE 6 with 50 ng of wild type ( WT ) , N102S , or G177R versions of pCMV-Myc-UBIAD1; the total amount of DNA/well was adjusted to 500 ng by the addition of pcDNA3 . 1 vector . 4 hr after transfection , cells received a direct addition of medium A containing 10% FCS ( final concentration ) . After 16 hr at 37°C , cells were fixed and analyzed by microscopy as described in ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 01610 . 7554/eLife . 05560 . 017Figure 8—figure supplement 1 . SCD-associated UBIAD1 ( N102S ) resists geranylgeraniol-mediated displacement from HMG CoA reductase in two independent experiments ( A and B ) . UBIAD1−/pCDNA3 . 1 , UBIAD1−/pMyc-UBIAD1 ( WT ) , and UBIAD1−/pMyc-UBIAD1 ( N102S ) cells were set up for experiments on day 0 at a density of 4 × 105 cells per 60-mm dish in medium A containing 10% FCS . On day 3 , cells were depleted of sterols as described in the legend to Figure 4 . After 16 hr at 37°C , cells received the identical medium containing 1 µg/ml 25-HC in the absence or presence of the indicated concentration of geranylgeraniol . After 45 min at 37°C , cells were harvested , lysed , and immunoprecipitated with polyclonal anti-reductase antibodies . Aliquots of the precipitated material and the lysates were subjected to SDS-PAGE and immunoblot analysis was carried out with IgG-A9 ( against reductase ) , IgG-H8 ( against UBIAD1 ) , and anti-calnexin IgG . Proteins corresponding to immunoprecipitated UBIAD1 were quantified using ImageJ software . The intensities of these signals in the absence of geranylgeraniol were arbitrarily set as 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 01710 . 7554/eLife . 05560 . 018Figure 8—figure supplement 2 . Subcellular localization of wild type and SCD-associated N102S UBIAD1 in transfected SV-589 cells . SV-589 cells were set up for experiments on day 0 , transfected on day 1 with pCMV-Myc-UBIAD1 ( WT ) or ( N102S ) in medium A containing 10% FCS , and subjected to immunostaining , followed by imaging as described in the legend to Figure 5—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 018 In the experiment of Figure 8B , we used immunofluorescence to examine the subcellular localization of wild type and SCD-associated mutants of UBIAD1 in SV-589 cells grown in FCS-containing medium . The results show that wild type UBIAD1 localized to Golgi membranes that also stained with antibodies against GM130 ( Figure 8B , panels 1–4 ) . UBIAD1 ( N102S ) primarily exhibited a diffuse , reticular localization corresponding to ER membranes and little , if any , of the protein appeared in the Golgi ( Figure 8B , panels 5–8 ) . The repeat experiment shown in Figure 8—figure supplement 2 confirms that wild type UBIAD1 localized to the Golgi , whereas UBIAD1 ( N102 ) primarily localized to ER membranes . Similar to the N102 mutant , UBIAD1 ( G177R ) was also primarily concentrated in membranes of the ER , but not the Golgi ( Figure 8B , panels 9–12 ) .
Results of the current experiments form the basis for the model shown in Figure 9 that depicts the role of UBIAD1 in sterol-accelerated ERAD of reductase . As previously proposed ( Morris et al . , 2014 ) , the reaction is initiated by accumulation of sterols , which triggers binding of Insigs to reductase and results in its gp78/Trc8-mediated ubiquitination ( Song et al . , 2005; Jo et al . , 2011 ) . We find in the current study that the oxysterol 25-HC and the 1 , 1-bisphosphonate ester Apomine , which mimics 25-HC in accelerating reductase ERAD ( Roitelman et al . , 2004; Sever et al . , 2004; Nguyen et al . , 2009 ) , trigger binding of reductase to UBIAD1 ( Figure 3B , C ) , a prenyltransferase that mediates synthesis of the vitamin K2 derivative MK-4 ( see Figure 1 ) . The sterol-induced binding of UBIAD1 to reductase appears to follow the action of Insigs as indicated by inhibition of the association by RNAi-mediated Insig knockdown ( Figure 4A ) . In contrast to the action of sterols , the nonsterol isoprenoid geranylgeraniol , which augments sterol-accelerated reductase ERAD ( Sever et al . , 2003a ) , inhibits formation of the UBIAD1-reductase complex ( Figure 5A and Figure 5—figure supplement 1B ) . Geranylgeraniol also regulates the subcellular localization of UBIAD1 . The prenyltransferase primarily localizes to Golgi membranes when cells are cultured in sterol-replete FCS-containing medium ( Figure 5C ) . The Golgi localization of UBIAD1 is disrupted when cells are switched to sterol-depleting medium containing LPPS and the reductase inhibitor compactin , which depletes cellular stores of nonsterol isoprenoids ( Brown and Goldstein , 1980 ) . Remarkably , Golgi localization of UBAID1 is restored by the addition of geranylgeraniol , but not of 25-HC or farnesol , to cells deprived of sterol and nonsterol isoprenoids ( Figure 5C and Figure 5—figure supplements 2 , 3 ) . We conclude from these observations that geranylgeraniol-mediated disruption of the UBIAD1-reductase complex allows for the translocation of UBIAD1 from ER membranes to the Golgi and for membrane extraction , cytosolic dislocation and proteasomal degradation of reductase ( Figure 9 ) . 10 . 7554/eLife . 05560 . 019Figure 9 . Proposed model for role of UBIAD1 in sterol-accelerated degradation of HMG CoA reductase . In sterol-deprived cells , both reductase and UBIAD1 localize to membranes of the ER . The intracellular accumulation of sterols in ER membranes triggers binding of reductase to Insigs , resulting in its ubiquitination by Insig-associated ubiquitin ligases gp78 and Trc8 and association with UBIAD1 . Geranylgeraniol becomes phosphorylated to produce geranylgeranyl pyrophosphate , which enhances reductase degradation by binding to UBIAD1 , causing its displacement from reductase-Insig . This displacement allows for transport of UBIAD1 to the Golgi and membrane extraction , cytosolic dislocation , and proteasomal degradation of reductase . We postulate that the SCD-associated N102S or G177R mutations in UBIAD1 abrogate binding of geranylgeranyl pyrophosphate . As a result , UBIAD1 ( N102S ) and ( G177R ) do not translocate to the Golgi and remain associated with reductase in the ER , thereby blocking its membrane extraction , cytosolic dislocation , and proteasomal degradation . DOI: http://dx . doi . org/10 . 7554/eLife . 05560 . 019 The model shown in Figure 9 predicts that UBIAD1 inhibits sterol-accelerated ERAD of reductase and that geranylgeraniol relieves this inhibition by triggering displacement of UBIAD1 from the protein . In support of this notion , we observed that RNAi-mediated knockdown and CRISPR/Cas9-mediated knockout of UBIAD1 eliminates the requirement of geranylgeraniol for maximal degradation of reductase ( Figure 6 ) . The continuance of accelerated reductase degradation in UBIAD1-deficient cells is consistent with the proposal that UBIAD1 mediates a step in the reaction downstream the action of Insigs . In future studies , efforts will be undertaken to delineate mechanisms through which UBIAD1 retards reductase ERAD . Geranylgeraniol is required for extraction of ubiquitinated reductase across the ER membrane through a reaction that is mediated by the AAA-ATPase VCP/p97 ( Morris et al . , 2014 ) . Thus , one likely mechanism for the inhibitory effects of UBIAD1 on reductase ERAD may involve inhibition of the association of ubiquitinated reductase with VCP/p97 or one of its ubiquitin-binding cofactors , thereby preventing its subsequent membrane extraction . In addition to revealing the molecular mechanism through which geranylgeraniol modulates reductase ERAD , the current results help to explain how cells maintain synthesis of nonsterol isoprenoids , while avoiding overproduction of cholesterol . In 1979 , Faust , Brown , and Goldstein found that in the absence of LDL , cells accumulate reductase to produce mevalonate for conversion into primarily cholesterol and other sterols ( Faust et al . , 1979 ) . The addition of LDL to cells partially suppressed reductase , limiting production of mevalonate . As a result , incorporation of mevalonate into cholesterol was reduced , whereas incorporation of the molecule into nonsterol isoprenoids such as CoQ10 was enhanced . We postulate that LDL failed to completely suppress reductase due to its binding to UBIAD1 and resultant inhibition of sterol-accelerated degradation . This residual reductase produces a small amount of mevalonate that is preferentially diverted into nonsterol isoprenoids . Once appropriate levels of geranylgeranyl pyrophosphate accumulate in cells , the isoprenoid binds to UBIAD1 , causing its release from reductase , which subsequently becomes degraded . Released UBIAD1 then translocates to the Golgi where it synthesizes PPHB ( Mugoni et al . , 2013 ) and perhaps MK-4 . Thus , substrate-regulated transport of UBIAD1 ensures that products of the enzyme are only produced when cellular demands for other nonsterol isoprenoids as well as sterols have been met . Studies are currently underway to appraise this notion and to determine whether Golgi-localized UBIAD1 is enzymatically active to synthesize MK-4 and PPHB , whereas the ER-localized enzyme is not . Moreover , the role of reductase in ER to Golgi trafficking of UBAD1 will be investigated . The significance of geranylgeraniol-mediated displacement of UBIAD1 from reductase is revealed through the study of SCD-associated mutants of the prenyltransferase . Consistent with the autosomal dominant phenotype of SCD , disease-associated N102S and G177R mutants of UBIAD1 inhibit sterol-accelerated ERAD of reductase ( Figure 7B , C , and Figure 7—figure supplement 1 ) . The molecular basis for this inhibition is suggested by Figure 8A and Figure 8—figure supplement 1 , which show that UBIAD1 ( N102S ) resists geranylgeraniol-mediated displacement from reductase . Subcellular localization studies show that in FCS-cultured cells , wild type UBIAD1 localizes to the Golgi , whereas the N102S and G177R mutants of the enzyme appear defective in Golgi trafficking and remain sequestered within the ER ( Figure 8B ) . Mutation of the asparagine residue in bacterial UbiA prenyltransferases that corresponds to N102 markedly diminishes substrate binding and enzymatic activity ( Cheng and Li , 2014; Huang et al . , 2014 ) . Thus , we postulate that binding of UBIAD1 to geranylgeranyl pyrophosphate produced by phosphorylation of geranylgeraniol triggers displacement of wild type UBIAD1 from reductase , allowing the protein to translocate to the Golgi . In contrast , the N102S and G177R mutants of UBIAD1 exhibit reduced affinity for geranylgeranyl pyrophosphate and thus , remain associated with reductase in the ER , thereby inhibiting its sterol-accelerated ERAD ( Figure 9 ) . This inhibition may contribute to the accumulation of cholesterol that is observed in corneas of SCD patients . It will be important in future studies to develop a geranylgeranyl pyrophosphate-binding assay for mammalian UBIAD1 and to determine the subcellular localization and effect of the other 18 SCD-associated UBIAD1 mutants on reductase ERAD . In conclusion , the current study has identified UBIAD1 as the elusive molecular target of geranylgeraniol in reductase ERAD . They have also revealed a potential mechanism whereby mutations in UBIAD1 cause the accumulation of cholesterol in corneas of SCD patients . Finally , results of this study indicate the existence of a novel link between the synthesis of MK-4 and cholesterol . Further investigation of this link is merited owing to the potential for development of novel therapies that accelerate reductase ERAD to lower plasma LDL-cholesterol and retard corneal accumulation of cholesterol that characterizes SCD .
We obtained MG-132 from Boston Biochem ( Cambridge , MA ) ; horseradish peroxidase-conjugated donkey anti-mouse and anti-rabbit IgGs ( affinity-purified ) were from Jackson ImmunoResearch Laboratories ( West Grove , PA ) ; digitonin was from Calbiochem ( San Diego , CA ) ; geranylgeraniol and farnesol from Sigma–Aldrich ( St . Louis , MO ) and Santa Cruz Biotechnology ( Dallas , TX ) ; biotin was obtained from Sigma–Aldrich ( St . Louis , MO ) ; and 25-hydroxycholesterol from Steraloids ( Newport , RI ) . Apomine was synthesized by the Core Medicinal Chemistry laboratory at the University of Texas Southwestern Medical Center . Other reagents , including new born calf lipoprotein-poor serum ( NC-LPPS , d > 1 . 215 g/ml ) , sodium compactin , sodium mevalonate , and stock solutions of digitonin were prepared or obtained from previously described sources ( Goldstein et al . , 1983; DeBose-Boyd et al . , 1999; Elsabrouty et al . , 2013 ) . The previously described expression plasmids pCMV-HMG-Red ( TM1-8 ) -T7 and pCMV-HMG-Red-T7 encode the membrane domain ( amino acids 1–346 ) and full-length ( amino acids 1–887 ) forms , respectively , of hamster reductase followed by three tandem copies of the T7 epitope tag under transcriptional control of the cytomegalovirus ( CMV ) promoter ( Sever et al . , 2003a , 2003b ) . The expression plasmid pCMV-HSV-HMG-Red ( TM1-8 ) -BirA* encodes the membrane domain of hamster reductase with two copies of an N-terminal epitope tag derived from Herpes Simplex Virus ( HSV ) glycoprotein D fused to the humanized R118G mutant of Escherichia coli BirA ( obtained from Addgene , Cambridge , MA and designated BirA* ) that exhibits promiscuous biotin ligase activity ( Roux et al . , 2012 ) . The cDNA encoding human UBIAD1 was purchased from Open Biosystems ( Lafayette , CO ) and cloned into the pcDNA3 . 1 ( + ) vector using standard PCR methods . The expression plasmid pCMV-Myc-UBIAD1 was generated by fusing one copy of the Myc epitope tag to the N-terminus of UBIAD1 . The plasmids pCMV-Myc-UBIAD1 ( N102S ) and ( G177R ) encode Myc-tagged human UBIAD1 harboring the SCD-associated asparagine-102 to serine ( N102S ) and glycine-177 to arginine ( G177R ) mutations , respectively , and were generated using the Quikchange Site-Directed Mutagenesis Kit ( Agilent Technologies , Santa Clara , CA ) and pCMV-Myc-UBIAD1 as a template . CRISPR plasmids hCas9 and gRNA Cloning Vectors were obtained from Addgene . Guide RNA constructs were designed using option B described by the Church laboratory ( Mali et al . , 2013 ) . ( See http://www . addgene . org/static/cms/files/hCRISPR_gRNA_Synthesis . pdf ) Guide RNA sequences unique to human UBIAD1 were selected from a published list ( Mali et al . , 2013 ) . ( See http://arep . med . harvard . edu/human_crispr ) . SV-589 cells are a line of immortalized human fibroblasts expressing the SV40 large T-antigen ( Yamamoto et al . , 1984 ) . Monolayers of SV-589 cells were maintained in medium A ( DMEM containing 1000 mg glucose/l , 100 U/ml penicillin , and 100 mg/ml streptomycin sulfate ) supplemented with 10% ( vol/vol ) fetal calf serum ( FCS ) at 37°C , 5% CO2 . Human embryonic kidney ( HEK ) -293S/pHMG-Red ( TM1-8 ) -BirA* cells were generated as follows: on day 0 , HEK-293S cells were set up at a density of 7 × 105 cells per 100-mm dish in medium A supplemented with 10% FCS . On day 1 , cells were transfected with 6 µg/dish of pCMV-HSV-HMG-Red ( TM1-8 ) -BirA* using FuGENE6 transfection reagent ( Promega , Madison , WI ) as previously described ( Sever et al . , 2003b; Jo et al . , 2011 ) . Following incubation for 16 hr at 37°C , cells were switched to medium A supplemented with 10% FCS and 700 µg/ml G418 . Fresh medium was added every 2–3 days until colonies formed after 2 weeks . Individual colonies were isolated using cloning cylinders , and expression of HSV-HMG-Red ( TM1-8 ) -BirA* was determined by immunoblot analysis . Cells from single colonies expressing high levels of HSV-HMG-Red ( TM1-8 ) -BirA* were selected and monolayers were maintained in medium B ( medium A supplemented with 10% FCS and 700 µg/ml G418 ) at 37°C , 5% CO2 . UBIAD1-deficient cells ( designated UBIAD1− ) were generated as follows: on day 0 , SV589 cells were set up at a density of 7 × 105 cells per 100-mm dish in medium A supplemented with 10% FCS . On day 1 , cells were transfected with 5 µg/dish each of hCas9 , hUBIAD1-gRNA12 and hUBIAD1-gRNA19 using FuGENE6 transfection reagent as described above . On day 2 and 3 the transfection above was repeated . On day 4 cell clones were isolated using serial dilution in 96-well plates . Clones were screened for the absence of UBIAD1 by immunoblot analysis using mouse monoclonal IgG-H8 and rabbit polyclonal antibodies against human UBIAD1 ( Santa Cruz Biotechnology , Dallas , TX ) . A homozygous 113 bp deletion/frameshift mutation ( starting at codon 60 ) of UBIAD1 was identified by PCR and sequencing of the PCR products by standard techniques . UBIAD1−/pcDNA3 . 1 , UBIAD1−/pMyc-UBIAD1 , and UBIAD1−/pMyc-UBIAD1 ( N102S ) are UBIAD1− cells stably transfected with pcDNA3 . 1 , pCMV-Myc-UBIAD1 , and pCMV-Myc-UBIAD1 ( N102S ) , respectively . These cells were generated as follows: on day 0 , UBIAD1−cells were set up at a density of 7 × 105 cells per 100-mm dish in medium A supplemented with 10% FCS . On day 1 , cells were transfected with 6 µg/dish of pcDNA3 . 1 , pCMV-Myc-UBIAD1 , or pCMV-Myc-UBIAD1 ( N102S ) using FuGENE6 transfection reagent as described above . Following incubation for 16 hr at 37°C , cells were switched to medium A supplemented with 10% FCS and 700 µg/ml G418 . Fresh medium was added every 2–3 days until colonies formed after 2 weeks . Individual colonies were isolated using cloning cylinders , and expression of Myc-UBIAD1 was determined by immunoblot analysis . Cells from single colonies of cells expressing moderate levels of Myc-UBIAD1 were selected and monolayers were maintained in medium B ( medium A supplemented with 10% FCS and 700 µg/ml G418 ) at 37°C , 5% CO2 . HEK-293S/pHMG-Red ( TM1-8 ) -BirA* cells were set up on day 0 at a density of 3 × 105 cells per 100-mm dish in medium B . On day 3 , cells were depleted of sterols through incubation in medium A supplemented with 10% ( vol/vol ) NC-LPPS , 10 µM compactin , and 50 µM mevalonate . After 16 hr at 37°C , the cells were treated for an additional 6 hr in sterol-depleting medium containing 50 µM biotin and 1 µg/ml 25-HC . Cells were subsequently harvested , washed with phosphate-buffered saline ( PBS ) , and resuspended in 0 . 3 ml/plate of buffer containing 10 mM HEPES , pH 7 . 4 , 10 mM KCl , 1 . 5 mM MgCl2 , 5 mM sodium EDTA , 5 mM sodium EGTA , and 250 mM sucrose supplemented with a protease inhibitor cocktail consisting of 5 mM dithiothreitol , 0 . 1 mM leupeptin , 1 mM phenylmethylsulfonyl fluoride , 0 . 5 mM Pefabloc , 5 µg/ml pepstatin A , 25 µg/ml ALLN , and 10 µg/ml aprotinin . The cell suspension was passed through a 22 . 5 gauge needle 25 times and centrifuged at 1000×g for 10 min at 4°C . The resultant post-nuclear supernatant was subjected to an additional round of centrifugation at 100 , 000×g for 30 min at 4°C . The membrane pellet of this spin was resuspended in 0 . 3 ml/10 plates of solubilization buffer containing 50 mM Tris-HCl , pH 7 . 4 , 150 mM NaCl , 5 mM EDTA , 5 mM EGTA , 8 M urea , and 2% SDS and subjected to dounce homogenization . The solubilized membrane pellet was then precleared with 1 ml/10 plate of agarose beads ( Sigma–Aldrich , St . Louis , MO ) by rotation for 6–16 hr at 4°C . After centrifugation at 300×g for 5 min at 4°C , the supernatants were mixed with 0 . 8 ml of streptavidin-coupled agarose beads ( Sigma–Aldrich , St . Louis , MO ) and rotated overnight at 4°C . The beads were collected by centrifugation at 2500×g for 1 . 5 min at 4°C , and washed three times with three volumes of solubilization buffer diluted 10-fold in Tris-HCl , pH 7 . 4 , 150 mM NaCl , 5 mM EDTA , and 5 mM EGTA . The streptavidin-coupled beads were then eluted three times through incubation for 10 min at 98°C in 1 volume of buffer containing 50 mM Tris-HCl , pH 7 . 4 , 150 mM NaCl , 5 mM EGTA , 5 mM EDTA , 5 mM biotin , 10% ( vol/vol ) dimethylsulfoxide , 2% SDS , and 0 . 4 M urea . The elutions were pooled and precipitated with 4 vol of ice-cold acetone at −20°C for 1 hr . Precipitated material was collected by centrifugation at 16 , 000 at 4°C and dissolved in a mixture consisting of 10 µl of buffer containing 10 mM Tris-HCl , pH 6 . 8 , 100 mM NaCl , 1% SDS , 1 mM EDTA , and 1 mM EGTA; 20 µl of buffer containing 62 . 5 mM Tris-HCl , pH 6 . 8 , 15% SDS , 8 M urea , 10% glycerol , and 100 mM dithiothreitol; 10 µl 4× SDS-PAGE loading buffer . The samples were incubated for 20 min at 37°C prior to SDS-PAGE . The gels were subjected to staining with colloidal coomassie as described ( Dyballa and Metzger , 2009 ) and three segments of the gel that contained visible bands ( see Figure 3—figure supplement 1A ) were excised , and the identities of the proteins were determined by tandem mass spectrometry in the Protein Chemistry Core Facility at the University of Texas Southwestern Medical Center . RNA interference ( RNAi ) was performed as previously described with minor modifications ( Jo et al . , 2011 ) . Duplexes of siRNAs were designed and synthesized by Dharmacon/Thermo Fisher Scientific . The siRNA duplexes against GFP , Insig-1 , and Insig-2 have been previously described ( Elsabrouty et al . , 2013 ) ; the sequence for siRNA duplexes targeting human UBIAD1 and reductase were 5′-UUAACAUCCUGUCGGGAGAUU-3′ and 5′-CCACAGAGGCUAUGAUUGAUU-3′ , respectively . SV-589 cells were set up for experiments on day 0 as described in the figure legends . On day 3 , the cells were incubated with 600 pmol of siRNA duplexes mixed with Lipofectamine RNAiMAX reagent ( Invitrogen , Grand Island , NY ) diluted in Opti-MEM I reduced serum medium ( Life Technologies , Grand Island , NY ) according to manufacturer's procedure . Following incubation for 6 hr at 37°C , the cells received a direct addition of medium A containing 10% FCS ( final concentration ) . On day 4 , the cells were switched to medium A containing 10% NC-LPPS , 10 µM compactin , and 50 µM mevalonate and incubated for 16 hr at 37°C . The cells were subsequently treated and analyzed as described in figure legends . Transfection of SV-589 cells with FuGENE6 transfection reagent was carried out as previously described ( Sever et al . , 2003b; Jo et al . , 2011 ) . Conditions of subsequent incubations are described in the figure legends . Following incubations , triplicate dishes of cells for each variable were harvested and pooled for analysis . For immunoprecipitations with polyclonal anti-reductase or anti-UBIAD1 antibodies , the cells were resuspended in PBS containing 1% digitonin , 5 mM EDTA , 5 mM EGTA , and the protease inhibitor cocktail . Following passage through a 22 . 5 gauge needle 15 times and rotation for 30 min at 4°C , the samples were clarified by centrifugation at 20 , 000×g for 10 min at 4°C . The detergent-solubilized material was then subjected to immunoprecipitation as described previously ( Elsabrouty et al . , 2013 ) . Subcellular fractionation of cells by differential centrifugation was performed as previously described ( Jo et al . , 2013 ) . Aliquots of detergent lysates ( 0 . 01 dish of cells ) and pellet fractions ( 0 . 125 dish of cells ) from immunoprecipitations and membrane fractions from subcellular fractionations were subjected to SDS-PAGE and immunoblot analysis . Primary antibodies used for immunoblot analysis included: mouse monoclonal anti-T7 Tag IgG and anti-HSV Tag ( Novagen , Darmstadt , Germany ) ; IgG-A9 , a mouse monoclonal antibody against the catalytic domain of reductase ( Liscum et al . , 1983 ) ; rabbit polyclonal anti-UBIAD1 IgG , mouse monoclonals IgG-H8 against UBIAD1 , IgG-F2 against human peroxiredoxin-4 , IgG-F3 against human progesterone receptor membrane component 2 ( PGRMC2 ) , IgG-C6 against human ERGIC-53 , IgG-ANNEX 5E4/1 against human annexin I , and IgG-Y20 against human lamina-associated peptide-2 ( Santa Cruz Biotechnology , Dallas , TX ) ; rabbit polyclonal anti-calnexin IgG ( Novus Biologicals , Littleton , CO ) ; and IgG-17H1 , a mouse monoclonal antibody against human Insig-1 . SV-589 cells were set up for experiments on day 0 as described in the figure legends . Following incubations described in the figure legends , cells were washed with PBS and subsequently fixed and permeabilized for 15 min in methanol at −20°C . Upon blocking with 1 mg/ml BSA in PBS , coverslips were incubated for 30 min at 37°C with primary antibodies ( IgG-H8 against UBIAD1 , rabbit polyclonal anti-GM130 IgG [Diao et al . , 2003] , and IgG-9E10 , a mouse monoclonal antibody against c-Myc purified from the culture medium of hybridoma clone 9E10 [American Type Culture Collection , Manassas , VA] ) diluted in PBS containing 1 mg/ml BSA . Bound antibodies were visualized with goat anti-mouse IgG conjugated to Alexa Fluor 488 and goat anti-rabbit Alexa Fluor 594 ( Life Technologies , Grand Island , NY ) as described in the figure legends . In addition , coverslips were stained for 5 min with 1 µg/ml Hoechst 33 , 342 ( Life Technologies ) to visualize nuclei . The coverslips were then mounted in Mowiol 4-88 solution ( Calbiochem/EMD Millipore , Billerica , MA ) and fluorescence analysis was performed using a Plan-Apochromat 63×/1 . 4 DIC objective ( Zeiss , Peabody , MA ) , an Axiovert 200M microscope ( Zeiss ) , an Orca 285 camera ( Hamamatsu , Houston , TX ) , and the software Openlab 4 . 0 . 2 ( Improvision , Coventry , England ) . | People with a rare genetic disorder called ‘Schnyder corneal dystrophy’ gradually lose their vision , because their corneas become increasingly cloudy . This cloudiness is caused by a build-up of excessive amounts of cholesterol , and the disorder itself is caused by mutations in a gene that encodes a protein called UBIAD1 . Researchers have previously discovered that the UBIAD1 protein is involved in making vitamin K2 , but it is not clear how this protein also helps to control cholesterol levels in the cornea . An enzyme called HMG CoA reductase makes a molecule that is used to make cholesterol and many other similar sterol molecules . A ‘feedback loop’ operates in cells to control the amount of the reductase and prevent cholesterol from becoming too high or too low . Sterol molecules , together with another molecule called geranylgeraniol , participate in this feedback loop by promoting the destruction of the reductase enzyme . Here , Schumacher et al . reveal a link between UBIAD1 and the reductase that may explain how UBIAD1 contributes to the production of excess cholesterol in patients with Schnyder corneal dystrophy . The experiments show that , in the presence of sterol molecules , UBIAD1 can bind to HMG CoA reductase to protect the reductase from being destroyed by other proteins . Geranylgeraniol—which stops the UBIAD1 protein from binding to the enzyme—is required to completely destroy the reductase enzyme . However , when UBIAD1 is missing , the reductase enzyme is destroyed even in the absence of geranylgeraniol . Furthermore , the experiments show that the genetic mutations linked to Schnyder corneal dystrophy lead to the production of versions of the UBIAD1 protein that bind to the reductase enzyme even when geranylgeraniol molecules are present . This prevents the normal breakdown of the reductase enzyme , which could lead to the build up of cholesterol in the cornea of individuals with the disorder . Schumacher et al . 's findings show that the UBIAD1 protein helps to control the levels of cholesterol in cells by protecting the HMG CoA reductase enzyme from destruction . These findings may aid the development of new therapies to lower cholesterol levels in cells , which may help patients with Schnyder's corneal dystrophy and other conditions caused by high cholesterol levels . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2015 | The prenyltransferase UBIAD1 is the target of geranylgeraniol in degradation of HMG CoA reductase |
The production of offspring is energetically costly and relies on incompletely understood mechanisms that generate a positive energy balance . In mothers of many species , changes in key energy-associated internal organs are common yet poorly characterised functionally and mechanistically . In this study , we show that , in adult Drosophila females , the midgut is dramatically remodelled to enhance reproductive output . In contrast to extant models , organ remodelling does not occur in response to increased nutrient intake and/or offspring demands , but rather precedes them . With spatially and temporally directed manipulations , we identify juvenile hormone ( JH ) as an anticipatory endocrine signal released after mating . Acting through intestinal bHLH-PAS domain proteins Methoprene-tolerant ( Met ) and Germ cell-expressed ( Gce ) , JH signals directly to intestinal progenitors to yield a larger organ , and adjusts gene expression and sterol regulatory element-binding protein ( SREBP ) activity in enterocytes to support increased lipid metabolism . Our findings identify a metabolically significant paradigm of adult somatic organ remodelling linking hormonal signals , epithelial plasticity , and reproductive output .
Reproduction is energetically costly . Mothers can adjust their energy balance to maximise their reproductive success through well-established neural mechanisms that match food intake to their enhanced energy requirements ( Roa and Tena-Sempere , 2014 ) . However , less well-understood changes also occur in many animals during reproduction; internal organs , such as the liver , pancreas , and gastrointestinal tract , increase their size and adapt their physiology , potentially contributing to an increased generation and delivery of nutrients ( Hammond , 1997; Speakman , 2008 ) . Establishment of a positive energy balance may be particularly important to animals with a reproductive strategy that involves rapid production of large numbers of progeny . Drosophila melanogaster females can lay up to 100 eggs per day at the peak of their fertility in early life ( David et al . , 1974; Klepsatel et al . , 2013 ) . We hypothesised that such demands may rely on major regulatory responses , which are amenable to genetic investigation in this model system . A network of organs and tissues in Drosophila perform many of the same basic functions as those found in mammals ( Padmanabha and Baker , 2014 ) , so we sought to explore the nature and significance of organ plasticity during reproduction .
Female flies undergo multiple post-mating adaptations including changes in digestive physiology ( Cognigni et al . , 2011 ) . This prompted us to characterise possible intestinal changes occurring during the phase of peak fertility ( David et al . , 1974; Klepsatel et al . , 2013 ) . We focused on the midgut epithelium because of its major digestive/absorptive roles ( Lemaitre and Miguel-Aliaga , 2013 ) . In the midgut epithelium , long-lived progenitors ( intestinal stem cells ( ISCs ) ) divide to self-renew and to give rise to committed progenitors ( called enteroblasts ( EBs ) ) , which directly differentiate into two types of progeny: absorptive enterocytes ( ECs ) and enteroendocrine cells ( EECs ) ( Jiang and Edgar , 2012 ) . We found that mating increases the number of both dividing and differentiating midgut cells , as revealed by phospho-Histone H3 ( pH3 ) stainings and temporal analyses of progenitors and their descendants using the dual-labelling system escargot-Repressible Dual Differential Marker ( esgReDDM , Antonello et al . , 2015 ) ( Figure 1A , C , E ) . The midgut of mated females also becomes visibly larger; gut diameter measurements were suggestive of a net increase in the number of postmitotic intestinal cells ( Figure 1B , D , Figure 1—figure supplement 1 ) : an increase that we confirmed by cell number and density counts ( Figure 1F , Figure 1—figure supplement 1 ) . Concurrent with midgut re-sizing , we observed mating-dependent activation of the single Drosophila homologue of the mammalian family of sterol regulatory element-binding proteins ( SREBPs [Theopold et al . , 1996; Shimano , 2001; Seegmiller et al . , 2002] , also known as HLH106 in flies , Figure 2A , B ) , using a reporter subject to the same physiologically regulated proteolytic processing as wild-type SREBP ( Kunte et al . , 2006 ) . SREBP activation after mating was accompanied by upregulation of midgut transcripts involved in fatty acid synthesis and activation ( SREBP , the long-chain fatty acid CoA ligases bubblegum ( bgm ) and Acyl-CoA synthetase long-chain ( Acsl ) and , depending on genetic background , Fatty acid synthase ( FAS ) and Acetyl-CoA carboxylase ( ACC ) ) ( Figure 2E , F ) , many of which are known SREBP targets in flies and/or mammals ( Seegmiller et al . , 2002; Horton et al . , 2003 ) . Immunohistochemical analyses using reporters pointed to the ECs located in the posterior midgut region ( R5 , Buchon et al . , 2013; Marianes and Spradling , 2013 ) as preferential sites of transcriptional and SREBP activity changes ( Figure 2A–D ) . Thus , in female flies actively engaged in reproduction , changes in both intestinal progenitors and their progeny parallel those observed in mammals leading to hyperplasia ( Hammond , 1997; Speakman , 2008 ) , increased organ size ( Hammond , 1997; Speakman , 2008 ) and upregulation of lipid gene expression ( Athippozhy et al . , 2011 ) . 10 . 7554/eLife . 06930 . 003Figure 1 . Mating increases ISC proliferation and gut size . ( A , A′ ) Using the esgReDDM tracer ( Antonello et al . , 2015 ) , intestinal progenitors ( intestinal stem cells ( ISCs ) and enteroblasts ) are labelled with GFP and RFP , whereas the postmitotic progeny ( enterocytes ( ECs ) and enteroendocrine cells ) that these progenitors give rise to in a defined time window is labelled with RFP only ( see Supplemental Information for additional details ) . At 3 days after mating , the posterior midgut of mated flies contains more newly generated postmitotic progeny ( A ) compared to age-matched virgins ( A′ ) . It has also become visibly larger ( B , B′ ) . At this time point , these guts also have a higher number of nuclei marked by the mitotic marker pH3 in both w1118 and OregonR backgrounds ( C , p = 0 . 008 , and E , p < 0 . 001 , negative binomial GLM ) , although the proliferation increase is transient ( data not shown ) . The size increase is quantified in the posterior midgut by measuring midgut diameter ( D , p < 0 . 001 , t-test ) and counting the number of cells labelled by the EC marker caudal-Gal4 ( F , p = 0 . 02 , t-test ) . See Table 1 for full genotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 06930 . 00310 . 7554/eLife . 06930 . 004Figure 1—figure supplement 1 . Mating re-sizes the Drosophila gut . The increase in gut size at 3 days after mating is also measurable ( A , A′ ) and significant ( B , p < 0 . 001 , t-test ) in the OregonR background . The esgReDDM tracing system reveals that mated guts contain more cells generated in the last 7 days if the fly had been mated in that time ( C , p < 0 . 001 , t-test ) than if it had not . The size increase is not due to stretching of the tissue , as the density of nuclei in the posterior midgut remains the same ( D , p = 0 . 77 , t-test ) . See Table 1 for full genotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 06930 . 00410 . 7554/eLife . 06930 . 005Figure 2 . Mating changes the activity and/or expression of lipid metabolism genes in the intestine . At 3 days after mating , increased expression of a reporter that replicates the transcriptional regulation and post-translational modification of sterol regulatory element-binding protein ( SREBP ) is apparent in the posterior midgut ( A , A′ , quantified in B , p < 0 . 001 , Mann–Whitney test ) . A bgm transcriptional reporter is also increased specifically in the ECs of the posterior midgut following mating ( C , C′ , quantified in D , p = 0 . 002 , Mann–Whitney test ) . Transcript abundance of SREBP , bgm , and the SREBP targets Acyl-CoA synthetase long-chain ( Acsl ) , Fatty acid synthase ( FAS ) , and Acetyl-CoA carboxylase ( ACC ) is increased by mating in either one or both of the w1118 and OregonR backgrounds ( E w1118: p = 0 . 02 SREBP , p = 0 . 02 bgm , p = 0 . 005 Acsl , p = 0 . 5 FAS , p = 0 . 3 ACC; F OregonR: p = 0 . 02 SREBP , p = 0 . 03 bgm , p = 0 . 03 Acsl , p = 0 . 01 FAS , p = 0 . 04 ACC , paired one-tailed t-test ) . See Table 1 for full genotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 06930 . 005 Female flies change their physiology and behaviour ( e . g . , by increasing egg production and food intake ) in response to male-derived peptides acquired during mating ( Carvalho et al . , 2006; Barnes et al . , 2008 ) . The synthesis of juvenile hormone ( JH ) in the corpus allatum , an endocrine gland , can be stimulated ex vivo by the male-derived Sex Peptide , suggesting regulation by mating ( Moshitzky et al . , 1996 ) . Using rapid direct analysis in real time ( DART ) mass spectrometry , we profiled haemolymph of both virgin and mated female flies and established that the levels of in vivo circulating JH are indeed increased after mating ( Figure 3A ) . The levels of two other juvenoid compounds , JH3-bisepoxide ( JH3B ) and methylfarnesoate ( MF ) , were too low to be detected . We cannot , however , rule out that they also are regulated by mating and contribute to signalling through the JH pathway ( Yin et al . , 1995; Tiu et al . , 2012; Wen et al . , 2015 ) . JH has been shown to sustain ovarian maturation through pleiotropic actions on adipose and reproductive tissues ( Flatt et al . , 2005 ) , but its intestinal roles remain to be established . Consistent with a possible intestinal role , we detected transcript upregulation of the JH target Kruppel homolog 1 ( Kr-h1 ) ( Jindra et al . , 2013 ) in guts of mated females ( Figure 3—figure supplement 1 ) . To explore the roles of JH signalling on intestinal remodelling , we first fed methoprene , a JH analogue ( JHa ) ( Cerf and Georghiou , 1972 ) , to virgin female flies . This led to effects on intestinal progenitors , gut diameter , and lipid metabolism comparable to those triggered by mating ( Figure 3B–E , H , Figure 3—figure supplement 1 ) . We next blocked endogenous JH production by mis-expressing the protein phosphatase inhibitor NiPp1 using the corpus allatum-specific driver Aug21-Gal4 ( Siegmund and Korge , 2001 ) : a genetic manipulation known to result in adult-specific ablation of the corpus allatum and a dramatic reduction of JH titres in the haemolymph ( Yamamoto et al . , 2013 ) . Depletion of systemic JH prevented mating-triggered remodelling: a phenotype that could be restored in these gland-ablated flies by JHa feeding ( Figure 3F , I ) . 10 . 7554/eLife . 06930 . 006Figure 3 . Systemic JH secreted after mating acts directly in the intestinal epithelium to drive reproductive remodelling . Circulating juvenile hormone ( JH ) is elevated after mating in the haemolymph of female flies ( A , p = 0 . 02 at 24 hr , p = 0 . 002 at 48 hr , t-test with Holm's correction ) . Increased tissue renewal ( B , B′ ) and SREBP activation ( C , C′ , quantified in D , p < 0 . 001 , Mann–Whitney test ) are apparent following a 3-day dietary supplementation with JH analogue ( JHa ) . JHa treatment is sufficient to increase mitoses ( E , p < 0 . 001 , negative binomial GLM ) and size ( H , p < 0 . 001 , t-test ) of the posterior midgut . Conversely , when the endogenous JH source is genetically ablated by means of Aug21 > NiPp1 ( Yamamoto et al . , 2013 ) , the proliferation and size increase that follow mating are abolished , although they can be reinstated by feeding JHa ( proliferation F , p < 0 . 001 between Aug21/+ and Aug21 > NiPp1 mated , p < 0 . 001 between Aug21 > NiPp1 and NiPp1/+ mated , p < 0 . 001 between Aug21 > NiPp1 and Aug21 > NiPp1 + JHa mated; all relevant comparisons between virgins are not significant , negative binomial GLM with Holm's correction; gut diameter I , p = 0 . 002 between Aug21/+ and Aug21 > NiPp1 mated , p < 0 . 001 between Aug21 > NiPp1 and NiPp1/+ mated , p < 0 . 001 between Aug21 > NiPp1 and Aug21 > NiPp1 + JHa mated; all relevant comparisons between virgins are not significant , t-test with Holm's correction ) . Downregulation of either gce or Met in adult progenitors abrogates post-mating proliferation ( G , p < 0 . 001 between esgReDDM/+ and esgReDDM > gce RNAi mated , p < 0 . 001 between esgReDDM/+ and esgReDDM > Met RNAi mated , negative binomial GLM with Holm's correction ) and gut size increase ( J , p < 0 . 001 between esgReDDM/+ and esgReDDM > gce RNAi mated , p < 0 . 001 between esgReDDM/+ and esgReDDM > Met RNAi mated , t-test with Holm's correction ) . The upregulation of bgm reporter upon mating is abolished by the downregulation of gce , but not Met , in ECs using the EC-specific driver Mex-Gal4 ( K–K′′′′ ) . See Table 1 for full genotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 06930 . 00610 . 7554/eLife . 06930 . 007Figure 3—figure supplement 1 . Intestinal JH signalling is relayed through Kr-h1 and underlies mating-dependent intestinal growth and gene expression phenotypes . JHa application in virgin females results in a net growth of the gut , as shown by the increase in caudal-marked cells ( A , p = 0 . 004 , t-test ) . The JH signalling pathway can be targeted using RNAi constructs against the receptors Met and gce , which decrease transcript abundance compared to a tubts control when expressed globally in larvae for 3 hr at 29°C ( B , p = 0 . 002 for Met , p = 0 . 005 for gce , paired one-tailed t-test ) . Consequently , using esgReDDM to specifically knockdown gce in adult intestinal progenitor cells abolishes the proliferative effect of JHa application ( C , C′ ) , but does not reduce the number of progenitors after 7 days of downregulation ( D , p > 0 . 05 , t test ) . Progenitors in which gce is downregulated can still proliferate normally to replenish a gut damaged by a 24 hr application of the toxin paraquat ( E , E′ with quantification of mitoses in F , p < 0 . 001 for both esgReDDM/+ and esgReDDM > gce RNAi , p > 0 . 05 for all other relevant comparisons , t test with Holm's correction ) and the number of progenitors is not reduced by this treatment ( G , p = 0 . 04 between esgReDDM/+ untreated control and esgReDDM > gce RNAi untreated control , p > 0 . 05 for all other relevant comparisons , t test with Holm's correction ) . The transcription factor Kruppel homolog 1 ( Kr-h1 ) , a well-established effector of JH responses , is transcriptionally upregulated after 3 days of mating ( H , p = 0 . 02 in w1118 , p = 0 . 02 in OregonR , paired two-tailed t-test ) . Kr-h1 function is necessary and sufficient for the re-sizing of the gut after mating , as its downregulation in intestinal progenitors through RNA interference using esgReDDM prevents the increase in proliferation ( I , p < 0 . 001 between esgReDDM/+ and esgReDDM > Kr-h1 RNAi mated , negative binomial GLM with Holm's correction ) and gut size ( J , p < 0 . 001 between esgReDDM/+ and esgReDDM > Kr-h1 RNAi mated , t-test with Holm's correction ) typically observed after 7 days of mating , while overexpression of Kr-h1 constructs from the same cells recapitulates the effect of mating in virgins ( proliferation , I , p < 0 . 001 between esgReDDM/+ and esgReDDM > Kr-h1GS virgin , p < 0 . 001 between esgReDDM/+ and esgReDDM > Kr-h1UAS virgin , negative binomial GLM with Holm's correction; gut size J , p < 0 . 001 between esgReDDM/+ and esgReDDM > Kr-h1UAS virgin , t-test with Holm's correction ) . RNAi constructs against Kr-h1 and SREBP are effective in downregulating these genes; they decrease transcript abundance compared to a tubts control when expressed globally in larvae for 3 hr at 29°C ( K , p = 0 . 02 for Kr-H1 , p < 0 . 001 for both SREBP constructs , paired one-tailed t-test ) . Downregulating gce constitutively from ECs using Mex-Gal4 significantly suppresses the transcriptional increase of the lipid metabolism gene bgm upon mating , as indicated by the intensity ranking of a gce reporter ( L , p = 0 . 004 between gce RNAi/+ and Mex > gce RNAi , p = 0 . 02 between Mex > gce RNAi and Mex/KK control , Mann–Whitney test with Holm's correction; relevant comparisons with Met RNAi are not significant ) . See Table 1 for full genotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 06930 . 007 To establish the cellular targets of JH and its mode of action , we interfered with JH signalling in a cell-autonomous manner in the intestine . We used the esgReDDM system , based on the widely used esg-Gal4 driver ( Micchelli and Perrimon , 2006 ) , to target both classes of intestinal progenitor cells ( ISCs and EBs ) . We first confirmed that expression of RNAi transgenes against either of the two previously identified JH receptors Methoprene-tolerant ( Met ) or germ cell-expressed bHLH-PAS ( gce ) ( Jindra et al . , 2013; Jindra et al . , 2015 ) resulted in a significant reduction in their transcript levels ( Figure 3—figure supplement 1 ) . We then confined expression of these RNAi transgenes against Met , gce , or their target Kr-h1 to adult intestinal progenitors using esgReDDM . Downregulation of any of these three genes fully prevented both the proliferative response to mating and midgut re-sizing , whereas overexpression of Kr-h1 led to mating-like responses in virgin females ( Figure 3G , J , Figure 3—figure supplement 1 ) , indicating direct actions of JH on intestinal progenitors . Intestinal progenitors with downregulated JH receptors were found in numbers comparable to those of controls in virgin females and were able to increase their proliferation in response to a JH-unrelated stimulus: the ROS-inducing compound paraquat ( Biteau et al . , 2008 ) ( Figure 3—figure supplement 1 ) . This indicates that they remain competent to respond to the well-studied homeostatic machinery that maintains gut integrity ( Jiang and Edgar , 2012 ) , and suggests that mating- and damage-induced proliferative mechanisms may differ and can be uncoupled . In ECs , targeted by the specific driver Mex-Gal4 ( Phillips and Thomas , 2006 ) , only downregulation of gce strongly reduced the mating-dependent upregulation of a bgm reporter ( Figure 3K , Figure 3—figure supplement 1 ) . Together , these findings show that intestinal remodelling results from a rise in systemic JH triggered by mating . JH signals directly to intestinal progenitors to yield a larger organ in a Met and gce-dependent manner . Acting predominantly through gce , JH also adjusts gene expression in ECs to support increased lipid metabolism . Intestinal remodelling during reproduction could result from increased nutrient intake ( O'Brien et al . , 2011 ) or utilisation by the developing offspring . Alternatively , it may occur in preparation for , but be uncoupled from , such nutritional demands . Consistent with the latter idea , the mating-triggered changes in proliferation , midgut size , and SREBP activity are all still apparent in sterile female ovoD1 mutant flies in which egg production is blocked prior to vitellogenesis and which do not increase food intake after mating ( Barnes et al . , 2008 ) ( Figure 4—figure supplement 1 ) . To investigate the significance of intestinal remodelling , we used several RNAi transgenes to downregulate either the JH receptors or SREBP , which is activated by mating , specifically in adult ECs . In all cases , EC-specific downregulation led to a reduction in the number ( but not viability ) of eggs produced ( Figure 4E , F and Figure 4—figure supplement 1 ) , indicating that JH signalling is required to specifically enhance the quantity ( fecundity ) , but not the quality ( viability ) , of reproductive output . Progenitor cell-specific downregulation may also be expected to reduce fecundity; however , we detected expression of several intestinal progenitor drivers outside the intestine ( data not shown ) , which could affect egg production independently of the intestine . More specific tools will be necessary to resolve this important issue . 10 . 7554/eLife . 06930 . 008Figure 4 . Metabolic remodelling of ECs by JH sustains reproduction . Lipid-harbouring tissues ( fat body , posterior midgut , and ovary ) are found in close proximity in the fly's abdomen ( represented schematically in A , and in confocal microscopy in D ) . The amount of stored triglycerides ( TAG ) in the carcass of 3-day mated sterile female flies is increased compared to virgins ( B , p = 0 . 003 in w1118 , p = 0 . 009 in OregonR , t-test ) , as quantified by thin-layer chromatography ( C ) . Adult-specific downregulation of JH receptors gce and Met or SREBP in ECs reduces the total progeny produced by females in the 6 days following their first mating ( E , p = 0 . 01 between Met RNAi/+ and Mexts > Met RNAi , p = 0 . 007 between Mexts > Met RNAi and Mexts/KK control , p < 0 . 001 between gce RNAi/+ and Mexts > gce RNAi , p = 0 . 007 between Mexts> gce RNAi and Mexts/KK control; F p = 0 . 04 between SREBP RNAi/+ and Mexts > SREBP RNAi , p = 0 . 04 between Mexts > SREBP RNAi and Mexts/TRiP control , t-test with Holm's correction ) . In the absence of the ovarian lipid sink in sterile ovoD1 virgin flies , treatment with JHa increases neutral lipid content , as revealed by Oil Red O staining , in the posterior midgut ( G , G′ , quantified in H: p = 0 . 002 , t-test ) . Acute block of lipid export by heat-shock activation of lpp > stop > LTP RNAi ( Palm et al . , 2012 ) in virgin females results in heavy accumulation of neutral lipid in this gut region , further indicating that this midgut region provides a net source of lipid in adult flies ( I , quantified in J: p < 0 . 001 between LTP RNAi/+ and lpp > stop > LTP RNAi , p < 0 . 001 between lpp > stop > LTP RNAi and lpp > stop>/+ , p < 0 . 001 between lpp > stop > LTP RNAi and lpp > stop > LTP RNAi heat shock control , t-test with Holm's correction ) . See Table 1 for full genotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 06930 . 00810 . 7554/eLife . 06930 . 009Figure 4—figure supplement 1 . Reproductive intestinal remodelling is uncoupled from germline demands and is needed to sustain reproduction . Sterile females carrying the ovoD1 mutation experience the post-mating increase in progenitor proliferation ( A , p < 0 . 001 for w1118 and ovoD1 , negative binomial GLM , visualised in C and C′ using the esgReDDM tracing system ) , gut size increase ( B , p < 0 . 001 for w1118 and ovoD1 , t-test ) , and SREBP reporter activation ( D , p < 0 . 001 for w1118 and p = 0 . 005 for ovoD1 , Mann–Whitney test , visualised in E and E′ ) . The role of intestinal remodelling in enhancing reproductive capacity is confirmed with additional RNA interference lines against the JH receptor gce ( chosen because of its larger effect in Figure 4E; F , p = 0 . 002 between GD11178/+ and Mexts> GD11178 , p = 0 . 008 between Mexts > GD11178 and Mexts/+ , p < 0 . 001 between GD47465/+ and Mexts> GD47465 , p = 0 . 003 between Mexts > GD11178 and Mexts/+ , t-test with Holm's correction ) and SREBP ( G , p = 0 . 008 between GD37641/+ and Mexts > GD37641 , p < 0 . 001 between GD37640/+ and Mexts > GD37640 , t-test ) . Despite these effect on fecundity , eggs laid by gce , Met , or SREBP RNAi mothers are viable ( H and I , mean hatched fraction >0 . 9 for all groups , p > 0 . 05 for all relevant comparisons , t-test ) . See Table 1 for full genotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 06930 . 009 The anatomical proximity between the ovary and the posterior midgut region where changes in lipid gene expression and activity take place ( Figure 4A , D ) raises the intriguing possibility that enhanced nutrient delivery from the intestine to the ovary may occur locally , to maximise loading into eggs . As the trafficked nutrients would be therefore released in the form of eggs , we used sterile ovoD1 females to quantify lipid content , reasoning that it might accumulate in gut and/or peripheral fat stores in the absence of the local ovarian sink ( Parra-Peralbo and Culi , 2011 ) . Consistent with this idea , sterile ovoD1 female flies accumulate peripheral fat after mating ( Figure 4B , C ) , and lipid accumulation in the posterior midgut could be induced in fertile flies by either treatment of these sterile flies with JHa or by knocking down lipid shuttling proteins acutely , thereby blocking all lipid circulation ( Palm et al . , 2012 ) ( Figure 4G–J ) . Together , these data show that the metabolic reprogramming of ECs by JH supports fecundity , thus confirming that intestinal plasticity is required to sustain reproductive output at the time of peak fertility . The importance of intestinal lipogenesis is becoming increasingly recognised in both flies and mammals ( Lodhi et al . , 2011; Palm et al . , 2012; Sieber and Thummel , 2012; Song et al . , 2014 ) , and here we show that it underpins reproductive output . Notably , upregulation of SREBP target genes has been reported in the small intestine of lactating rats ( Athippozhy et al . , 2011 ) , suggesting that our findings may be conserved beyond insects .
The onset of reproduction involves a significant shift in metabolic demands , now routed towards the growing offspring as well as the mother . Drosophila may experience a particularly extreme example of this shift after mating: an event that enhances egg production tenfold and triggers multiple metabolic and behavioural adaptations ( Kubli , 2003; Rogina et al . , 2007; Avila et al . , 2011; Cognigni et al . , 2011 ) . These changes are in large part brought about by signals delivered by the male during copulation , in particular the Sex Peptide ( Kubli , 2003; Avila et al . , 2011 ) . Several reports connect SP to the corpus allatum and JH production ( Moshitzky et al . , 1996; Bontonou et al . , 2015 ) , suggesting that the systemic effects of mating via SP could be carried out through this pathway . Intriguingly , both JH knockdown in females ( Yamamoto et al . , 2013 ) and SP deficiency in males ( Wigby and Chapman , 2005 ) extend female lifespan while reducing reproductive output and/or peak fertility . This ‘cost of mating’ on lifespan is not relieved by sterility ( Ueyama and Fuyama , 2003 ) , suggesting that physiological effects in non-reproductive tissues are responsible . The intestinal reprogramming that we describe here represents a novel physiological target of postmating plasticity ideally placed at the interface between nutrition and reproduction . Ageing in flies is accompanied by reduced fertility ( Economos et al . , 1979 ) and intestinal dysplasia ( Biteau et al . , 2008; Choi et al . , 2008 ) , and genetic manipulations that affect intestinal progenitors can affect lifespan ( Biteau et al . , 2010; Rera et al . , 2011 ) . Thus , it will be informative to explore the links between JH-triggered postmating responses in lifespan , age-dependent intestinal and reproductive dysfunction , and lifetime fertility . In larvae , the corpora allata integrate age and metabolic status information to optimise developmental progression ( Mirth and Shingleton , 2012; Sarraf-Zadeh et al . , 2013 ) . Increasing evidence is revealing that , in adults , this insect endocrine organ acts as a nexus that detects changes in the organism's circumstances and condition to alter its metabolic and/or reproductive set points . It does so through regulated release of JH , with pleiotropic effects on ovarian maturation , adipose tissue , learning and memory , diapause , innate immunity , and ageing ( Nijhout and Riddiford , 1974; Jowett and Postlethwait , 1980; Fahrbach and Robinson , 1996; Wyatt and Davey , 1996; Flatt et al . , 2005; Riddiford , 2012; Yamamoto et al . , 2013 ) . Some of these effects may be modulated by crosstalk between JH and other systemic signals such as insulin-like peptides and ecdysteroids ( Jindra et al . , 2013; Koyama et al . , 2013; Mirth et al . , 2014; Rauschenbach et al . , 2014 ) , but the cellular and molecular targets of JH action remain incompletely understood . Our findings have uncovered a direct and functionally significant effect on adult organ plasticity by showing that JH promotes proliferation and resets gut size through its actions on intestinal progenitors and activates expression of lipid metabolism genes in ECs . Based on structural and functional similarities , insect JH has been compared to mammalian thyroid hormones ( Flatt et al . , 2006 ) : key energy balance regulators often associated with gastrointestinal alterations when pathologically dysregulated ( Middleton , 1971 ) . Given the well-established changes in thyroid function during human pregnancy ( Glinoer , 1997 ) , it will therefore be of interest to explore their contribution to reproductive intestinal remodelling . Downstream of its receptor ( s ) , relay of JH signalling in the intestine may differ from the classical model in which Met and gce act redundantly ( Abdou et al . , 2011; Jindra et al . , 2013 ) . Indeed , downregulation of either gene alone is sufficient to prevent the mating-induced changes in intestinal progenitors: a finding that we confirmed by observing that viable Met27 mutants also fail to undergo mating-induced remodelling ( data not shown ) . The actions of Met and Gce may also be cell-specific , as suggested by a preferential requirement for gce in ECs , and may result from different Met/gce expression levels ( our unpublished observations ) and/or interacting partners . Candidates to consider include Taiman , homologous to the mammalian steroid receptor coactivator 1 ( SRC-1 ) /NCoA-1/p160 ( Charles et al . , 2011; Li et al . , 2011; Zhang et al . , 2011 ) and , more intriguingly , circadian clock proteins: Met-binding partners recently shown to coordinate the switch from diapause to reproduction in other insects ( Shin et al . , 2012; Bajgar et al . , 2013 ) . Adult organ plasticity is not a peculiarity of Drosophila reproduction; examples of changes in intestinal size and nutrient utilisation are widespread across the animal kingdom in response to both environmental and internal challenges ( Carey , 1990; Hammond , 1997; Piersma and Lindstrom , 1997; Speakman , 2008; O'Brien et al . , 2011 ) . Although intestinal remodelling has not been assessed in human pregnancy , it could be one of the major drivers for the changes in gut microbiota observed during pregnancy ( Koren et al . , 2012 ) and could contribute to changes in gastrointestinal physiology , common during pregnancy ( Keller et al . , 2008 ) . Resetting of anatomical or metabolic features of internal organs may thus be a broadly used strategy to achieve a positive energy balance which , when matched to the developing offspring's demands , will contribute to reproductive success . However , if deployed in the absence of such demands , organ remodelling could contribute to the weight gain and increased fat mass that has been observed upon gonadectomy of multiple species including mice , rats , cats , monkeys , and other mammals ( Hansen et al . , 2013 and references therein ) . In a more physiological context , inappropriate persistence of such metabolic remodelling beyond pregnancy and lactation could similarly contribute to post-pregnancy weight retention in humans—a phenotype that , at least in mice , is correlated with enhanced intestinal function ( Casirola and Ferraris , 2003; Gore et al . , 2003 ) . Similarly , inappropriate persistence of JH-like mechanisms that change the homeostatic set point of adult stem cells and their progeny to transform an organ may also help explain why pregnancy changes the susceptibility to certain cancers ( Gwinn et al . , 1990 ) .
For wild-type experiments , the genetic backgrounds w1118 , OregonR , and CantonS were used as indicated in the figures and/or full genotypes list ( Table 1 ) . The following transgenic and mutant stocks were used: esg-Gal4 ( Bloomington , unknown insertion ) , tub-Gal80 ( Bloomington 7018 , McGuire et al . , 2003 ) , UAS-mCD8::GFP ( Bloomington 5130 , Lee and Luo , 1999 ) , UAS-H2B::RFP ( presumed from Langevin et al . , 2005 ) , caudal-Gal4 ( insertion used in Ryu et al . , 2008 ) , SREBP-Gal4 ( Bloomington 38395 , Kunte et al . , 2006 ) , bgm-lacZ ( Bloomington 28120 , Min and Benzer , 1999 ) , Aug21-Gal4 ( Bloomington 30137 , Siegmund and Korge , 2001 ) , UAS-NiPp1 ( Bloomington 23712 , Parker et al . , 2002 ) , tub-Gal4 ( Bloomington 5138 , Lee and Luo , 1999 ) , Mex-Gal4 ( Phillips and Thomas , 2006 ) , UAS-Kr-h1 ( DGRC 120052 , referred to as UAS-Kr-h1 ) , ovoD1 ( Busson et al . , 1983 ) , hs-FLP; lpp-Gal4 and UAS > stop > LTP RNAi stocks ( both from Palm et al . , 2012 ) . RNAi constructs were obtained from VDRC for gce ( KK101814 , GD11178 and GD47465 ) , Met ( KK100638 ) , Kr-h1 ( KK107935 ) , and SREBP ( GD37641 and GD37640 ) , as well as the genetically matched KK control ( KK60100 ) ; and from the Bloomington TRiP collection for SREBP ( 34073 ) and the genetically matched TRiP control ( GFP in valium10 , 35786 ) . Because the control stocks are generated in the same background as the RNAi lines used , the Gal4 parental control ( e . g . , yv; Mex-Gal4/+; tub-Ga80ts/UAS-GFP ) is genetically matched to the experimental genotype ( e . g . , yv; Mex-Gal4/+; tub-Gal80ts/UAS-SREBP RNAi TRiP ) . The line referred to as UAS-Kr-h1GS is GS ( 2 ) 73ES2b , which was isolated in a genetic screen for enhancer/suppressors of a large-eye phenotype caused by Dl overexpression in the Dominguez lab . Genomic DNA flanking the P-element insertions in the GS ( 2 ) 73ES2b stock were recovered by inverse PCR and sequenced . A BLAST search with the obtained sequence produced perfect matches to the genomic region upstream of the Kr-h1 gene ( 26B5 Chromosome 2L: 6 , 082 , 603 , . . . , 6 , 096 , 498 ) . 10 . 7554/eLife . 06930 . 010Table 1 . Full genotypesDOI: http://dx . doi . org/10 . 7554/eLife . 06930 . 010Genotype in text/figureFull genotypeFigure panel ( s ) esgReDDMw; esg-Gal4 , UAS-mCD8::GFP/+; tub-Gal80ts , UAS-H2B::RFP/+; +Figure 1A , Figure 1—figure supplement 1C , D , Figure 3Bw1118 backgroundw1118; +; +; +Figure 1B–D , Figure 2E , Figure 3E , H , Figure 3—figure supplement 1H , Figure 4—figure supplement 1A , BOregonR background+; +; +; +Figure 1E , Figure 1—figure supplement 1A , B , Figure 2F , Figure 3—figure supplement 1Hcaudal > H2B::RFPw; caudal-Gal4/+; UAS-H2B::RFP/+; +Figure 1F , Figure 3—figure supplement 1ASREBP > CD8::GFPw/+; SREBP-Gal4/+; UAS-CD8::GFP/+; +Figure 2A , B , Figure 3C , D , Figure 4D , Figure 4—figure supplement 1Dbgm-lacZw/+; bgm-lacZ/+; +; +Figure 2C , DCantonS background+; +; +; +Figure 3AAug21/+w; Aug21-Gal4/+; +; +Figure 3F , IAug21 > NiPp1w; Aug21-Gal4/+; UAS-NiPp1/+; +Figure 3F , INiPp1/+w; +; UAS-NiPp1/+; +Figure 3F , IesgReDDM/+w; esg-Gal4 , UAS-mCD8::GFP/+; tub-Gal80ts , UAS-H2B::RFP/+; +Figure 3G , J , Figure 3—figure supplement 1D , F , G , I , JesgReDDM > gce RNAiw; esg-Gal4 , UAS-mCD8::GFP/UAS-gce RNAi KK101814; tub-Gal80ts , UAS-H2B::RFP/+; +Figure 3G , J , Figure 3—figure supplement 1C–GesgReDDM > Met RNAiw; esg-Gal4 , UAS-mCD8::GFP/UAS-Met RNAi KK100638; tub-Gal80ts , UAS-H2B::RFP/+; +Figure 3G , JesgReDDM > Kr-h1 RNAiw; esg-Gal4 , UAS-mCD8::GFP/UAS-Kr-h1 RNAi KK107935; tub-Gal80ts , UAS-H2B::RFP/+; +Figure 3—figure supplement 1I , JesgReDDM > Kr-h1GSw; esg-Gal4 , UAS-mCD8::GFP/UAS-Kr-h1GS;tub-Gal80ts , UAS-H2B::RFP/+; +Figure 3—figure supplement 1IesgReDDM > Kr-h1UASw; esg-Gal4 , UAS-mCD8::GFP/UAS-Kr-h1UAS; tub-Gal80ts , UAS-H2B::RFP/+; +Figure 3—figure supplement 1I , Jtubts> Met RNAiw; tub-Gal80ts/UAS-Met RNAi KK100638; tub-Gal4 , UAS-mCD8::GFP/+; +Figure 3—figure supplement 1Btubts > gce RNAiw; tub-Gal80ts/UAS-gce RNAi KK101814; tub-Gal4 , UAS-mCD8::GFP/+; +Figure 3—figure supplement 1Btubts > Kr-h1 RNAiw; tub-Gal80ts/UAS-Kr-h1 RNAi KK107935; tub-Gal4 , UAS-mCD8::GFP/+; +Figure 3—figure supplement 1Ktubts > SREBP RNAi GDw; tub-Gal80ts/UAS-SREBP RNAi GD37640; tub-Gal4 , UAS-mCD8::GFP/+; +Figure 3—figure supplement 1Ktubts > SREBP RNAi TRiPw; tub-Gal80ts +; tub-Gal4 , UAS-mCD8::GFP/UAS-SREBP RNAi TRiP34073; +Figure 3—figure supplement 1Ktubts/+w; tub-Gal80ts/+; tub-Gal4 , UAS-mCD8::GFP/+; +Figure 3—figure supplement 1B , K ( control ) Met RNAi/+w; bgm-lacZ/UAS-Met RNAi KK100638; +; +Figure 3K , Figure 3—figure supplement 1LMex > Met RNAiw; Mex-Gal4 , bgm-lacZ/UAS-Met RNAi KK100638; +; +Figure 3K , Figure 3—figure supplement 1Lgce RNAi/+w; bgm-lacZ/UAS-gce RNAi KK101814; +; +Figure 3K , Figure 3—figure supplement 1LMex > gce RNAiw; Mex-Gal4 , bgm-lacZ/UAS-gce RNAi KK101814; +; +Figure 3K , Figure 3—figure supplement 1LMex/KK controlw; Mex-Gal4 , bgm-lacZ/attp40; +; +Figure 3K , Figure 3—figure supplement 1LovoD1 ( w1118 background ) w1118; +; ovoD1/+; +Figure 4C , Figure 4—figure supplement 1A , BovoD1 ( OregonR background ) +/w1118; +; ovoD1/+; +Figure 4C , G , HMet RNAi/+w; UAS-Met RNAi KK100638/+; +; +Figure 4E , Figure 4—figure supplement 1HMexts> Met RNAiw; Mex-Gal4/UAS-Met RNAi KK100638; tub-Gal80ts/+; +Figure 4E , Figure 4—figure supplement 1Hgce RNAi/+w; UAS-gce RNAi KK101814/+; +; +Figure 4E , Figure 4—figure supplement 1HMexts > gce RNAiw; Mex-Gal4/UAS-gce RNAi KK101814; tub-Gal80ts/+; +Figure 4E , Figure 4—figure supplement 1HMexts/KK controlw; Mex-Gal4/attp40; tub-Gal80ts/+; +Figure 4E , Figure 4—figure supplement 1HSREBP RNAi/+w/y , v; +; UAS-SREBP RNAi 34073; +Figure 4FMexts> SREBP RNAiw/y , v; Mex-Gal4/+; tub-Gal80TS/UAS-SREBP RNAi 34073; +Figure 4FMexts/TRiP controlw/y , v; Mex-Gal4/+; tub-Gal80ts/UAS-GFP; +Figure 4Fhs-FLP lpp > stop > LTP RNAiw , hs-FLP/w; lpp-Gal4/+; UAS > stop > LTP RNAi/+; +Figure 4I , JUAS > stop > LTP RNAi/+w; +; UAS > stop > LTP RNAi/+; +Figure 4Jhs-FLP lpp/+w , hs-FLP/w; lpp-Gal4/+; +; +Figure 4JesgReDDM/ovoD1ovoD1/w; esg-Gal4 , UAS-mCD8::GFP/+; tub-Gal80ts , UAS-H2B::RFP/+; +Figure 4—figure supplement 1CSREBP > CD8::GFP/ovoD1w/+; SREBP-Gal4/+; UAS-CD8::GFP/ovoD1; +Figure 4—figure supplement 1D , Egce RNAi GD11178/+w; UAS-gce RNAi GD11178; +; +Figure 4—figure supplement 1FMexts> gce RNAi GD11178w; Mex-Gal4/UAS-gce RNAi GD11178; tub-Gal80ts/+; +Figure 4—figure supplement 1Fgce RNAi GD47465/+w; +; UAS-gce RNAi GD47465/+; +Figure 4—figure supplement 1FMexts> gce RNAi GD47465w; Mex-Gal4/+; tub-Gal80ts/UAS-gce RNAi GD47465; +Figure 4—figure supplement 1FMexts/+w; Mex-Gal4/+; tub-Gal80ts+; +Figure 4—figure supplement 1FSREBP RNAi GD37641/+w; UAS-SREBP RNAi GD37641/+; +; +Figure 4—figure supplement 1G , IMexts> GD37641w; Mex-Gal4/UAS-SREBP RNAi GD37641; tub-Gal80ts/+; +Figure 4—figure supplement 1G , ISREBP RNAi GD37640/+w; UAS-SREBP RNAi GD37640/+; +Figure 4—figure supplement 1G , IMexts> RNAi GD37640w; Mex-Gal4/+; tub-Gal80ts/UAS-SREBP RNAi GD37640; +Figure 4—figure supplement 1G , I Fly stocks were reared on a standard cornmeal/agar diet ( 5 . 5% cornmeal , 6% dextrose , 1 . 3% yeast , 0 . 55% agar supplemented with 0 . 18% nipagin and 2 . 9 ml/l propionic acid ) or ‘Iberian’ diet ( 4 . 4% wheat flour , 6% brown sugar , 3% yeast , 1% agar supplemented with 0 . 04% nipagin and 7 . 6 ml/l of propionic acid ) . All experimental flies were kept at 25°C expect for those containing temperature-sensitive regulation ( tub-Gal80ts ) , which were set up at 18°C ( restrictive temperature ) and transferred to 29°C ( permissive temperature ) at the time when activation was needed in the specific experiment . For all experiments , experimental and control flies were handled in parallel and experienced the same temperature shifts and treatments . For the analysis of mating and JHa effects , virgin female flies were collected at eclosion , aged for 4–5 days on standard food and then transferred for 3 days ( 7 days for flies harbouring the esgReDDM transgenes , as these flies show a delay in mating responses at 29°C ) into new tubes in the presence of wild-type males ( typically 4–5 females + 6 males ) or food supplemented with 1 . 5 mM methoprene ( Sigma-Aldrich , St Louis , MO , PESTANAL 33375 , racemic mixture ) , added to freshly prepared food when still liquid but <50°C . This concentration was chosen in a pilot dilution test from 0 . 5 to 7 . 5 mM as the one that induced activation of the SREBP-Gal4 reporter to levels comparable to mating , and corresponds to approximately half of the concentration used in a previous study ( Flatt and Kawecki , 2007 ) . Controls were age-matched virgin females , also transferred to new tubes for the same time but without the addition of males and/or methoprene . For the paraquat experiments , virgin female flies were raised at 18°C and aged for 4–5 days after eclosion , at which point they were starved for 4 hr without water . The flies were then transferred to vials containing filter paper soaked in 5% sucrose with or without 10 mM paraquat dichloride ( Sigma-Aldrich ) . After spending 24 hr at 29°C in these vials , their midguts were dissected and stained for pH3 as described before . The following antibodies were used: rabbit anti-pH3 ( 1:2000 , Upstate , Merck-Millipore , Germany ) , sheep anti-GFP ( 1:1000 , Biogenesis , for esgReDDM staining ) , goat anti-GFP ( 1:1500 , Abcam , UK , for SREBP > mCD8::GFP staining ) , rabbit anti-β-galactosidase ( 1:5000 , MP Biomedicals , Santa Ana , CA ) ; secondary antibodies were either FITC/Cy3 conjugates from Jackson ImmunoResearch ( 1:200 , West Grove , PA , for SREBP > mCD8::GFP and bgm-lacZ ) or Alexa488/647 conjugates from Invitrogen Life Technologies ( 1:1000 , Carlsbad , CA , for esgReDDM and caudal > H2B::RFP ) . Preparations for proliferation analysis were counterstained with DAPI ( Sigma-Aldrich ) and mounted in Fluoromount-G ( Southern Biotech , Birmingham , AL ) . Preparations for reporter analysis were mounted in Vectashield with DAPI ( Vector Labs , Burlingame , CA ) . Quantification of mitoses in wild-type and ovoD1 female flies was carried out by counting individual nuclei marked by the mitotic marker pH3 using a Nikon Eclipse 90i Fluorescence microscope through a 40× objective . For the acquisition of gut images in these samples , a single 1392 × 1040 field was acquired posterior to the midgut–hindgut boundary using QCapture software ( QImaging ) . Progeny dynamics were analysed using the esgReDDM system ( Antonello et al . , 2015 ) , which has the genetic makeup esg-Gal4 , UAS-mCD8::GFP; tub-Gal80ts , UAS-H2B::RFP . At the permissive temperature of 29°C , the GFP reporter is expressed in esg-Gal4 positive cells ( ISCs and EBs ) , but due to the perdurance of the RFP-tagged histone H2B::RFP the esg-Gal4-negative progeny ( including ECs and EECs ) generated from these progenitors since the shift to permissive temperature is additionally labelled in red . To restrict progeny analysis to mating-induced changes , esgReDDM flies were maintained at 18°C , such that Gal4 expression is suppressed by tub-Gal80ts , and moved to 29°C only at the time of mating . After 3 days of mating at 29°C , guts were dissected and stained for GFP and pH3 ( the endogenous RFP signal was detected directly ) . EC number in the posterior midgut was assessed by imaging the entire gut of caudal > H2B::RFP flies and counting the number of RFP-marked cells . Confocal images were obtained with a Leica TCS SP5 inverted confocal microscope using a 20× air objective for esgReDDM and a 10× air objective for caudal > H2B::RFP . Stacks were typically collected every 1 µm , and the images ( 1024 × 1024 ) were reconstructed using maximum projection . Bright-field images or confocal maximum projections were loaded into ImageJ ( Schneider et al . , 2012 ) and the line tool used to quantify the width of the gut across the centre of the image . ImageJ was also used to outline the guts of esgReDDM flies using the polygon tool before analysing the resulting region of interest ( ROI ) with a custom MATLAB ( The MathWorks , Inc . ) script optimised for the ReDDM method . Extended details about this analysis are available from ( Antonello et al . , 2015 ) . Briefly , maximum projections were adjusted for levels and offsets and filtered to remove noise ( using always the same parameters for scans within one experiment ) , then the area of the gut was identified by background staining and the cell nuclei by DAPI signal . The size of nuclei can be used to discriminate between diploid cells ( ISCs , EBs , and EECs ) and polyploid ECs . The red-labelled nuclei ( persistent H2B::RFP ) and green-labelled cells ( mCD8::GFP ) were identified by segmentation and compared to the pattern of nuclei defined by DAPI to generate a report of total ECs ( large DAPI cells ) , total progenitors + progeny ( RFP signal ) , total ISCs and EBs ( GFP signal ) , and total area . The same script was also used to count the number of caudal > H2B::RFP cells . For SREBP > mCD8::GFP and bgm-lacZ experiments , confocal images were obtained with a Leica SP5 upright confocal microscope using a 20× glycerol immersion objective . A single 20× field ( 1024 pixels wide ) immediately posterior to the midgut–hindgut boundary was acquired with a Z resolution of 1 . 5 µm . ImageJ was used to generate a maximum projection for each sample and all images pertaining to one experiment were loaded as separate layers into a single Adobe Photoshop CS6 file . The layers were then ranked blindly on the basis of their relative intensity in the relevant channel . To quantify mating-induced changes in gene expression , posterior midguts from at least 10 adult female flies were dissected , discarding Malphigian tubules and the hindgut . To determine the knockdown efficiency of the RNAi transgenes , tub-Gal80ts; tub-Gal4 , UAS-GFP was used to downregulate them ubiquitously . 8–10 third instar larvae were collected from crosses kept at 21°C and were shifted to 29°C for 3 hr to allow RNAi transgene expression . Samples ( posterior midguts or whole larvae ) were directly stored on dry ice and at −80°C in RNAlater TissueProtect Tubes ( Qiagen , the Netherlands ) until total RNA was extracted using RNeasy Mini Kit ( Qiagen ) , from which cDNAs were prepared with SuperScript First-Strand Synthesis System ( Invitrogen Life Technologies ) using oligo-dT primers . Quantitative PCR was performed using the SYBR Green PCR Master Mix ( Applied Biosystems Life Technologies ) in a 7500 Real-Time PCR System ( Applied Biosystems ) using the housekeeping gene rp49 as a control . All qPCRs were performed in triplicate and the relative expression was calculated using comparative Ct method . Primers used:Forward 5′–3′Reverse 3′–5′SREBPGCAAAGTGCGTTGACATTAACCAGTGTCGTGTCCATTGCGAAbgmGCAATCGATTTGCGTGACCAGGCCCAGGACGATTGTAGAGAcslCGGAGATCCGACAAAGCAGTTGAGCACAGCTCCTCAAAGGFASGACATTCGATCGACGCCTCTGCTTTGGCTTCTGCACTGACACCAATTCTCCAAGGCTCGTCCCCATGCCGCAATTGTTTTCGCKr-h1ACAATTTTATGATTCAGCCACAACCGTTAGTGGAGGCGGAACCTGgceAGCTGCGTATCCTGGACACTTCGAGAGCTGAAACATCTCCATMetCCGCCGTCCTTAGATTCGCGTTCCCTTGAGGCCGGTTTrp49TGTCCTTCCAGCTTCAAGATGACCATCCTTGGGCTTGCGCCATTTGTG Haemolymph was extracted from virgin or mated females using pulled glass microcapillary needles ( 10 μl vol , #2-000-010; Drummond Scientific , PA , USA ) . The needle tip was placed into the gap between the anepisternum and anepimeron of anesthetised flies , and haemolymph was collected using a slight vacuum ( 0 . 2–1 . 0 mPa ) for ∼30 s . Haemolymph from 45 to 50 flies was collected in the same needle . The contents were ejected into a 0 . 1 ml glass vial insert ( Thermo Fisher Scientific , MA , USA ) by applying pressurised air ( ∼5–6 kPa ) with a Femtojet microinjector ( Eppendorf , NY , USA ) , and weighed prior to extraction . 20 μl of MeOH was added to the haemolymph followed by extraction with 20 μl of hexane , repeated four times . Pooled hexane extract was evaporated under a gentle stream of N2 and reconstituted in 10 μl of hexane . All extracts were prepared and measured immediately after collection . Mass spectra were acquired with an atmospheric pressure ionisation time-of-flight mass spectrometer ( AccuTOF-DART , JEOL USA , Inc . ) equipped with a DART interface and operated with a resolving power of 6000 ( FWHM definition ) . The RF ion guide voltage was set at 600 V . The atmospheric pressure ionisation interface potentials were as follows: orifice 1 = 15 V , orifice 2 = 5 V , ring lens = 5 V . Mass spectra were stored at a rate of one spectrum per second with an acquired m/z range of 60–1000 . The DART interface was operated in positive-ion mode using helium gas with the gas heater set to 200°C . The glow discharge needle potential was set to 3 . 5 kV . Electrode 1 was set to +150 V , and electrode 2 was set to +250 V . Helium gas flow was set to 2 . 0 l/min . Calibration for exact mass measurements was accomplished by acquiring a mass spectrum of polyethylene glycol ( average molecular weight 600 ) as a reference standard in every data file . Analysis was done with JEOL MassCenter software ( version 1 . 3 . 0 . 1 ) . Accurate mass measures and isotope pattern matching by MassMountaineer ( FarHawk Marketing Services , Ontario , CA ) were used to support elemental composition assignments . 2 μl of the haemolymph hexane extract was placed on the tip of a borosilicate glass capillary . The capillary was introduced to the DART ion source with a micromanipulator , thus allowing for reproducible placement of the sample . Each extract was measured 4–5 times . The averaged signal intensity was normalised to the total weight of the haemolymph and converted to absolute quantities after establishing a calibration curve with a JHIII standard ( Santa Cruz Biotechnology , CA , USA , CAS 24198-95-6 ) . Analysis of JHIII by DART produces two signature ions at m/z 267 . 20 ( intact molecule ) and at m/z 249 . 18 ( loss of water ) , consistent with a previous study ( Navare et al . , 2010 ) . The abundance of the [M-H2O + H]+ signal peak was used for all measurements because the parent ion at m/z 267 . 20 could not be consistently resolved due to interference from other signals . To detect other juvenoid compounds , the following mass signatures were used: methylfarnesoate ( [M + H]+ 251 . 20 ) and JHIII Bisepoxide ( [M + H]+ 283 . 19 ) . DART MS was previously shown to be an effective method for quantitative and high-sensitivity measurements of JHIII ( Navare et al . , 2010 ) . Flies for fecundity and egg viability experiments were raised at 18°C to prevent the expression of the RNAi transgenes during development , then shifted to 29°C in late pupariation ( after ∼20 days ) . Virgin females were collected upon eclosion and after 4 days mated overnight to OregonR males ( 10 males , 10 females per vial ) . Males were then removed , individual female flies were transferred to a new single vial of yeast-supplemented standard food ( cornmeal/agar diet with 5% yeast content ) every 48 hr , and eggs were counted from the vacated vial to quantify fecundity . To assess egg viability , a fraction of the egg-containing vials were then maintained at 29°C , and the number of eclosed progeny was counted and compared with egg counts . Each genotype cross was performed three times , and egg production from each fly was assessed over three 48 hr repeats , covering a total of 6 days of egg laying . Ovaries and guts were removed from flies immobilised on ice and the remaining carcasses ( three flies per sample ) were immediately homogenised in a mixture of methanol ( 60 μl ) , chloroform ( 150 μl ) , and water ( 75 μl ) , following previously described procedures ( Al-Anzi et al . , 2009; Hildebrandt et al . , 2011 ) . After an extraction period ( 1 hr at 37°C ) , aqueous and organic phases were separated by the addition of a 1:1 mixture of 1 M potassium chloride and chloroform ( 75 μl each ) . Samples were briefly centrifuged and 120 μl of the organic phase was transferred to fresh Eppendorf tubes and left to air-dry for 3 hr . The resulting dessicated lipids were resuspended in 16 μl of a 1:1 chloroform:methanol mixture . 3 μl of each sample was applied to TLC plates ( Merck Millipore116487 ) and lipid species were separated by standing the plates in ∼1 cm of a mobile phase consisting of 69 . 5% hexane , 29 . 5% diethyl ether , and 1% acetic acid . Once the mobile phase had traversed the plates , they were briefly dried and then dipped in a cerium-ammonium-molybdate stain ( ammonium heptamolybdate tetrahydrate 2 . 5 g , cerium ( IV ) sulphate hydrate complex with sulphuric acid 1 g , water 90 ml , sulphuric acid 10 ml ) . The TLC plates were developed at 80°C for 25 min and then imaged on a digital scanner . The TAG content was quantified by analysing the resulting TIFF images using the densitometry tool in ImageJ software . All reagents were purchased from Sigma-Aldrich . The shuttling of lipids between organs was abolished by downregulating the apolipoprotein lipid transfer protein ( LTP ) through heat-shock-activated acute RNAi expression based on the pFRiPE system ( Marois and Eaton , 2007 ) . In the larva , this manipulation prevents the loading of gut-originated medium-chain diacylglycerides , which are a dominant component of circulating lipids , onto the haemolymph carrier Lipophorin ( Lpp ) and leads to the accumulation of stored lipid in the larval gut in triglyceride form ( Palm et al . , 2012 ) . The downregulation of LTP from the fat body driver lpp-Gal4 was triggered in virgin females by 1 hr heat-shock at 37°C; after 6 hr , the guts were dissected for neutral lipid detection using Oil Red O staining . Fly guts were dissected from flies immobilised on ice and were then fixed in a solution of 4% formaldehyde for 45 min . Guts were washed in consecutive applications of phosphate buffered saline ( PBS ) , double-distilled water , and a 60% isopropanol solution . Oil Red O ( Sigma-Aldrich ) stock was prepared as a 0 . 1% solution in isopropanol , then a freshly prepared working solution ( a 6:4 dilution in water ) was added for 20 min to the guts , then washed in 60% isopropanol and water . The preparations were mounted in glycerol for analysis , and the posterior midgut was imaged using either a Zeiss Axioplan stereo microscope equipped with Nomarski optics or an Olympus BX53 phase contrast microscope equipped with a 4×/0 . 13 UPlanFLN lens through CellSens software ( Olympus , Japan ) . The resulting TIFF files were analysed quantitatively using a custom ImageJ script: the gut was manually outlined as a ROI using the polygon tool , then the RGB channels were split and the red channel subtracted from the green to eliminate background ( grey ) signal . The mean intensity of the resulting signal within the ROI was calculated with the built-in Analyse Particles function . All statistical analyses were carried out in the R environment ( R Development Core Team , 2014 ) . Comparisons between normally distributed groups were carried out using Student's t-test ( R function t . test ) , unpaired , two-tailed and incorporating Welch's correction to account for unequal variances , followed by Bonferroni-Holm correction when multiple comparisons were applied . qPCR data were analysed comparing the housekeeping-subtracted Cts of experimentally matched virgin and mated samples , thus using paired t-test , one-tailed when confirming previous reporter experiments ( Figure 2E , F ) , and two-tailed when no prediction could be made ( panel H in Figure 3—figure supplement 1 ) . Count data with a distinctly non-normal distribution ( specifically , pH3 counts ) were fitted with a negative binomial model ( R function glm . nb from MASS package , Venables and Ripley , 2002 ) followed by likelihood ratio tests ( R function anova . negbin from MASS package ) . Rank-based experiments were analysed with the Mann-Whitney-Wilcoxon rank sum test ( R function wilcox . test ) . All graphs were generated in R using a custom script based on the base boxplot function superimposed with individual data points plotted with the beeswarm function ( package beeswarm ) . Confocal and bright-field images shown in conjunction were always acquired with the same settings as part of a single experiment . For visualisation purposes , level and channel adjustments were applied using Adobe Photoshop CS6 to the confocal images shown in figure panels ( the same correction in all comparable images ) , but all quantitative analyses were carried out on unadjusted raw images or maximum projections . In all figures , * indicates 0 . 05 > p ≥ 0 . 01 , ** indicates 0 . 01 > p ≥ 0 . 001 , and *** indicates p < 0 . 0001 . | Producing offspring places extra energy demands on individuals . Female animals—which generally invest more time and resources than the males—need to ensure that sufficient nutrients reach their offspring during pregnancy and lactation . The small intestines of many female animals increase in size during this period , but it is not clear to what extent these changes help to maximise reproduction , or how they are regulated . Reiff , Jacobson , Cognigni , Antonello et al . investigated what happens to the middle section of the gut in female fruit flies after mating . A fly's ‘midgut’ performs a similar role to the small intestine in humans and other mammals . The experiments show that mating increases the numbers of cells in the midgut so that it increases in size . These changes are due to a hormone called ‘juvenile hormone’ , which is released after the fly mates . In particular cells of the midgut , juvenile hormone also regulates some genes involved in the metabolism of lipids . If the activity of juvenile hormone is blocked in these cells , the female flies produce fewer eggs . These changes in the midgut still happen in mutant flies that cannot produce eggs and don't increase their food intake after they mate . Therefore , the changes appear to prepare flies for the increased nutritional demands rather than being caused by it . Altogether , these findings reveal that changes in the midgut play an important role in the ability of female fruit flies to reproduce . Similar changes to the gut may also increase reproductive success in humans and other mammals . However , if the changes are maintained after reproduction , it is possible that they may contribute to weight gain and an increased risk of cancer in females after pregnancy . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2015 | Endocrine remodelling of the adult intestine sustains reproduction in Drosophila |
Massive zygotic transcription begins in many organisms during the midblastula transition when the cell cycle of the dividing egg slows down . A few genes are transcribed before this stage but how this differential activation is accomplished is still an open question . We have performed ChIP-seq experiments on tightly staged Drosophila embryos and show that massive recruitment of RNA polymerase II ( Pol II ) with widespread pausing occurs de novo during the midblastula transition . However , ∼100 genes are strongly occupied by Pol II before this timepoint and most of them do not show Pol II pausing , consistent with a requirement for rapid transcription during the fast nuclear cycles . This global change in Pol II pausing correlates with distinct core promoter elements and associates a TATA-enriched promoter with the rapid early transcription . This suggests that promoters are differentially used during the zygotic genome activation , presumably because they have distinct dynamic properties .
The development of a fertilized egg is initially under the control of maternal products and then becomes under zygotic control when transcription begins . In animals such as Xenopus , zebrafish and Drosophila , development begins with rapid , synchronous cell divisions without gap phases ( Lamb and Laird , 1976; Newport and Kirschner , 1982; Tadros and Lipshitz , 2009 ) . During the midblastula transition ( MBT ) , cells switch to prolonged asynchronous divisions and this coincides with a massive increase in zygotic transcription . Before the MBT , only a small set of ‘pre-MBT genes’ are expressed . How this differential transcription is set up globally during these early stages of development , including the role of histone modifications and the recruitment of RNA Polymerase II ( Pol II ) , has been the subject of considerable interest ( Akkers et al . , 2009; Vastenhouw et al . , 2010; Lindeman et al . , 2011 ) . In mammalian embryonic stem cells , as well as early Drosophila embryos , paused Pol II is frequently found at developmental control genes ( Guenther et al . , 2007; Zeitlinger et al . , 2007; Min et al . , 2011 ) , but it is not known when pausing is first established in the embryo . Pol II pausing prior to activation may promote the rapid and synchronous induction of genes ( Boettiger and Levine , 2009; Adelman and Lis , 2012 ) , but it is unclear whether Pol II can be recruited and paused during the rapid early nuclear cycles prior to the MBT ( Kim and Jinks-Robertson , 2012; Petruk et al . , 2012 ) .
We first probed the status of Pol II in the early embryo by immunostainings ( Figure 1A ) . In Drosophila , the MBT mainly occurs in the interphase of nuclear cycle 14 , just before cellularization and subsequent gastrulation ( Foe et al . , 1993 ) , although there is some evidence that this may already occur in nuclear cycle 13 ( Harrison et al . , 2011 ) . Before the MBT , a small fraction of genes ( De Renzis et al . , 2007; ten Bosch et al . , 2006 ) may be transcribed as early as nuclear cycle 8 . 10 . 7554/eLife . 00861 . 003Figure 1 . Global recruitment of Pol II during the Drosophila midblastula transition . ( A ) Immunostainings of embryos during pre-blastoderm stages ( nc 1–7 ) , pre-MBT ( nc 8–12 ) and MBT ( nc 13–14 ) suggest that the initiated form of Pol II ( serine-5-phosphorylation of the CTD repeats—Ser5-P ) , as well as TBP are only detectible in the nuclei ( outlined by the Lam 0 in red ) of pre-MBT embryos when zygotic transcription begins ( scale = 20 μm ) . ( B ) Outline of the hand-sorting of embryo collections for ChIP-Seq experiments . ( C ) Heat map of ChIP-seq enrichments across all genes that are significantly bound by Pol II during MBT . Pre-MBT genes are also significantly bound in the pre-MBT sample; MBT-maternal genes also have maternally provided transcripts in the early embryo ( RPKM > 1 during nc 10 ) ; the remaining genes are MBT-zygotic genes . Among the latter group , MBT active genes are expressed during the MBT ( RPKM > 5 at nc 14D ) , while the transcript levels of MBT poised genes are below this threshold . Each line shows the normalized enrichments for a gene from −200 bp to +800 bp from the TSS . Note that Pol II is only bound to few pre-MBT genes before the MBT and that there is massive de novo recruitment of Pol II during the MBT . AED = after egg deposition , nc = nuclear cycle . DOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 00310 . 7554/eLife . 00861 . 004Figure 1—figure supplement 1 . Standards for staging pre-MBT and MBT embryos . Embryos from a pre-MBT collection ( 1–2 hr ) were screened for embryos in nuclear cycle 13–14 or later and removed . Using DIC , these embryos were recognized by the cellularization or gastrulation furrow ( marked by arrows in the post-MBT embryos ) , were removed . Under UV light , these embryos were recognized based on their number of nuclei with DAPI staining . Likewise , for MBT collections ( 2–3 hr ) , embryos with visible furrow formation or high DAPI staining were removed . The small number of embryos that were younger than the desired nuclear cycles were tolerated in both pre-MBT and MBT collections since they do not contribute a significant fraction of nuclei . DOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 00410 . 7554/eLife . 00861 . 005Figure 1—figure supplement 2 . High reproducibility of pre-MBT and MBT Pol II ChIP-seq data and agreement with previous mRNA data . ( A ) Scatterplot comparing two pre-MBT Pol II and two MBT Pol II replicates . For each replicate , the enrichment over background in each TSS region ( first 200 bp ) is displayed . Genes that qualified as pre-MBT genes or MBT genes based on all replicates are shown in red . ( B ) Pearson correlations between all Pol II ChIP replicates in pre-MBT and MBT embryos . The TSS read counts from each Pol II replicate ( numbered from r1 to r4 ) are compared with each other and with the input ( starting material for the ChIP ) for each stage . Note that the replicates have high correlations with each other , which far exceed the correlation with the input . ( C ) Venn diagram showing the strong overlap between pre-MBT active genes ( those in the ‘dual’ and ‘not paused’ group ) and early zygotic genes identified previously by mRNA microrarray analysis by Wieschaus et al . ( De Renzis et al . , 2007 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 00510 . 7554/eLife . 00861 . 006Figure 1—figure supplement 3 . ChIP-seq occupancy of Pol II and TBP at pre-MBT genes with complex patterns . The reads were normalized to the total read count . ( A ) MED11 , shown in red , is an example of a false pre-MBT gene due to read-through signal from an upstream gene , CG6885 . ( B ) CG11929 , shown in red , is an example of a false pre-MBT gene due to signal from an overlapping gene , Bsg25A . ( C ) Taf4 has an un-annotated more proximal TSS ( light grey ) that is used during the pre-MBT stage . The two distal known TSSs are used during MBT . ( D ) hb , a pre-MBT dual gene has two known alternative TSSs that are differentially used during early development . While the most proximal TSS is preferentially used in pre-MBT and MBT embryos , the distal TSS is used mostly in the post-MBT stage . DOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 00610 . 7554/eLife . 00861 . 007Figure 1—figure supplement 4 . Pre-MBT genes identified by ChIP-Seq show high conservation scores among insect genomes . Boxplot comparing conservation scores between gene groups . Each transcript’s score is the mean of the phastCons scores ( http://genome . cshlp . org/content/15/8/1034 ) along its annotated location from Flybase . DOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 007 Our immunostainings show that unphosphorylated Pol II can be detected in nuclei from the earliest cleavage stages on , thus before the beginning of transcription . However , both TATA-box binding protein ( TBP ) , the key subunit of TFIID , which binds to the promoter prior to Pol II recruitment , as well as Serine 5-phosphorylation of the C-terminal domain of Pol II ( Ser5-P Pol II ) , which marks transcriptional initiation , can only be first detected during nuclear cycles 8–12 , when significant transcription of pre-MBT genes occurs . This was the first indication that Pol II is recruited to promoters de novo during the zygotic genome activation . We next performed chromatin immunoprecipitation experiments coupled to deep sequencing ( ChIP-seq ) to analyze the occupancy of Pol II , TBP and histone modifications in pre-MBT embryos ( nuclear cycles 8–12 ) , MBT embryos ( nuclear cycles 13–14 ) , as well as post-MBT embryos as control ( Figure 1B ) . Although the large amount of Drosophila embryos required for ChIP-seq can be collected by conventional means , such collections always contain a fraction ( 5–20% ) of older embryos due to maternal egg holding and thus cannot be used to study very early stages of embryogenesis ( Harrison et al . , 2011 ) . To eliminate this contamination , we stained our embryo collections with DAPI and removed ‘out-of-stage’ embryos under a microscope with a pipette ( Figure 1B , Figure 1—figure supplement 1 , and see ‘Materials and methods’ ) . The ChIP-seq data from these hand-sorted embryos have robust and reproducible signals in replicates ( Figure 1—figure supplement 2A , B ) . Despite the high Pol II signal in the pre-MBT sample , Pol II only occupies around a hundred genes before the MBT ( Figure 1C , Supplementary file hosted by Dryad [7 . 6 Mb; Chen et al . , 2013] ) . These genes include previously described pre-MBT genes , as defined by in situ hybridization ( ten Bosch et al . , 2006 ) and microarray data ( De Renzis et al . , 2007 ) ( Figure 1—figure supplement 2C ) . In contrast , Pol II and TBP are recruited de novo to 4007 promoters during the MBT , which equates to roughly a third of all genes ( Figure 1C , Supplementary file hosted by Dryad [7 . 6 Mb] ) . This shows that there is massive de novo recruitment of Pol II during the MBT . To obtain a complete list of pre-MBT genes occupied by Pol II before the MBT , we identified all genes with at least twofold enrichment of Pol II over input at the transcription start site across four Pol II ChIP-seq replicates . From this list , we removed 12 genes that were likely false positives as a result of Pol II read-through from a nearby gene and added 10 genes that were missed due to un-annotated alternative start sites ( examples in Figure 1—figure supplement 3A , B , C ) . This yielded 117 pre-MBT genes , many of which have known functions in sex determination , cellularization , anterio-posterior patterning and dorso-ventral patterning ( Table 1 ) . Among them are also 14 precursors of non-coding RNAs , which are involved in maternal RNA degradation , dosage compensation , and RNA splicing , as well as many genes whose function is unknown but that are well conserved among insect species ( Figure 1—figure supplement 4 , Supplementary file hosted by Dryad [7 . 6 Mb] ) . 10 . 7554/eLife . 00861 . 008Table 1 . Classification of pre-MBT genesDOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 008FunctionGene namesSex determinationDpn , sisA , Sxl , osCellularizationnullo , Sry-alpha , kuk , bnk , slamAnterio-posterior patterningcad , hb , gt , kni , tll , eve , h , run , slp1 , odd , ftz , EgfrDorso-ventral patterningsna , esg , Nrt , glec , ac , l ( 1 ) sc , Tom , BobA , m4 , zen , zen2 , tsg , tld , scw , Neu2 , sc , fd19B , bnb , Bro , Brd , Ocho , amos , ato , sim , leaOther functionTaf4 , wech , Corp , toc , spri , Z600 , halo , SNCF , CG4570 , spo , hrg , sca , Lac , RpL3 , btsz , αTub84BNon-coding RNAmir-9a , mir-309 , roX1 , snRNA:U5:34A , snRNA:U4atac:82E , snRNA:U1:82 Eb , snRNA:U5:23D , snRNA:U5:38ABb , snRNA:U5:14B , snRNA:U4:38AB , snRNA:U1:95CcUnknown function Localized expressiongk , CG9894 , CG5059 , sala , term , CG14427 , CG8960 , CG13711 , CG13713 , CG15876 , CG6885 , CG7271 , CG14014 Ubiquitous expressionBsg25A , Bsg25D , CG15634 , CG15382 OthersCG2201 , CG42666 , CG43659 , CG13716 , CG13712 , CG13000 , CG13465 , CG14561 , CG18269 , CG14915 , CG16813 , CG15479 , CG15480 , CG4440 , CG14317 , CG13427 , CG34137 , CG34214 , CG34224 , CG34266 , CG16815 , CG42762 , CG43184 , CG 9775 , CG9883 , CR43887 , CG9821 , CG33232Bold marks the pre-MBT dual genes and italic marks the pre-MBT paused genes . Inspection of the Pol II occupancy revealed that most pre-MBT genes have no notable enrichment of Pol II at the pause site ( +30–50 bp downstream from the transcription start site [Zeitlinger et al . , 2007] ) when they are initially transcribed , while TBP occupancy is found upstream ( on average −20 bp from the transcription start site ) ( Figure 2 ) . When quantifying the degree of pausing with the pausing index ( see ‘Materials and methods’ ) , pre-MBT genes are indeed much less paused than MBT genes ( Figure 2A ) . However , a small number of pre-MBT genes have a higher pausing index , lack enrichment of Pol II within the gene body and thus may be paused . 10 . 7554/eLife . 00861 . 009Figure 2 . Minimal Pol II pausing before the MBT . ( A ) Violin plot of the Pol II pausing index distribution shows that pre-MBT genes ( during pre-MBT stages ) display less Pol II pausing than MBT genes ( during the MBT stage ) . The width of a violin plot is equivalent to a density curve showing the distribution of values ( here pausing indices ) within a dataset . ( B ) Median RNA-seq expression data ( Lott et al . , 2011 ) of the three pre-MBT groups and the two MBT groups shows that paused genes are expressed at lower levels and tend to be induced later . ( C ) Examples and average enrichment of Pol II occupancy ( blue ) and TBP occupancy ( pink ) for the three pre-MBT gene groups . Examples are shown as normalized reads while average enrichment is normalized to input . Note that pre-MBT not-paused genes have a non-paused Pol II profile since they do not show elevated Pol II levels at the pause site during pre-MBT and MBT stages . Pre-MBT dual genes switch from an initial non-paused profile during pre-MBT stages to a paused profile during the MBT . Pre-MBT paused genes appear to be paused even during pre-MBT stages . AED = after egg deposition , nc = nuclear cycle . DOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 009 We then classified the pre-MBT genes based on their Pol II occupancy during development and identified three distinct groups ( Figure 2B , C , Supplementary file hosted by Dryad [7 . 6 Mb] ) . Genes in the first group ( ‘pre-MBT not-paused’ , n = 77 ) have highest expression during cellularization ( nuclear cycle 14 ) and tend to diminish in expression thereafter ( Figure 2B ) . These genes appear never to become paused during early development ( see SNCF and the average profile in Figure 2C ) . Genes in the second group ( ‘pre-MBT dual’ , n = 30 ) initially show little evidence of pausing; however , higher levels of Pol II gradually accumulate at the pause site during and after the MBT ( see ac and the average profile in Figure 2C ) . Finally , there is a small group of genes ( ‘pre-MBT paused’ , n = 10 ) that appear to have Pol II that is paused or non-productive even at the pre-MBT stages ( see sim and the average profile in Figure 2C ) . This is consistent with the expression of these genes; their transcript levels rise much later during pre-MBT stages as compared to the first two groups ( Figure 2B ) . This suggests that Pol II pausing exists during pre-MBT stages but that most genes are non-paused . As a control , we analyzed Pol II pausing at genes that are newly occupied by Pol II during the MBT ( ‘MBT genes’ , Supplementary file hosted by Dryad [7 . 6 Mb] ) . We first subtracted from them the 3163 genes that are also maternally expressed ( ‘MBT-maternal genes’ ) because they are known to be enriched for broadly expressed housekeeping genes ( Rach et al . , 2009 ) . The remaining 844 genes ( ‘MBT-zygotic genes’ ) frequently have high Pol II occupancy at the pausing site and a high pausing index ( Figure 3 ) , suggesting that Pol II pausing is widespread during the MBT . 10 . 7554/eLife . 00861 . 010Figure 3 . Widespread Pol II pausing of MBT genes and poising for later activation . ( A ) Examples and average normalized enrichment ( as in Figure 2C ) of Pol II occupancy ( blue ) and TBP occupancy ( pink ) for the MBT-zygotic genes that are significantly transcribed during MBT ( MBT active ) or not ( MBT poised ) . Both groups show widespread Pol II pausing . ( B ) Analysis of large scale in situ hybridizations ( ImaGO database , see ‘Materials and methods’ ) confirms the earlier initial expression of MBT active genes ( mostly stage 4–6 , peri-cellularization ) and shows that many MBT poised genes are first transcribed at later stages of embryogenesis ( mostly stage 9–10 , post-gastrulation ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 010 When we analyzed the expression of these MBT-zygotic genes ( Lott et al . , 2011 ) , we found that 251 genes ( 30% ) are expressed at significant levels during late nuclear cycle 14 ( ‘MBT active genes’ , see brk and the average profile in Figure 3A ) , while the remaining 593 genes ( ‘MBT poised genes’ , see Dr and the average profile in Figure 3A ) are expressed at very low levels typical for paused genes poised for activation ( Zeitlinger et al . , 2007; Adelman and Lis , 2012 ) . This difference in expression was confirmed by analyzing their in situ hybridization patterns: while the MBT active genes are expressed very early , the MBT poised genes tend to be first detected at later embryonic stages ( stages 9–10 or later ) ( Figure 3B ) . Thus , many genes become newly paused during the MBT and are poised for later activation . The significant difference seen in Pol II pausing between pre-MBT and MBT genes is likely to be biologically meaningful . Genes expressed before the MBT have to be transcribed particularly fast because the nuclear cycle is extremely short ( 8 min in nuclear cycle 10 , increasing to 13 min in nuclear cycle 12 [Foe et al . , 1993] ) and progression through mitosis causes abortion of nascent transcripts ( Rothe et al . , 1992; Shermoen and O’Farrell , 1991 ) . As previously noted ( De Renzis et al . , 2007 ) , we also found that the pre-MBT genes are particularly short ( median of 1228 bp vs 6024 bp , Mann–Whitney test p<10−20 ) and more often intronless ( 54 . 6% vs 9 . 2% , Fisher test p<10−23 ) compared to MBT-zygotic genes . We also noticed that pre-MBT genes frequently use the transcription start site that yields the shortest transcript ( 62 . 9% vs 30 . 6% , Fisher test p<0 . 0003 ) ( Table 2 , and examples in Figure 1—figure supplement 3C , D ) . This supports the idea that rapid transcription is important during the fast cleavage cycles before the MBT and makes it plausible that the lack of Pol II pausing is advantageous for pre-MBT transcription . 10 . 7554/eLife . 00861 . 011Table 2 . Size and intron difference between pre-MBT and MBT zygotic genesDOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 011Transcript sizeShortest transcript usageIntron contentGene groupGene countMedian width ( bp ) Genes with multiple TSSGenes using shortest transcript ( % ) Gene count ( protein-coding ) Genes with no introns ( % ) All pre-MBT1171228*3522 ( 62 . 9% ) †9753 ( 54 . 6% ) ‡MBT maternal316333221125308 ( 27 . 4% ) 3163228 ( 7 . 2% ) MBT zygotic844604228487 ( 30 . 6% ) 73668 ( 9 . 2% ) active251542219150 ( 26 . 2% ) 25135 ( 13 . 9% ) poised59367719337 ( 39 . 8% ) 48533 ( 6 . 8% ) *Mann–Whitney test for pre-MBT vs MBT zygotic transcript size: p<10−20 . †Fisher test for pre-MBT vs MBT zygotic: p<0 . 0003 . ‡Fisher test for pre-MBT vs MBT zygotic: p<10−23 . We next explored potential mechanisms that could explain the difference between pre-MBT and MBT genes . Histone modifications such as H3K4me3 and H3K27me3 are present in embryonic stem cells ( Bernstein et al . , 2006 ) , in human sperm ( Hammoud et al . , 2009 ) , and may be present in zebrafish embryos before gene activation ( Vastenhouw et al . , 2010; Lindeman et al . , 2011 ) . However , we did not find these marks in Drosophila nuclei prior to gene activation . In both immunostainings and ChIP-seq data , the signal of H3K4me3 , a modification associated with gene activation , only starts to be detectable during the MBT ( Figure 4A , see also the ChIP-seq results in Figure 1C ) . H3K27me3 , a marker of Polycomb-mediated gene silencing and possibly epigenetic memory , can be detected in nuclei and in polar bodies at the earliest cleavage stages but is then undetectable in somatic nuclei until after the MBT ( Figure 4C ) . Consistent with this , the ChIP-seq H3k27me3 signal at Polycomb response elements ( PREs ) increases over time ( Figure 4B ) . Thus , H3K27me3 is likely present in oocytes but may be diluted or erased during replication consistent with a recent study ( Petruk et al . , 2012 ) . Accordingly , it seems unlikely that H3K4me3 or H3K27me27 have a direct role in regulating the zygotic genome activation in Drosophila . 10 . 7554/eLife . 00861 . 012Figure 4 . Absence of bivalent domains in pre-MBT embryos . ( A ) Lack of detectable H3K4me3 immunostaining in nuclei of embryos before MBT . H3K27me3 is observed in nuclei and polar bodies ( pb ) of early pre-blastoderm nuclei but not during pre-MBT or MBT stages . H3K27me3 first becomes detectable again in post-MBT embryos ( scale = 20 μm ) . ( B ) ChIP-seq experiments also suggest that H3K27me3 is absent during the MBT but that the levels increase soon after . Shown is the average pattern of H3K27me3 signal over input surrounding 441 previously identified Polycomb response elements ( PREs ) . ( C ) Despite high levels of Pol II occupancy , H3K4me3 average gene enrichments are low at pre-MBT genes ( top left ) . This is in contrast to MBT-maternal genes , which have high H3K4me3 enrichment that peaks ∼200 bp downstream of the TSS ( bottom left ) . All enrichments are calculated over input and normalized ( see ‘Materials and methods’ ) . The overall nucleosome occupancy , as measured by MNase digestion , shows more pronounced nucleosome positioning at MBT maternal genes but the overall nucleosome occupancy is not dramatically different ( compare top and bottom panels at the right ) . Average read counts from a Micrococcal Nuclease ( MNase ) experiment are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 012 We also noticed that pre-MBT genes tend to have particularly low levels of H3K4me3 ( Figure 4C ) . Even when pre-MBT genes continue to be transcribed during the MBT , H3K4me3 tends to be low and to accumulate to highest levels at further downstream nucleosomes at pre-MBT genes . This is in stark contrast to the MBT-maternal group , which has a sharp peak of H3K4me3 at the +1 nucleosome . This difference cannot be explained by lower levels of nucleosomes since the nucleosome occupancy based on our Micrococcal Nuclease experiments coupled to deep sequencing ( MNase-seq ) does not show a dramatic difference between the two gene groups ( Figure 4C ) . Since histone modification levels tend to be dependent on the promoter types ( Rach et al . , 2011 ) , we next analyzed whether pre-MBT genes are enriched for specific core promoter elements . We analyzed well-studied sequence motifs associated with either focused or dispersed transcription initiation in Drosophila ( Table 3 ) . Focused transcription initiates within a very narrow window and often at a single nucleotide ( also called peaked promoters ) , while dispersed transcription initiates from several weak transcription start sites within a ∼50–100 nucleotide region ( also called broad promoters ) ( Juven-Gershon and Kadonaga , 2010 ) . In Drosophila , dispersed initiation is typically found at broadly expressed housekeeping genes with constitutive promoters . Consistently , we found that MBT-maternal genes are strongly enriched for core promoter elements associated with dispersed initiation: Ohler1 , Ohler6 , Ohler7 and Dref response element ( DRE ) ( Figure 5A ) . Core promoter elements associated with focused initiation such as Initiator ( Inr ) , downstream promoter element ( DPE ) , Motif Two Element ( MTE ) and Pause Button ( PB ) have previously been associated with paused genes ( Hendrix et al . , 2008; Lee et al . , 2008 ) . As expected , MBT-zygotic genes are highly enriched for these elements , as well as GAGA , which is consistent with reports that GAGA factor ( GAF ) promotes the recruitment of paused Pol II ( Lee et al . , 1992; Leibovitch et al . , 2002; Lee et al . , 2008 ) . We noticed that these genes are not significantly enriched for the TATA box , although they are usually occupied by TBP ( Figure 1C , and Figure 3A ) . 10 . 7554/eLife . 00861 . 013Table 3 . Drosophila promoter elements analyzed in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 013Motif nameIUPAC consensusDirectionalWindow ( bp from TSS ) Transcript countReferenceNoteDREWATCGATWYes−100 to 02111 ( Hochheimer et al . , 2002 ) Dispersed initiationOhler1YGGTCACACTRYes−100 to 50609 ( Ohler et al . , 2002 ) Ohler6YRGTATWTTYYes−150 to 25840 ( Ohler et al . , 2002 ) Ohler7CAKCNCTRYes−100 to 502190 ( Ohler et al . , 2002 ) TATASTATAWAWRYes−100 to 01503 ( Goldberg , 1979 ) Focused initiationInrTCAKTYYes−50 to 505965 ( Smale and Baltimore , 1989 ) DPEKCGGTTSKYes0 to 75537 ( Burke and Kadonaga , 1996 ) PBKCGRWCGYes−50 to 1002093 ( Hendrix et al . , 2008 ) MTECSARCSSAYes0 to 30212 ( Lim et al . , 2004 ) GAGAGAGANo−100 to 09559 ( Stark et al . , 2007 ) Other motifsZeldaYAGGTARNo−2000 to 09798 ( Liang et al . , 2008; ten Bosch et al . , 2006 ) 10 . 7554/eLife . 00861 . 014Figure 5 . Differential usage of core promoter elements during the zygotic genome activation . ( A ) Promoter analysis of all previously identified gene groups . Shown is the enrichment of known core promoter elements found in promoters with dispersed initiation or focused initiation , as well as in the binding motifs for GAGA factor ( GAF ) and Zelda . Only occurrences close to the known location of the motif relative to the TSS were scored ( see Table 3 ) . The star indicates significant enrichment ( orange ) or depletion ( black ) . Note that the three pre-MBT groups with different Pol II pausing patterns are enriched for distinct core promoter elements . ( B ) The top two known motifs identified by de novo motif analysis for active pre-MBT genes and active MBT genes . The analysis was performed with MEME on the 200 bp long region centered on the TSS . The number of occurrences , p-value and the density distributions relative to the TSS of the identified motifs are shown on the right . Note that all motifs are found with the highest frequency at the expected location but that the DPE/MTE/PB is less specific and more frequently found at positions where it is unlikely to be functional . DOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 01410 . 7554/eLife . 00861 . 015Figure 5—figure supplement 1 . High frequency of Zelda motifs in the promoter region of pre-MBT genes . Histograms showing the distribution of the number of Zelda motifs found in each promoter region ( 2 kb upstream of the TSS , see ‘Materials and methods’ ) among all genes , all MBT genes , all pre-MBT genes , and MBT active genes . Pre-MBT genes are significantly enriched for Zelda motifs ( p<0 . 005 ) . MBT active genes may also show a small enrichment ( p<0 . 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 01510 . 7554/eLife . 00861 . 016Figure 5—figure supplement 2 . Analysis of the co-occurrence of promoter elements among all genes and early genes . The relative enrichment or depletion of the co-occurrences compared to the frequency expected by random chance is shown . Significant values ( p<0 . 05 , Fisher test ) are marked with a star . The order of the motifs is based on hierarchical clustering of the enrichment values shown in C . Groups of co-occurring motifs are highlighted in red . Our analysis for all genes produces similar results to those published previously by FitzGerald et al . ( 2006 ) . The results from the FitzGerald analysis ( which included 8 , 289 transcripts available at the time ) are graphically visualized by our method ( A ) . Our updated analysis using all current FlyBase promoters ( 19 , 845 transcripts ) and additional motifs are shown in comparison ( B ) . The co-occurrence of pausing elements ( MTE , PB , DPE , Inr , GAGA ) and dispersed promoter elements ( DRE , Ohler7 , Ohler1 , Ohler6 ) , respectively , are marked with a red box . Note that the motifs in the FitzGerald analysis deviate in some cases from our promoter elements ( e . g . , GAGA ) , which explains some differences . We then compared the results to our ‘pre-MBT and MBT’ set ( C ) . The co-occurrence of TATA , Inr and Zelda ( as well as with GAGA , which is found in dual genes ) is significant in this set ( marked with a red box ) , suggesting that promoters with this combination of motifs are preferentially found among early genes . DOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 016 In contrast , the promoters of pre-MBT genes are significantly enriched for TATA ( Figure 5A , Supplementary file hosted by Dryad [7 . 6 Mb] ) . Interestingly , only pre-MBT genes that initially show the non-paused profile ( pre-MBT not-paused , pre-MBT dual ) are significantly enriched for Inr and TATA . Furthermore , only pre-MBT genes that are paused at some point ( pre-MBT dual , pre-MBT paused ) show enrichment for GAGA , Inr and PB ( Figure 5A ) . Thus , the presence of specific core promoter elements correlates well with the Pol II occupancy profile across the gene body . Previous studies have shown that the transcription factor Zelda and its binding motif ( known as the TAGteam motif ) regulates the onset of zygotic gene expression ( De Renzis et al . , 2007; Harrison et al . , 2011; Liang et al . , 2008; ten Bosch et al . , 2006 ) . Notably , the binding levels of Zelda in the promoter region correlate well with the onset of gene expression , although Zelda is also abundantly found at enhancers ( Harrison et al . , 2011 ) . Consistent with this , we found that Zelda motifs are highly enriched in the promoter region of pre-MBT genes ( Figure 5—figure supplement 1 ) . However , this enrichment is found in all three pre-MBT classes and does not correlate with the Pol II pausing pattern ( Figure 5A ) . To further consolidate the differences between pre-MBT and MBT genes , we performed de novo motif analysis with MEME on the 200 bp centered on the transcription start site ( Figure 5B ) . For pre-MBT stage non-paused genes ( pre-MBT not-paused , pre-MBT dual ) , the top two motifs were Zelda and TATA ( Figure 5B ) . In contrast , the top two known motifs for the most comparable MBT group ( MBT active genes , which are also early-expressed developmental genes ) , were GAGA and a motif that resembles DPE , MTE and PB ( Figure 5B ) . This confirms that pre-MBT and MBT genes differ in their core promoter sequences . Finally , an analysis of the co-occurrences of core promoter elements similar to previous analyses ( FitzGerald et al . , 2006 ) also supports our finding ( Figure 5—figure supplement 2 ) . For example , Zelda , Inr and TATA significantly co-occur among all our Pol II-bound genes ( pre-MBT and MBT genes ) but not among all annotated genes , suggesting that these motifs preferentially function together during early development . This suggests a model in which rapid pre-MBT transcription without Pol II pausing is mediated by Zelda bound close to a TATA-enriched promoter ( Figure 6 ) . In contrast , paused Pol II is typically established through GAF during the MBT at promoters with pausing elements such as DPE , MTE or PB . Thus , there are two principle modes by which zygotic genes are activated but genes may also have elements of both and show a dual behavior . 10 . 7554/eLife . 00861 . 017Figure 6 . Proposed model for the two main modes of Pol II recruitment and elongation behavior during the zygotic genome activation . Before the midblastula transition , when the cell cycle is fast , efficient transcription occurs through a TATA promoter with multiple Zelda sites upstream . This combination leads to transcription without pausing , presumably due to fast re-initiation . During the midblastula transition , Pol II is recruited de novo to many genes and pausing is established with the help of GAGA factor and core promoter elements associated with pausing such as DPE , MTE and PB . Note that genes can have core promoter elements of both modes ( e . g . , TATA and PB ) , leading them to switch from a non-pausing behavior to a pausing behavior during the midblastula transition . It is likely that transcription factors in addition to Zelda and GAF also influence the Pol II behavior at genes . DOI: http://dx . doi . org/10 . 7554/eLife . 00861 . 017
Our results suggest that a large number of genes first recruit Pol II during the MBT with widespread Pol II pausing , while the small number of genes transcribed before that are not paused . Presumably , these pre-MBT genes are required for very early developmental events such as sex determination ( Erickson and Cline , 1993; Barbash and Cline , 1995 ) or cellularization ( Schejter and Wieschaus , 1993; Pritchard and Schubiger , 1996 ) , or may be involved in early patterning events that require feedback regulation over time ( for example ftz [Edgar et al . , 1986; Pritchard and Schubiger , 1996] or sna [De Renzis et al . , 2006; Reeves et al . , 2012] ) . Thus , while Pol II pausing is commonly found at developmental genes and may be advantageous for their precise and synchronous expression in response to localized extracellular signals ( Boettiger and Levine , 2009 ) , a different mode of transcription is used during pre-MBT stages . Due to the short nuclear cycles at this stage , it is likely that transcription is optimized to achieve high levels of transcripts in a very short time period . The fact that we find the TATA box enriched among pre-MBT genes is consistent with the known properties of TATA-containing promoters . TATA is a strong core promoter element that efficiently supports transcription in vitro ( Aso et al . , 1994 ) , mediates efficient re-initiation in vitro ( Yean and Gralla , 1997 , 1999 ) , and its presence in vivo correlates with ‘bursts’ of transcription that produce many transcripts within a short time ( Zenklusen et al . , 2008 ) . Furthermore , it has been shown that TATA promotes pTEFb activity , leading to more efficient elongation rates in vitro and in vivo ( Amir-Zilberstein et al . , 2007; Montanuy et al . , 2008 ) . This suggests that TATA-enriched promoters and paused promoters have different transcription dynamics and serve different purposes during development . While this difference is particularly evident during the zygotic genome activation as reported here , we propose that this difference is general and also applies to later development . For example , we have analyzed promoters during muscle development and find that many genes are induced during late stages of embryogenesis without prior Pol II pausing and that these genes are also enriched in TATA ( Gaertner et al . , 2012 ) . Consistent with this , statistical analyses suggest that the combination of Inr and TATA represents a separate class of promoters that is often found among genes expressed in adult tissues ( FitzGerald et al . , 2006; Lenhard et al . , 2012 ) . Since the properties of TATA are not specific to Drosophila , it is likely that differences among promoter types and their propensity for Pol II pausing are conserved across animals .
Wild-type embryos ( Oregon R ) were collected from six population cages ( 28 × 17 × 17 cm , with 10 , 000–12 , 000 flies each , maintained in fly incubators at 25°C and 60% humidity ) on 15 cm apple juice plates with yeast paste after pre-clearing . The collection windows were 0–4 hr after egg deposition ( AED ) for immunostainings , and 1–2 hr , 2–3 hr , 2–4 hr , 6–8 hr AED for ChIP . Embryos were dechorionated with bleach and cross-linked with 1 . 8% formaldehyde in 2 . 5 ml Hepes buffer and 7 . 5 ml heptane , while vortexing at medium speed for 15 min . Embryos were devitellinized in methanol/heptane and kept at −20°C in methanol for up to 3 months until needed . Immunostainings were performed by standard methods with Alexafluor Dyes and the embryos were analyzed by confocal imaging ( Zeiss LSM-510-VIS , Carl Zeiss Microscopy , LLC , Thornwood , NY ) . The following antibodies were used for immunostainings: mouse monoclonal antibody CTD4H8 against Pol II CTD ( 05-623; EMD-Millipore , Billerica , MA ) , mouse monoclonal antibody H14 against Ser5-phosophorylated Pol II ( MMS-134R; Covance , Princeton , NJ ) , rabbit polyclonal antibodies against H3K4me3 ( 9751S; Cell Signaling Technology , Danvers , MA ) , rabbit polyclonal antibodies against H3K27me3 ( 39155; Active Motif , Carlsbad , CA ) , rabbit polyclonal antibodies against dTBP ( a kind gift from J Kadonaga ) , and mouse monoclonal antibody against lamin 0 ( ADL101 developed by P Fisher , obtained from the Developmental Studies Hybridoma Bank ) . Embryos were visualized by confocal imaging ( LSM-510-VIS ) . The following antibodies were used for ChIP: antibodies against dTBP ( a kind gift from J Kadonaga ) , H3K4me3 ( 9751S; Cell Signaling Technology ) , H3K27me3 ( 39155; Active Motif ) , mouse monoclonal Pol II antibody 8WG16 ( MMS-126R; Covance ) and rabbit polyclonal antibodies against H3ac ( 07-360; Millipore ) . For embryo sorting , wild-type embryos collected within 1–2 hr ( pre-MBT ) and 2–3 hr ( MBT ) were fixed in 1 . 8% formaldehyde and stained with DAPI . Embryos were sorted in PBT on ice under an inverted contrasting microscope ( Leica DMIL , Buffalo Grove , IL ) . All embryos were screened once for morphology with a DIC filter , and twice for DNA content under UV light ( Figure 1—figure supplement 1 ) . Out-of-stage embryos were removed with a 10 μl pipette tip connected to a Cell Tram Vario ( 920002111; Eppendorf , Hauppauge , NY ) . After practice , 200 μl embryos ( ∼5000 embryos ) could be screened in 1 day . ChIPs were performed with whole cell extracts ( WCE ) from embryos as previously described ( Zeitlinger et al . , 2007 ) using 200 μl embryos for pre-MBT embryos and 50 μl for older embryos . Libraries from the immunoprecipitated DNA and WCE DNA were prepared using the Paired-End DNA Sample Preparation Kit ( PE-102-1001; Illumina , San Diego , CA ) but with a modified protocol . To remove adapter dimers , biotin-labeled dATP ( 19524-016; Invitrogen , Grand Island , NY ) was added in the A-tailing reaction after end-repair . After ligation to the PE adaptor , the samples were incubated with streptavidin beads in 500 μl binding buffer ( 650-01; Invitrogen ) at room temperature for 15 min . DNA bound to the beads was then washed twice with 800 μl binding buffer with 0 . 05% Tween 20 , twice with NEB buffer 2 ( New England Biolabs , Ipswich , MA ) and resuspended in 31 μl NEB buffer 2 . The PCR reaction was then performed according to the Illumina protocol . 50 μl sorted 2–3 hr embryos were homogenized as previously described ( Zeitlinger et al . , 2007 ) and digested based on a previously published protocol ( Mavrich et al . , 2008 ) . Briefly , homogenized chromatin in NPS buffer was digested with an MNase ( LS004798; Worthington , Lakewood , NJ ) gradient of 20 U , 10 U , 5 U , 5/2 U , 5/4 U , 5/8 U , 5/16 U , to 5/32 U , and a negative control for 30 min at 37°C . Mono-nucleosome size DNA was extracted from the lane with two clear bands in a 1 . 7% agarose gel , and prepared for paired-end sequencing . All sequencing reads were aligned to the UCSC Drosophila melanogaster dm3 genome with Bowtie v0 . 12 . 8 ( Langmead et al . , 2009 ) using the following parameters: -k 1 –m 1 –l 40 –n 2 –best –strata The MBT MNase-seq library was paired-end sequenced and alignment was performed with an allowable insert size of 47 bp to 297 bp . After alignment , single-end reads were extended to the estimated insert size of the library as determined by a Bioanalyzer . To identify alignment and amplification artifacts , custom R scripts were used to analyze the aligned reads of all single-end libraries with more than 10 duplicates ( defined as having the same chromosome , start and strand values ) . These ‘stacks’ of identical reads were removed unless a corresponding number of reads were present on the opposite strand approximately one fragment length away in the 3′ direction . For all libraries , genome-wide coverage was calculated by assigning an integer score to each genomic coordinate representing the number of extended reads that overlapped that location . To obtain gene expression measurements at different nuclear cycles , we downloaded single-embryo RNA-seq datasets from http://eisenlab . org/dosage/ ( Lott et al . , 2011 ) . One female and one male replicate were downloaded for nuclear cycles 10 , 11 , 12 , 13 and 14 ( A-D ) . The male and female datasets were combined for each nuclear cycle . In addition , we downloaded 4–6 hr and 6–8 hr staged whole-embryo RNA-seq datasets ( Graveley et al . , 2011 ) . We processed all samples using TopHat v2 . 0 . 4 ( Trapnell and Salzberg , 2009 ) by aligning against the FlyBase r5 . 47 genome and its corresponding gene annotations using the following parameters: Single-embryo samples ( 40 bp reads ) : -G fb-r5 . 47 . gtf –I 5000 –segment-length 20 fb547_genome 4–6 hr and 6–8 hr embryo samples ( 75 bp reads ) : -G fb-r5 . 47 . gtf–I 5000 –segment-length 37 fb547_genome Next , we used the cuffdiff tool from Cufflinks v2 . 0 . 2 ( Trapnell et al . , 2010 ) to obtain gene expression values ( RPKMs ) for all samples using the following non-default parameters: -u –b fb547_genome . fa fb-r5 . 47 . gtf For the four Pol II replicates in the pre-MBT embryo and the three Pol II replicates in the MBT embryo , enrichment ratios were calculated for the TSS region ( first 200 bp of the transcript ) , a region immediately downstream of the TSS ( +201 to +400 bp ) , and the transcription unit ( TU ) region ( +401 to the end of the transcript ) of each unique FlyBase r5 . 47 transcript . For transcripts less than 600 bp in length , the TU region was defined as the entire transcript . Total signal for each region was found for each Pol II and WCE sample . Enrichment in each region was calculated after normalizing for both fragment length and total read count:Enrichment= ( IP signal/[IP read count×IP fragment length] ) / ( WCE signal/[WCE read count×WCE fragment length] ) The stalling index for each gene was defined as: log2 Pol IITSS–log2 Pol IIDownstream TSS after flooring both Pol II enrichment values at 1 ( background ) . Stalling indexes for all replicates were averaged . To identify genes bound by Pol II in the pre-MBT embryo , we first identified all transcripts with Pol IITSS enrichment twofold above WCE in all four replicates . To ensure that these enrichments were due to high Pol II signal , we also required the Pol II signal portion of the enrichment calculation ( the numerator in the above equation ) to be in the 99th percentile of all transcripts in all four replicates . Manual inspection of some of these transcripts showed that the Pol II signal originated from a different gene’s TSS ( see examples in Figure 1—figure supplement 4 ) . To eliminate these false positives , we used MACS to identify peaks in our best pre-MBT TBP sample and manually examined all Pol II-enriched pre-MBT transcripts that did not have a detected TBP peak within 500 bp of the TSS . We used the default parameters of MACS v2 . 0 . 10 . 20120703 ( Zhang et al . , 2008 ) , specifying only the preset alignable genome size for Drosophila melanogaster using the ‘-g dm’ argument . This identified 12 transcripts in which the Pol II signal did not appear to originate from the TSS . These transcripts were removed from our pre-MBT list and are marked as ‘rejected pre-MBT genes’ in ( Supplementary file hosted by Dryad [7 . 6 Mb] ) . We next checked for possible pre-MBT genes with missing or mis-annotated transcription start sites . To do this , we used MACS to call peaks on all four of our pre-MBT Pol II samples using the same default parameters as described above . We then identified all regions that were called as peaks in at least two of the four replicates . These regions were assigned to the nearest gene within 5 kb and all regions assigned to a gene not already considered a pre-MBT gene were manually examined . This revealed ten possible additional pre-MBT genes where the Pol II signal originated from an un-annotated transcription start site . As all of these genes also had at least some TBP signal upstream of the Pol II signal , we defined custom transcript entries for these genes by setting the transcript start site to 19 bp downstream of the location of the maximum TBP signal . To ensure these custom transcripts met our existing enrichment criteria , we performed the same calculations as described above in the Calculating Pol II enrichments section . All ten of the custom transcripts were sufficiently enriched in Pol II and were added to our pre-MBT gene list . We classified the 117 pre-MBT genes into three groups . First , the ‘paused’ group was defined as those pre-MBT genes having a mean ( among all four replicates ) Pol IITU ratio less than 1 . The ‘dual’ group was defined as any pre-MBT gene not in the paused group that had Pol IITSS enrichment in the top 20% of all genes in 6–8 hr Mef2-positive muscle cells ( Gaertner et al . , 2012 ) . The remaining pre-MBT genes were classified as the ‘not paused’ group . To identify genes bound by Pol II in the MBT embryo , we selected all transcripts with Pol IITSS enrichment at least twofold above WCE in all three replicates . If multiple transcripts for the same gene met these criteria , we selected the one with the highest Pol IITSS signal ( breaking ties using the mean Pol IITU enrichment ) . MBT genes were classified into three groups using gene expression values calculated from previously published single-embryo RNA-seq experiments ( see ‘Analysis of RNA-seq expression data’ section ) . We classified as ‘maternal’ all MBT genes with an RPKM of at least 1 in nuclear cycle 10 . We classified as ‘MBT active’ all non-maternal MBT genes with an RPKM of at least 5 in nuclear cycle 14D . The remaining MBT genes were classified as ‘MBT paused’ . PhastCons scores ( Siepel et al . , 2005 ) from the alignment of 14 insect species’ genome assemblies to the Drosophila melanogaster genome were downloaded from http://hgdownload-test . cse . ucsc . edu/goldenPath/dm3/phastCons15way/ . For each mRNA transcript in Flybase r5 . 47 , the mean phastCons score along the transcript’s length was used as a relative measure of conservation . These transcripts were then organized into groups as described in the ‘Identification and classification of pre-MBT genes’ and ‘Identification and classification of MBT genes’ sections . For Figure 1C , enrichment values were first calculated in a 100 bp sliding window across all samples . Replicates were combined by taking the minimum enrichment value at each base . Samples were then independently normalized by defining ‘minimum’ as an enrichment value of 1 ( background ) and ‘maximum’ as the 99th percentile enrichment value encountered among all displayed bases . As there was no significant ChIP enrichment in the pre-MBT H3K4me3 sample , it was normalized to the maximum enrichment value of the MBT H3K4me3 sample to avoid amplifying noise . For Figure 2 , Figure 3C and Figure 1—figure supplement 3 , both read counts and enrichment values for each sample were independently scaled by dividing the values at each base by the maximum value encountered among the displayed genes or gene groups across stages after normalizing for both read count and fragment size differences . Using supplemental Table 17 from Schuettengruber et al . ( Schuettengruber et al . , 2009 ) , the 441 regions were selected based on Ph ChIP-chip enrichment ( p<0 . 0001 ) . The regions were then aligned at their midpoints and extended by 80 kb in both directions . Average region graphs were constructed showing the average enrichment value at each base for three H3K27me3 samples . The enrichment for each sample was defined by dividing read-count normalized IP signal by the read-count normalized WCE control signal . Sequences surrounding all FlyBase r5 . 47 transcription start sites ( plus our ten additional custom pre-MBT transcripts ) were scanned for the core promoter elements listed in Table 3 . A core promoter element was scored as present if found with no mismatch within a specified window relative to the transcription start site . For Zelda , we also counted the number of motifs found in each transcript’s window . For each group of transcripts analyzed for promoter element composition , an enrichment and p-value were calculated for each promoter element . Enrichment was calculated as follows , where G is the group of transcripts tested and PE is a particular promoter element:Observed= ( Number of transcripts in G with element PE ) / ( Number of transcripts in G ) Expected= ( Number of transcripts in the genome with element PE ) / ( Number of transcripts in the genome ) Enrichment=Observed/Expected For enrichment values less than one , the negative reciprocal of the enrichment value was used ( indicating depletion instead of enrichment ) . To calculate a p-value for the observed frequency of each promoter element in each group of transcripts , a Fisher test was performed . Enrichments and depletions with a p<0 . 05 ( after correcting for multiple testing with the Benjamini and Hochberg method ) were deemed significant . For Zelda , we calculated enrichment and p-values via random sampling . Enrichment values for each group of transcripts were calculated by dividing the number of Zelda sites per transcript in each group by the average number of Zelda sites per transcript in the genome . To calculate p-values for each group of transcripts , we randomly selected an equal number of transcripts from the entire genome 10 , 000 times and calculated the enrichment value for each random sample . The p-value was then calculated as the portion of random samples with higher Zelda enrichment than the transcript group . The enrichments ( Observed/Expected ) and p-values ( Fisher test ) for the co-occurrence of promoter elements were calculated for all pairs of core promoter elements in three sets of genes . For the ‘FitzGerald’ set of genes , we extracted the gene and overlap counts for a subset of motifs listed in Table 1 of FitzGerald et al . , 2006 . For the ‘all promoters’ set , we included all unique promoters in FlyBase r5 . 47 as well as our additional ten custom transcripts . For the ‘pre-MBT and MBT’ set , we included only the promoters of our pre-MBT and MBT gene groups . The order of motifs shown in Figure 5—figure supplement 2 was determined by hierarchically clustering the enrichment values with the R v3 . 0 . 1 hclust function using Euclidean distance . For the Zelda motif , which can occur multiple times in a single promoter region , the presence of at least one motif within 2 kb upstream was scored . Fasta sequence files were generated for each of the classified Pre-MBT and MBT groups , based on regions ±100 bp surrounding the FlyBase r5 . 47 transcription start sites ( plus our ten additional custom pre-MBT transcripts ) . Using MEME v4 . 8 . 1 ( Bailey et al . , 2009 ) the fasta files were processed using the following parameters: -mod zoops -dna -nmotifs 50 -revcomp -maxw 12 -maxsize 5000000 -oc meme/ The resulting motifs were then compared against the TRANSFAC 2011 . 4 database and the Table 3 listed above using the TOMTOM tool also from the MEME suite . To calculate enrichment of gene groups for their first expression in specific tissues , we downloaded the Berkeley Drosophila Genome Project in situ expression database from http://insitu . fruitfly . org/ ( Tomancak et al . , 2002 , 2007 ) . We removed the ‘maternal’ and ‘no staining’ annotation entries and then removed all but the first ( in stage order ) annotation of each gene . For each gene group analyzed , we calculated enrichments and p-values using the same method as the promoter element enrichment analysis described above . For all gene groups plotted in Figure 2B and Figure 1—figure supplement 4 , we first removed any genes with evidence of maternally deposited mRNA using the following criteria:RPKM expression > 16 in nuclear cycle 10 , orMaternal expression at least twofold above zygotic expression in nuclear cycle 10 [F10 sample in Dataset S1 of S . E . Lott et al . ( Lott et al . , 2011 ) ] All ChIP-seq and MNase-seq data have been deposited with the NCBI Gene Expression Omnibus under accession number GSE41703 . In addition , we have replicated our analysis environment ( including software tools , analysis source code , and raw data ) in a Linux virtual machine hosted by Amazon Web Services . Instructions for accessing the virtual machine can be found at http://research . stowers . org/zeitlingerlab/data . The analysis code is available on GitHub at https://github . com/zeitlingerlab/chen_elife_2013 . A spreadsheet summarizing our classification and analysis of pre-MBT and MBT gene groups is available via Dryad digital repository . The first sheet gives an explanation for all column headings . The second sheet lists all data for all our annotated genes , including our ten custom transcripts . It includes the classifications into pre-MBT and MBT gene groups , the Pol II ChIP-seq enrichment values at the transcription start site ( TSS ) and transcription unit ( TU ) for all replicates , phastCon conservation scores , and the presence or absence of all core promoter motifs analyzed in this study , as well as the presence of the TATA element identified by de novo motif analysis . | Fertilized eggs—zygotes—develop into embryos via several distinct stages . In many animals , the zygote initially undergoes rapid rounds of genome replication; however , this hectic activity is not controlled by the zygote itself . Instead , the mother deposits RNA molecules in the egg as it forms inside her , and after the egg has been fertilized , these RNA molecules are translated into proteins that guide the development of the early embryo . Only at a stage called midblastula transition does the zygote take over control by transcribing its own RNA molecules . Fruit flies start to transcribe their own genes en masse after completing thirteen rounds of DNA replication . However , some genes are already transcribed during the rapid cycles of DNA replication earlier in development . How these early genes are transcribed , and how the embryo shifts to more widespread transcription during the midblastula transition , are not well understood . In particular , it is not known if the molecular machinery needed to transcribe the genes is recruited a long time before transcription starts , or if it is recruited ‘just in time’ . Here , Chen et al . explore how genes are switched on in the fruit fly zygote . Genes are transcribed by a protein complex called RNA polymerase , which binds to DNA sequences , called promoters , within the genes . Chen et al . used a technique called ChIP-Seq to determine how much RNA polymerase was bound to the DNA before , during and after the midblastula transition . Before the transition—from about eight rounds of DNA replication onward—RNA polymerase was bound to only about 100 genes , and was active in most of these cases . In contrast , after the transition , RNA polymerase had been recruited to the promoters of around 4000 genes ( fruit flies have a total of about 14 , 000 genes ) . However , it was often found in a paused , rather than active , form , at these genes , which is thought to help ensure that their transcription can occur on a precise schedule . Chen et al . then used computer analyses to test the theory that differences in the DNA sequences of the gene promoters might determine which genes the RNA polymerase bound to , and whether or not the polymerase underwent pausing or became active immediately . Strikingly , there were clear differences in the sequence motifs that recruited RNA polymerase to the promoters of genes that were transcribed immediately and those that showed pausing of the polymerase . Moreover , genes that were transcribed before the midblastula transition were shorter , on average , than those transcribed after . This suggests that transcription during the rapid genome replication cycles has to occur quickly and therefore lacks pausing . Together , these findings present a biological rationale for differences in how genes are first transcribed during fruit fly development . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"developmental",
"biology"
] | 2013 | A global change in RNA polymerase II pausing during the Drosophila midblastula transition |
Poly ( ADP-ribose ) polymerase 1 ( PARP1 ) is both a first responder to DNA damage and a chromatin architectural protein . How PARP1 rapidly finds DNA damage sites in the context of a nucleus filled with undamaged DNA , to which it also binds , is an unresolved question . Here , we show that PARP1 association with DNA is diffusion-limited , and release of PARP1 from DNA is promoted by binding of an additional DNA molecule that facilitates a ‘monkey bar’ mechanism , also known as intersegment transfer . The WGR-domain of PARP1 is essential to this mechanism , and a point mutation ( W589A ) recapitulates the altered kinetics of the domain deletion . Demonstrating the physiological importance of the monkey bar mechanism for PARP1 function , the W589A mutant accumulates at sites of DNA damage more slowly following laser micro-irradiation than wild-type PARP1 . Clinically relevant inhibitors of PARP1 did not alter the rate or mechanism of the release of PARP1 from DNA .
Poly ( ADP-ribose ) polymerase 1 ( PARP1 ) serves as a first responder to DNA damage and is the founding member and most abundant representative of the large family of diphtheria toxin-like ADP-ribosyltransferases ( ARTDs ) ( Bai , 2015; Bock and Chang , 2016; Beck et al . , 2014; Daniels et al . , 2015; De Vos et al . , 2012; Mashimo et al . , 2014; Morales et al . , 2014 ) . Binding to either single or double-strand DNA breaks ( SSBs or DSBs ) enzymatically activates PARP1 to use NAD+ in polymerizing long chains of poly ( ADP ) -ribose ( PAR ) onto itself and other nuclear acceptor proteins such as histones and DNA repair proteins . These PAR chains then recruit the appropriate DNA repair machinery containing PAR-binding motifs ( Karlberg et al . , 2013; Teloni and Altmeyer , 2016 ) . PARP1 is of special interest because it is a validated target for cancer therapy ( Tangutoori et al . , 2015; Liu et al . , 2014 ) . Most notably , olaparib and rucaparib are in clinical use for treatment of ovarian and/or breast cancer in BRCA1/2 negative patients , and there are many on-going phase III clinical trials for inhibitors of PARPs either as monotherapy or in combination with chemo- or radiotherapy . Overall , the domain structures of the 16 members of the ARTD family are quite diverse , but they all share a common catalytic core domain ( ~40 kDa ) ( Barkauskaite et al . , 2015 ) . Clinically relevant inhibitors of PARP1 bind in the catalytic domain . The N-terminal region of PARP1 contains five additional domains; three Zn-finger domains , an automodification domain that contains a BRCT-fold , and a WGR domain ( Figure 1A ) . Seminal work from the Pascal laboratory has provided a molecular understanding of how Zn1 , Zn2 , Zn3 , and the WGR domain collaborate to recognize DNA strand breaks in a structure-specific and sequence-independent manner , and subsequently activate the catalytic activity of PARP1 . Zn1 and Zn2 separately ( Langelier et al . , 2011a ) , and together in the context of an SSB ( Eustermann et al . , 2015 ) , bind one DNA end each using two points of contact , termed the phosphate backbone grip and the base stacking loop . In the context of a DSB , this grip-loop interaction mode is maintained by Zn1 , while the Zn3 and WGR domains make additional contacts to the DNA ( Figure 1B ) ( Langelier et al . , 2012 ) . Importantly , stepwise assembly of the different domains of PARP1 on DNA leads to the destabilization of the helical subdomain ( HD ) of the catalytic domain , which results in activation of its ADP-ribosyl transferase activity ( Eustermann et al . , 2015; Langelier et al . , 2018; Dawicki-McKenna et al . , 2015 ) . In cells , PARP1 is known to contribute to many types of DNA repair mechanisms , including base excision repair , homologous recombination , nucleotide excision repair , and alternative non-homologous end-joining ( de Murcia et al . , 1997 ) . In vitro , PARP1 is activated by a wide variety of DNA damage models including nicks , gaps , blunt ends , 5’- or 3’- extensions , all with or without a 5’-phosphate ( Langelier et al . , 2014 ) . There is now clear evidence from multiple laboratories that PARP1 also binds tightly to undamaged DNA . For example , the Kraus laboratory has shown that PARP1 binds to and condenses intact chromatin , represses Pol II-dependent transcription , and is activated for auto-PARylation ( Kim et al . , 2004 ) . We have previously shown that PARP1 serves as a chromatin architectural protein and interacts tightly ( Kd ~nM ) with and is activated by various nucleosome constructs ( Clark et al . , 2012; Muthurajan et al . , 2014 ) . Additionally , atomic force microscopy has shown that PARP1 binds not only to DNA ends or specific nicks , but also has significant affinity for undamaged DNA ( Sukhanova et al . , 2016 ) . Most recently , single molecule tightrope assays have demonstrated that PARP1 interacts with and moves along undamaged DNA ( Liu et al . , 2017 ) . Thus PARP1 faces a similar ‘speed-stability’ paradox ( Mirny et al . , 2009; Zandarashvili et al . , 2015; Halford and Marko , 2004 ) as transcription factors that need to find their target recognition site in an overwhelming excess of non-specific sites for which they also have significant affinity . PARP1 must rapidly search the genome for damaged DNA , yet it has significant affinity for the billions of base pairs of undamaged DNA that are present at concentrations of ~100 mg/mL in the nucleus ( Krebs et al . , 2017 ) . In fact , laser micro-irradiation experiments in live cells have shown that PARP1 significantly accumulates at DNA damage sites in less than 10 s ( Mortusewicz et al . , 2007 ) . The conundrum is that repeated cycles of release of PARP1 from undamaged DNA , random diffusional collisions , and rebinding to a different location may not be fast enough to explain how PARP1 can rapidly localize to sites of DNA damage . Various models have been put forth and tested for explaining how ‘facilitated diffusion’ could accelerate this search process , all of which recognize the importance , as opposed to hindrance , of non-specific binding to DNA for efficient site localization ( Halford and Marko , 2004; Berg et al . , 1981; Iwahara et al . , 2006; Doucleff and Clore , 2008 ) . These models include binding followed by one-dimensional sliding along DNA , hopping to a near-by site in the same chain , and intersegment transfer via an intermediate loop that is formed when the protein binds two different DNA sites at the same time . While some localized sliding along DNA has been reported for PARP1 ( Liu et al . , 2017 ) , a more thorough kinetic characterization of binding to and dissociation from DNA is needed in order to address how PARP1 can efficiently localize to sites of DNA damage to initiate repair . PARP1 has been found to associate more tightly with DNA in vivo in the presence of clinically relevant inhibitors . This phenomenon , known as PARP ‘trapping' ( Tangutoori et al . , 2015; Brown et al . , 2016; Shen et al . , 2015; Murai et al . , 2012; Pommier et al . , 2016 ) is thought to be in part responsible for the clinical effects of PARP inhibitors and has been used to explain the numerous discrepancies between in vitro inhibition of PARP1 vs . potency in preclinical models . For example , talazoparib is 100-fold more potent at trapping PARP1 on DNA and >50 fold more potent at killing cancer cells than rucaparib and olaparib , although the apparent IC-50’s for all three compounds are quite similar ( 1–5 nM ) ( Shen et al . , 2015 ) . Further complicating matters , extensive biochemical investigations of PARP1 trapping failed to provide evidence for an allosteric interaction between DNA- and inhibitor-binding ( Hopkins et al . , 2015 ) , suggesting that trapping is due solely to inhibition of catalytic activity ( but see [Langelier et al . , 2018] ) . Thus , an evaluation of PARP inhibitors in a quantitative assay that measures DNA binding and release has the potential to shed further light on this controversial issue . Here , we report on the kinetics of association and dissociation of PARP1 with DNA . We find that association of PARP1 with DNA is extremely fast , and that dissociation depends on the formation of a ternary complex where a second DNA molecule binds before release of the original DNA . We find that the WGR-domain , more specifically the conserved residue Trp589 , is essential for triggering DNA-dependent release of DNA from PARP1 , and we demonstrate the importance of this mechanism of DNA release for the accumulation of PARP1 at sites of DNA damage in the cell . Finally , we find that clinically relevant inhibitors do not perturb the rate or mechanism of release of DNA from PARP1 .
We began our investigations by measuring the rate of association of PARP1 with DNA . Varying concentrations of PARP1 ( 60–250 nM ) were mixed in a stopped-flow apparatus with fixed concentrations ( 30 nM ) of a fluorescently labeled model of a double-strand break with a 5’-phosphate ( p18mer* ) . Addition of protein results in an increase in fluorescence anisotropy that is not observed by addition of buffer alone ( Figure 2A ) . The data at all concentrations of PARP1 could be fitted with a single exponential to yield kobs with very good residuals ( Figure 2A ) . Under idealized experimental conditions wherein the concentration of PARP1 greatly exceeds the concentration of p18mer* , one would expect a replot of kobs vs . the concentration of PARP1 to yield a straight line , as was indeed observed here ( Figure 2A , inset ) . The slope of such a line equals the apparent second order rate constant of association , whereas the y-intercept equals the first-order rate constant of dissociation . To analyze the data more rigorously , we used Kintek Explorer , a powerful fitting program that allows for global model-dependent fitting that does not require adherence to limiting conditions ( Figure 2B ) . Our analysis yields a k1 of 3 . 1 nM−1s−1 ( Scheme 1 , Table 1 ) , which is significantly greater than previously reported for PARP1 associating with DNA as measured using surface plasmon resonance ( Jorgensen et al . , 2009 ) . The rate of dissociation ( k-1 ) could not be determined from this experiment since no significant dissociation occurs over the 25 ms time course of the experimental observation . Using global fitting , we could derive an upper bound for k-1 of 10 s−1 . Thus , the true equilibrium dissociation constant ( KD ) of a double-strand break under these conditions is <3 nM ( Figure 4—figure supplement 2 ) , lower than the previously reported KDs of 31 nM ( Clark et al . , 2012 ) , 14 nM ( Langelier et al . , 2010 ) , and 97 nM ( Langelier et al . , 2018 ) ( see Discussion ) . Because we were unable to determine the rate of DNA dissociation from PARP1 in the previous experiment , we designed an experiment to explicitly measure this rate using competition . Here , we pre-form a complex between PARP1 and fluorescently labeled DNA and use an excess of unlabeled DNA to compete away the labeled DNA and prevent its re-association with PARP1 . We began these investigations by first performing a label-swap experiment to ensure that unlabeled p18mer behaves similarly to fluorescein labeled p18mer* . Since the experimental read-out is based on the change in fluorescence anisotropy of p18mer* , we used a fixed and limiting concentration of total labeled DNA such that no excess p18mer* is present . PARP1 ( 37 nM ) , pre-bound to either p18mer or p18mer* ( 25 nM ) was mixed with 25 nM p18mer* or p18mer ( respectively ) in a stopped-flow apparatus . Dissociation of p18mer or p18mer* ( followed by binding of the competitor ) was monitored by an increase or decrease in fluorescence anisotropy , respectively ( Figure 3—figure supplement 1 ) . The similarity of these two experiments is best visualized by plotting the sum of the signal to generate a flat line equal to the probe concentration ( 25 nM ) , a pseudo-residual indicating that p18mer and p18mer* are kinetically indistinguishable in our assay . In order to probe the mechanism of DNA dissociation from PARP1 , we next varied the concentrations of competitor DNA . Under ideal experimental conditions , where the concentration of competitor DNA ( >500 nM ) greatly exceeds the probe concentration ( 25 nM ) , and assuming the simplest model wherein the rate of dissociation is rate-determining ( k’1 [DNA]>>k-1 , Scheme 2 ) , we expect that kobs would be independent of the concentration of competitor DNA . PARP1 ( 37 nM ) , pre-bound to p18mer* ( 25 nM ) , was mixed with various concentrations of competitor DNA ( p18mer , 500 nM – 4000 nM ) in a stopped-flow apparatus and dissociation of p18mer* was monitored by a decrease in fluorescence anisotropy ( Figure 3 ) . The data could be fitted to a single exponential to yield kobs with very good residuals ( Figure 3 ) . However , as seen in the data in Figure 3 by comparing dissociation in the presence of 2 . 2 vs 4 µM DNA , and in the replot of kobs vs . multiple concentrations of competitor DNA , kobs increases at increasing concentrations of competitor DNA ( Figure 3 , inset ) . Additionally , attempts to fit these data with Scheme 2 in Kintek Explorer yielded very poor fits and highly skewed residuals ( Figure 4—figure supplement 2 ) . Thus , a different kinetic scheme is needed to fit these data , one where competitor DNA is actively contributing to the dissociation of the pre-bound p18mer* . The simplest model to explain active participation of a competitor DNA in the dissociation of an already bound DNA is formation of a ternary complex wherein the competing DNA binds to PARP1 prior to the dissociation of the pre-bound DNA ( Scheme 3 ) . This model consists of four rate constants: k2 , ( formation of the ternary complex ) , k-2 ( release of the competing DNA to regenerate the pre-bound complex ) , k3 ( release of the pre-bound DNA to generate PARP1 only bound to the competing DNA ) , and k-3 ( re-formation of the ternary complex ) . Experimentally , both the starting pre-bound complex and the ternary complex are assigned a high anisotropy , whereas the final complex bound only to competing , unlabeled DNA is assigned a low anisotropy . In order to best constrain the four rate constants required to describe Scheme 3 , we used a broader range of competing DNA concentrations ( 50 nM – 4000 nM ) . Also , each concentration series was independently determined and fitted using Kintek Explorer at least three times . Representative fits of this model to the data are shown in Figure 4 , and the residuals indicate very good agreement between the data and this model , even at low concentrations of competitor DNA where kobs does not fit to a simple exponential and the apparent extent of exchange is significantly lower than at high concentrations . The aggregated rate constants are shown in Table 1 and the derived dissociation constants are shown in Figure 4—figure supplement 1 . The quality of the fits with the kinetic model in Scheme 3 provides strong support for the requisite formation of a ternary complex in the dissociation of DNA from PARP1 . The second order rate constant of association for the second DNA molecule is 0 . 043 nM−1s−1 is almost two orders of magnitude lower than that for association of the first DNA oligomer . The KD for the second DNA strand is 2600 nM , explaining why this complex would be rarely if ever detected under typical experimental conditions performed at nanomolar concentrations of PARP1 . Note that the rates of association and dissociation for the second DNA are not ‘symmetrical’ ( i . e . k2 ≠ k-3 and k-2 ≠ k3 ) . This asymmetry is most pronounced in the comparison between k-2 and k3: the pre-bound DNA is less likely to dissociate than the second competitor DNA . This observation makes intuitive sense in that the newly incoming DNA presumably binds to a different ( weaker ) site than the originally more tightly bound DNA . Although there is a lack of symmetry in the rate constants , the kinetically derived dissociation constants ( KDs ) are quite similar ( Figure 4—figure supplement 1 ) . Formation of a ternary complex with two different DNA molecules bound simultaneously requires two separate DNA binding sites . PARP1 has four domains that are known to contribute to DNA binding: Zn1 , Zn2 , Zn3 , and WGR ( Figure 1 ) . In order to identify if one or more of these domains selectively contributes to the formation of the ternary complex required for efficient DNA release , we generated constructs of PARP1 lacking each of these individual domains . To facilitate proper assembly of the remaining domains , we inserted a flexible 30 amino acid linker into each deletion , except for the N-terminal deletion of Zn1 . All mutants were purified to near homogeneity and were tested for DNA-dependent PARylation activity ( Figure 1—figure supplement 1 ) . As previously reported ( Langelier et al . , 2012 ) , Zn1 , Zn3 , and WGR are essential for catalytic activity , and thus deletion of these domains disrupts PARylation activity . On the other hand , the deletion of the non-essential Zn2 domain does not affect PARylation activity . We next measured the rates of association to , and dissociation from p18mer* for each of the individual deletions of the DNA-binding domains , using the stopped-flow anisotropy assays described above . As for wild-type PARP1 , each deletion construct was assayed at multiple different concentrations of protein or DNA , and the data were analyzed using global fitting in Kintek Explorer . Despite each construct missing one DNA-binding domain , all four bound to DNA with similar second-order rate constants ( k1s , Table 1 , Figure 2—figure supplements 1–4 ) . In the dissociation experiment , the Zn deletions ( ΔZn1 , ΔZn2 , and ΔZn3 ) behaved essentially like wild-type PARP1: increasing competitor DNA concentrations yielded increasing kobs , and the data were best described by global fitting of the kinetic model of Scheme 3 with the formation of a ternary complex ( Figure 4—figure supplements 3–5 , Table 1 ) . In contrast , the ΔWGR mutant behaved dramatically differently; increasing concentrations of competitor DNA did not yield higher kobss , and globally the data were best described by Scheme 2 ( Figure 5A , Table 1 ) . In the structure of PARP1 bound to a DSB , Trp589 in the WGR domain stacks against the ribose sugar of the 5’-end of the DNA ( Figure 1 ) ( Langelier et al . , 2012 ) . Since deletion of the entire WGR domain disrupted formation of the ternary complex , we tested whether the more conservative W589A substitution could recapitulate this effect . PARP1-W589A was prepared ( Figure 1—figure supplement 1 ) and tested in both the association and dissociation assays . The W589A point mutation is properly folded , as the mutant and wild-type PARP1 have identical melting temperatures ( 43 . 9 ± 0 . 3 vs . 43 . 3 ± 0 . 3°C ) . Similar to what was observed with the deletion of the entire WGR domain , the W589A mutant also bound to free DNA rapidly ( Figure 2—figure supplement 5 ) , and released DNA via the simple mechanism in Scheme 2 that is not dependent on binding a second DNA molecule ( Figure 5B , Table 1 ) . If PARP1 movement around the nucleus is indeed facilitated by the monkey-bar mechanism , then undamaged DNA , not just a short oligomer , should also promote release of pre-bound DNA . To address this question , we used plasmid DNA as a competitor of p18mer* pre-bound to PARP1 . Intact plasmid ( 4 . 5 kb;~90% supercoiled and 10% nicked ) at 5 nM is a surprisingly effective competitor of a pre-formed 25 nM complex , yielding a kobs comparable to what is observed with 1 µM p18mer ( Figure 6A ) . For comparison , 5 nM of p18mer yields no observable release of p18mer ( Figure 6A ) . These results demonstrate that undamaged DNA is an effective trigger for the release of pre-bound p18mer* . To validate these results further , we prepared increasing numbers of free ends ( models for DSBs ) by restriction digests of the plasmid with different enzymes ( Figure 6—figure supplement 1 ) . If DNA ends are the actual triggers for release of pre-bound p18mer* , we would expect the purposeful increase in the numbers of ends ( using the same amount of total plasmid ) to yield increasing rates of release ( see insert to Figure 3 ) . Instead , we observe an unchanged kobs , regardless of whether the concentration of ends is 10 nM ( cut once ) , 20 nM ( cut twice ) , 30 nM ( cut thrice ) , or 240 nM ( cut 24 times ) ( Figure 6A ) . In the hallmark experiment of inter-strand transfer , we demonstrate the concentration-dependence of intact competitor plasmid on the apparent rate of release of p18mer* , wherein the data were analyzed using global fitting in Kintek Explorer . As seen for p18mer , increasing competitor DNA concentrations of plasmid DNA yielded increasing kobs , and the data were best described by global fitting of the kinetic model of Scheme 3 with the formation of a ternary complex ( Figure 6B ) . In fact , the k-2 , k3 , and k-3 are all similar to the values seen previously using p18mer ( Table 1 ) . In order to have 5 nM plasmid release all the 25 nM pre-bound p18mer* , we can assume there are minimally five binding sites per plasmid . At the other extreme , estimating that one can place one PARP1 every 10 bp , there are maximally ~450 binding sites . Thus , one can place limits on the true value for k2 of 0 . 04–3 . 7 nM−1s−1 . Interestingly , the lower limit of this rate of association is similar to that measured for p18mer ( Table 1 ) , suggesting that undamaged DNA is a very effective competitor of damaged DNA . We conclude that intact DNA can also engage the monkey-bar mechanism to facilitate the movement of PARP1 around the nucleus . In order to test the physiological relevance of the mechanism of DNA-dependent release of DNA from PARP1 revealed in our in vitro experiments with p18mer and intact plasmid , we compared the rate and magnitude of accumulation of wild-type PARP1 with the W589A mutant at sites of DNA damage in cells . Mouse embryo fibroblasts were transiently transfected with GFP-tagged PARP1 ( wild-type or W589A ) , and DNA damage was induced by laser microirradiation at a designated region of interest ( ROI ) within the nucleus . Accumulation of PARP1 in the ROI was monitored by confocal microscopy for 1–5 min and the diffusion coefficient ( Deff ) and magnitude of PARP1 accumulation ( F ) were derived as recently described ( Mahadevan and Rudolph , 2018 ) . As seen in Table 2 , the W589A mutant accumulated to a lower level and with a significantly slower diffusion coefficient than wild-type PARP1 . Given the uncertain experimental basis for PARP1 trapping on DNA in cells treated with clinically relevant inhibitors of PARP1 ( Shen et al . , 2015 ) , we used the rigorous in vitro assay described above to investigate whether these inhibitors lead to a change in the rate or mechanism of DNA dissociation . We monitored the dissociation of p18mer* from PARP1 by fluorescence anisotropy in the presence of four different tight-binding inhibitors of PARP1 , using 1 µM competitor p18mer , conditions which lead to the inhibition of auto-PARylation ( Figure 7—figure supplement 1 ) . The observed dissociation curves were fit with a first-order exponential and the calculated rates were essentially identical to the DMSO control for olaparib , veliparib , niraparib , and talazoparib ( Figure 7A ) . To ensure that in the presence of inhibitor , DNA dissociation was still dependent on binding of competitor DNA ( Scheme 3 ) , we investigated the DNA concentration dependence of p18mer* dissociation in the presence of talazoparib , the most potent PARP1-trapping compound ( Shen et al . , 2015 ) . The dissociation of p18mer* in the presence of talazoparib ( 50 nM ) was measured at varying concentrations of competitor p18mer ( 1–4 µM ) and a concentration-dependent increase in kobs was observed just as for the control without inhibitor ( Figure 7B ) . We conclude that these inhibitors do not change the rate or mechanism of DNA dissociation from PARP1 .
Our results regarding the mechanisms of association and dissociation of PARP1 to and from DNA have important implications for our understanding of how PARP1 can move around the nucleus to scan for DNA damage . In vitro , PARP1 binds to DNA at or above the commonly accepted diffusion-limited rate ( Record et al . , 1991 ) of 1–2 nM−1s−1 , consistent with its extremely rapid accumulation at sites of DNA damage in vivo ( Mortusewicz et al . , 2007; Aleksandrov et al . , 2018 ) . In fact , compared to other model systems using DNA oligomers , PARP1 – DNA association is among the fastest previously reported . For comparison , the rates of association of eukaryotic mismatch repair complex Msh2-Msh6 ( Biro et al . , 2010 ) , RNA polymerase ( Johnson and Chester , 1998 ) , 8-oxoguanine-DNA-glycosylase ( Lukina et al . , 2017 ) , and papillomavirus E2 protein ( Ferreiro and de Prat-Gay , 2003 ) with double-stranded DNA are 0 . 002 , 0 . 004 , 0 . 13 , and 0 . 6–1 . 4 nM−1s−1 , respectively , effectively spanning three orders of magnitude . The fast association of PARP1 with DNA means that when dissociation does occur , re-association is most likely to the same site on the same DNA , as association is faster than diffusion carrying PARP1 away from its original binding site . This observation suggests that PARP1 , like other DNA-binding proteins such as transcription factors ( Mirny et al . , 2009 ) , must have a mechanism for moving around the genome that does not rely on simple dissociation and re-association . Although protein sliding along DNA in one-dimension has previously been invoked as a potential mechanism for accelerating the search for specific binding sites ( Berg et al . , 1981 ) , more recent publications point out potential difficulties with such a long-distance sliding model ( Halford , 2009 ) , which is even more difficult to envision given the organization of DNA into nucleosomes in the eukaryotic genome . Instead , we have found that PARP1 dissociation from DNA is triggered by binding of an additional DNA oligomer or undamaged plasmid prior to dissociation from the first DNA oligomer . We envision a ‘monkey bar’ model ( Vuzman et al . , 2010a; Vuzman et al . , 2010b ) , wherein PARP1 moves from one DNA molecule to another DNA molecule , much like a child swings from bar to bar , transferring one hand at a time . This mechanism allows PARP1 to effectively scan the genome , moving to new and different sections of DNA . In the absence of competing DNA , PARP1 would remain effectively stuck at or near one site given its fast rate of association . Interestingly , undamaged DNA appears to be very effective at promoting the monkey-bar mechanism ( Table 1 ) , raising the question of how PARP1 remains stationary at sites of DNA damage . We have found that the WGR domain provides the other weaker ‘hand’ to facilitate the movement from one DNA strand to the next . Based on the structure of PARP1 bound to a DSB ( Langelier et al . , 2012 ) , the WGR domain would need to first dissociate from the bound DNA prior to association with a second different molecule of DNA . Such a model is consistent with an NMR study that provides evidence for the stepwise assembly of PARP1 on a site of DNA damage where the Zn1 , Zn2 , and Zn3 domains engage a DNA molecule prior to final engagement of the WGR and catalytic domains ( Eustermann et al . , 2015 ) . Thus , one can readily imagine a partial reversal of this process wherein the WGR domain releases the original DNA to bind the incoming DNA , followed by release of the original DNA and subsequent rearrangement of the domain around the newly bound DNA ( Figure 8 ) . A recent study of PARP1 using single-molecule DNA tightrope assays provides strong evidence for such a monkey bar mechanism ( Liu et al . , 2017 ) . It was shown that micro-dissociation of one of PARP1’s multiple DNA-binding domains from DNA allows it to bind to a free 37 bp fragment , thus preventing rebinding of the domain to the tightrope and accelerating overall macro-dissociation . Our mechanism for DNA-dependent DNA dissociation also provides a compelling explanation for the wide diversity and significantly weaker dissociation constants previously reported for PARP1 with DNA ( Langelier et al . , 2018; Clark et al . , 2012; Langelier et al . , 2010 ) : the measured apparent KD depends strongly on the experimental conditions ( i . e . [DNA] ) under which the experiment is performed , since higher concentrations of DNA promote release of DNA . Strong in vitro evidence for a monkey bar mechanism , also known as intersegment transfer , exists for various other proteins , all of them transcription factors . Kinetic experiments similar to ours have demonstrated strong dependence of DNA release on additional DNA for the lac repressor ( Ruusala and Crothers , 1992 ) , cAMP receptor protein ( Fried and Crothers , 1984 ) and glucocorticoid receptor ( Lieberman and Nordeen , 1997 ) . NMR methods have demonstrated that both HoxD9 ( Iwahara et al . , 2006 ) and Oct1 ( Doucleff and Clore , 2008 ) can bridge one DNA strand to another . A combination of methods including NMR , rate measurements , and computational modeling have elegantly demonstrated how EgrI uses its three Zn fingers to both bind and scan other DNA fragments as it moves between different recognition sites on different DNA molecules ( Zandarashvili et al . , 2015 ) . Interestingly , mutagenesis of specific residues in EgrI was used to shift the equilibrium between binding and scanning modes . One major caveat to any biochemical investigation of the mechanism of DNA – protein interactions is the artificial nature of a DNA oligomer compared to intact chromatin . This limitation affects studies of PARP1 interactions with DNA in particular because it is quite difficult to prepare completely intact DNA without ends or nicks , preferably wrapped around nucleosomes . Thus , it was important to test the significance of the monkey bar mechanism in a more physiologically relevant model . We have demonstrated the validity of interstrand transfer in vivo by demonstrating that the point mutant W589A , which disrupts DNA-dependent release of DNA , accumulates slower and to a lesser amount than wild-type PARP1 at sites of laser microirradiation ( Table 2 ) . The slower accumulation of W589A in cells is a particularly powerful demonstration of the importance of the monkey bar mechanism for PARP1 in finding sites of DNA damage for two reasons . First , the rate of dissociation for W589A from DNA is greater than for wild-type ( 20 s−1 vs . < 10 s−1 ) . Second , the apparent KD of W589A for DNA is weaker than that of wild-type PARP1 ( 5 nM vs . < 3 nM , respectively ) . Simplistically , these two observations might suggest that W589A should arrive at sites of DNA damage more rapidly than wild-type PARP1 since its interaction with DNA is not as tight or as long-lived ( i . e . it spends less time occupying irrelevant sites ) ; yet we observe the opposite . The monkey bar mechanism provides the explanation for these results: high concentrations of intranuclear DNA allow PARP1 to explore the nucleus rapidly . A dysfunctional monkey ( W589A ) surrounded by a lot of DNA does not move as rapidly . Thus , we have quantitatively demonstrated the importance of intersegment transfer in the accumulation of a DNA-binding protein at its target destination in vivo . Finally , our results also provide further insight into the much-discussed topic of PARP1 ‘trapping’ , wherein cells treated with inhibitors of PARP1 exhibit a shift of PARP1 from the soluble fraction to a chromatin-associated insoluble fraction ( Murai et al . , 2012 ) . Many cell-based studies have since confirmed the phenomenon of PARP-trapping ( Pommier et al . , 2016 ) . Our data showing no effect of four different tight-binding inhibitors of PARP1 on the release of DNA agree with a previous thorough biochemical analysis that also could not find any effects of inhibitors on DNA binding constants or rates of dissociation ( Hopkins et al . , 2015 ) . Thus , the mechanistic basis for PARP-trapping is more complex than can be reconstituted in vitro . Our results are in agreement with recent findings that PARP inhibitors lead to defective fork recovery and/or homologous recombination-mediated repair , and thus an increase in DNA damage where PARP1 is bound to DNA , and lacking activity due to inhibitor binding ( Maya-Mendoza et al . , 2018 ) .
NAD+ was obtained from Sigma . Olaparib , veliparib , niraparib , and talazoparib were obtained from Selleck . DNA oligonucleotides and their complementary strands were obtained from IDT: p18mer: 5’-phosphate-GGGTTGCGGCCGCTTGGG-3’ . Labeled oligonucleotides with a 5’-fluorescein dye ( * ) were also obtained from IDT . Double-stranded fragments were prepared by annealing at 100 µM DNA in 10 mM Tris-HCl ( pH 7 . 5 ) , 100 mM NaCl , and 0 . 1 mM EDTA . The DNA was heated to 95°C for 5 min and then slowly cooled at 0 . 1 °C/second to 4°C . Annealing was confirmed by 10% ( wt/vol ) native TBE-PAGE at 200 V for 30 min . Intact supercoiled plasmid ( pUC19-601-147-12copy ) is a pUC derivative that was prepared as described ( Dyer et al . , 2004 ) . Restriction enzymes were obtained from New England Biolabs . The pET28a vector encoding cDNA of full-length human PARP1 was used to design constructs lacking various domains of PARP1 following the method outlined in Hansson et al ( Hansson et al . , 2008 ) . Briefly , primer one was designed as a reverse complement of the sequence that corresponds to 20–25 bases upstream of the DNA sequence to be deleted , followed by 20–25 bases corresponding to the downstream sequence . Primer two corresponds to the complementary strand . These primers were used in a PCR reaction to loop out the DNA encoding individual domains of PARP1: ΔZn1 ( M1-K97 ) , ΔZn2 ( G96-linker-K207 ) , Δn3 ( G215-linker-A367 ) , and ΔWGR ( N517-linker-L655 ) . After PCR , DpnI digestion was used to degrade the template plasmid and was then transformed to generate clones . Next , a linker DNA sequence encoding amino acids LLA ( GS ) 4GAAL was inserted in place of the deleted domain using partially overlapping primers comprising the entire sequence of the insert followed by 20–25 bases of the downstream sequence . Thereafter , another step of insertion of linker DNA sequence encoding amino acids ALA ( GS ) 5GLAL upstream of the previous insert was performed in a similar manner . The plasmids used to express various domain deletion PARP1 mutants eventually all contained the 30 amino acid linker ALA ( GS ) 5GLALLLA ( GS ) 4GAAL in place of the deleted PARP1 domain . The W589A mutant of PARP1 was generated using QuikChange Mutagenesis ( Agilent ) following the manufacturer’s instructions . All constructs were verified by DNA sequencing of the entire PARP1 gene . Wild-type PARP1 , all deletion constructs , and the W589A mutant of PARP1 were expressed and purified from E . coli as previously described ( Clark et al . , 2012; Langelier et al . , 2011b ) with the minor modification that PARP1 was eluted from the nickel-NTA column using a gradient from 20 to 400 mM imidazole . PARylation activity was evaluated by incubating 0 . 5 µM PARP1 with 1 µM p18mer and 500 µM NAD+ in 50 mM Tris- HCl ( pH 7 . 5 ) , 50 mM NaCl , and 1 mM MgCl2 , for 5 min . Reactions were quenched in Laemmli buffer , boiled for 5 min , and then resolved on SDS-PAGE ( 4–20% ) . PARP1 stability was evaluated using the Protein Thermal Shift Dye Kit from Applied Biosystems and a BioRad C1000 ThermalCycler with a CFX96 RealTime module . A SX20 Stopped-Flow Spectrometer ( Applied Photophysics ) was used for measuring fluorescence anisotropy using an excitation wavelength of 485 nm and cut-off filters in the parallel and perpendicular detectors at 515 nm . Association reactions were measured by mixing equal volumes of p18mer* ( 60 nM ) with three to eight different concentrations of PARP1 ( 60–250 nM ) and monitoring the anisotropy at 20°C for 25 ms . All indicated concentrations are after mixing . Although PARP1 can bind to both ends of p18mer* ( and p18mer ) simultaneously ( Langelier et al . , 2012 ) , we treat each DNA oligomer as one equivalent ( not two ) because fluorescence anisotropy detects only the first binding event . Control reactions used for determining background signal lacked PARP1 . For measuring dissociation , a pre-formed complex of PARP1 ( 37 nM ) and p18mer* ( 25 nM ) was mixed with 5–15 different concentrations of p18mer ( 100 nM – 8 µM ) and anisotropy was monitored at 20°C for 1–5 s . Control reactions for determining background signal lacked p18mer . All reactions were performed in 50 mM Tris-HCl ( pH 7 . 5 ) , 50 mM NaCl , 1 mM MgCl2 , 0 . 1 mM EDTA , and 0 . 01% IGEPAL . For all stopped-flow reactions , data were collected in log mode , and 10–12 shots were averaged for each different concentration of reagents . All experiments consisting of series of different concentrations of PARP1 ( for association ) or of p18mer ( for dissociation ) were performed on at least three different days with a least two different preparations of protein . Plasmids for dissociation experiments were either untreated or digested at 0 . 25 mg/mL using 50–1000 U of the appropriate restriction enzyme at 37°C overnight . SacI , DrdI , EarI , and EcoRV were used to generate 1 , 2 , 3 , and 24 cuts , respectively . ( Digestion with EcoRV yields the parent plasmid and 12 identical inserts of 147 bp DNA ) . Dissociation experiments in the presence of inhibitors ( 50 nM ) of PARP1 were compared to DMSO controls ( <2 %v/v ) . Data were initially analyzed for fitting to single exponential kinetics using the software from Applied Photophysics ( ProDataTSV ) . Global analysis incorporating multiple different concentrations of protein or competing DNA were performed using KinTek Explorer ( KinTek Corporation ) . For association kinetics , control reactions in the absence of protein were used to determine the baseline , and the maximum anisotropy signal ( identical at all protein concentrations ) was used to convert anisotropy units to concentration values . For dissociation kinetics , control reactions in the absence of DNA were used to determine the baseline , and the maximum anisotropy signal at high concentrations of p18mer ( 1–4 µM ) were used to convert anisotropy units to concentration values . For dissociation kinetics using plasmid , global fitting was performed by assuming 5–450 binding sites/4 . 5 kb plasmid , which as consequence implies that the value of k2 is subject to this assumption ( Scheme 3 ) . However , values for k-2 , k3 , and k-3 are independent of this assumption . For clarity , only a subset of the concentrations used experimentally are shown in the figures . Mammalian expression plasmid ( pEGFP-C3 , 250 ng/20 , 000 cells ) encoding full-length GFP-tagged human PARP1 was transfected using jetPEI ( Polyplus Transfection ) into wild-type mouse embryo fibroblasts ( MEFs ) cultured in DMEM supplemented with 50 μg/ml of gentamicin and 10% FBS , as previously described ( Mahadevan et al , in preparation ) . Cells were sensitized with Hoechst 33342 ( Invitrogen ) ( 10 μg/ml ) for 10 min prior to induction of DNA damage using a 405 nm diode laser ( 100% power for 1 s ) . Cells were imaged for 1–5 min using excitation at 488 nm within a heated environmental chamber set to 37°C and 5% CO2 ( Nikon A1R confocal laser scanning; frame size of 512 × 512 ) . Analysis of the fluorescent images was carried out using custom codes in MATLAB and Mathematica to allow derivation of the diffusion coefficient ( Deff ) and mobile fraction of PARP1 ( F ) ( Mahadevan and Rudolph , 2018 ) . These codes are provided as DNA Repair Analysis Toolbox . mtlbx ( Source code 1 ) , Bioformats Image Toolbox ( v1 . 0 . 4b ) . mltbx ( Source code 2 ) , and Q-FADD 0 . 13 . nb ( Source code 3 ) . | Our cells constantly withstand damage that can lead to breaks in the strands of our DNA . These cuts need to be fixed for the cell to stay healthy . When a break happens , one of the first responders to the scene is a protein known as PARP1 . It binds to the ruptured strand ( or strands ) and then it recruits other repair agents to that location . But first , PARP1 needs to scan for cuts and notches amongst an overwhelming amount of DNA that is still intact . This is a complicated task , especially since the protein tends to bind both broken and unbroken DNA . How does it not stay ‘stuck’ on an undamaged portion of the genome ? Here , Rudolph et al . use a combination of biochemical techniques and cell biology to show that PARP1 travels through our genome by swinging from one DNA location to another , the way a child grabs onto monkey bars . One of the DNA-binding domains of PARP1 , known as the WGR-domain , acts like an arm and initiates the movement by gripping onto a new segment of DNA . In fact , chopping off the WGR-domain or disabling it through mutations makes PARP1 worse at finding DNA breakages in the cell . Unfixed DNA damage can lead to a cell becoming cancerous; ultimately , if the breakages keep accumulating , the cell does not survive . This makes PARP1 an important target for cancer treatment . Indeed , certain drugs already rely on trapping the protein so that tumor cells die . Understanding how cells cope with DNA damage and exactly how PARP1 works could help in the fight against cancer . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology"
] | 2018 | Poly(ADP-ribose) polymerase 1 searches DNA via a ‘monkey bar’ mechanism |
Tumors defective for DNA polymerase ( Pol ) ε proofreading have the highest tumor mutation burden identified . A major unanswered question is whether loss of Pol ε proofreading by itself is sufficient to drive this mutagenesis , or whether additional factors are necessary . To address this , we used a combination of next generation sequencing and in vitro biochemistry on human cell lines engineered to have defects in Pol ε proofreading and mismatch repair . Absent mismatch repair , monoallelic Pol ε proofreading deficiency caused a rapid increase in a unique mutation signature , similar to that observed in tumors from patients with biallelic mismatch repair deficiency and heterozygous Pol ε mutations . Restoring mismatch repair was sufficient to suppress the explosive mutation accumulation . These results strongly suggest that concomitant suppression of mismatch repair , a hallmark of colorectal and other aggressive cancers , is a critical force for driving the explosive mutagenesis seen in tumors expressing exonuclease-deficient Pol ε .
Human cancers share common features of genome instability and mutagenesis ( Hanahan and Weinberg , 2011 ) that are the sources of the 103 to 106 somatic mutations observed in the genomes of most types of adult tumors ( Stratton , 2011; Wheeler and Wang , 2013 ) . The total mutation burden in a tumor is the result of multiple mutational pathways operating within the cells at varying rates over time . This can complicate attempts to assign the relative contributions of each pathway to the mutation spectrum of a tumor . One essential tool to our understanding of how mutations accumulate and influence tumor progression is using computational means to extract multiple individual signatures from many tumor genomes ( Alexandrov et al . , 2013a; Alexandrov and Stratton , 2014; Haradhvala et al . , 2016 ) . This is proving to be instrumental in resolving the relative extents to which pathways contribute to the ultimate mutation spectrum in a tumor ( Nik-Zainal et al . , 2016; Roberts et al . , 2013 ) . Comparing these tumor mutation signatures to those generated in experimental cell lines is another critical tool to understanding the relative rates and causality of mutation acquisition ( Fox et al . , 2016; Helleday et al . , 2014 ) . Traditionally , these measurements have relied on assays using reporter genes , which necessarily look at a tiny fraction of the genome and may miss global contributions to genome instability . Advances in next generation sequencing now allow for detailed genome-wide analyses of mutation accumulation over defined periods of cellular growth . Since each nucleotide in the genome is subject to the three major determinants of replication fidelity - nucleotide selection , proofreading and mismatch repair ( MMR ) - during every round of replication , tumors and cells with defects in replication fidelity are uniquely poised to address these issues . Proofreading defects are now known to occur in a wide variety of tumors , with significant enrichment in colorectal and endometrial tumors ( Cancer Genome Atlas Network , 2012; Kandoth et al . , 2013; Heitzer and Tomlinson , 2014; Rayner et al . , 2016 ) . Mutations in DNA polymerase ( Pol ) ε cluster in the exonuclease proofreading domain and the tumors are clinically characterized by several criteria , including being ultrahypermutated , having a unique mutation spectrum , containing a heterozygous Pol ε mutation with no evidence of loss of heterozygosity ( LOH ) and being microsatellite stable ( MSS ) ( Briggs and Tomlinson , 2013; Church et al . , 2013; Palles et al . , 2013; Zhao et al . , 2013; Henninger and Pursell , 2014; Shinbrot et al . , 2014; Shlien et al . , 2015; Barbari and Shcherbakova , 2017 ) . Whole genome and whole exome analyses of tumors have been the primary means to establish the ultrahypermutated ( >100 Mutations per megabase ) unique mutational signature that distinguish Pol ε tumors from other cancers ( Alexandrov et al . , 2013a; Alexandrov and Stratton , 2014; Shinbrot et al . , 2014; Shlien et al . , 2015; Alexandrov et al . , 2013b; Campbell et al . , 2017 ) . While there is a rich history of studies on the effects of exonuclease defects on mutagenesis in model organisms , the extent to which Pol ε proofreading-deficiency by itself drives each of these criteria remains poorly understood . It is clear from studies in model organisms that complete , biallelic inactivation of Pol ε proofreading activity causes mutagenesis and carcinogenesis in model organisms , where mutation rates have been precisely measured using reporter genes . For example , mutation rates are increased in haploid or diploid yeast strains expressing only proofreading-deficient alleles of Pols ε ( Morrison et al . , 1991; Morrison and Sugino , 1994; Shcherbakova et al . , 2003 ) or δ ( Morrison et al . , 1993; Simon et al . , 1991; Herr et al . , 2011a ) . These rates are further elevated when combined with defects in mismatch repair , indicating that these errors are made during replication ( Morrison and Sugino , 1994; Tran et al . , 1999; Tran et al . , 1997; Kennedy et al . , 2015 ) . In mouse models , homozygous inactivation of both copies of either Pol ε or δ exonuclease activity ( Pol εexo-/exo- or Pol δexo-/exo- ) causes increased mutation rates and cancer ( Albertson et al . , 2009; Goldsby et al . , 2002; Goldsby et al . , 2001 ) . Interestingly , their tumor spectra are different , with gastrointestinal tumors predominant in Pol εexo-/exo- mice while thymic lymphomas are the major tumor in Pol δexo-/exo- mice . However , mice with a heterozygous inactivation of a single Pol ε proofreading allele ( the monoallelic Pol εwt/exo- genotype ) fail to develop tumors when mismatch repair is functional ( Albertson et al . , 2009 ) . The equivalent diploid heterozygous Pol ε exonuclease mutant in yeast is also a mutator , but the effect is modest and partially dominant to the wild type allele and lacks the unique mutation spectrum seen in human tumors ( Morrison and Sugino , 1994; Shcherbakova et al . , 2003; Morrison et al . , 1993; Kane and Shcherbakova , 2014 ) . These results raise critical questions as to the source of the unique , ultrahypermutant phenotype in human tumors with heterozygous Pol ε exonuclease-deficiency . Mismatch repair is responsible for the recognition and removal of replication errors and deficiencies in this activity cause genome instabilities that can lead to cancer ( Kunkel and Erie , 2005; Li , 2008; Jiricny , 2013; Modrich , 2006 ) . Mismatch repair is normally an extremely efficient process , correcting more than 99% of replication errors . However , genome-wide studies have recently shown that MMR efficiencies can vary by over two orders of magnitude and are influenced by a number of factors , including the strand on which the mismatch occurs , the polymerase that made the error , the nature of the mismatch , local sequence context , distance from the origin and replication timing ( Hawk et al . , 2005; Hombauer et al . , 2011; Lujan et al . , 2014; Lujan et al . , 2012; Supek and Lehner , 2015 ) . Patients with biallelic mismatch repair disorder ( bMMRD ) have biallelic germline inactivating mutations in a mismatch repair gene and are completely lacking mismatch repair and develop a number of early-onset tumors in which microsatellite instability ( MSI ) is readily detectable ( Durno et al . , 2017; Wimmer et al . , 2014 ) . A subset of these patients acquires a later somatic mutation in a single allele of Pol ε , leading to very aggressive tumor development . Mutation rates from these Pol εwt/exo- MMR−/− tumors have been estimated on the order of several hundred per genome duplication ( Shlien et al . , 2015 ) . This is consistent with results from model systems as mice with the equivalent genotype ( heterozygous Pol εwt/exo- combined with homozygous MMR−/− ) develop tumors within 1–2 months ( Treuting et al . , 2010 ) . The equivalent yeast strains are strong mutators as well ( Shcherbakova et al . , 2003; Morrison et al . , 1993; Kennedy et al . , 2015 ) . However , since sporadic POLE tumors are generally microsatellite stable , the role of MMR in Pol ε proofreading-deficiency in the development of these MSS tumors remains a critical unanswered question . Whether MMR and POLE defects together are required for ultramutation , elevated mutation rates or for establishing the unique mutation signature is unknown . Understanding how MMR function or dysfunction affects proofreading-dependent mutagenesis is essential to understanding the mechanisms of mutagenesis during cancer development . In the current study , we constructed a human cell line model system to address the roles of Pol ε proofreading in driving the clinical characteristics that define Pol ε tumors . Critically , we used a targeted knock-in approach to inactivate one copy of Pol ε 3'−5' exonuclease activity , since human tumors contain heterozygous , monoallelic Pol ε mutations . Using mutation rates measured at a reporter gene in combination with whole-exome and whole-genome sequencing we found a rapid accumulation of large numbers of Pol ε-specific mutations in mismatch repair-deficient cells . This confirms results suggested by observations in Pol ε mutant bMMRD tumors . We further show that mismatch repair is able to suppress exonuclease-deficient Pol ε-induced mutation rates back to wild type levels using a combination of reporter gene and whole-exome sequencing ( WES ) . These results support the idea that additional unique features beyond a single exonuclease active site inactivation are helping facilitate the massive mutation acquisition seen in microsatellite stable tumors containing mutant Pol ε .
Tumors with mutations in the exonuclease domain of POLE are generally microsatellite stable and show no or low loss of heterozygosity , suggesting that inactivation of exonuclease activity in one allele is sufficient to drive mutagenesis and tumor development , though this has not been directly tested previously . To test whether inactivation of a single allele of Pol ε proofreading was sufficient to cause a mutator phenotype in human cells , we used recombinant adenoassociated virus ( rAAV ) -mediated gene targeting to engineer a diploid human cell line to express one allele of Pol ε with the D275A/E277A double substitution ( Figure 1—figure supplements 1–2; Figure 1—source data 1 ) . We chose the D275A/E277A mutation because it inactivates exonuclease proofreading in vitro ( Shcherbakova et al . , 2003; Korona et al . , 2011 ) . The parental cell line , HCT-116 , is constitutively mismatch repair-deficient due to an inactivating mutation in Mlh1 , thus allowing us to first define the contributions of proofreading deficiency separately to mutagenesis . We then measured the mutation rate at the hypoxanthine-guanine phosphoribosyltransferase ( HPRT1 ) locus using 6-thioguanine ( 6-TG ) resistance and a fluctuation assay . The measurements were repeated in clones derived from independent exonuclease-deficient ( exo- ) allele integration events . A moderate mutator effect was seen in Pol εwt/exo- heterozygotes ( Figure 1A ) , indicating the exo- allele was partially dominant over the endogenous exo + allele , similar to what is seen in a mismatch repair-deficient diploid cell line heterozygous for a Pol ε proofreading mutation , pol2-4/+pms1/pms1 ( Pavlov et al . , 2004 ) . Mutation rates were not measured in cells from the comparable heterozygous Pol εwt/exo- mice lacking mismatch repair ( Albertson et al . , 2009 ) . To begin measuring the effect of inactivating a single Pol ε exonuclease allele on mutation rates in cells , we sequenced the HPRT1 gene from twenty and twenty-five independently derived 6-TGR ( and thus HPRT1 mutant ) clones from mismatch repair-deficient Pol εwt/wt and Pol εwt/exo- cells , respectively ( Figure 1—source data 2 ) . This allowed comparison to previously measured mutation rates from different groups using the same parental cell line . Mutation rates from the Pol εwt/wt cells were similar to the spontaneous mutation rates reported by three previous studies ( Bhattacharyya et al . , 1995; Glaab and Tindall , 1997; Ohzeki et al . , 1997 ) . These results suggest that the baseline rates of mutagenesis are an accurate measure of comparison for the Pol εwt/exo- cell lines . The increase in mutation rate seen in the Pol εwt/exo- mismatch repair-deficient cells was primarily due to base pair substitutions ( Figure 1B ) . Frameshift error rates did not change , in agreement with previous findings in vitro that Pol ε proofreading primarily strongly corrects base-base mispairs with little effect on frameshift fidelity ( Korona et al . , 2011 ) . However , the number of mutational events scored by this method is insufficient to make statistical claims regarding individual mutations , reinforcing the need for genome sequencing to examine mutations in all possible sequence contexts . Using an in vitro lacZ reversion substrate that specifically measures TCT→TAT transversions ( Shinbrot et al . , 2014; Shlien et al . , 2015 ) , the D275A/E277A mutant made these errors at a significantly higher rate in vitro than the wild type exonuclease-proficient Pol ε enzyme ( Figure 1C ) . We used a construct comprised of the N-terminal 140 kDa of Pol ε , which contains the DNA polymerase and exonuclease domains and has similar fidelity and catalytic activity to the complete four subunit holoenzyme ( Aksenova et al . , 2010; Ganai et al . , 2015; Zahurancik et al . , 2015 ) . Importantly , the elevated TCT→TAT error rate we observed with the D275A/E277A mutant was not statistically different from those measured with the S459F and S461P Pol ε cancer mutants previously ( Shinbrot et al . , 2014; Shlien et al . , 2015 ) , suggesting a common mechanism of mutagenesis for these hotspot mutations . Mutation rates calculated using reporter genes ( μL ) can be used to extrapolate to genome-wide per base pair mutation rates ( μBS ) ( Drake , 1991; Lynch , 2010 ) . The availability of high-throughput DNA sequencing now allows for empirical validation of these calculations in addition to providing insight into the influence of genomic context on mutagenesis . To address this we performed whole-genome sequencing ( 2 . 8 × 109 bp at an average depth of 36 . 1x ) on genomic DNA prepared from Pol εwt/exo- cells . Based on our measured mutation rate for HPRT1 ( μL ) in Pol εwt/exo- cells lacking mismatch repair ( 180 × 10−7 ) , we calculated a μBS value of 0 . 23 × 10−7 mutations per base pair per genome duplication . Because the parental HCT-116 cell line already carries a significant number of single nucleotide variants ( SNVs ) relative to the human reference sequence ( [Abaan et al . , 2013] and see Discussion ) , we needed a way of measuring de novo mutations resulting from Pol ε-dependent replication errors . To do this we first performed whole genome sequencing ( WGS ) on genomic DNA prepared from mismatch repair-deficient Pol εwt/exo- cells , which we then used as a matched normal control . We termed this mutation spectrum P0 . We then passaged these cells through a calculated 13 . 9 population doublings and then performed WGS again on the passaged population , which we termed P14 . Mutations unique to P14 arose during the defined number of population doublings . The P0 and P14 samples contained 140 . 3 and 141 . 4 Mut/Mb , respectively . Given the calculated μBS and the 2 . 8 × 109 bp sequenced , we predicted the accumulation of 906 novel genome-wide mutations after 14 population doublings . Whole-genome sequencing revealed 5 , 282 SNVs unique to the P14 population , 5 . 8-fold higher than that predicted from the μL at HPRT1 . Mutations observed in HPRT1 in this cell line may thus slightly underrepresent those found genome-wide . This difference is consistent with what is seen in microbes , where reporter gene mutation rates are consistently 6–8-fold lower than concurrently measured whole-genome mutation rates , likely due to phenotypic lag , strong selective pressure and transcription in the reporter gene ( Drake , 2012; Jee et al . , 2016; Lee et al . , 2012 ) . C→A transversions exceeding 20% of all base pair substitutions is a primary characteristic of mutation spectra from tumors containing Pol ε exonuclease domain mutations ( Rayner et al . , 2016; Shinbrot et al . , 2014 ) . C→A transversions were increased significantly in the Pol εwt/exo- cells as compared to the control Pol εwt/wt spectrum , accounting for 46% of all base pair substitutions ( Figure 2A , χ2 = 11 . 874 , p<0 . 0001 ) . These were not cell line artifacts , as whole exome sequencing from HCT-116 cells from two independent studies ( [Abaan et al . , 2013] and this study ) showed roughly 10% C→A transversions ( Figure 2A , p>0 . 5 ) . HCC2998 cells , which harbor the Pol εwt/P286R mutation , also showed a significant increase in C→A transversions relative to Pol εwt/wt cells ( Figure 2A , p<0 . 0001 ) . Two sequence context mutational hotspots were observed that are consistent with Pol ε exonuclease domain mutant spectra: C→A transversions in TCT context and T→G transversions in TTT context and , to a lesser extent , ATT and GTT contexts ( Figure 2B ) . These hotspots are seen in Pol ε tumors from patients with bMMRD ( Shlien et al . , 2015 ) , colorectal and endometrial cancer ( Alexandrov and Stratton , 2014; Cancer Genome Atlas Network , 2012; Kandoth et al . , 2013; Shinbrot et al . , 2014 ) , as well as in the Pol ε-P286R HCC2998 cells ( Figure 2—figure supplement 1 , data extracted from [Abaan et al . , 2013] ) . These are not mutational hotspots in HCT-116 cells , which contain wild type Pol ε ( Figure 2—figure supplement 1 ) . The largest number of mutations that arose during the 14 doublings were C→A transversions in triplet contexts containing adjacent cytosines: CCA , CCT , CCC and CCG . Triplet nucleotide occurrences can vary in the regions captured by WGS and WES . In order to address this we reanalyzed each sample relative to the number of times each trinucleotide is found in the relevant sample and found the hotspot patterns are all retained ( Figure 2—figure supplements 3–4 ) . The increase in C→A mutations in the CCT context was also seen in Pol ε exonuclease domain ( EDM ) tumors from bMMRD patients ( Shlien et al . , 2015 ) , suggesting a link between Pol ε replication errors left uncorrected by mismatch repair . C→A mutations in CCA , CCC and CCG contexts are slightly elevated in Mutation Signature 20 , which has been associated with loss of mismatch repair ( Alexandrov et al . , 2013b ) . These transversions were seen in the HCT116 cell line with wild type Pol ε ( Figure 2—figure supplement 1 ) , though to a lesser extent . The lack of C→T transitions in TCG contexts is significantly different from colorectal and endometrial Pol ε tumors , but consistent with their absence from bMMRD tumors with Pol ε EDM mutations ( Cancer Genome Atlas Network , 2012; Kandoth et al . , 2013; Shinbrot et al . , 2014 ) . While it is clear that Pol ε-dependent mutagenesis in the absence of functional MMR accounts for the ultramutated phenotype in bMMRD tumors with Pol ε mutations , the role of MMR in Pol ε somatic tumors is less clear . In order to measure the effects of MMR on Pol ε exonuclease-dependent replication errors , we wanted to measure error rates in both the presence and absence of MMR . Previous studies have restored MMR by stably adding Mlh1-expressing chromosome 3 to cells ( Glaab and Tindall , 1997 ) . We made Mlh1-encoding lentivirus and used this to infect Mlh1-deficient HCT-116 cells containing wild type and mutant Pol ε ( Figure 3A ) . Lentiviral Mlh1 expression reduced mutation rates at the HPRT1 locus by 14- to 20-fold in the wild type polymerase background ( Figure 3B ) , similar to the 12-fold reduction reported when the Mlh1-encoding chromosome 3 was added back to HCT-116 cells ( [Glaab and Tindall , 1997; Tindall et al . , 1998]; 73 × 10−7 and 5 . 9 × 10−7; 12 . 4-fold reduction ) , indicating that the expressed Mlh1 is functional . Mlh1 expression in Pol εwt/exo- cells caused an over 50-fold decrease in the mutation rate ( to 2 . 3 and 3 . 0 × 10−7 , Figure 3B ) , making them indistinguishable from those measured in Pol εwt/wt cells with Mlh1 expressed ( Figure 3B ) . This result also suggests that Msh3 is unlikely to play a significant role in correcting the exonuclease-deficient Pol ε errors since HCT-116 cells are deficient in this factor and it was not added back in these experiments ( Papadopoulos et al . , 1994 ) . When fluctuation assay mutation rates are very low due to a significant number of independent isolates giving rise to zero HPRT1-mutant colonies , as was the case here , an alternative method to measure mutation rates can be used . We chose to periodically measure HPRT1 mutant frequencies at increasing population doubling level ( PDL ) , where the slope of the plotted line is equal to the mutation rate ( Glaab and Tindall , 1997 ) . We measured HPRT1 mutant frequencies at several population doublings from PDL = 0 to PDL = 70 or 71 in Pol εwt/wt and Pol εwt/exo- cells expressing Mlh1 , respectively ( Figure 4A ) . At each PDL we scored between 1 and 19 6-TG-resistant colonies . However , when we sequenced the HPRT1 ORF from all 6-TG-resistant colonies we saw many instances of repeat mutations in a collection from a single PDL , indicative of a single mutational event that expanded throughout the population . Plotting mutant frequency values calculated for the indicated PDL using only the unique HPRT1 mutations ( Figure 4—source data 1 ) returned a line with slope of ~1 , suggesting that the mutation rates were at or near the level of detection of this assay . The Pol εwt/exo- mutant frequencies were consistently higher than those from the Pol εwt/wt cells , but this difference was not statistically significant ( Figure 4A ) . To determine if this phenomenon held throughout the genome , we carried out whole-exome sequencing to an average depth of 100x on the early ( PDL = 0 ) and late ( PDL = 70 ) samples from both Pol εwt/wt and Pol εwt/exo- mismatch repair-proficient cell lines ( Figure 4B ) . Using the PDL = 0 samples as matched normal controls , we measured similar low mutation rates in Pol εwt/wt and Pol εwt/exo- cells ( 13 × 10−9 Mut/bp/doubling and 18 × 10−9 Mut/bp/doubling , respectively ) . The total numbers of all mutations acquired were essentially no different than with wild type Pol ε . Interestingly , there was a statistically significant increase in C→A transversions ( p=0 . 0002 ) between the mismatch repair-proficient Pol εwt/exo- cells and the mismatch repair-proficient Pol εwt/wt cells , while no statistically significant difference was found in any other class of base pair substitution ( p>0 . 2 for each of the six classes , Fisher’s Exact Test ) . Further , all triplet context mutations were observed in insufficient numbers to evaluate statistically . C→A mutations were , however , observed in all triplet contexts seen as hotspots in the MMR-deficient cells ( CCA , CCC , CCG , CCT and TCT , Figure 4—figure supplement 1 ) . Mutation signature 10 , the unique Pol ε mutation signature , was extracted from Pol ε exonuclease-deficient mutation spectra from cells with and without mismatch repair ( Figure 2—figure supplement 2 and Figure 4—source data 2 ) . The relative contribution of signature 10 in Pol ε exo-deficient cells is closer to that seen in bMMRD patients ( Figure 2—figure supplement 5 ) , most likely due to the relative absence of C→T transitions in TCG context . These results indicate that the majority of replication errors made by the Pol ε-D275A/E277A mutant are in fact corrected by mismatch repair .
In the current study we examined the relative contributions of two essential determinants of replication fidelity , proofreading and mismatch repair , on mutagenesis in human cells . We used a combination of gene editing , reporter gene studies and next generation sequencing to measure mutation rates and specificities in human cells engineered to model proofreading-deficient tumors with and without mismatch repair . This is the equivalent to what occurs in human tumors with mutations in the Pol ε exonuclease domain and genomic mutation frequencies exceeding 100 mutations per Mb ( Cancer Genome Atlas Network , 2012; Rayner et al . , 2016; Shinbrot et al . , 2014; Shlien et al . , 2015 ) . We show that large and rapid mutation accumulation occurs when Pol ε exonuclease domain mutations occur along with inactivation of mismatch repair . Most of these are specific transversion mutations known to be hotspots of exonuclease-deficient Pol ε mutagenesis . We further show that this large increase in mutation rate is largely suppressed by functional mismatch repair . Taken together , these results suggest that the mechanism of replication error mutagenesis in sporadic tumors with heterozygous Pol ε mutations likely requires an additional feature , several of which are described below , including suppression of MMR and alternative effects on Pol ε activity . We used rAAV-mediated gene targeting to replace two exonuclease active site residues , D275 and E277 , with alanines on a single POLE allele . The single allelic inactivation was chosen to model the case in tumors with heterozygous Pol ε mutations . This double amino acid substitution has been shown to inactivate exonuclease proofreading in vitro and cause increased reporter gene mutation rates in yeast and mammalian cells ( Morrison et al . , 1991; Morrison and Sugino , 1994; Tran et al . , 1999; Albertson et al . , 2009; Korona et al . , 2011; Shcherbakova and Pavlov , 1996; Agbor et al . , 2013 ) . Next generation sequencing on these cells in the presence or absence of mismatch repair over defined numbers of population doublings allowed us to compare genome-wide mutation rates and spectra to the mutation spectra from patient tumors . Unbiased whole-genome sequencing confirmed the rapid accumulation of Pol ε-specific mutations seen in POLE tumors lacking functional mismatch repair ( Shlien et al . , 2015 ) . The total number of measured SNVs suggests a mutation rate of 380 mutations per population doubling , similar to the 608 mutations per cell cycle calculated for a mismatch repair-deficient brain tumor harboring a Pol ε exonuclease domain mutation . Our cellular mutation rate values possibly underestimate the true Pol ε exonuclease-deficient mutation rate for several reasons . Our data were generated from a cancer cell line with a large number of pre-existing mutations ( Abaan et al . , 2013 ) , as well as additional mutations that have assuredly arisen during passaging in the laboratory . These could conceivably include suppressor mutations functioning to restrain elevated mutation rates ( Morrison and Sugino , 1994; Herr et al . , 2011a; Williams et al . , 2013 ) . Importantly , no additional mutations in POLE were sequenced , suggesting that viability of this cell line is not due to an acquired mutation elsewhere in POLE acting to suppress the mutation rate , as occurs frequently in yeast ( Herr et al . , 2011a; Williams et al . , 2013; Herr et al . , 2011b; Dennis et al . , 2017 ) . While we cannot formally exclude the possibility that a de novo mutation in another gene acted to suppress the mutation rate in trans , no obvious candidates were identified . An additional reason that our mutation rates may underestimate the true mutation rate is that mutations that arise in the last several rounds of replication and those that fall below 5% allele frequency would not meet the threshold for scoring as a true SNV . The genome data was generated using an instrument with high accuracy ( <1% error rate ) and variants were called using an established algorithm , however there are indeed a small number of areas in the genome that are inaccessible – either due to gaps in the reference assembly , or excessive numbers of repeats that prevent proper alignment . Experiments using single-cell sequencing could address these issues , ideally by selecting single cells , expanding subclones and then measuring mutations at higher stringency values than used here . These rates are also similar to the per base pair mutation rates in haploid yeast with complete Pol ε exonuclease deficiency and disrupted MMR ( Kennedy et al . , 2015 ) . This similarity is striking considering our measurements were made in a heterozygous diploid human cell line . A key finding from the yeast study was that individual cell mutation rates could vary by an order of magnitude . We are currently unable to measure mutation rates in individual cells , but this remains a critical issue to address in future studies . The unique mutation spectrum seen in POLE tumors was recapitulated in our gene-targeted cell lines , with one notable exception . In tumors , many C→A transversions occur in a highly specific triplet sequence context , TCT , which we also see in the cell lines , though not to the same proportion as in the tumor genomes . Interestingly , this particular mutation is also enriched in yeast with the P286R equivalent allele ( Barbari and Shcherbakova , 2017 ) . We also observe increased T→G transversions in TTT ( and to a lesser extent ATT and CTT ) context , similar to Pol ε tumors . Because of the limited number of mutational target sites we cannot at this time draw conclusions as to Pol ε strand usage during replication . Experiments designed to assess strand bias in these errors are currently underway . What is notable , however , is the lack of TCG→TTG transitions in our dataset . This is the second most frequent Pol ε-specific mutation in the TCGA database . TCG→TTG transitions were also not found elevated in the Pol ε bMMRD brain tumor mutation spectrum . This difference may reflect interesting , but as-yet undefined tissue differences . Another possible explanation for these differences is that the Pol ε mutants found in tumors are somehow intrinsically different biochemically from the double alanine substitution mutant used in the current study . Depending on the reporter gene used , the monoallelic Pol ε-P286R mutant is a 2 . 3- to 12-fold stronger mutator than the pol2-4 mutant ( equivalent to the human Pol ε D275A/E277A studied here ) when measured in a diploid yeast strain ( Kane and Shcherbakova , 2014 ) . However , a number of direct biochemical comparisons of activity and fidelity ( Figure 1C , ( Shinbrot et al . , 2014; Shlien et al . , 2015 ) and unpublished observations ) between several cancer mutant constructs and the D275A/E277A construct have not yet shown any significant differences that could account for this . Certain DNA Pol mutants , including some found in human tumors , can cause increased mutagenesis by inducing expansions of normal dNTP pools in yeast and human cells ( Dennis et al . , 2017; Mertz et al . , 2015; Williams et al . , 2015 ) . Interestingly , the pol2-4 allele has no effect on dNTP pools in yeast , suggesting a possible explanation for possible allelic differences with functional MMR . In heterozygous Pol εwt/exo- cells with functional mismatch repair , mutation rates were suppressed to the levels seen in cells with wild type Pol ε . These rates would be insufficient to give rise to ultrahypermutated tumors in a matter of months . In addition , there is no explosive accumulation of triplet context-specific mutations in the MMR-proficient Pol εwt/exo- cells that is seen in these tumors . Given that the HCT-116 cells used in these studies are mutators themselves , it is possible that pre-existing deficiencies in other DNA repair or replication proteins could contribute to the observed mutagenesis . While direct contribution is unlikely given the absence of POLE mutation spectrum in the wild type Pol ε cells , cooperation with exonuclease-deficient Pol ε remains a formal possibility . To address this we used gene ontology to identify 58 DNA repair and replication proteins mutated in our HCT-116 cells , including 38 non-synonymous and 20 indel mutations . While several interesting candidates with known links to mutagenesis were identified , all have been shown by other groups to be expressed in this cell line and each , when tested , is functional ( e . g . ATM , SETD2 , Pol η , Pol ζ [Bhat et al . , 2013; Hahn et al . , 2011; Nicolay et al . , 2012; Zhou et al . , 2013; Zhu et al . , 2009] ) . No mutation that arose during the population doubling experiments showed any obvious link to mutagenesis . Our results support a model in which simple heterozygous loss of two Pol ε exonuclease metal chelating residues on a single allele of POLE is insufficient to drive Pol ε ultramutational specificity . Additional factors are likely required to help drive the ultramutated phenotype observed in POLE tumors , including suppression of mismatch repair , discussed below . In bMMRD , the complete lack of mismatch repair prior to Pol ε mutation leads to the moderate accumulation of Pol ε-independent replication errors ( Figure 5 ) . Mutation rates then increase dramatically upon loss of proofreading in one allele , with the Pol ε error signature representing a smaller fraction of the total errors , which is seen in these tumors ( Shlien et al . , 2015 ) . Our results suggest that Pol ε mutations in somatic tumors can occur first and early , but later suppression of MMR would then accelerate overall mutation rates to that seen in the ultramutated tumors , while the signature mutation proportion remains high ( Cancer Genome Atlas Network , 2012; Kandoth et al . , 2013 ) . Analysis of the mutational status of all mismatch repair genes in Pol ε tumors sequenced by TCGA supports the model of mismatch repair loss dramatically accelerating the acquisition of Pol ε-specific mutations . 85% ( 22/26 ) of the TCGA Pol ε tumors also have a mutation in at least one mismatch repair gene , most of which ( 18/22 ) harbor at least one nonsense mutation , which are more likely to be inactivating mutations ( Figure 5—figure supplement 1 ) . This predicts that at least some tumors would show evidence of MSI . In the original TCGA studies , several POLE tumors were actually first classified as MSI ( three as MSI-H; five as MSI-L ) ( Cancer Genome Atlas Network , 2012; Kandoth et al . , 2013 ) . Analysis of sequencing reads from 46 homonucleotide runs in the POLE endometrial tumors showed no evidence of instability , so the POLE tumors were then reclassified as MSS ( Shinbrot et al . , 2014 ) . However , the initial TCGA studies used both homo- and di-nucleotide loci to score MSI , raising the possibility that a subset of POLE tumors have a microsatellite instability defect at repeats more complex than homonucleotides . Indeed , the repeat unit size , the number of repeats and the repeat sequence composition are known to have very strong influences on the variability of microsatellite mutagenesis ( Shah et al . , 2010 ) . Curiously , however , most ( 15/18 ) of the MMR gene nonsense mutations are the result of TCT→TAT transversions , raising the possibility that Pol ε mutation occurs first and possibly even promotes subsequent mutational inactivation of MMR . Of all the Pol ε mutant colorectal and endometrial tumors sequenced in the TCGA studies , 15% ( 4/26 ) lacked a mutation in any mismatch repair gene and also showed no evidence of MLH1 promoter hypermethylation , demonstrating that the ultramutated phenotype can arise when mismatch repair is intact at both the genetic and epigenetic level . An alternative possibility is that mismatch repair activity is suppressed at some point during POLE tumor development . In this scenario , mutations introduced by the mutant Pol ε could accumulate slowly even in the presence of genotypically and epigenetically wild type mismatch repair . A number of conditions have been shown to transiently and reversibly lower mismatch repair protein levels and inhibit mismatch repair activity , including hypoxia , oxidative damage , inflammation , reduced pH , exposure to adriamycin or cadmium and treatment with mutagenic dNTP analogs ( Banerjee and Flores-Rozas , 2005; Francia et al . , 2005; Larson and Drummond , 2001; Mihaylova et al . , 2003; Chang et al . , 2002; Hile et al . , 2013; Iwaizumi et al . , 2013; Lu et al . , 2014; Negishi et al . , 2002 ) . The variable nature and duration of such a suppression event would be expected to result in a complex effect on microsatellite instability . Perhaps even more intriguingly , transient mismatch repair suppression has been seen in the context of proofreading-deficiency in E . coli ( Fijalkowska and Schaaper , 1996; Schaaper and Radman , 1989 ) . While replication errors made by the proofreading-deficient allele tested here were clearly insufficient to suppress MMR , it is possible that the nature and rate of errors made by cancer-associated alleles might be sufficient to saturate and overwhelm MMR pathways . Our results support the idea that loss of a single Pol ε proofreading allele is sufficient to drive a subset of the observed clinical characteristics of Pol ε tumors , provided mismatch repair is compromised in some way . These observations further support the idea that in the presence of fully functional MMR the appearance of the ultrahypermutated mutation signature may be more directly related to some as yet uncharacterized additional defect in the mutant polymerase ( Barbari and Shcherbakova , 2017 ) . These ideas are not mutually exclusive of one another . Given the recent success of immune checkpoint therapies in treating tumors with high mutation burden ( Shlien et al . , 2015; Bouffet et al . , 2016; Hodi et al . , 2010; Le et al . , 2015; Santin et al . , 2016 ) , it is of great interest to understand the mechanisms that result in ultrahypermutated tumors harboring DNA polymerase mutations .
Trypsin-EDTA was from Life Technologies and Geneticin was from Invitrogen . Antibodies against Mlh1 ( mouse α-human Mlh1 , G168-728 ) and β-actin ( mouse α-human beta-actin , A1978 ) were from Pharmingen and Sigma , respectively . The human colorectal cancer cell line HCT-116 ( a kind gift from Dr . Prescott Deininger ) was grown in HyClone MEM/EBSS ( Thermo Scientific ) supplemented with 10% fetal bovine serum ( Atlanta Biologicals ) , 1% sodium pyruvate ( Invitrogen ) and 1% MEM-NEAA ( Invitrogen ) . The HCT-116 cells used in this study were validated via analysis of genome-wide mutation signature , microsatellite instability and biomarker . HCT-116 cells lack Mlh1 resulting in a well-characterized MSI phenotype ( Lynch et al . , 1993; Parsons et al . , 1993; Boland and Goel , 2010 ) . They further have a unique mutational spectrum that can be evaluated via next-generation sequencing ( Abaan et al . , 2013 ) . Western blot analyses ( Figure 3A ) showed a lack of Mlh1 protein . The mutation spectrum from our whole-exome sequencing of HCT-116 cells ( Figure 2A and Figure 2—figure supplement 1 ) is identical with that reported by Abaan ( Abaan et al . , 2013 ) . Lastly , we performed microsatellite stability analysis in our HCT116 cells at five mononucleotide homopolymeric run loci ( NR27 , NR21 , NR24 , BAT25 , BAT26 ) using capillary electrophoresis , which showed instability at these loci providing a phenotypic readout consistent with the lack of Mlh1 expression in our cells ( data not shown ) . The HCT-116 cell line is also not in the 488 commonly misidentified cell lines from the most recent ICLAS database ( Version 8 . 0 ) and tested negative for mycoplasma . In order to target the proofreading inactivating mutations to the POLE locus in vivo , we used rAAV with a synthetic exon promoter trap ( Rago et al . , 2007 ) . A 1045 bp fragment containing POLE exons 7 and 8 along with intron 7 ( termed HA1 ) was PCR amplified from HCT-116 genomic DNA using primers designed to add unique NotI and SacI sites to the 5' and 3' ends , respectively . A 1057 bp fragment containing exons 9 , 10 and 11 along with introns 9 and 10 ( termed HA2 ) was PCR amplified from HCT-116 genomic DNA using primers designed to add unique EcoRI and NotI sites to the 5' and 3' ends , respectively . Both HA1 and HA2 were first cloned into pCR-TOPO and sequence verified . The catalytic exonuclease DIE residues located in HA2 ( exon 9 ) were changed to AIA using site-directed mutagenesis and sequence verified . The Pol ε rAAV shuttle vector was assembled by four-way ligation using the restriction enzyme-digested gene-specific HA1 and HA2 fragments , along with the SEPT/loxP cassette digested with NotI-EcoRI and the ITR-containing pAAV shuttle vector digested with NotI ( SEPT/loxP cassette and pAAV shuttle vectors were kind gifts of Dr . Fred Bunz , Johns Hopkins University ) . The Exo-targeting vector was used to package high-titer ( 1 . 6 × 106 PFU/ml ) recombinant adeno-associated virus into AAV2 serotype capsids . Cells were grown in 100 mm dishes and infected with rAAV when ~75–80% confluent . At the time of infection , cells were washed with 1x Hanks buffered saline solution ( Invitrogen ) before adding 3 ml of media containing 75 μl of a 1:250 dilution of rAAV lysate . 3 hr after infection an additional 6 ml of media was added to plates and allowed to incubate at 37°C for 48 hr . After 48 hr , media was changed and Geneticin was added to a final concentration of 400 μg/ml . Plates were then incubated under selection for an additional 14 days . At the end of the selection period , colonies from plates were isolated using glass cloning rings and 0 . 05% trypsin ( Invitrogen ) was used to transfer colonies to 6-well plates for subsequent expansion . Genomic DNA was extracted from expanded clones using DNeasy Blood and Tissue kit ( QIAgen ) according to the manufacturer’s protocol and eluted in 100 μl of elution buffer . Locus-specific integration was assessed by PCR using a primer that annealed outside the homology region and another that annealed within the neo cassette . To remove the SEPT cassette from correctly targeted clones , cells were infected in a 25 cm2 flask with adenovirus that expresses the Cre recombinase ( 1 . 0 × 106 PFU/ml , Vector Biolabs , Philadelphia , PA ) . Cells were plated at a limiting dilution in nonselective medium 24 hr after infection . 12 days after infection , single cell colonies were plated in duplicate and geneticin was added to one set of wells at a final concentration of 400 μg/ml to test for sensitivity . During this time , genomic DNA was extracted as previously described and screened using primers that annealed across both homology arms . PCR products were digested with SacI to distinguish between the wild type and recombinant locus . Genomic DNA was harvested from the knock-in cell lines using the DNeasy Blood and Tissue Kit ( Qiagen ) , and double digested with SacI and SalI . Hoechst fluorimetry was used to determine the concentration of DNA samples for accurate loading of samples . 4 μg of each sample was run on a 0 . 8% agarose gel in TBE . DNA was transferred to Hybond N + membrane ( Amersham ) , blotted with a probe to HA2 at 65°C overnight , and washed at 65°C . To make the probe , a 300 bp sequence was amplified from the HA2-pCR-TOPO clone using the primers: 5ʹ-GCATCTGCCCCACTGTTAGT-3ʹ and 5ʹ-CTCCCTGTTGGTGATGAGGT-3ʹ . The PCR product was labeled using the Prime-It II Random Primer Labeling Kit ( Agilent ) and α-32P-dCTP ( Perkin Elmer ) . Membrane was blocked in Denhardt’s pre-hybridization buffer [6x SSC , 0 . 5% SDS , 0 . 1% Ficoll 70 , 0 . 1% Ficoll 400 , 0 . 2% PVP , and 0 . 2%] at 65°C for 1 hr . The probe was added to hybridization buffer [6x SSC , 0 . 5% SDS , and 10% Dextran Sulfate] and incubated overnight at 65°C . To wash off excess probe , the blot was washed for 2 × 15 min washes in wash 1 [10x SSC , 0 . 5% SDS] , 2 × 15 min washes in wash 2 [1x SSC , 1% SDS] , and 2 × 30 min washes in wash 3 [0 . 1x SSC , 1% SDS] . The gel was exposed to a PhosphorImage screen and scanned on a Typhoon Imager . An expression vector encoding residues 1–1189 of the catalytic subunit of human Pol ε containing the D275A/E277A substitution was prepared as described ( Korona et al . , 2011 ) . Briefly , the human Pol ε was coexpressed in autoinduction medium with pRK603 , which allows coexpression of TEV protease , at 25°C until the culture was saturated . Peak fractions from the HisTrap column were pooled , dialyzed into 50 mM HEPES , pH 7 . 5 , 1 mM DTT , 5% glycerol and bound to SP sepharose . Bound protein was eluted with a 0–1 M with NaCl gradient . Peak fractions were pooled , dialyzed into 50 mM Tris , pH 7 . 5 , 1 mM DTT , 5% glycerol , 100 mM NaCl and bound to Q Sepharose . Bound protein was eluted with a 100 mM–M M NaCl gradient . Peak fractions were pooled , concentrated and passed through a pre-equilibrated Superdex200 size exclusion column . Fractions containing the purified 140 kDa protein were pooled , dialyzed into 50 mM Tris , pH 8 . 0 , 1 mM DTT , 5% glycerol and aliquots were frozen and stored at −80°C . We previously reported that the lacZ forward mutation assay template lacks sites at which TCT→TAT transversions are phenotypically detectable ( Shlien et al . , 2015 ) . To overcome this limitation we previously made a reversion substrate that reports only this mutation by using site-directed mutagenesis to change A-11 to C-11 . Double-stranded M13mp2 DNA containing the TC-11T sequence was used as a substrate in reactions containing 0 . 15 nM DNA , 50 mM Tris-Cl , pH 7 . 4 , 8 mM MgCl2 , 2 mM DTT , 100 μg/ml BSA , 10% glycerol , 250 μM dNTPs and 1 . 5 nM Pol ε at 37°C . Completely filled product was transfected into Escherichia coli cells , which were used to determine the frequency of dark blue revertant plaques that occurred as a result of TCT→TAT transversions arising during DNA synthesis . In this assay , accurate DNA synthesis yields colorless plaques . Error rates were calculated according to the following equation: error rate ( per nucleotide synthesized ) = ( ( number of mutants of a particular class ) × ( mutant frequency ) ) / ( ( number of mutations sequenced ) × ( 0 . 6 ) × ( number of detectable sites ) ) . Mlh1 ORF was PCR amplified using the pCMV-XL5-Mlh1 vector ( kindly provided by Victoria Belancio , Tulane University ) , forward and reverse primers ( fwd 5'-TCGACTCGAGTCCACCATGTCGTTCGTGGCAGG-3'; rev 5'-TCGAGGATCCGTTACTTAACACCTCTCAAAGAC-3' ) and Q5 DNA polymerase ( NEB ) . After gel purification , dA was added to the 3' ends with Taq and the Mlh1 ORF was cloned into pLenti6 . 3/V5-TOPO ( Invitrogen ) . Mlh1 was found to have a common I219V SNP that does not affect Mlh1 function ( Plotz et al . , 2008 ) . Mlh1 Lentiviral particles were made using the ViraPower Lentiral Expression System ( Invitrogen ) . Briefly , 293FT cells were transfected with pLenti6 . 3/V5-TOPO-Mlh1 and a mixture of plasmids encoding lentiviral packaging factors . Viral supernatant was harvested 48 hr after transfection , filter sterilized and stored in aliquots at −80°C . After titering , HCT-116 cells were transduced with Mlh1 lentivirus at MOI of 1 . 0 . Cells were selected for 1 week in 10 μg/ml blasticidin . Blasticidin-resistant clones were identified and cells were harvested , lysed and probed by Western blot ( mouse α-human Mlh1 , G168-728 , Pharmingen ) to confirm Mlh1 expression . Prior to mutation rate measurements , preexisting HPRT1 mutants were eliminated from cell populations by incubating cells in HAT medium ( 1x Hypoxanthine-Aminopterin-Thymidine ) for five passages . For each cell line analyzed , 500 cells were seeded and grown to confluence in 12 wells across two 6-well plates . Cells from one well were harvested and counted to estimate cell number in the remaining 11 wells . For mutation rate measurement , 500 cells from each of the remaining eleven wells were seeded per dish in 3 × 100 mm dishes in media lacking 6-TG to be used to measure plating efficiency . At the same time , 5 × 105 cells from each of the remaining eleven wells were plated in 5 × 100 mm dishes in media containing 6-TG . After 7 days , colonies on the plating efficiency wells were stained with crystal violet and counted . After 12–14 days , the 6-TG resistant colonies were also stained with crystal violet and counted . Mutation rate was calculated using the Ma-Sandri-Sarkar Maximum Likelihood Estimator ( MSS-MLE ) method ( Rosche and Foster , 2000 ) . For mutant frequency measurement , 500 cells per clone were seeded in duplicate in 6-well plates in media lacking 6-TG and allowed to grow for 5–7 days to determine plating efficiency . The remaining wells were seeded with 5 × 104 cells in media containing 6-TG and allowed to grow for 12–14 days . After the indicated time , colonies were stained with crystal violet and counted . Mutant frequency was calculated by the following equation: ( # 6-TG resistant colonies ) / ( [ ( # colonies scoredPE ) / ( # cells seededPE ) ] x ( # cells seeded6-TG ) ) . PE refers to plating efficiency . Colonies were defined as ≥50 cells . HCT116 and HCT116 + Mlh1 cells were seeded into T75 flasks and grown at 37°C/5% CO2 until 80% confluency was reached . Cells were counted using the Countess Automated Cell Counter ( Invitrogen ) and 1 × 106 cells were seeded into new T75 flasks and incubated until 80% confluency was reached . The above protocol was repeated at regular intervals ( 3–4 days ) and population doubling ( PDL ) numbers calculated using the following equation: PDL = [ln ( Nt ) -ln ( N0*PE ) ]/ln2 . Nt = Number of viable cells counted after passage; N0 = Number of cells seeded prior to passage; PE = plating efficiency . At PDL ~ 6 , 44 and 69 cells were trypsinized and counted . For mutant frequency measurement , 300 cells were seeded into each of 3 × 100 mm dishes in media lacking 6-TG to be used to measure plating efficiency . Concurrently , 2 × 105 cells were seeded into each of 10 × 100 mm dishes in media supplemented with 6-TG to a final concentration of 5 μg/mL . After 7 days , colonies on the plating efficiency dishes were stained with crystal violet and counted . After 12–14 days , 6-TG resistant colonies were isolated using glass cloning rings and 0 . 05% trypsin and transferred into 24-well plates for expansion and RNA isolation . Additionally , at the above PDLs an aliquot of cells were harvested , lysed and probed by Western blot ( mouse α-human Mlh1 , G168-15 , Abcam ) to confirm maintenance of Mlh1 expression . Genomic per base pair mutation rates ( μBS ) were calculated using the method of Drake ( Drake , 1991 ) with modifications as applied in Lynch ( Lynch , 2010 ) . The equation used was: μBS = ( μL • fT • fBS ) / ( L • fL • [x ( nm + nn ) /nn] ) , where μL is the measured mutation rate at the HPRT1 reporter gene , fT is the fraction of mutants found after sequencing , fBS is the fraction of mutations due to base pair substitutions , L is the length ( in nt ) of the reporter gene , fL is the fraction of HPRT1 that gives rise to detectable mutations , x is the fraction of mutations that would give rise to chain terminator mutations , nm is the observed number of missense mutations and nn is the observed number of nonsense mutations . We used 126 HPRT1 mutations from three independent studies ( Bhattacharyya et al . , 1995; Glaab and Tindall , 1997; Ohzeki et al . , 1997 ) to calculate μBS . The values used were: fT = 1 . 0 , fBS = 79/126 = 0 . 627; L = 627 nt; fL = 1; x = 3/64 = 0 . 047; nm = 74; nn = 5 . The μL value for Pol ε mutant cell lines was determined empirically using fluctuation analysis . Total RNA was isolated using the Qiagen RNeasy kit ( Qiagen ) according to the manufacturer’s protocol . RT-PCR was performed with SuperScript III Reverse Transcriptase ( Invitrogen ) according to the manufacturer’s protocol using 1 μg of RNA as a template . Primer-specific cDNA was amplified for 32 cycles at an annealing temperature of 60°C using the following HPRT1 primers: 123 ( fwd ) CTTCCTCCTCCTGAGCAGTC and 1041 ( rev ) GCCCAAAGGGAACTGATAGTC . From the HPRT1 sequencing of 6-TG resistant colonies , one clone was found to have exon 2 completely deleted . Exon deletions in HPRT1 have been shown to be caused by splice site mutations ( Bhattacharyya et al . , 1995 ) . We therefore amplified exon 2 and its flanking region from genomic DNA prepared from the appropriate clone using the following primers: Forward: TTGTTTTCTTACATAATTCATTATCATACC; Reverse: TTACTTTGTTCTGGTCCCTACAGAG . Next generation sequencing was performed as per the published protocols . Whole genome sequencing ( WGS ) was performed on an Illumina HiSeq Xten instrument with libraries prepared using the manufacturer’s TruSeq Nano DNA Library Prep kit and sequenced to a depth of 36 . 1x . For exome sequencing , DNA was enriched using Agilent SureSelect Human Exome Library Preparation V5 kit , then sequenced to a depth of 101 . 38x ( 96 . 61x-108 . 19x ) . All samples were processed from raw reads ( FASTQ files ) from paired end libraries . The reads were aligned to the human reference ( GRCh37 with decoy sequences ) using BWA-MEM v0 . 7 . 8 ( Li and Durbin , 2009 ) . Duplicate reads were identified and marked using Picard v1 . 108 ( https://broadinstitute . github . io/picard/ ) . The Genome Analysis Toolkit ( GATK ) v2 . 8 . 1 ( McKenna et al . , 2010 ) was used to locally realign reads to known indels and recalibrate base quality scores . Quality metrics were generated from the final BAM files to ensure high quality alignment . This includes: Limitations in the genome due to low-complexity regions and incomplete areas in the genome ( Li , 2014 ) prevent proper alignment resulting in sources of error . Somatic point mutations between the tumour and matched normal were identified using MuTect v1 . 1 . 4 ( Cibulskis et al . , 2013 ) . In addition , we used MuTect v1 . 1 . 4 in single sample mode to detect all mutations in each sample . All mutations were annotated using ANNOVAR v20130823 ( Wang et al . , 2010 ) . Subsequent filtering was performed to reduce potential false positives and allow only high confidence mutations in the dataset using a custom R package ( ShlienLab . Core . SNV v0 . 09 ) . Mutations were retained if they met the following criteria: To investigate the quality of somatic mutations , we also identified key metrics including: DNA sequencing data from this study have been submitted to the NCBI Gene Expression Omnibus ( GEO; http://www . ncbi . nlm . nih . gov/geo/ ) under accession number PRJNA327240 . | New cells are made when an existing cell divides in two . Each time a cell divides , it duplicates its DNA so that each new cell inherits a complete copy . Molecular machines called DNA polymerases make these DNA copies . The main DNA polymerases , known as delta and epsilon , can “proofread” the new DNA , which ensures that the genetic information stored in the DNA is correctly copied . Cells also use another system , called mismatch repair , to catch any errors that get missed by the polymerases . Cancer cells contain many mutations in genes that regulate the growth and production of new cells , which is why cancers grow out of control and produce tumors . Research shows that many cancer cells with high numbers of mutations have lost their proofreading ability . Yet it is not clear if the loss of proofreading is enough to cause cancers , or if other systems , such as mismatch repair , must also be defective . Hodel , de Borja , Henninger et al . examined human cells grown in the laboratory to understand the importance of proofreading in cancer . It turns out that even the partial loss of polymerase epsilon proofreading could lead to distinctive mutations . Yet , these mutations were repaired by mismatch repair , so they actually are only found in cells when mismatch repair is also defective . This result demonstrates that the lack of proofreading is not enough to cause a large number of mutations . These cancers only happen when other systems are damaged too . These new findings add to the current understanding of the origins of mutations in cancers and how mutations accumulate over time . It should lead scientists to further investigate the patterns of mutations that happen in the absence of proofreading . It may also enhance our knowledge of proofreading-deficient cancers . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cancer",
"biology"
] | 2018 | Explosive mutation accumulation triggered by heterozygous human Pol ε proofreading-deficiency is driven by suppression of mismatch repair |
Pain is a prevalent symptom of Parkinson’s disease , and is effectively treated by deep brain stimulation of the subthalamic nucleus ( STN ) . However , the link between pain and the STN remains unclear . In the present work , using in vivo electrophysiology in rats , we report that STN neurons exhibit complex tonic and phasic responses to noxious stimuli . We also show that nociception is altered following lesions of the STN , and characterize the role of the superior colliculus and the parabrachial nucleus in the transmission of nociceptive information to the STN , physiologically from both structures and anatomically in the case of the parabrachial nucleus . We show that STN nociceptive responses are abnormal in a rat model of PD , suggesting their dependence on the integrity of the nigrostriatal dopaminergic system . The STN-linked nociceptive network that we reveal is likely to be of considerable clinical importance in neurological diseases involving a dysfunction of the basal ganglia .
Pain is highly prevalent in Parkinson’s disease ( PD ) and includes primary symptoms assumed to originate from a dysfunction of the central nervous system . Patients describe bizarre and unexplained painful sensations such as burning , stabbing , aching , itching or tingling sensations , predominating on the more affected side ( Schestatsky et al . , 2007 ) . These symptoms are not directly related to the pain caused by the motor symptoms ( Ha and Jankovic , 2012 ) . Sensitivity to noxious stimulation is also increased in patients with PD , with or without pain symptoms ( Berardelli et al . , 2012; Brefel-Courbon et al . , 2013; Tinazzi et al . , 2008 ) , and their nociceptive threshold is altered ( Chudler and Dong , 1995; Conte et al . , 2013; Djaldetti et al . , 2004 ) . Although it is well known that PD affects the basal ganglia , there is to date no clear description of a link between the basal ganglia and the cerebral network involved in pain . Interestingly , deep brain stimulation of the subthalamic nucleus ( STN-DBS ) in PD , a valuable and effective therapeutic technique for motor symptoms ( Krack et al . , 2003; Limousin et al . , 1998 ) , has also been shown to reduce pain ( Cury et al . , 2014; Hanagasi et al . , 2011; Kim et al . , 2008; Klingelhoefer et al . , 2014 ) . By contrast , the effects of dopamine replacement therapy on the pain symptoms in PD are controversial , with some investigations reporting an improvement , no effect or an aggravation of pain symptoms or nociceptive thresholds ( Conte et al . , 2013; Dellapina et al . , 2011 ) . These variable results of dopamine replacement therapy indicate a role for other systems in the pain symptoms observed in PD . Importantly in the present context , it has been demonstrated that pain relief following STN-DBS is superior to that following dopaminergic treatment , further positioning STN as a crucial structure in pain symptoms in PD and their relief ( Sürücü et al . , 2013 ) . The mechanism by which STN-DBS improves pain in PD patients remains unclear , which raises a fundamental question about the link between the subthalamic nucleus ( STN ) and pain . Preliminary evidence suggests that noxious stimulation can modulate background activity in the STN , at least in the Parkinsonian brain ( Belasen et al . , 2016; Heise et al . , 2008 ) . As a consequence , the STN could be linked to a nociceptive network involved in the perception of noxious stimuli , although this has yet to be explored fully . Despite the classical description of a sensory territory in the STN ( Alexander et al . , 1986 ) and the functional impact that STN sensory responses could have on the basal ganglia ( Baunez et al . , 2011 ) , there is a paucity of information in the literature regarding the type of sensory stimuli that activate this structure and the afferent sensory source ( s ) ( Coizet et al . , 2009; Hammond et al . , 1978; Matsumura et al . , 1992 ) . Two major subcortical central targets for ascending nociceptive information from the spinal cord are potential relays for nociceptive information to the STN: the superior colliculus ( SC ) and the parabrachial nucleus ( PBN ) ( Hylden et al . , 1989; Craig , 1995; Klop et al . , 2005; McHaffie et al . , 1989 ) . We have shown that the SC , a highly conserved but evolutionarily ancient subcortical multi-sensory structure , directly projects to the STN and is the primary , if not exclusive , source of visual input into the STN ( Coizet et al . , 2009; Tokuno et al . , 1994 ) . This projection shows that subcortical hyper-direct pathways exist between brainstem sensorimotor structures and the STN which predate , from an evolutionary perspective , those from the cortex — thus reinforcing the position of the STN as a critical input structure processing short latency visual signals in the basal ganglia ( Baunez et al . , 2011 ) . The substantia nigra pars compacta ( SNc ) is also on the input side of the basal ganglia and we have shown that SNc receives nociceptive-related afferents from the PBN ( Coizet et al . , 2010 ) , raising the possibility that the STN may also receive such inputs . It is interesting to note that the STN is well known to project heavily to basal ganglia output structures such as the substantia nigra pars reticulata ( SNr ) ( Alexander et al . , 1986; Gurney et al . , 2001 ) , which in turn projects to the SC and PBN ( Schneider , 1986; Deniau and Chevalier , 1992 ) , linking the STN , SNr and SC/PBN anatomically . With the STN in a position to modulate a nociceptive network involving the SC/PBN , the elevated activity in the STN in Parkinsonism ( Bergman et al . , 1994; Albin et al . , 1995 ) could underlie some of the unexplained pain symptoms in this disease , with STN-DBS acting ( at least in part ) locally to achieve its analgesic effects . Nociceptive processing in the STN would also be consistent with the nucleus’s role as part of the brain’s interrupt circuitry ( Jahanshahi et al . , 2015 ) , terminating behaviors that achieve negative outcomes , of which pain is a clear example . Therefore , the main objective of the present work was to characterize the link between STN and nociception , answering the following questions: We present convergent evidence that the STN is functionally linked to a nociceptive network and that STN nociceptive responses are affected in Parkinsonism . The objectives above are summarized in Figure 1 .
In a classical rat model of PD induced by an injection of 6-hydroxydopamine ( 6-OHDA ) , a neurotoxin thattargets dopaminergic neurons ( DA ) , in the SNc , we tested whether STN nociceptive responses were dysfunctional . TH immunohistochemistry was used to assess the extent of the dopamine denervation induced by 6-OHDA in the DA-lesioned rats . TH-labeled neurons on the lesioned side were reduced to an average of 6 . 13 ± 0 . 71% ( mean ± SEM ) of those on the unlesioned side , with the remaining neurons located in the medial part of the SNc . The reduced number of dopaminergic cells led to an average decrease of 65 . 22 ± 2 . 08% in the dopaminergic innervation of the striatum . Totals of 34 and 43 cells were recorded in the control and DA-lesioned rats , respectively . As expected in a model of PD , STN cells had a significantly higher firing rate in DA-lesioned rats compared to controls ( mean ± SEM: control = 8 . 00 ± 1 . 04 ms; DA-lesioned = 12 . 96 ± 2 . 02 ms; p<0 . 05 , Figure 3A ) . STN responses to nociceptive stimulation were abnormal in the DA-lesioned group . Analysis revealed that STN cells in PD rats exhibited significantly longer responses ( mean ± SEM: control = 84 . 39 ± 16 . 75 ms; DA-lesioned = 175 . 38 ± 39 . 85 ms , p<0 . 05 ) with a greater amplitude ( mean ± SEM: control = 27 . 81 ± 3 . 58 ms; DA-lesioned = 40 . 88 ± 5 . 91 ms; p<0 . 05 , Figure 3C ) and magnitude ( mean ± SEM: control = 8 . 13 ± 1 . 38 ms; DA-lesioned = 13 . 11 ± 2 . 47 ms;p<0 . 05 ) compared to the sham control animals . The proportion of cells exhibiting the three levels of STN baseline firing rate upon introduction of the stimulation ( up , down or no change ) was not altered in the PD rat groups ( χ2 = 0 . 32; p=0 . 85 ) . Therefore , STN phasic nociceptive responses in PD rats were exacerbated , while the tonic modulation of the firing rate was preserved .
The present study demonstrates for the first time that a large majority of STN neurons exhibit various mono- or multi-phasic responses to noxious stimulation , consistent with the hypothesis that one of the functions of the STN is to interrupt behavior when appropriate ( Jahanshahi et al . , 2015 ) ; in this case , to select a more appropriate action to try to relieve the noxious sensation . STN nociceptive responses mainly had a short latency ( ~20–40 ms ) and could be recorded all over the structure . We have shown that most of the responsive cells are nociceptive-specific , as only few of them also respond to non-noxious somatosensory stimulation . In addition , we found that we could differentiate three types of STN neurons , which showed an increase , a decrease or no change , respectively , in their baseline firing rate upon introduction of the noxious stimulation . The spontaneous firing rate of STN ‘down’ cells was significantly higher than that of the two other groups , suggesting the possibility of a separate group of cells . Furthermore , we have shown that STN responses to nociceptive stimuli were abnormal in a rat model of PD , suggesting that the nociceptive responses recorded in STN depend on the integrity of the nigrostriatal DA system . When determining the afferent source of nociception-related influence on STN activity , we have revealed a crucial role for two brainstem structures , the PBN and the SC , by demonstrating the effects of their inactivation on nociceptive responses in the STN and also by highlighting the existence of an anatomical direct pathway from the PBN to the STN . This parabrachio-subthalamic projection represents a second example of a subcortical hyperdirect pathway to the STN from a sensori-motor structure , in addition to the tecto-subthalamic pathway described previously ( Coizet et al . , 2009; Tokuno et al . , 1994 ) . Finally , we have shown that these anatomico-electrophysiological findings translate into a functional role for the STN in mediating nociception , in that nociceptive behavioral responses were affected by lesions of the STN . However , a note of caution is required since we used only male rats , and thus care should be taken in extrapolating our results to females . Using noxious electrical stimulation of the hindpaw allowed us to record precisely the timing of responses in our structures of interest with controlled parameters of stimulation . It is interesting to note that the majority of STN nociceptive responses had short latencies ( 76/79 ) and were monophasic ( 40/79 ) , a pattern of responses that is similar in proportion to the pattern of STN responses following visual stimulation ( Coizet et al . , 2009 ) but dissimilar to the pattern of STN responses following stimulation of the frontal ( Magill et al . , 2004 ) , sensorimotor ( Fujimoto and Kita , 1993 ) or motor cortex ( Kolomiets et al . , 2001 ) . The latter — in the majority of cases — are multi-phasic , with two excitatory phases ( equivalent to the present biphasic +/+ ) often separated by an inhibition ( equivalent to the present triphasic +/–/+ ) . It has been hypothesized previously ( Magill et al . , 2004; Kitai and Deniau , 1981 ) that the short latency excitation following cortical stimulation appears to be driven by a hyperdirect pathway to the STN , whereas the later phases of the response arise from polysynaptic interactions , which are manifested more slowly . The inhibition following the first excitation has been hypothesized to involve the reciprocally connected STN–globus pallidus ( GP ) network ( Fujimoto and Kita , 1993 ) . STN excitation may activate GP GABAergic neurons , which in return inhibit the STN . Overall , our data suggest that when the rat is subjected to noxious stimulation , the main pathway to the STN that is activated is a fast hyperdirect pathway , originating in part in the PBN . The fact that we only have a few cells showing an inhibitory second phase ( 10 biphasic and 5 triphasic , 15/79 ) indicates that STN nociceptive cells are rarely closely connected to the GP and lack GP-STN feedforward control . Functionally antagonistic STN neuronal subpopulations have been found in the STN . Specific GO and STOP cells have been described in PD patients performing a stop signal paradigm , during motor execution or response inhibition , respectively ( Benis et al . , 2016 ) . Sub-populations of STN cells have also been shown to code exclusively for reward magnitude ( 4% vs 32% sucrose [Lardeux et al . , 2009] ) , error-related activity ( ‘Oops neurons’ [Lardeux et al . , 2009] ) , reward value ( cocaine or sucrose [Lardeux et al . , 2013] ) and for positive and aversive reinforcers ( Breysse et al . , 2015 ) . A major finding of our work is that we were able to differentiate a subpopulation of STN neurons that had a higher spontaneous firing rate on the basis of the effect of noxious stimuli on general tonic activity . This finding is important as glutamatergic tone from the STN is likely to have a strong impact on the tonic level of activity in the basal ganglia network , especially in the output structures ( as we hypothesized that nociceptive-responding STN cells may not be densely connected to GP ) . The results from previous computational studies ( Gurney et al . , 2001 ) suggest that tonic control by the STN may adjust the general level of activity of the inhibitory GABAergic neurons of the basal ganglia output structures , which are known to project to the SC and PBN ( Schneider , 1986; Deniau and Chevalier , 1992 ) . This tonic STN control is hypothesized to be used by the basal ganglia to optimize selection of the most appropriate action . The identification of separate subpopulations of cells in the STN according to their spontaneous firing rate , and the orientation of the change of their firing rate following the occurrence of noxious stimulation , suggests that the tonic excitatory effects of the STN may not be uniform , although further work is required to elucidate the connectivity of the STN subpopulations . This mechanism is important in the context of PD in which STN activity is pathologically increased ( Bergman et al . , 1994; Albin et al . , 1995 ) , probably disrupting this control . This possibility is further supported by our results showing enhanced phasic nociceptive responses in a PD rat model , with an increase in the latency to make nocifensive responses in the hotplate test following lesions of the STN . As well as interfering with action selection , disrupted STN control in PD would probably have an impact on the SC and PBN and their role in sensory signal processing . In addition to demonstrating that STN neurons process nociceptive information , we also assessed whether two subcortical sensori-motor structures from the brain stem transmit nociceptive signals to the STN . Despite SC nociceptive responses having shorter latencies than those of STN neurons to the same stimulus , chemical suppression and lesions of SC had relatively minor effects on the responses of STN to noxious footshock . Lesions of the SC with acid ibotenic reduced the number of cells that do not respond to noxious stimulation , suggesting that the SC gates the pool of responding cells in the STN . Our previous work has demonstrated that the SC is a critical relay for short-latency visual input into DA neurons ( Dommett et al . , 2005 ) but not for short-latency nociceptive input ( Coizet et al . , 2006 ) transmitted by the PBN ( Coizet et al . , 2010 ) . The current results suggest the same organization when considering visual and nociceptive input into the STN . The SC is a crucial structure to transmit visual information while the PBN strongly contributes to the relay of nociceptive signals . PBN lesions significantly reduced the number of STN cells responding to noxious stimuli , sparing a group of cells with short latency short duration responses ( Figure 7C ) , which are possibly activated by nociceptive information relayed by the thalamus ( Dostrovsky , 2000a ) . This nociceptive network linked to the STN is the probable substrate underlying the successful analgesic effects of STN deep-brain stimulation . Kim et al . ( 2012 ) hypothesized that STN-DBS improves secondary pain symptoms in PD because this stimulation decreases the abnormally increased muscle tone in patients and may alleviate the primary nociception processing in the central nervous system . DBS effects are complex and despite the success of DBS in treating a variety of psychiatric and neurological disorders , the mechanisms underpinning its therapeutic efficacy remain unclear ( McIntyre et al . , 2004; Ashkan et al . , 2017 ) . DBS is hypothesized to induce a ‘functional lesion’ of the STN ( Follett , 2000 ) , via depolarization blockade and synaptic inhibition ( Beurrier et al . , 2001; Dostrovsky et al . , 2000b ) , which would lead to a suppression of the activity of STN neurons . We hypothesize that these mechanisms would reduce the pathologically increased firing rate in the STN in PD ( and thus the pain symptoms ) , as well as nociceptive responses . Our current work using anterograde tract-tracing neuroanatomy coupled with 3D reconstruction indicates that a small ascending bundle leaves the PBN and then splits into three massive projections traveling toward the SC , the thalamus and the SNc/STN . Some fibers from this last ascending pathway continue rostrally to terminate in the amygdala and the cortex . Comparison of the size of the bundle leaving the PBN and the size of the bundles traveling to their targets indicate that the number of labelled fibers clearly increase , suggesting that PBN axons have collaterals . This PBN-STN projection is possibly interconnected with other PBN efferents , such as the PBN–amygdala projection , which partly travels through the STN and has cells of origin that are of a similar type to those of the PBN cells projecting to STN ( Sarhan et al . , 2005 ) . With a DBS effect on axons and fibers ( Chiken and Nambu , 2014 ) , the characterization of this projection and network are important in the context of the effects of STN-DBS on pain symptoms . Overall , STN-DBS would not only impact STN and PBN nociceptive processing but would also modulate PBN-amygdala fibers and possibly other PBN efferents via the collaterals . The effect of STN-DBS would therefore impact many aspects of pain such as , for example , pain-related emotional reactions when activating the PBN-amygdaloid connection or neuroendocrine homeostatic regulation in response to pain by activating the PBN-hypothalamic pathway ( Gauriau and Bernard , 2002 ) . Further experiments are now needed to fully characterize the effect of STN-DBS on nociceptive processing in our rat models and how aspects of that network are modulated to achieve a DBS-related analgesic effect . Pain is a multifaceted experience that can be understood in terms of somatosensory , affective and cognitive dimensions . DBS therapies that are focused on a single facet of pain , originally targeting somatosensory networks or more recently targeting affective regions ( Schneider , 1986 ) . The STN is a small structure with functional territories such as the limbic , cognitive and sensory , in close proximity to each other . This would allow the potential modulation of different modalities of pain and , in the future , the best placements of DBS electrodes within those territories would have to be tested to maximize the analgesic effect . Finally , numerous non-motor symptoms can worsen or improve depending on the electrical stimulation parameters , as well as the location of the electrode ( Kim et al . , 2015 ) . The best parameters of stimulation for nociception would need to take into account the effect of those parameters on other symptoms of PD . Finally , non-neuropathic pain , recently recognized as a frequent and disabling symptom in PD , is a complaint affecting many patients who have numerous neurodegenerative disease such as Alzheimer’s disease and other dementias , motor neuron disease , Huntington’s disease , spinocerebellar ataxia and spinal muscular atrophy ( de Tommaso et al . , 2016 ) . Our findings on the involvement of the STN in nociceptive processing and its link to a nociceptive network open a new direction for research to explore a possible role of this structure in other pain syndromes , especially extra-pyramidal ones like Huntington’s disease , which is characterized by a dysfunction of the basal ganglia . It also opens up the possibility of developing therapeutic strategies using DBS . A variety of brain sites have been identified for chronic stimulation procedures to attenuate pain ( Davis et al . , 1998 ) . These targets include the thalamus , the periventricular gray nucleus , the cingulate cortex and the motor cortex ( Gorecki et al . , 1989; Davis et al . , 1998 ) . With the involvement of the STN in a nociceptive network as demonstrated in our work , the STN-DBS technique can thus be considered in the future as a new target for the treatment of pain not only in pharmaco-resistant patients suffering from previously described neurodegenerative disease , but also , for example , in those with chronic pain disease or pharmaco-resistant patients with certain forms of migraine that have been shown to activate the STN ( Schwedt et al . , 2014 ) .
Fourteen rats received a unilateral ibotenic acid lesion of the SC or the PBN . Each rat was anesthetized with isofluorane ( 5% for the induction and 1–2% for maintenance ) and placed in a stereotaxic instrument . A 30-gauge metal injector needle filled with ibotenic acid ( 20 µg/µl in phosphate buffered saline ) was introduced using the same coordinates as for the electrophysiological procedure . The injections in the PBN were made according to a previously published procedure by Reilly and Trifunovic ( 2000 , 2001 ) with electrophysiological guidance to improve the accuracy of the location of the lesion . The microinjections were made ( 0 . 5 µl/min ) in the SC ( 0 . 5–0 . 65 µl ) and the PBN ( 0 . 3–0 . 5 µl ) as for the muscimol injections ( see above ) . The cannula remained in situ for a further 10 min to minimize the spread of neurotoxin back along the track before the cannula was removed and the incision was closed . Twenty Long Evans rats were anesthetized with ketamine ( 100 mg/Kg , s . c . , Imalgène 1000 , Merial , Lyon , France ) and medetomidine ( 0 . 85 mg/Kg , s . c . , Domitor , Orion Pharma , Espoo , Finland ) . Rats were secured in Kopf stereotaxic apparatus . Then , a unilateral 30-gauge stainless-steel injector needle connected by Tygon tubing ( Saint Gobain performance plastics ) with a 10 µL Hamilton microsyringe ( Bonaduz , Switzerland ) fixed on a micropump ( CMA , Kista , Sweden ) was positioned into the STN . Coordinates for the aimed site were ( with tooth bar set at −3 . 3 mm ) : anteposterior −3 . 72 mm; lateral ±2 . 4 mm from bregma; dorsoventral −8 . 4 mm from skull ( Paxinos and Watson , 2005 ) . Rats received a bilateral injection of either ibotenic acid ( 9 . 4 µg/µL , AbCam Biochemical , Cambridge , UK; STN-lesioned group , n = 12 ) or vehicle solution ( phosphate buffer , 0 . 1M; Sham control group , n = 8 ) . The volume injected was 0 . 5 µL per side infused over 3 min . The injectors were left in place for 3 min to allow diffusion . At the end of surgery , medetomidine was reversed by 0 . 2 mg ( 4 . 28 mg/Kg , s . c . ) atipamezole ( Antisedan , Orion Pharma , Espoo , Finland ) . Three weeks after the surgery , all the animals were subjected to the hotplate test . Each rat was placed on a heated metal plate ( 53° ) surrounded by a transparent cylinder . The experimenter was constantly watching the rat’s behaviour during the test to measure the latency of the first sign of paw licking or jumping and to remove the animals from the apparatus quickly . The maximum time on the hot place was set to 30 s . The rat’s behaviour was also video recorded online on the computer for a second finer analysis . Rats were anesthetised with an intraperitoneal injection of a mixture of ketamine-xylazine ( 0 . 765/1 . 1 ml; 1 ml/kg , i . p . ) and placed in a stereotaxic frame with the skull level . All the microinjections were made via a sharpened 30G injection cannula connected with polyethylene tubing to a 10 µl Hamilton syringe driven by an infusion pump ( 0 . 5 µl/min ) . After the injection , the cannula was left in place for a further 5 min to allow diffusion . Animals were divided into two groups: i ) those with a total dopaminergic lesion ( n = 9 ) , in which 3 µl of 6-OHDA ( Sigma-Aldrich , 3 mg/ml in sterile 0 . 9% NaCl and 0 . 1% ascorbic acid ) was injected into the left SNc using the following stereotaxic coordinates: AP:+3 . 0 mm; ML:+2 . 1 mm and DV:+2 . 4 mm from interaural zero mm; and ii ) a control group with no injection of the toxin ( n = 9 ) . The extent of the DA denervation following the 6-OHDA injections in the SNc was determined using tyrosine hydroxylase ( TH ) immunohistochemistry . To reveal TH , the sections were washed and incubated in a blocking solution containing 0 . 1M PB with 0 . 3% of triton X-100 ( TX ) , 2 . 5% of Bovine Serum Albumin ( BSA ) and 5% normal horse serum ( NHS ) for 2 hr before being transferred overnight to a 0 . 1M PB-TX 0 . 3% with 1% BSA and 2% NHS containing the primary mouse monoclonal TH antibody , diluted 1:3 , 000 ( Chemicon , Hampshire , UK ) . The following day , sections were washed in 0 . 1M PB and incubated with the secondary antibody , biotinylated antimouse made in horse ( in a dilution of 1:1 , 000 in 0 . 1M PB-TX 0 . 3% with 2% NHS ) for 2 hr . Following further washes in 0 . 1M PB , the sections were exposed to the elite Vectastain ABC reagent ( Vector Laboratories , Burlingame , CA , USA ) diluted 1:100 in PB-TX 0 . 3% , for 2 hr . Again following washes in 0 . 1M PB , immunoreactivity was revealed by exposure to VIP ( Vector Laboratories ) for 2 min , which produced a purple reaction product . Sections were then mounted onto gelled slides , dehydrated through alcohols and cleared in xylene before being coverslipped with DPX . TH-immunolabelling of DA neurons and terminals was evaluated using a light microscope ( Nikon , Eclipse 80i , TRIBVN , Chatillon , France ) coupled to the ICS Framework computerized image analysis system ( TRIBVN , 2 . 9 . 2 version , Chatillon , France ) . For quantification , TH-labeled coronal sections of SNc ( AP −5 . 3 mm to −5 . 8 mm from Bregma ) and striatum ( AP 0 . 20 mm to −0 . 30 mm from bregma ) were digitized using a Pike F-421C camera ( ALLIED Vision Technologies Stradtroda , Germany ) . Optical densities ( OD ) were measured for the denervated and non-denervated territories of the lesioned animals for each section and were compared to those in the homologous regions of the sham-operated animals . The statistical reliability of differences between response latencies for the SC/PBN and STN , and comparisons of response duration , amplitude and magnitude before and after SC/PBN injections of muscimol was made using parametric ( ANOVA , t-test ) or non-parametric ( Wilcoxon , Mann-Whitney ) statistical tests according to the normality of the data . STN baseline firing rate change before and after the noxious stimulation was assessed during the 500 ms before the sham and noxious stimulations . The data were imported in MATLAB , bined and compared using a Wilcoxon test . STN firing pattern was also assessed using MATLAB according to the methodology developed by Piallat et al . ( 2011 ) . Neurons were classified as irregular , regular or bursting according to the interspike distributions and autocorrelograms . Burst activity showed a wide or bimodal interval interspike distribution and a significant single peak on the autocorrelation function . Irregular activity was characterised by a wide interval interspike distribution and a flat autocorrelogram . Regular activity was characterised by a narrow interval interspike distribution and an autocorrelogram with multiple regular peaks . | Parkinson’s disease is a condition affecting the human brain that becomes worse over time . The most common symptoms are tremors , muscle spasms and movements that are much slower than normal; all of which decrease an individual’s quality of life . Although there is currently no cure , the brain structures involved in Parkinson’s disease are known . These are collectively termed the basal ganglia , and are often targeted to treat the symptoms of Parkinson’s disease . For example , electrically stimulating the subthalamic nucleus ( STN ) , one part of the basal ganglia , reduces muscle tremors and stiffness . Pain is another common symptom in Parkinson’s disease . Patients often report strange burning or stabbing sensations with no obvious physical cause . They are also likely to be more sensitive to painful stimuli and have a lower pain threshold than normal . This suggested that the brain circuits that allow us to perceive and process pain could be somehow involved in Parkinson’s disease . Indeed , stimulating the STN is known to relieve pain in Parkinson’s disease , as well as the muscle symptoms , but exactly how the STN might link up with the brain’s ‘pain network’ remains poorly understood . Pautrat et al . therefore set out to explore the connection between pain networks and the STN , and determine its potential role in Parkinson’s disease . First , the electrical activity of nerve cells in the STN of rats was measured , which revealed that these cells do respond to mildly painful sensations . Experiments using dyes to label cells in both the STN and brain structures known to transmit painful signals showed that the STN was indeed directly linked to the brain’s pain network . Moreover , rats with a STN that did not work properly also responded abnormally to painful stimuli , confirming that the STN did indeed influence their perception of pain . Finally , Pautrat et al . repeated their measurements of electrical activity in the STN , this time using rats that lacked the same group of nerve cells affected in the basal ganglia of patients with Parkinson’s disease . Such rats are commonly used to model the disease in laboratory experiments . In these rats , the STN cells responded very strongly to painful stimuli , suggesting that problems with the STN could be causing some of the pain symptoms in Parkinson’s disease . This work reveals a new role for the STN in controlling responses to pain , both in health and disease . Pautrat et al . hope that their results will inspire research into more effective treatments of nerve pain in both Parkinson’s disease and other neurodegenerative conditions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Revealing a novel nociceptive network that links the subthalamic nucleus to pain processing |
Flexible neural networks , such as the interconnected spinal neurons that control distinct motor actions , can switch their activity to produce different behaviors . Both excitatory ( E ) and inhibitory ( I ) spinal neurons are necessary for motor behavior , but the influence of recruiting different ratios of E-to-I cells remains unclear . We constructed synthetic microphysical neural networks , called circuitoids , using precise combinations of spinal neuron subtypes derived from mouse stem cells . Circuitoids of purified excitatory interneurons were sufficient to generate oscillatory bursts with properties similar to in vivo central pattern generators . Inhibitory V1 neurons provided dual layers of regulation within excitatory rhythmogenic networks - they increased the rhythmic burst frequency of excitatory V3 neurons , and segmented excitatory motor neuron activity into sub-networks . Accordingly , the speed and pattern of spinal circuits that underlie complex motor behaviors may be regulated by quantitatively gating the intra-network cellular activity ratio of E-to-I neurons .
Many behaviors are based on circuits with flexible activity capable of switching their output ( Bargmann and Marder , 2013; Garcia-Campmany et al . , 2010 ) . Although connectomes and functional roles for the neuronal subtypes that comprise circuits have begun to be defined , the output of large multicellular networks are difficult to predict from the input pattern because the mechanisms that coordinate and regulate these complex systems remain poorly understood . A remarkable attribute of many CNS networks with extensive interconnections among the cells is their ability to default into a sustained self-organized oscillatory activity ( Buzsaki , 2006 ) . If only excitatory ( E ) neurons comprised these networks it is thought that inputs would trigger an avalanche of epileptic activity ( Buzsaki , 2006 ) . Thus , inhibitory ( I ) cells are considered important components of dynamic neural networks , because they can impose a regulated pattern on the system ( Brown , 1911; Buzsaki , 2006; Goulding et al . , 2014; Grillner and Jessell , 2009; Isaacson and Scanziani , 2011; Marder and Bucher , 2001 ) . Within neuronal networks , inhibitory neurons can increase coding efficiency , sharpen contrasts , and specify the topography of active circuits ( Arevian et al . , 2008; Buetfering et al . , 2014; Denève and Machens , 2016 ) . While the physiological roles of rhythmic brain activity are not always apparent , oscillatory central pattern generator ( CPG ) networks underlie breathing , chewing , scratching , and locomotion ( Grillner and Jessell , 2009; Marder and Bucher , 2001 ) . Vertebrate CPGs associated with locomotion are distributed networks of interconnected excitatory and inhibitory neurons that represent autonomous units capable of generating precisely patterned rhythmic activity ( Feldman and Smith , 1989; Nishimaru et al . , 2000; Whelan et al . , 2000; Cowley and Schmidt , 1995; Grillner , 2006 ) . Importantly , spinal CPGs have dynamic features that endow them with the ability to switch their frequency of rhythmicity in order to modulate locomotor speed and change the inter-coordination of motor pool firing patterns to drive different motor behaviors . The basis for producing flexible CPG activity remains unclear ( Grillner , 2006; McCrea and Rybak , 2008 ) , but it is likely founded in the differential recruitment of excitatory and inhibitory neurons comprising the CPG via selective inputs from descending and sensory systems ( Cowley and Schmidt , 1995; Gosgnach et al . , 2006; Grillner , 2006; Zhang et al . , 2014; Zhong et al . , 2011 ) . However , the contribution of a precisely controlled E/I cell ratio in the context of rhythmic circuit dynamics has been difficult to establish using cell ablation approaches that eliminate entire cell populations or pharmacological applications that silence entire types of synaptic transmission . The molecular and functional characterization of spinal interneuron subtypes that contribute to vertebrate CPG activity has led to the identification of cells with local and long-range connections , with ipsilateral and contralateral projections , and with inhibitory and excitatory properties ( Alaynick et al . , 2011; Bikoff et al . , 2016; Gabitto et al . , 2016; Garcia-Campmany et al . , 2010; Goulding and Pfaff , 2005; Grillner and Jessell , 2009; Stepien and Arber , 2008 ) . A common trait among each class of spinal interneurons comprising the locomotor CPG is that they form an extensive network of interconnections within their subclass , between the interneuron subclasses , and onto motor neurons ( Alvarez et al . , 2005; Crone et al . , 2008; Lanuza et al . , 2004; Zhang et al . , 2008 ) . In addition to their connectivity , the functional role of interneuron subclasses has been investigated by genetic methods such as cell killing or silencing . Ablation of inhibitory V1 neurons slows the burst frequency of the CPG ( Gosgnach et al . , 2006 ) , whereas ablation of both V1 and V2b inhibitory neurons disrupts the coordination of flexor-extensor alternation ( Zhang et al . , 2014 ) . Removing the V2a or V3 excitatory neurons from the network results in more irregular CPG bursting ( Crone et al . , 2008; Zhang et al . , 2008 ) . These studies elegantly demonstrate that V1 , V2b , V2a , and V3 interneurons each represent a necessary component of the CPG , but they do not address whether the cell types play instructive roles in gating the dynamic features of the circuitry . To begin to address this difficult issue , we used genetically labeled neurons to construct synthetic rhythmically active neuronal networks composed of defined cell subtypes . This approach allowed us to investigate whether quantitative differences in the cellular E/I ratio of specific neuronal subtypes are sufficient to alter the oscillatory speed and intra-network coordination of bursts . Embryonic stem cells with genetic reporters for individual spinal cord interneuron subtypes and motor neurons were isolated and differentiated with inductive factors that mimic embryonic development of the spinal cord ( Kutejova et al . , 2016; Peljto et al . , 2010; Wichterle and Peljto , 2008; Wichterle et al . , 2002 ) . We found that synthetic networks comprised of only excitatory V2a or V3 interneurons were sufficient to produce stable rhythmic bursts of activity at a frequency similar to drug-evoked fictive locomotor CPG activity . In contrast , V1 inhibitory neurons lacked the ability to produce regular bursts in isolation , but they did affect the activity of other cells . The addition of increasing numbers of V1 ( I ) neurons to fixed V3 ( E ) networks quantitatively accelerated the burst frequency , demonstrating that the E/I cell ratio instructively sets the oscillation speed . Interestingly , we found that V1 ( I ) neurons influenced the activity of motor ( E ) neuron bursting differently . Motor neuron networks , whose interconnected cells fire in unison , were uncoupled into sub-networks by the addition of increasing numbers of V1 cells . The formation of subnetworks is a form of patterning we define as segmentation . Taken together , our data lead to a model in which input-specific shifts in E/I cellular activity can flexibly tune the speed and pattern of an autonomous oscillatory circuit . By extension , this simple strategy for controlling the oscillatory properties that naturally emerge from interconnected neuronal networks might serve as the basis for generating complex , yet highly flexible , motor behaviors .
We constructed circuits with neurons that were generated de novo from mouse embryonic stem ( ES ) cells rather than using neurons isolated from spinal cords that may have acquired functional properties through complex interactions with their environment . Stem cell lines were isolated from blastula embryos harboring fluorescent reporters that indelibly label cardinal populations of spinal neurons as they develop postmitotically and form mature circuits . This allowed us to accurately and sensitively monitor ES cell differentiation into spinal neuron subtypes in real time . We isolated four independent ES cell lines containing the V1 reporter En1:Cre/tdTomato , two lines with the V2a reporter Chx10:Cre/tdTomato , seven lines with the V3 reporter Sim1:Cre/tdTomato , and two lines with the motor neuron reporter tgHb9-GFP ( Figure 1A ) . These ES cell lines were differentiated into neurospheres containing ~50 , 000 aggregated cells using retinoic acid ( RA ) and smoothened agonist ( SAG ) following procedures similar to those described for generating motor neurons ( Peljto et al . , 2010; Wichterle and Peljto , 2008; Wichterle et al . , 2002 ) . These ES-cell-derived neurospheres could be maintained for weeks in culture and were found to contain glia and multiple neuronal subtypes ( Figure 1B ) . These cellular aggregates did not appear to adopt a morphological organization that resembled the neural tube or spinal cord , rather neurons and glia seemed to be randomly distributed within the spheres ( Figure 1B , data not shown ) . 10 . 7554/eLife . 21540 . 003Figure 1 . Spontaneous activity emerges from networks created from ES cell-derived spinal neurons . ( A ) Mouse ES cell lines were derived from embryos with genetic tags for defined spinal neuron subclasses . The number of individual ES cell lines generated for each genotype is shown in parentheses . ( B ) ES cells ( phase contrast ) were differentiated ( diff . ) into neurospheres ( dark field ) with retinoic acid and smoothened agonist ( SAG ) for six days . After maturation ( mat . , 17 days post-ES cell ) , a circuitoid was immunostained to identify cell types ( neurotrace: white; V3 tomato reporter: red; astrocytes , GFAP: green; nuclei , DAPI: blue ) . Network activity was recorded ( rec . ) from circuitoids using a suction electrode and calcium imaging ( tgCAG:GCaMP3 ES cell line , 1000 nM SAG , 15 days post-ES ) . Vertical scale bar is 8% dF/F and 280 μV , and horizontal 1 min . ( C ) The V3 reporter line generates tomato +V3 interneurons ( red ) with 1000 nM SAG , whereas the V1 ES cell line generates tomato +V1 interneurons ( red ) with 5 nM SAG 10 days post-ES cell differentiation . DAPI ( blue ) . ( D ) Quantification of neuronal subtypes from ES cell-fluorescent reporter lines in ( A ) generated with increasing SAG concentrations using FACS to quantify cell numbers . Mean ± standard error of the mean ( SEM ) . Differentiations: n = 8 for each ES cell-reporter line tested at each SAG concentration . Unpaired t test: *p<0 . 05; **p<0 . 01; ***p<0 . 001; ****p<0 . 0001 . ( E and F ) Circuitoids composed of different neuronal subtypes produce distinct spontaneous network burst activity . ( E ) Calcium dye traces from spontaneously active circuitoids ( Cir ) generated at low ( L ) or high ( H ) SAG concentration: 5 nM SAG , CirL , red trace; 1000 nM SAG , CirH , green trace . Fluorescent images shown for indicated time points . ( F ) Quantification of bursting parameters using calcium imaging . CirL spheres produce low-amplitude high-frequency bursts compared to CirH spheres 16–17 days post-ES . Mean ± SEM; 1000 nM n = 21; 5 nM n = 19; unpaired t test ***p<0 . 001 , ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 00310 . 7554/eLife . 21540 . 004Figure 1—figure supplement 1 . Neurosphere differentiation and composition . ( A and B ) 10 days-post ES cell differentiation neurospheres were dissociated and cell composition was quantified using FACS . Plots of V3 , V2a , and V1 interneurons and motor neurons from Sim1:Cre;R26/C:LSL:Tomato , Chx10:Cre;R26/C:LSL:Tomato , En1:Cre;R26/C:LSL:Tomato , and Hb9:GFP ES cell reporter lines , respectively are shown after differentiation with ( A ) 1000 nM and ( B ) 5 nM SAG . ( C ) Two Chx:10:Cre;R26/C:LSL:Tomato ES cell lines were differentiated 10 days under a range of SAG concentrations . Tomato+V2a neuron number was quantified using FACS . ( D ) Chx:10:Cre;R26/C:LSL:Tomato lines #1 and #2 were differentiated with 15 nM and 40 nM SAG , respectively . 5 μM DAPT was applied days 6–10 . Standardized tomato+ V2a neurons relative to no DAPT control . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 00410 . 7554/eLife . 21540 . 005Figure 1—figure supplement 2 . ES-cell-derived neurons express neuronal subtype markers . ( A–B ) Motor neurons and V2a interneurons were differentiated from Hb9:GFP and Chx10:Cre;R26/C:LSL:Tomato ES cell lines , respectively . FACS-purified cells were gene-expression profiled using next generation RNAseq . ES-cell-derived motor neurons and V2a interneurons express markers corresponding to motor neurons and V2a cells purified from mouse embryos , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 00510 . 7554/eLife . 21540 . 006Figure 1—figure supplement 3 . Circuitoid activity and synaptic structures . ( A ) Plated ES-cell-derived V3 interneurons labeled with calcium indicator dye ( OGB-1 , green ) and gradient contrast transmitted light ( Dodt ) . Cells simultaneously recorded with electrode and calcium imaging circled ( red , blue ) . ( B ) Simultaneous calcium imaging and electrical recording similarly reveals neuronal activity . ( C ) V3 interneurons ( red ) differentiated from Sim1:Cre;LSL:tdTomato ES cells immunostained for VGLUT2 ( magenta ) and PSD95 ( green ) form numerous pre- and post-synaptic structures . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 006 ES cells with each neuronal-subtype reporter were differentiated with increasing SAG concentrations and the resulting neurospheres were dissociated and quantified using FACS ( fluorescent activated cell sorting ) ( Figure 1—figure supplement 1A , B ) . At 5nM SAG , dorsally located V1 interneurons were enriched compared to ventral V3 interneurons and motor neurons ( Figure 1C and D ) . V3 interneurons were preferentially generated at the expense of V1 interneurons in 1000 nM SAG , whereas motor neurons were most enriched using 50–200 nM SAG ( Figure 1C and D ) . In each case , up to 30% of the cells within a neurosphere were V1 , V3 , or motor neurons using the optimal SAG concentration for each neuronal subtype , with similar results observed for each of the independent stem cell lines . V2a interneurons proved more difficult to generate and variability among our two reporter lines for these cells was noted ( Figure 1—figure supplement 1C ) . We found , however , that blocking the notch-delta signaling pathway with the gamma secretase inhibitor DAPT improved V2a interneuron generation consistent with previous developmental studies on specification of this interneuron subtype ( Del Barrio et al . , 2007; Brown et al . , 2014; Crone et al . , 2008 ) ( Figure 1—figure supplement 1D ) . In addition to the cardinal marker used to label each cell type ( Figure 1A ) , RNA-sequencing of FACS purified ES-cell-derived motor neurons and V2a interneurons revealed that each expressed a battery of genes consistent with their identity assignment ( Al-Mosawie et al . , 2007; Brown et al . , 2014; Kimura et al . , 2006; Kutejova et al . , 2016; Lundfald et al . , 2007; Saueressig et al . , 1999; Wichterle et al . , 2002; Xu and Sakiyama-Elbert , 2015; Zhang et al . , 2008 ) ( Figure 1—figure supplement 2 ) . ES-cell-derived V1 and V3 interneuron identity was monitored by assaying for expression of their cardinal marker gene ( En1 and Sim1 , respectively ) , and confirming their neurotransmitter identity ( glycinergic/GABAergic and glutamatergic , respectively ) . The neuronal subtypes within the ventral spinal cord form extensive interconnections with each other and are known to fire spontaneous bursts of activity ( Alvarez et al . , 2005; Barry and O'Donovan , 1987; Crone et al . , 2008; Lanuza et al . , 2004; O’Donovan and Landmesser , 1987; Zhang et al . , 2008 ) . We observed that ES-cell-derived neurons formed numerous excitatory synapses in culture ( Figure 1—figure supplement 3C ) . To investigate whether matured neurospheres likewise produce spontaneous bursts , we monitored network output using both an extracellular suction recording electrode and calcium imaging dyes and found the cell permeable calcium dyes provided an accurate measure of neuronal activity ( Figure 1B , Figure 1—figure supplement 3A and B ) . Likewise , genetically encoded ubiquitously expressed , GCaMP3 was also a reliable reporter of activity ( Figure 1B ) ; however , most experiments were performed with cell-permeable calcium indicator dyes because they were easier to use . Imaging neuronal activity revealed that bursting within the spheres was highly synchronized across many neurons , suggesting the cells were interconnected ( Video 1 ) . We designated these de novo networks ‘circuitoids’ because they displayed a highly organized and behaviorally relevant pattern of activity . This nomenclature is intended to help distinguish these microphysical systems from neurospheres defined more broadly by the presence of neurons , or embryoids and organoids , which are defined by their morphological and cellular organization rather than their neuronal activity ( Lancaster et al . , 2013; Sato et al . , 2009; Stevens , 1960 ) . 10 . 7554/eLife . 21540 . 007Video 1 . Mature circuitoids display spontaneous activity . Heterogeneous circuitoids display spontaneous bursts of network activity that appear to be synchronous throughout each sphere . Here , an En1:Cre;R26/C:LSL:Tomato ES cell line was differentiated with 1000 nM SAG and allowed to mature until 16 days post-ES cell . Calcium intensity change ( dF/F ) was pseudocolored ( scale from black to white ) . About 15 individual circuitoids with 50 , 000 cells each are in the field of view . Movie plays at 2x speed . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 007 Next , we examined whether the cellular composition of the circuitoids influenced the spontaneous activity by comparing bursts recorded from circuitoids produced with high SAG concentrations ( 1000 nM , CirH , cell composition: V3>MN>V1 ) versus low SAG ( 5 nM , CirL , cell composition: V1>MN>V3 ) ( cell types quantified Figure 1D ) . We found that CirL spheres produced bursts with lower amplitude and higher frequency than CirH spheres ( Figure 1C–F ) . Since CirH produced reliable high-amplitude bursts , these circuitoids were used to examine the development of and physiological basis for this neuronal activity . CirH spheres are enriched in V3 interneurons but are nonetheless heterogenous mixtures of ventral-spinal cord cell types . We found that these circuitoids were spontaneously active at two weeks and bursting increased in frequency at week 5 ( Figure 2A ) . Acetylcholine receptor antagonists did not significantly alter the spontaneous bursting of circuitoids ( Figure 2B ) , whereas the glutamatergic AMPA receptor antagonist , CNQX , blocked the activity suggesting that excitatory synaptic drive is necessary for coordinated bursts ( Figure 2C ) . Next , we examined how these circuitoids respond to NMA and 5-HT , which trigger rhythmic activity in CPG circuits ( evoked condition , see Materials and methods , Kudo and Yamada , 1987; Smith and Feldman , 1987; Whelan et al . , 2000 ) . Remarkably , these drugs activate a long-lasting highly regular ( rhythmic ) pattern of activity from circuitoids that is similar in frequency to fictive locomotor preparations ( Figure 2D ) . Taken together , these results indicate that circuitoids acquire network properties that allow the interconnected cells to fire rhythmic bursts . 10 . 7554/eLife . 21540 . 008Figure 2 . Physiological properties of circuitoids . Circuitoids were generated with 1000 nm SAG ( CirH ) for analysis using calcium imaging with a ubiquitously expressed GCaMP3 ES cell line . ( A ) Spontaneous bursting frequency increased from week 2 to 5 . Mean ± SEM , Week 2 n = 78; Week 3 n = 60; Week 4 n = 46; Week 5 n = 48; ****p<0 . 0001 , unpaired t test . ( B–D ) Activity recorded 15–17 days post ES cell differentiation . ( B ) Cholinergic antagonists DHβE and mecamylamine ( red ) do not block activity . Mean ± SEM , n = 17 circuitoids , ns p=0 . 31 , paired t test . ( C ) Glutamatergic antagonist CNQX ( brown ) abolishes bursting . Mean ± SEM , n = 16 circuitoids; ****p<0 . 0001 , paired t test . ( D ) CPG evoking drugs NMA and 5-HT increase the frequency and rhythmicity of circuitoids compared to spontaneous activity . Rhythmicity measured using interval coefficient of variation ( I . C . V . ; low values indicate high rhythmicity ) . Mean ± SEM , n = 18 circuitoids; ****p<0 . 0001 , **p<0 . 01 , paired t test . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 008 Circuitoids generated from low or high SAG concentrations during differentiation are biased toward distinct sets of neuronal subtypes , which correlate with specific alterations in functional output ( see Figure 1 ) . We noted , however , that while most CirH spheres produced regular activity in the evoked condition , a minority had irregular bursting ( Figure 3A and D ) . This led us to consider whether subtle differences in circuitoid composition may cause variability in their activity characteristics . We subdivided CirH spheres into two populations for further analysis: CirH-R ( rhythmic , interval coefficient of variation ( I . C . V . ) <0 . 2 ) and CirH-NR ( non-rhythmic , I . C . V . >0 . 2 ) . When picrotoxin and strychnine , two inhibitory antagonists , were applied , CirH-NR spheres burst in a more regular pattern ( Figure 3A–C ) , suggesting their irregular bursting was caused by the presence of inhibitory neurons . Consistent with this , CirH-R spheres remained rhythmic in inhibitory antagonists; however , we noted that their burst frequency slowed ( Figure 3D–F , see below ) . Thus , under baseline conditions CirH-R and CirH-NR spheres may differ slightly in their level of inhibitory neuron activity , accounting for their different bursting characteristics . 10 . 7554/eLife . 21540 . 009Figure 3 . Cellular composition influences circuitoid rhythmicity and burst speed . ( A–F ) Circuitoids generated with 1000 nM SAG ( CirH ) were recorded in drugs that evoke CPG activity ( evoked , NMA +5 HT , green ) 16–17 days post ES cell differentiation . Activity was classified as either rhythmic ( CirH-R , I . C . V . <0 . 2 ) or non-rhythmic ( CirH-NR , I . C . V . >0 . 2 ) . ( A ) Irregular bursting CirH-NR sphere . ( B ) CirH-NR bursting after application of inhibitory antagonists strychnine and picrotoxin ( purple ) . ( C ) CirH-NR bursting becomes more rhythmic ( lower I . C . V . ) when inhibitory synaptic transmission is blocked . Mean ± SEM , n = 14 circuitoids , paired t test ****p<0 . 0001 . ( D ) Bursting activity of CirH-R sphere . ( E ) Bursting activity of CirH-R after application of inhibitory antagonists strychnine and picrotoxin ( purple ) . ( F ) CirH-R burst frequency decreases when inhibitory synaptic transmission is blocked . Mean ± SEM , n = 9 circuitoids , paired t test **p<0 . 01 . ( G–N ) Circuitoids generated from the V3DTA ES cell line ( derived from Sim1:Cre;R26:LSL:DTA ) with 1000 nM SAG ( CirH ) were recorded in drugs that evoke CPG activity ( evoked , NMA +5 HT , green ) 16–17 days post differentiation . ( G ) Control ( Sim1:Cre , WT , green ) circuitoid bursting . ( H ) V3DTA bursting ( grey ) in circuitoids lacking V3 interneurons . ( I ) V3DTA bursting is less rhythmic and ( J ) more frequent . ( K–N ) V3DTA bursting becomes more rhythmic and less frequent with inhibitory antagonists strychnine and picrotoxin ( purple ) . ( G–J ) WT n = 5 independent differentiations n = 69 circuitoids . V3DTAn = 5 independent differentiations n = 85 circuitoids . Mean ± SEM , unpaired t test: *p<0 . 05 , ****p<0 . 0001 . ( K–N ) n = 18 circuitoids , mean ± SEM , paired t test: ***p<0 . 001; ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 009 To test how circuitoid rhythmic bursting responds to a reduction in the cellular level of excitatory drive , we employed a genetic approach to ablate V3 excitatory neurons . Transgenic floxed-diphtheria toxin subunit-A ( DTA ) mice were crossed to Sim1-Cre animals and used to isolate ES cells in which V3 interneurons are killed upon differentiation ( V3DTA , Figure 1A ) ( Zhang et al . , 2008 ) . We found that V3DTA circuitoids failed to burst regularly in drugs that evoke rhythmic CPG activity compared to controls , and they produced these irregular bursts at a higher frequency ( Figure 3G–J ) . Interestingly , by lowering the inhibitory drive in V3DTA circuitoids with inhibitory antagonists the networks switched to a slower and more regular activity pattern ( Figure 3K–N ) . These findings reveal that the cellular composition within heterogeneous circuitoids influences the network's activity . Despite extensive functional characterization of neuron subtypes within the spinal CPG , it remains unclear whether a particular cell type acts as a pacemaker , or if the rhythmicity arises from the emergent properties of interconnected excitatory and inhibitory neurons ( Feldman et al . , 2013; Harris-Warrick , 2010; Marder and Bucher , 2001 ) . We sought to determine whether synthetic networks generated from specific neuronal subtypes were sufficient to produce rhythmic activity . CirH spheres were generated using the V3 interneuron reporter line , dissociated and tomato+ V3 interneurons were purified with FACS ( Figure 4A , Figure 4—figure supplement 1A , B ) . We reconstituted the purified V3 neurons into a synthetic network ( purity > 98% ) by aggregating these cells around an astrocyte core ( Figure 4A , Figure 4—figure supplement 1C ) . Similar to heterogeneous circuitoids , large spontaneous bursts of activity were observed in synthetic networks comprised of only V3 neurons ( Figure 4B ) . Likewise , NMA and 5-HT evoked a rhythmic pattern of activity with a frequency higher than the spontaneous burst rate ( Figure 4B , C , S , T ) . Consistent with the glutamatergic properties of V3 interneurons , inhibitory antagonists did not significantly perturb the evoked pattern of bursting , whereas the glutamatergic antagonist , CNQX , blocked activity ( Figure 4C–E and R–T ) . If gap junctions are present between cells , they appear to be insufficient to sustain rhythmic network activity in the presence of AMPA receptor blockers . As V3 circuitoids mature from 17 to 45 days after the initiation of ES cell differentiation , we determined that their bursting frequency increases ( Figure 4—figure supplement 2A ) . We also found that circuitoids ranging in size from 5000 to 100 , 000 cells exhibited similar bursting parameters , suggesting they form highly scalable networks ( Figure 4—figure supplement 2B ) . All together , these findings indicate that synthetic networks comprised of synaptically coupled , purified V3 interneurons are sufficient to produce rhythmic bursting in response to drugs that evoke CPG activity . 10 . 7554/eLife . 21540 . 010Figure 4 . Rhythmic activity in networks with purified neuron subclasses . ( A ) Circuitoids of defined cellular composition were created by re-aggregating FACS purified ES cell-derived neurons ( red ) with astrocytes ( blue ) following the outlined scheme . ( B–Q ) Bursting activity of neuronal subclasses measured with calcium dyes under various conditions: spontaneous ( blue ) , evoked ( NMA +5 HT , green ) , evoked + inhibitory antagonists ( NMA +5 HT + strychnine + picrotoxin , purple ) , and evoked + glutamatergic antagonist ( NMA +5 HT + CNQX , brown ) . ( R ) Spontaneous burst amplitude of V1 interneurons is lower than V3 , V2a , and motor neuron circuitoids . Evoked burst amplitudes were unaffected by inhibitory antagonists with V3 , V2a , and motor neuron networks . Mean ± SEM , {number} of independent differentiations and [number] of circuitoids indicated after each cell type . V3 interneurons ( INs ) {22}[175] , V2a interneurons {5}[49] , motor neurons ( MNs ) {11}[49] , V1 interneurons {12}[42] . Unpaired t test: ( ns ) p>0 . 05; **p<0 . 01; ****p<0 . 0001 . ( S ) V3 and V2a interneuron rhythmicity increased ( I . C . V . decreased ) under the evoked condition compared to spontaneous bursts . Mean ± SEM , {number} of independent differentiations and [number] of circuitoids indicated after each cell type . V3 {22}[175] , V2a {5}[49] , MN {11}[49] , V1 {12}[42] . Paired t test: ( ns ) p>0 . 05; *p<0 . 05; ***p<0 . 001; ****p<0 . 0001 . ( T ) Burst frequency increases in the evoked condition and decreases with glutamatergic antagonist CNQX . Mean ± SEM , {number} of independent differentiations and [number] of circuitoids indicated after each cell type . V3 {22}[175] , V2a {5}[49] , MN {11}[49] , V1 {12}[42] . Paired t test: ( ns ) p>0 . 05; *p<0 . 05; ***p<0 . 001; ****p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 01010 . 7554/eLife . 21540 . 011Figure 4—figure supplement 1 . Cell purifications to generate synthetic neural networks . ES cells were differentiated , dissociated , and neuronal subtypes were FACS purified to evaluate differentiation efficiency and purity . ( A ) Wild-type non-fluorescent cell line , negative control . ( B ) V3 interneuron purification ( Sim1:Cre;R26/C:LSL:Tomato ) . ( C ) Resorting of purified V3 interneurons from ( B ) reveals 98 . 8% purity . ( D ) V2a interneuron ( Chx10:Cre;R26/C:LSL:Tomato ) sort . ( E ) Motor neuron ( Hb9:GFP ) sort . ( F ) V1 interneuron ( En1:Cre;R26/C:LSL:Tomato ) sort . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 01110 . 7554/eLife . 21540 . 012Figure 4—figure supplement 2 . Frequency and rhythmicity of V3 interneuron networks . ( A ) Sim1:Cre;R26/C:LSL:Tomato ES cell lines were differentiated and cell sorted . 50 , 000 V3 interneurons were aggregated and calcium imaged at weekly intervals . Spontaneous and evoked bursting frequency increases as networks mature . Evoked bursting is rhythmic by D17 and stable at D24 , roughly corresponding to the gestational period of a mouse embryo . Mean ± SEM , four independent differentiations and ≥19 circuitoids measured for each time point . Unpaired t test **p<0 . 01 , ***p<0 . 001 . ( B ) Sim1:Cre;R26/C:LSL: Tomato ES cell lines were differentiated and cell sorted . 5000 to 100 , 000 V3 interneurons were reaggregated and bursting activity was recorded three weeks later using calcium dyes . Network cell number does not affect spontaneous burst frequency or the regularity of evoked rhythmicity ( I . C . V . ) , but burst frequency decreases slightly as network size increases in the evoked condition . Mean ± SEM , four independent differentiations and ≥23 circuitoids measured for each network size . Unpaired t test ( ns ) p>0 . 05 , *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 01210 . 7554/eLife . 21540 . 013Figure 4—figure supplement 3 . Cholinergic antagonists do not affect motor neuron networks . ( A ) Hb9:GFP ES cell lines were differentiated and purified with FACS to create pure motor neuron networks . Three weeks after reaggregation spontaneous and evoked bursts were detected with calcium dye imaging . Neither cholinergic antagonists ( DHβE and mecamylamine ) nor inhibitory antagonists ( strychnine + picrotoxin ) disrupt the bursts , whereas glutamatergic antagonist ( CNQX ) blocks activity . Mean ± SEM , four independent differentiations and 47 circuitoids measured , paired t test ( ns ) p=0 . 52 . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 013 Like V3 interneurons , V2a cells represent another major class of excitatory glutamatergic interneurons found to contribute to robust CPG activity ( Crone et al . , 2008 ) . We purified V2a interneurons ( Figure 4—figure supplement 1A , D ) and created synthetic networks that were spontaneously active and became rhythmic in the evoked condition ( Figure 4F , G and S ) . The amplitude , rhythmicity , and frequency of V2a and V3 networks were similar ( Figure 4R–T ) . As expected , the burst patterns of V2a excitatory networks were not affected by glycine/GABA antagonists but were disrupted by AMPA receptor blockers ( Figure 4G–I and R–T ) . Networks comprised of purified motor neurons ( Figure 4—figure supplement 1E ) , which have recurrent collaterals that synapse onto neighboring motor neurons and interneurons ( Eccles et al . , 1954; Nishimaru et al . , 2005; Renshaw , 1946 ) , also displayed spontaneous bursting and responded to drugs that evoke CPG activity ( Figure 4J and K ) . However , the bursts were more irregular and in general had greater variability between experiments than those produced by the V2a and V3 interneurons networks ( Figure 4S ) . The activity in motor neuron circuitoids was dependent upon glutamatergic signaling rather than acetylcholine ( Figure 4K–M and T , Figure 4—figure supplement 3 ) . These findings are consistent with the intrinsic burst properties of motor neurons ( Hochman et al . , 1994; Kiehn et al . , 2000; MacLean et al . , 1997 ) , their responsiveness to glutamatergic input ( Jahr and Yoshioka , 1986; Rekling et al . , 2000 ) , and their co-release of cholinergic and glutamatergic neurotransmitters ( Herzog et al . , 2004; Lamotte d'Incamps and Ascher , 2008; Meister et al . , 1993; Mentis et al . , 2005; Nishimaru et al . , 2005 ) . To test whether networks of inhibitory neurons can produce rhythmic activity , we generated circuitoids of V1 interneurons ( Figure 4—figure supplement 1A , F ) . While spontaneous activity can be observed in V1 networks , the burst amplitudes are significantly lower than excitatory networks and fail to burst rhythmically in drugs that evoke CPG activity ( Figure 4N , O , R and S ) . Although cell purifications are >98% pure ( see Figure 4—figure supplement 1C ) , the low levels of activity found in V1 networks could be due to contamination with small amounts of excitatory neurons because inhibitory antagonists increased burst amplitude and rhythmicity , and a glutamatergic AMPA receptor blocker , CNQX , stopped the activity ( Figure 4N–Q and T ) . These findings reveal a marked difference between the rhythmic activity of excitatory versus inhibitory neuron networks , and indicate that multiple subclasses of excitatory interneurons are sufficient to form rhythmically active networks . Dynamic circuits can flexibly switch how they use different cellular components within the larger network in order to meet the changing behavioral demands of the animal ( Ampatzis et al . , 2014; McLean et al . , 2007 , 2008 ) . This suggests that the active cellular E/I relationships within complex networks are not fixed , prompting us to explore how network activity is affected by different E/I cellular ratios . We differentiated neurons from ES cells and purified fluorescently tagged subtypes with FACS . Specific combinations of excitatory and inhibitory neuronal subtypes were plated onto an astrocyte layer ( Figure 5A ) where they formed interconnected networks that were spontaneously active . We found that excitatory V3 interneurons produced a reliable pattern of spontaneous bursts that continually increased in frequency when co-cultured with increasing numbers of inhibitory V1 interneurons ( Figure 5B and D ) . 10 . 7554/eLife . 21540 . 014Figure 5 . V3 network burst speed is tuned by inhibitory V1 interneurons . ( A ) ES cell lines with fluorescent reporters for neuronal subtypes were differentiated , and tomato +V3 and V1 interneurons were purified by FACS . Monolayer circuitoids were created by plating 100 , 000 V3 interneurons with 0–40 , 000 V1 cells onto astrocytes . Activity was recorded using calcium indicator dye to image co-culture networks 14 days after plating . ( B ) Spontaneous network bursting increases as the number of inhibitory V1 interneurons increases in the V3 interneuron cultures ( green ) . Network frequency returns to baseline levels in the presence of inhibitory antagonists strychnine + picrotoxin ( purple ) . ( C ) Motor neuron burst frequency does not increase as the concentration of V1 interneurons increases , but the amplitude of the bursts decreases ( brown ) . Inhibitory antagonists strychnine + picrotoxin return motor neuron network burst amplitude to normal levels ( red ) . ( D ) Quantification of burst frequency from traces in ( B ) . For each trial and condition , the standardized rate of activity was calculated by dividing the burst frequency of each network by the control network’s ( lacking V1 cells ) rate . Mean ± SEM , n = 10 networks for each V1 concentration in the spontaneous condition , n = 4 networks with inhibitory antagonists for each V1 concentration . Unpaired t test: *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001 . ( E ) Quantification of relative burst rate from traces in ( B , C ) . Standard rate of activity was calculated by taking the ratio of a network’s frequency in the inhibitory antagonist condition to its frequency in the spontaneous condition . Mean ± SEM , n = 6 motor neuron networks and n = 4 V3 interneuron networks at 0 and 40 , 000 V1 interneurons , unpaired t test: ( ns ) p=0 . 63; **p<0 . 01 . ( F ) Quantification of relative burst amplitude from traces in ( B , C ) . Mean ± SEM , n = 10 V3 networks vs n = 7 MN networks for each V1 concentration tested , unpaired t test: **p<0 . 01; ***p<0 . 001; ****p<0 . 0001 . ( G ) Amplitude coefficient of variation ( C . V . , Materials and methods ) calculated from network activity in ( B , C ) . Motor neuron burst amplitude C . V . increased as V1 inhibitory cell number increased . Mean ± SEM , n networks in the ( spontaneous ) and [inhibitory antagonist] condition for V3 networks ( 10 ) [4] and MN networks ( 8 ) [≥4] at all V1 interneuron concentrations . MN spontaneous vs MN inhibitory antagonist ( * ) and MN spontaneous vs V3 spontaneous ( Δ ) . Δ/*p<0 . 05; ΔΔ/**p<0 . 01; ***p<0 . 001; ΔΔΔΔ/****p<0 . 0001; unpaired t test . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 01410 . 7554/eLife . 21540 . 015Figure 5—figure supplement 1 . Generation of synthetic networks comprised of defined neuronal subtypes . Hb9:GFP ES cell lines were differentiated and GFP+ motor neurons purified using FACS to generate base networks of 100 , 000 motor neurons . In parallel , En1:Cre;R26/C:LSL:Tomato ES cell lines were differentiated and tomato +V1 inhibitory neurons purified using FACS . 0 to 40 , 000 V1 interneurons were added to the base motor neuron networks . ( A ) Fluorescent images of synthetic networks confirming graded numbers of tomato+ cells ( red ) in the co-cultures . ( B ) Tomato+ signal quantification in the co-cultures . Mean ± SEM , n = 7 networks per condition . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 015 To test whether the inhibitory neurons entrained the bursting of V3 interneurons , we acutely blocked all inhibitory synaptic transmission in these cultures with picrotoxin and strychnine . After application of the inhibitory antagonists , the cultures returned to the bursting pattern observed without V1 neurons ( Figure 5B , D , F and G ) . These findings reveal three aspects of the synthetic networks . First , the effect of inhibitory neurons is mediated by direct synaptic activity rather than causing changes in the development , survival , or physiological properties of V3 cells because picrotoxin and strychnine quickly restored V3 activity to its normal frequency . Second , the addition of V1 cells to V3 networks unlikely changes V3-activity by displacing cells or altering the connectome of the excitatory network because drug inhibitors of V1 interneurons restore normal V3-activity . Third , excitatory networks may tune their burst rate by flexibly recruiting different numbers of inhibitory neurons . Because networks comprised solely of purified motor neurons display different patterns of activity compared to V3 interneurons ( see Figure 4 ) , we next examined how increasing the ratio of inhibitory V1 interneurons influenced motor neuron activity ( Figure 5—figure supplement 1 ) . Unlike V3 networks , V1 interneurons had little influence on the burst frequency of motor neurons , but instead caused greater burst amplitude variation ( Figure 5C and E–G ) . To test whether the amplitude variability observed in motor neuron-V1 interneuron networks arose by converting the coordinated bursting of the network into subunits of desynchronized activity ( termed segmentation ) we refined how we monitored network output . Rather than measuring average neuronal bursts across the entire network we measured activity in sub-regions . In networks with a 10:1 motor neuron/V1 ratio , we found that separate regions of interest ( ROI ) could burst independently ( Figure 6A , Video 2 ) . Changes in the coordination of neuronal activity across the network were not due to depletion of excitatory interconnections among motor neurons or a physical alteration of the circuit's connectome , because inhibitory antagonists caused the network to rapidly switch back into a synchronous bursting pattern ( Figure 6B , Video 3 ) . The desynchronizing effect of inhibitory neurons on motor neuron networks was quantified as network complexity ( Figure 6C , Materials and methods ) . Interestingly , inhibitory V1 interneurons lacked the ability to create complexity within V3 networks ( Figure 6C , Figure 6—figure supplement 1 ) . 10 . 7554/eLife . 21540 . 016Figure 6 . V1 interneurons control the segmentation of motor neuron network activity . Circuitoids were established with either 100 , 000 motor neurons or 100 , 000 V3 interneurons combined with 0–40 , 000 V1 interneurons , plated on astrocytes . Networks were imaged 14 days after sorting with a calcium indicator dye and the static frames show pseudocolored calcium intensity ( dF/F ) images ( scale from black to white ) . ( A ) Neurons within plated circuitoids ( 10 , 000 V1 interneurons and 100 , 000 motor neurons ) fire asynchronously . Static frames during Burst A and B reveal different areas of activity in the network . Traces from two regions of interest ( ROI 1 and 2 ) on opposite corners of the field of view are displayed . ( B ) Inhibitory antagonists ( strychnine + picrotoxin ) applied to the network in ( A ) lead to synchronous bursts across the entire network . ( C ) Quantification of network complexity ( Materials and methods ) calculated from network activity . V1 interneurons increase network complexity of motor neuron networks . Median ± bootstrap standard error , n networks in the ( spontaneous ) and [inhibitory antagonist] condition for V3 networks ( 10 ) [4] and MN networks ( 8 ) [≥4] at all V1 interneuron concentrations . MN spontaneous vs MN inhibitory antagonist ( * ) and MN spontaneous vs V3 spontaneous ( Δ ) . Δ/*p<0 . 05; ΔΔ/**p<0 . 01; ΔΔΔp<0 . 001; ΔΔΔΔp<0 . 0001 Kolmogorov–Smirnov test . ( D ) 100 , 000 purified GFP+ motor neurons were cocultured with 10 , 000 tomato +V1 inhibitory neurons . Calcium imaging was used to detect bursts from five individual GFP+ motor neurons . Individual cells displayed coupled activity ( brown ) , but the entire cohort of motor neurons only became synchronously active when V1 inhibition was blocked with strychnine + picrotoxin ( red ) . ( E ) 100 , 000 purified V3 interneurons were cocultured with 100 , 000 V1 cells , and five individual cells were calcium imaged . Individual interneurons consistently fired together whether V1 cells were active ( green ) or silenced with inhibitory antagonists ( purple ) . ( F ) Quantification of neuronal synchrony using pair-wise cross-correlation analysis of motor neurons in motor neuron-V1 networks and interneurons in V3-V1 networks . The activity within motor neuron networks became increasingly less synchronized as V1 cell number increased . Mean±SEM , sample size n= ( network number ) [neuron number] . V3 network , spontaneous burst condition: 0–40 , 000 V1 cells ( 10 ) [200] , 100 , 000 V1 cells ( 4 ) [80] . V3 network , inhibitory antagonist condition: 0–40 , 000 V1 cells ( 4 ) [80] , 100 , 000 V1 cells ( 2 ) [40] . MN network , spontaneous burst condition , each V1 concentration tested ( ≥7 ) [≥140] . MN network , inhibitory antagonist condition , each V1 concentration tested ( 6 ) [120] . MN spontaneous vs MN inhibitory antagonist ( * ) and MN spontaneous vs V3 spontaneous ( Δ ) all compared at same V1 concentration , except for spontaneous 100 , 000 V1 in V3 network vs 40 , 000 V1 in MN network . Δ/*p<0 . 05; ΔΔ/**p<0 . 01; ΔΔΔp<0 . 001; ΔΔΔΔp<0 . 0001 , unpaired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 01610 . 7554/eLife . 21540 . 017Figure 6—figure supplement 1 . Activity across different network configurations . Motor neuron networks , V3 interneuron networks , and V3:V1 mixed networks were created by differentiating and sorting 100 , 000 motor neurons ( using Hb9:GFP ES cell lines ) , 100 , 000 V3 interneurons ( using Sim1:Cre;R26/C:LSL:Tomato ES cell lines ) , and 100 , 000 V1 interneurons ( using En1:Cre;R26/C:LSL:Tomato ES cell lines ) and plating them on astrocytes . Network activity was imaged using calcium indicator dye . ( A ) Pure motor neuron networks burst synchronously across the network ( ROIs are co-active during Burst A and B ) . ( B ) Pure V3 interneuron networks burst synchronously across the network ( ROIs are co-active during Burst C and D ) . ( C ) Networks with a 1:1 mixture of V3:V1 interneurons burst synchronously across the network ( ROIs are co-active during Burst E and F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 01710 . 7554/eLife . 21540 . 018Video 2 . V1 interneurons generate subnetwork activity in motor neuron-V1 networks . A network of 100 , 000 motor neurons and 10 , 000 V1 interneurons displays segmented activity . Calcium intensity change ( dF/F ) was pseudocolored ( scale from black to white ) , showing different active regions within a network . Movie plays at 10x speed . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 01810 . 7554/eLife . 21540 . 019Video 3 . Inhibitory antagonists synchronize motor neuron-V1 networks . The same network ( see Video 2 ) displays synchronous network activity after the application of inhibitory antagonists ( strychnine+picrotoxin ) , suggesting that synaptic activity from V1 inhibitory neurons patterns motor neuron networks . Calcium intensity change ( dF/F ) was pseudocolored ( scale from black to white ) , showing coordinated activity across the network . Movie plays at 10x speed . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 019 To further explore the functional interactions between V1-V3 interneurons and V1-motor neurons , we used cell-type-specific reporters to monitor the activity of individual identified neurons within the mixed networks . In V1-motor neuron co-cultures we noted that motor neurons ( GFP+/Tomato- ) were co-active on some bursts but also had the freedom to fire independently ( Figure 6D ) . When synaptic inhibition was blocked in these cultures all the motor neurons became synchronized ( Figure 6D and F ) . In contrast , V3 interneurons always fired in synchrony within mixed V3-V1 cultures , however the frequency slowed in the presence of inhibitory antagonists ( Figure 6E and F ) . Our results indicate that the cellular E/I ratio functions in a cell-type-dependent manner—namely , V1 interneurons accelerate the rate of V3 interneuron bursts in a coordinated fashion across the entire network and uncouple the activity of motor neurons to establish separate units without changing overall burst frequency . Flexibility within motor behaviors requires the ability to dynamically switch the combination of motor pools that are co-active when synergistic muscles are recruited , while separately regulating the burst frequency to control speed . To understand how this might arise , we examined the activity of tripartite circuitoids comprised of motor neurons , V3 and V1 interneurons . Equal numbers of motor neurons and excitatory V3 interneurons were combined with increasing numbers of V1 inhibitory cells . We found that the tripartite circuitoids maintained their synchrony across the network regardless of V1 cell number ( Figure 7A–C ) . The activity of tripartite networks resembled that observed for V3-V1 co-cultures studied above , as the relative bursting speed increased as V1 cells were added ( Figure 7D ) . This prompted us to examine the activity of motor neurons within the tripartite networks using the GFP+/Tomato- status of the cells to distinguish them from V1 and V3 interneurons in the culture . We found that within tripartite circuitoids motor neurons displayed the same coordinated pattern of activity as the interneurons and fired synchronously as a single unit within the culture across a range of V1 cell numbers ( Figure 7E–G ) . 10 . 7554/eLife . 21540 . 020Figure 7 . Motor neuron burst frequency is set by V3-V1 network activity . ( A–D ) Tripartite networks of 50 , 000 motor neurons , 50 , 000 V3 interneurons , and 0–40 , 000 V1 interneurons were formed , and burst activity monitored using calcium dyes ( see Figure 5A ) . ( A ) Spontaneous activity of 50 , 000:50 , 000 V3-MN network with 40 , 000 V1 neurons . ( B ) Spontaneous activity of network in ( A ) with inhibitory antagonists ( strychnine + picrotoxin ) . ( C ) Quantification of network complexity ( Materials and methods ) shows that unlike MN-V1 networks ( see Figure 6 ) MN-V1-V3 networks burst synchronously . Median ± bootstrap standard error , n = 5 for each V1 concentration and condition . ( D ) Increasing numbers of V1 interneurons in V3-MN networks increases burst frequency . Inhibitory antagonists reduce burst frequency of V1-V3-MN networks . For each trial , the frequency of the networks in each condition was standardized to the burst rate of control V3-MN networks ( lacking V1 cells ) . Mean ± SEM , n = 5 for each V1 concentration and condition . Paired t test: *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . ( E–G ) Burst analysis of individual fluorescent-labeled cell types in V1-V3-MN networks prepared as described in ( A–D ) . ( E ) Motor neuron bursting in a 50 , 000:50 , 000 V3-MN network with 40 , 000 V1 neurons . Spontaneous bursts ( brown ) and activity with inhibitory antagonists ( strychnine + picrotoxin , red ) are shown . ( F ) Interneuron bursting in the same network as ( E ) . Spontaneous burst ( green ) and activity with inhibitory antagonists ( purple ) are shown . ( G ) Neuronal synchrony quantification using pair-wise cross-correlation analysis of MN-MN , IN-IN , and MN-IN activity . The activity of each neuronal combination is highly correlated in V1-V3-MN networks . Mean±SEM , sample size n= ( network number ) [neuron number] ( 5 ) [50] for all cell populations and conditions at each V1 concentration . ( H–L ) Recordings of miniature post-synaptic currents ( mEPSC ) in 0 . 5 μM TTX . ( H ) Synaptic activity in pure motor neuron network . ( I ) Synaptic activity in pure V3 interneuron network . ( J ) Motor neuron mEPSCs in MN-V3 network . ( H-J ) Insets show synaptic drive associated with network bursting prior to TTX application ( Burst , left ) and averaged miniature events from one neuron ( mEPSC , right ) . ( K ) Frequency of mEPSCs is reduced in pure motor neuron networks . ( L ) Amplitudes of motor neuron mEPSCs are smaller than V3 interneurons . Mean ± SEM , n = 6 MNs from MN networks; n = 4 V3 INs from V3 networks; n = 6 MNs from MN-V3 networks . *p<0 . 05; **p<0 . 01; ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 21540 . 020 We considered the possibility that V3 interneurons can more strongly bind neurons within a network than motor neurons . To test this , we recorded miniature post-synaptic currents ( mEPSCs ) in networks composed of different neuronal subtypes . Within V3 networks , V3 interneurons displayed higher frequencies of miniature events than motor neurons in a pure motor neuron network ( Figure 7H , I , K ) . Interestingly , the identified motor neurons in mixed V3-motor neuron circuitoids had similar mEPSC frequencies as the V3 interneurons in V3 networks ( Figure 7H–K ) . This analysis indicates that the V3 cell-type may strongly bind the activity of neurons within a network due to a greater probability of neurotransmitter release or by providing an increased number of synaptic release sites .
Rhythmic neuronal activity that is coordinated across circuits comprised of interconnected neurons is a general property found in many areas of the CNS ( Buzsaki , 2006 ) . This activity pattern probably arises when irregular spikes randomly produced by interconnected excitatory neurons become integrated across dendritic trees , thereby converging into a regular oscillatory pattern of output shared by cells across the network ( Softky and Koch , 1993 ) . This raises the paradoxical issue of understanding how diverse patterns of activity necessary for driving complex behaviors are produced by networks that default into stable patterns of rhythmicity . Theoretical considerations of this problem argue that inhibitory neurons are necessary in such systems ( Buzsaki , 2006 ) . Although prevalent , the function of neural oscillations remains obscure in many cases . A wide range of frequency bands are observed in cortical structures ranging from 1 to 70 Hz and have been considered as possible substrates for binding information through synchronization and providing a time parameter to neural codes . Rhythmic neural activity likely plays roles in cognition and memory , and when unregulated produces tremors and seizures ( Buzsaki and Draguhn , 2004; Fries , 2005; Wang , 2010 ) . Aside from neural development ( Katz and Shatz , 1996 ) , the circuits where rhythmic activity plays the clearest role have central pattern generator attributes such as those that control repetitive movements linked to respiration , chewing , scratching , and walking for example . The cellular components of the mouse spinal cord CPG have been investigated through functional studies using pharmacological agents and molecular tools to target distinct neuronal subclasses . Despite the extensive characterization of this circuit , it has remained unclear what drives the pacemaker system within the CPG . Several studies have found that perturbing the function of excitatory interneurons such as V2a or V3 cells degrades the rhythm . Likewise , we found that V3 interneurons are necessary for well-coordinated rhythmicity in heterogeneous networks . In addition , we found that V3 cells , V2a interneurons , and to a limited degree motor neurons were sufficient as isolated cell populations to establish rhythmic networks . These homogenous networks displayed somewhat regular spontaneous bursting patterns , but fired with increased frequency and higher regularity in drugs that evoke central pattern generator activity . It has been suggested that the pacemaker is an emergent property of interconnected excitatory neurons within the CPG , and that the default frequency is defined by the biophysical properties of the membrane , which cyclically convert cells from on-fire to off-fire states ( Grillner , 2006; Harris-Warrick , 2010; Kiehn et al . , 2000; Marder and Bucher , 2001 ) . The inherent ability of a network to integrate neural properties across the cell population may represent one mechanism for helping to control for the intrinsic biophysical variability among neurons and thereby ensure a reliable and consistent output ( Marder et al . , 2015 ) . Our findings support this view of pacemaker control and further suggest that robustness in the system is derived from the ability of multiple excitatory spinal cord cell types to default into regular oscillatory patterns . We generated subclasses of spinal-like neurons from ES cells harboring genetic reporters for specific neuronal types functionally implicated as components of the mouse spinal cord central pattern generator . Although the ES-cell-derived neuron subtypes used in this analysis without exception displayed features similar to their in vivo counterparts , we cannot exclude that there are differences between the in vitro and in vivo cells . The de novo generated neurons were either cultured as spheres or plated and allowed to form interconnections . Remarkably , after culturing we found that these synthetic networks produced rhythmic bursts of activity that resembled the activity of the spinal cord central pattern generator . Because these ES-cell-derived neural networks produced a behaviorally relevant activity pattern , we termed these microphysical systems circuitoids to distinguish them from neurospheres defined solely by their cellular composition . Pharmacological experiments and recordings of EPSCs indicated that circuitoid activity is synaptically driven , and that the synchronized bursting of the neurons comprising each circuitoid is likely a byproduct of the extensive interconnections among the cells . In addition to synaptic connections , gap junctions may also form among neurons in circuitoids . Nevertheless , antagonists of glutamatergic synapses disrupted bursting , indicating that gap junctions alone are insufficient to produce the oscillatory patterns of activity we observed . We found no evidence for migration of neurons into particular groupings or layers within the oscillatory-circuitoids in either aggregated or plated conditions; nor did we find evidence for a specific connectome among cells , other than cell-type-specific differences in network binding strength revealed in mEPSC recordings . Instead , the oscillatory activity we observed under a variety of conditions seemed to emerge simply from the intrinsic synaptic and membrane channel characteristics of the neuron subtypes and their ability to form extensive synaptic connections with one another . Our findings indicate that the coordinated oscillatory behavior of a rhythmic network can arise without a stringent requirement for a particular circuit architecture that relies upon a specific physical arrangement of pre- and post-synaptic cells . Interestingly , motor neurons acquire rhythmic activity even when ectopically located within the spinal cord ( Hinckley et al . , 2015; Machado et al . , 2015 ) . Can circuitoids comprised of spinal-like neurons be compared to spinal cord central pattern generators ? The oscillatory networks studied in this report lacked the structured inhibition that underlies left-right and flexor-extensor coordination found in complex CPG networks such as the lumbar spinal cord ( Kullander et al . , 2003; Zhang et al . , 2014 ) . Rather , the oscillatory activity of circuitoid networks was more akin to a half-center CPG or respiratory CPG , which produce regular on-off bursts ( Garcia-Campmany et al . , 2010 ) . The similarities between circuitoids and simple CPGs were several fold . ( 1 ) Circuitoid oscillations were synaptically driven by glutamate , like CPGs ( Beato et al . , 1997; Hägglund et al . , 2010; Kiehn et al . , 2000 ) . ( 2 ) Circuitoids displayed spontaneous activity that resembled the spontaneous bursts recorded from isolated spinal cords ( Myers et al . , 2005; O’Donovan and Landmesser , 1987; Whelan et al . , 2000 ) . ( 3 ) Circuitoids became rhythmically active in drugs that evoke CPG activity ( Jiang et al . , 1999; Kudo and Yamada , 1987; Smith and Feldman , 1987; Whelan et al . , 2000 ) . ( 4 ) The frequency of bursts produced by circuitoids was in a similar range to those produced by the spinal CPG ( Talpalar and Kiehn , 2010; Whelan et al . , 2000 ) . ( 5 ) The ablation of V3 interneurons from circuitoids degraded the rhythmicity analogous to the phenotype caused by ablating V3 cells in vivo ( Zhang et al . , 2008 ) . ( 6 ) Like spinal CPGs , circuitoid rhythms could occur in the absence of inhibition ( Bracci et al . , 1996; Cowley and Schmidt , 1995; Kremer and Lev-Tov , 1997 ) . ( 7 ) In both circuitoids and CPGs blocking inhibitory V1 interneuron function reduced the burst speed ( Cowley and Schmidt , 1995; Gosgnach et al . , 2006 ) . Although there are likely important differences between circuitoids produced in vitro from ES cells and CPGs , the ability to create networks with defined cell types and numbers represents a unique approach for testing the sufficiency and instructive qualities of cell types for particular activity patterns . Functional studies of neurons within oscillatory circuits have been performed by titrating pharmacological agents; however , this global alteration of synaptic signaling strength may not be comparable to the way different cell combinations are dynamically selected for activity within a network by higher control centers . Consequently , we designed our experiments around using defined cell mixtures to study network dynamics , rather than using drug titrations , allowing us to have tight control over the cellular E/I makeup of networks , an aspect that cannot currently be controlled as precisely with in vivo manipulation . While the underpinnings of circuit-switches are poorly understood , rhythmic networks within the spinal cord are clearly capable of dynamically changing the speed and pattern of motor neuron activation to produce an elaborate repertoire of motor actions . Functional studies have established inhibitory neurons as necessary components of the circuitry that patterns motor output ( Cowley and Schmidt , 1995; Gosgnach et al . , 2006 ) ; however , it is not possible to determine if particular neuron subtypes are instructive components involved in CPG flexibility by eliminating a cell type from the system . To explore the consequences of changing the E/I ratio within oscillatory circuits , we made synthetic networks with identified neuronal subtypes mixed in controlled numbers . Base networks of purified excitatory neurons with strong interconnections , such as those comprised of V3 interneurons , reliably produced regular rhythmic bursts ( Figure 8A ) . Networks of weakly interconnected excitatory neurons , such as motor neurons , burst with more irregular patterns , but adopted a regular synchronized activity-pattern in the presence of interneurons with strong connections ( Figure 8B and C ) . Thus , under the cell mixing conditions we tested , neurons with the intrinsic ability to form strong connections establish dominance in the network . One manner in which different behaviors might emerge from a highly interconnected network like the spinal cord , is by descending inputs recruiting different proportions of inhibitory and excitatory neurons within the network and thereby changing the E/I balance . Although we could not mimic the structured recruitment of cell types by descending inputs , we could investigate the consequence of changing the E/I balance of networks by mixing different combinations of excitatory and inhibitory neuron subtypes . V1 interneurons increased the burst frequency of strongly interconnected V3 networks in proportion to the E/I cell ratio ( Figure 8D ) , suggesting V1 interneuron influence on network speed can occur upstream of motor neurons . When V1 inhibitory neurons were combined with weakly connected motor neurons they had little influence on burst frequency , but they caused sub-networks to emerge in proportion to the E/I cell ratio ( Figure 8E ) . Although our findings do not explain how the activity of different motor pools is coordinated , our results support the possibility that the selective recruitment of increasing numbers of V1 cells could switch motor pools from a coupled ( synchronous ) to an uncoupled ( segmented ) state in order to facilitate the differential recruitment of muscles during complex behaviors ( Figure 8E ) . In tripartite circuits with weakly- and strongly-connected excitatory neurons in combination with inhibitory neurons , we found that the entire network adopted a synchronous pattern of activity whose frequency increased as the number of inhibitory neurons increased ( Figure 8F ) . The selective recruitment of V1 cells into the CPG may be facilitated by the large genetic diversity that has been uncovered within this cell group ( Bikoff et al . , 2016; Francius et al . , 2013 ) . Our observations suggest that flexibility within rhythmic circuits can be achieved by constructing networks with neurons that have unequal synaptic interactions , and by regulating the activity-ratio of excitatory and inhibitory neurons . Together these features are sufficient to control the speed and segmentation of bursting . It has not escaped our notice that circuitoids could be used to study neurological diseases that affect circuits and might represent transplantable modules for nervous system repair .
The generation and genotyping of the En1:Cre , Chx10:Cre ( Chx10 is encoded by the Vsx2 gene ) , Hb9:GFP , and Sim1:Cre alleles in mice has previously been described ( Azim et al . , 2014; Gosgnach et al . , 2006; Lee et al . , 2004; Sapir et al . , 2004; Zhang et al . , 2008 ) . The Gt ( ROSA ) 26Sortm9 ( CAG-tdTomato ) ( R26/C:LSL:Tomato ) and Gt ( ROSA ) 26Sortm1 ( DTA ) ( R26:LSL:DTA ) lines were obtained from Jackson Laboratory ( 007905 and 010527 , respectively ) . Transgenic CAG:GCaMP3 mice were generated by using restriction enzymes to cleave the promoter+reporter fragments from the bacterial plasmid , and injecting the purified DNA into mouse oocyte pronuclei . After microinjection , founders were genotyped by PCR with the GFP primers and screened for ubiquitous presence of GCaMP3 . All ES cell lines were derived as novel lines for the experiments in this paper . Blastocysts were flushed 3 . 5 days after fertilization using M2 media ( MR-015-D , Millipore ) . Each individual blastocyst is placed in one well of a 96-well plate containing primary mouse embryonic fibroblasts ( pMEF - GlobalStem ) with 2i media ( SF016-100 , Millipore ) . After 5 days of incubation , the 2i media is aspirated and each hatched blastocyst is dissociated using accutase and passaged to one well of a 24-well plate with pMEF and 2i media . Colonies are visible after 1 or 2 days . Every second passage with accutase decreases the concentration of 2i media from 100% to 75% , 50% , 25% and finally to 0% with FCS media [Knockout DMEM ( Life Technologies , now Thermofischer Scientific , Waltham MA ) , 1X HEPES ( Life Technologies ) , 1X non-essential amino acids ( Life Technologies ) , 200 mM L-glutamine ( Life Technologies ) , 10% ES-qualified fetal bovine serum ( Millipore ) , 0 . 1 mM β-mercaptoethanol ( Sigma ) , 1 , 000–2 , 000 units of leukemia inhibitory factor ( LIF ) ( Calbiochem ) , 1X Antibiotic-Antimycotic ( Life Technologies ) ] making up the other fraction . After colonies were established , ES cells were passaged as needed using 0 . 25% trypsin ( Life Technologies ) and plated into FCS media . At times , 2x the concentration of LIF was used to improve ES cell colony morphology . All ES cell lines were genotyped by Transnetyx using the same protocols to genotype the mouse lines from which they were derived . All lines were negative for mycoplasma contamination , as verified with a PCR screen . ES cells are differentiated in suspension in 10 cm petri dishes . 1 × 106 dissociated ES cells are resuspended in 10 ml ADFNK media [Advanced D-MEM/F-12 ( Life Technologies ) : Neurobasal medium ( Life Technologies ) ( 1:1 ) , 10% Knockout Serum Replacement ( Life Technologies ) , 200 mM L-Glutamine ( Life Technologies ) , and 0 . 1 mM β-mercaptoethanol ( Sigma ) ] . Two days later , embryoid bodies ( EBs ) were allowed to settle to the bottom of a 15-mL conical tube . Media was aspirated , and a third to a tenth of the EBs were transferred to a new 10-cm plate with fresh ADFNK media that was supplemented with 1 μM all-trans retinoic acid ( RA , Sigma ) and 5 nM to 1000 nM smoothened agonist ( SAG , Calbiochem ) . Two days later , freshly supplemented media was exchanged ( Peljto et al . , 2010; Wichterle and Peljto , 2008; Wichterle et al . , 2002 ) . For DAPT application , following the 6 days of differentation , 5 μM N-[N- ( 3 , 5-difluorophenacetyl-l-alanyl ) ]- ( S ) -phenylglycine t-butyl ester ( DAPT; Sigma ) , a Notch inhibitor , was applied for four days prior to FACS . On day 6 , if to be used for sorting , heterogeneous neurospheres were maintained in non-supplemented ADFNK media . To maximize sorting efficiency ( greatest fluorescent+ population and ease of dissociation ) for generation of pure or mixed circuitoids through FACS , Hb9:GFP ES cell lines were sorted on days 6–7 and all Cre-dependent tomato+ lines were sorted on days 10–11 . If used for imaging , heterogeneous neurospheres were switched to a neuronal media [Neurobasal medium ( Life Technologies ) , 2% ES-qualified fetal bovine serum ( Millipore ) , 200 mM L-Glutamine ( Life Technologies ) , 1X B-27 supplement ( Life Technologies ) , L-glutamic acid ( Sigma ) , 1X Antibiotic-Antimycotic ( Life Technologies ) , 10 ng/ml Human Brain Derived Neurotrophic Factor ( BDNF , Peprotech 450–02 ) and 10 ng/ml Recombinant Murine Glial-Derived Neurotrophic Factor ( GDNF , Peprotech 450–44 ) ] . Half the media was exchanged three times a week until activity was recorded . Activity of these heterogeneous networks , unless otherwise noted , was recorded 15–17 days from ES cells . Neurospheres , 6–11 days from ES cells , were dissociated ( Papain , Worthington ) , and then counted with a BD FACScan to determine the percentage of neurons composing neurospheres at different SAG concentrations . For sorting and generating purified networks , the BD FACSDIVA and BD Influx were used to sort neurons into low-adherent , u-bottomed 96-well dishes ( Corning 7007 ) . To derive cortical astrocytes for reaggregated de novo networks we used a similar protocol to McCarthy and DeVellis procedure described previously , adapted for mice ( Ullian et al . , 2001 ) . P0-P3 mouse cortices were dissected and dissociated ( Papain , Worthington ) . They were grown in a T-75 flask for 3–4 days with AGM media [DMEM + GlutaMAX ( Life Technologies ) , 10% ES-qualified fetal bovine serum ( Millipore ) , 1 mM Na-Pyruvate ( Life Technology ) , 5 μg/ml insulin ( Sigma I1882 ) , 5 μg/ml n-acetylcysteine , 1 μM hydrocortisone ( Sigma – H0888 ) , 1X Antibiotic-Antimycotic ( Life Technologies ) ] . Following a 1X PBS wash , contaminating cells were shook off and then the media was exchanged . 1–3 days later , once confluent , 10 μM AraC ( Sigma 1768 ) was added and subsequently removed 2 days later with three washes with 1x PBS . To use , cells were dissociated in 0 . 05% Trypsin ( Life Technologies ) . For astrosphere formation 50 , 000 astrocytes were placed in each well of a 96-well ultra-low adherent u-bottomed plate ( Corning – 7007 ) and spun at 300 g to aggregate the cells . If compact spheres were not formed 24 hr later , light trituration was used to break the aggregate apart , and then the plate was respun . For plated assays , 100 , 000 astrocytes per 100–200 µl were plated onto the glass coverslip of the 35 mm dish ( Corning 354077 ) for 2 hr . After cell adhesion , an addition 2 ml of AGM media was added to the dish . At times , to aid in adherence , prior to plating the astrocytes , dishes were recoated with 10 µg/ml poly-d-lysine ( Sigma P6407 ) or 10 µg/ml laminin ( Life Tech 23017–015 ) . Neurons were plated onto astrospheres or confluent astrocytes 2–4 days after astrocyte dissociation . To generate reaggregated circuitoids , differentiated heterogeneous neurospheres were dissociated between days 6 and 11 and specific neuronal subtypes were sorted directly into a well of a 96-well ultra-low adherent u-bottomed plate ( Corning 7007 ) . After sorting , these plates were spun to pellet the neurons . FACS sheath was removed without disrupting the loose pellet at the bottom of the well and neuronal media was added back . two more washes were conducted , combining wells if necessary ( some experiments would have caused wells to overflow during FACS if neurons for one network were not split between multiple wells ) . For sphere assays , the neurons were resuspended and transferred into a well of a u-bottomed 96-well plate that was already filled with an astrosphere . These plates were now spun to increase contact of neurons and astrosphere . The following day , light trituration was used to remove any debris from the main reaggregated sphere . A few smaller satellite spheres may have also formed , so the plate was spun once more . 24 hr later , all the neurons and astrocytes formed one coherent sphere . All the media were carefully removed from the cells and new media was added . Unless otherwise noted , the circuitoids were incubated for 3 weeks after FACS before their network activity was imaged , with half of their media being exchanged three times a week . For plated networks , the AGM from the 35 mm dish was aspirated sufficiently to dry the plastic surrounding the 10 mm coverslip , and then media on the coverslip was aspirated . Neurons for each network were resuspended in 100–200 µl of neuronal media and plated onto the confluent astrocyte layer . After 2 hr the neurons had adhered and 2 ml of neuronal media were added to the 35 mm dish . The plated networks were incubated for 2 weeks after FACS before their network activity was imaged , with half of their media being exchanged three times a week . During recordings , samples were perfused in ACSF ( 128 mM NaCl; 4 mM KCl; 21 mM NaHCO3; 0 . 5 mM NaH2PO4; 1 mM MgSO4; 30 mM D-glucose; and 2 mM CaCl2 ) bubbled with a 95/5/% O2/CO2 mixture . Calcium signal was recorded using either a ubiquitously expressed GCaMP3 ( tgCAG:GCaMP3 ) or Oregon Green 488 BAPTA-1-AM ( Life Technologies ) . For dye application , Oregon Green 488 BAPTA-1-AM ( Life Technologies ) was applied at 10 μM in ACSF for 1 hr in a 37°C incubator . Dye was washed out with perfusion of ACSF for 15 min prior to recording . Unless otherwise stated , image series were acquired on an upright epifluorescent Olympus microscope ( BX51WI ) using a Hamamatsu C9100‐13 camera and ImageJ plugin: μManager software ( RRID:SCR_000415 ) , capturing 20 frames/second at 128 × 128 using a 4 × 0 . 28 NA air objective ( Olympus ) with a 0 . 63x camera mount , and an X-Cite exacte light source at 2% power . For Figures 6D–F and and 7E–G cellular resolution data of plated networks was acquired at 512 × 512 using a 20 × 1 . 0 NA water-immersion objective ( Olympus ) with a 0 . 63x camera mount . For electrical , population recordings , activity was measured using a suction electrode with a multiclamp 700B amplifier , filtered at 300 Hz to 1 kHz . For whole-cell patch clamping , cells were visualized for whole-cell patch recordings using a BX51WI ( Olympus ) microscope equipped to allow both differential interference contrast ( DIC ) and epifluorescence imaging . Recordings were performed using a Multiclamp 700B amplifier ( Molecular Devices ) . Signals were filtered at 6 kHz , sampled at 50 kHz through an Axon Digidata 1550A interface device ( Molecular Devices ) and recorded using Clampex 10 software ( Molecular Devices ) . Electrodes were pulled using a P-97 flaming-browning micropipette puller ( Sutter Instruments ) from thick-walled borosilicate glass GC150F capillaries ( Harvard Apparatus ) to a resistance of 2–5 MΩ . All cells were voltage-clamped at −60 mV . Series resistance of 4–10 MΩ was compensated by 20–40% and recordings were abandoned if it increased by more than 20% . During recording , cell cultures were continuously perfused at 5–8 ml/min with ACSF bubbled with a 95/5/% O2/CO2 mixture . The pipette solution consisted of ( in mM ) 140 Cs-gluconate , 4 CsCl , 2 CaCl2 , 10 HEPES , 5 EGTA , 2 MgATP , 3 QX-315 Br , pH 7 . 3 with CsOH , and osmolarity of 290–310 mOsm . Burst activity was recorded in the first cell of each cell culture prior to the wash in of 500 nM TTX . Miniature post-synaptic potentials were detected using WinEDR 3 . 2 . 4 ( Strathclyde Electrophysiology Software ) and analyzed using Clampfit 10 . 2 ( Molecular Devices ) with further analysis completed in R . Drugs used were in the following final concentrations: 20 μM N-Methyl-DL-aspartic acid ( NMA Sigma M2137 – discontinued , 10 μM NMDA M3262 equivalent ) , 40 μM serotonin creatinine sulfate monohydrate ( 5-HT , Sigma H7752 ) , 50 µM dihydro-β-erythroidine hydrobromide ( DHβE , Tocris 2349 ) , 10 μM mecamylamine hydrochloride ( MLA , Tocris 2843 ) , 10 μM CNQX disodium salt ( Tocris 1045 ) , 1 μM strychnine hydrochloride ( Sigma S8753 ) , and 10 μM picrotoxin ( Sigma P1675 ) , 500 nM tetrodotoxin citrate ( TTX , Cayman Chemicals 14964 ) . Neurospheres and circuitoids were fixed in 4% PFA for an hour and then washed in 1x PBS and prepared for cryosectioning . Cryosections ( 60 µm ) were stained in 1x PBS containing 1% BSA and 0 . 1% triton with NeuroTrace 640/660 Deep-Red Fluorescent Nissl Stain ( Life Technologies N-21483 , RRID:AB_2572212 ) , DAPI , rabbit anti-PSD95 ( 1:1000 , Life Technologies , RRID:AB_2533914 ) , guinea pig anti-VGLUT2 ( 1:3000 , Millipore , RRID:AB_1587626 ) , goat anti-dsRed ( 1:500 , Santa Cruz ) , rat anti-RFP ( 1:1000 , Chromotek , RRID:AB_2336064 ) , and rabbit anti-GFAP ( 1:500 , Dako , RRID:AB_10013482 ) . Imaging was conducted on an Olympus FV1000 confocal . Days 6–8 neurospheres and e12 . 5 spinal cords , micro-dissected ( Leica stereomicroscope ) from mice , were dissociated with papain ( papain dissociation kit , Worthington Biochemical ) . After 45 min , tissue was triturated and centrifuged at 1000 rpm for 5 min . Dissociated cells were resuspended in 1:1 Neurobasal:DMEM/F12 ( without phenol red ) with 3% Horse Serum ( Invitrogen ) and DNase ( Worthingon Biochemical ) . Before sorting , cells were passed through a 35-µm cell strainer ( 08-771-23 , BD Falcon ) . Sorting was conducted on a Becton Dickinson FACS Vantage SE DiVa using Coherent Sapphire 488 nm and 568 nm solid state lasers ( 200 mW ) . Cells were collected directly into miRvana RNA lysis buffer and stored at −80C . RNA was isolated using the miRvana miRNA isolation kit ( Ambion AM1560 ) or RNeasy Mini Kit ( Qiagen ) . Each in vivo sample is a pool of isolated cells from one to three spinal cords , as necessary , to obtain sufficient RNA ( quantified by Agilent TapeStation ) for RNA sequencing . mRNA sequencing libraries were prepared using the Illumina TruSeq RNA Library Preparation Kit ( v2 ) according to the manufacturer’s instructions . Briefly , RNA with polyA+ tails was selected using oligo-dT beads . Then , mRNA was fragmented and reverse-transcribed into cDNA . cDNA was end-repaired , index adapter-ligated and PCR amplified . Nucleic acids were purified with AMPure XP beads ( Beckman Coulter ) after each step . Libraries were then quantified , pooled , and sequenced using either the Illumina HiSeq 2500 or Illumina HiSeq 2000 platforms at the Salk NGS Core and Beijing Genomics Institute . Sequencing libraries were either 50 bp single-end or 100 bp paired-end . To help control library type bias in the quantifications , raw FASTQ reads were trimmed to 50 bp and processed as single-end . Reads with <15 minimum average base quality were filtered out . Trimmed and filtered reads were quantified using kallisto against the most recent release of the Refgene annotation for mm10 ( downloaded from UCSC Genome Browser ) ( Bray et al . , 2016 ) . Gene level TPM expression values were obtained by summing isoform level TPM at each gene loci . Data were exported from ImageJ ( RRID:SCR_003070 ) using the ROI manager after manually drawing regions of interest around spheres , or for plated networks , ROIs were drawn around individual neurons ( cellular analysis ) , or the full or partial field of view ( network analysis ) . Burst detection was carried out in Igor Pro ( RRID:SCR_000325 ) with TaroTools , Dr Taro Ishikawa , ( https://sites . google . com/site/tarotoolsregister ) or a custom pipeline generated in R . To analyze network complexity , movies were converted to 8 bit grey-scale and resized to 128 × 128 pixels and exported as TIFF stacks from ImageJ . TIFF stacks were imported into R for analysis . Stacks were transformed into 196 ‘signals’ where each signal is the averaged intensity vs time from an 11 × 11 pixel oval shaped ROI . ROIs were taken at nine pixel spacing to allow for a small amount of overlap between adjacent ROIs . The size of the ROIs was selected manually based on training data to maintain good detail while still averaging out imaging noise . Signals were smoothed with a 1 s wide running mean filter and delta-f / f transformed . The full set of ROIs from a movie was then PCA transformed . The principal components were separated into ‘signal’ and ‘noise’ using the scree-plot method of significant component estimation ( finding the component number at the elbow of the plot ) . Based on training data , we occasionally found that information was lost by selecting the component at the elbow of the scree-plot so , instead , selected two past the elbow . The significant components were recombined to form the signal matrix and all non-significant components were recombined to form the noise matrix . The sum of these two matrices is equal to the original input to PCA . For each signal , the standard deviation of the noise version was used to threshold the signal version for burst location calling . Burst locations were identified as rising edges in the signal version of an ROI with change in intensity greater than approximately 1 . 96 times the standard deviation of the noise version of the same ROI . This threshold was selected manually based on training data and gave a good balance between false-positive and false-negative burst calling . Change in intensity was measured from the base to peak of the rising edge of the burst . Burst positions from all ROIs within a movie were cataloged to create a set of distinct burst positions within ±1 s windows ( the burst catalog ) . Each ROI was then re-described as a sequence of the cataloged bursts . We then filtered out bursts from the catalog that appeared in less than 4% of the ROIs . This threshold was set manually based on training data to balance out Type I and Type II error as fair as possible and then used for all primary analysis . After filtering , the ROIs were re-described again . This re-described set of ROIs represents all robust bursting activity within a single movie . To quantify the complexity , taken as a deviation from the state of all ROIs having identical activity , we built a directed graph of the burst sequences within a movie . Briefly , a graph is a type of data structure with one or more nodes connected by edges . They are often used to describe networks . If the edges are assigned a direction then the graph is a directed graph . In our case , the nodes are the cataloged bursts . Edges were added to the graph between any two bursts that occurred adjacently in time in any ROI . The direction of the edge is in the direction of occurrence in time . If , in one ROI , burst Y followed burst X and was adjacent to burst X ( no bursts occurred between them ) then an edge would be added connecting from X to Y . From the point of view of a node in the graph , edges may be incoming or outgoing . The foregoing connection of X to Y has a single edge . From the point of view of node X it is an outgoing edge while from the point of view of Y it is an incoming edge . Complexity of a movie was then defined as the average of both the ratio of nodes with more than one outgoing edge and ratio of nodes with more than one incoming edge . If all ROIs in a movie have the same burst sequence then , in terms of the graph , each node would only be connected to one other node resulting in a baseline score of 1 for the movie . When the ROIs share some bursts , but not all , then the nodes of the graph begin to have more than one incoming or outgoing edge , which increases the final complexity score . We did not consider a case for ROIs that do not share any bursts because we did not observe any such cases in our data . The complexity statistic was found to be non-normal across the different testing groups and to have unequal variances so the two-sample Kolmogorov-Smirnov test was used to perform statistical tests between groups . All other data were compared using a student t-test . | The nerve cells or neurons within an animal’s nervous system connect with one another like the wires in a complex circuit . Each neuron can send and receive signals and a major challenge in neuroscience is to understand how these circuits of neurons behave . To do this , researchers often use genetic tools and computer modeling to map the connections between the cells in a nervous system . However , it remains difficult to predict how an input signal will appear at the output after it passes through a network made of different types of neuron . Brains contain many networks of interconnected neurons . Some of these networks send signals with a rhythmic pattern and typically drive repetitive movements such as breathing and walking . The networks are called central pattern generators ( or CPGs for short ) . They contain both excitatory and inhibitory neurons and can generate rhythmic activity without any additional input . Nevertheless CPGs are not rigid , but can flexibly control when and how fast the muscles are activated to suit the animal's needs . It is thought the circuits are flexible because of the way excitatory and inhibitory neurons interact , but it is not known how these interactions define the behavior of the circuit . Sternfeld et al . have now developed a new method to examine how the neurons that make up a circuit influence its activity . First , embryonic stem cells from mice were coaxed to develop into a number of subtypes of both excitatory and inhibitory neurons in the laboratory . These neurons were used to grow networks of neurons in a dish , named “circuitoids” . The precise combination of subtypes of neuron was deliberately varied between each circuitoid , and Sternfeld et al . then studied how the different circuitoids behaved . Several subtypes of excitatory neurons showed rhythmic bursts of activity , just like simple CPGs . Moreover , the ratio of excitatory to inhibitory neurons in the circuitoids was critical for establishing how fast and synchronized the bursts of activity were across the network . It is possible that the brain also uses this simple strategy of varying the ratio of excitatory to inhibitory neurons in circuits of neurons to generate complex , yet highly flexible , circuits with rhythmic activity . Further work will be needed to test this idea . Finally , other researchers will hopefully be able to use this new approach to construct circuitoids and learn more about how the brain generates and controls rhythmic activity . It might also be possible to one-day transplant similar circuitoids into people to repair injured or diseased parts of a nervous system , or use circuitoids that resemble specific neurological disorders to screen for new treatments . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2017 | Speed and segmentation control mechanisms characterized in rhythmically-active circuits created from spinal neurons produced from genetically-tagged embryonic stem cells |
Gliomas are highly malignant brain tumors with poor prognosis and short survival . NAD+ has been shown to impact multiple processes that are dysregulated in cancer; however , anti-cancer therapies targeting NAD+ synthesis have had limited success due to insufficient mechanistic understanding . Here , we adapted a Drosophila glial neoplasia model and discovered the genetic requirement for NAD+ synthase nicotinamide mononucleotide adenylyltransferase ( NMNAT ) in glioma progression in vivo and in human glioma cells . Overexpressing enzymatically active NMNAT significantly promotes glial neoplasia growth and reduces animal viability . Mechanistic analysis suggests that NMNAT interferes with DNA damage-p53-caspase-3 apoptosis signaling pathway by enhancing NAD+-dependent posttranslational modifications ( PTMs ) poly ( ADP-ribosyl ) ation ( PARylation ) and deacetylation of p53 . Since PARylation and deacetylation reduce p53 pro-apoptotic activity , modulating p53 PTMs could be a key mechanism by which NMNAT promotes glioma growth . Our findings reveal a novel tumorigenic mechanism involving protein complex formation of p53 with NAD+ synthetic enzyme NMNAT and NAD+-dependent PTM enzymes that regulates glioma growth .
Glioma is the most common intrinsic tumor of the central nervous system ( CNS ) and derives from the neoplastic glial cells or neuroglia ( Goodenberger and Jenkins , 2012 ) . Based on pathological criteria , gliomas are classified from WHO grade I to IV , among which the high-grade gliomas generally have a much poorer prognosis ( Wesseling and Capper , 2018 ) . Several major cellular signaling pathways associated with glioma have been well studied , including RTK/Ras/PI3K , p53 , and RB signaling pathways ( Cancer Genome Atlas Research Network , 2008 ) . In addition , metabolism factors , such as IDH1/2 , were found to play important roles in glioma ( Yan et al . , 2009 ) . IDH1 is an enzyme of tricarboxylic acid ( TCA ) cycle in glucose metabolism and the main producer of NADPH ( Molenaar et al . , 2014 ) . However , drugs targeting these pathways showed a limited clinical response , indicating a critical need for the mechanistic understanding of the metabolic requirement for glioma tumorigenesis . Nicotinamide adenine dinucleotide ( NAD+ ) is an essential signaling cofactor that regulates cancer metabolism through its co-enzymatic function for many bioenergetic pathways , including glycolysis , TCA cycle , and oxidative phosphorylation ( Hanahan and Weinberg , 2011 ) . Multiple processes associated with NAD+ signaling are dysregulated in cancer , including DNA repair , cell proliferation , differentiation , and apoptosis ( Chiarugi et al . , 2012 ) . Inherited polymorphisms and epigenetic repression of DNA damage repair genes are significantly correlated with the risk of gliomas , indicating that abnormal DNA damage repair plays important roles in glioma formation and progression ( Chen et al . , 2010; Qi et al . , 2017 ) . One of the key initiation events of DNA damage response is poly ( ADP-ribose ) polymerase ( PARP ) -mediated poly ( ADP-ribosyl ) ation ( PARylation ) , the main process that consumes nuclear NAD+ ( Amé et al . , 2004 ) . Moreover , NAD+-dependent SIRTs-mediated deacetylation regulates many oncogenes and tumor suppressor genes in cancer cells ( Brooks and Gu , 2009 ) . Consistently , a high level of NAD+ is observed in gliomas ( Reddy et al . , 2008; Tso et al . , 2006 ) , and 90% of gliomas are susceptible to NAD+ depletion ( Tateishi et al . , 2015 ) . Therefore , it is critical for rapidly proliferating glioma cells to replenish the NAD+ pool for survival . In the past years , targeting NAD+ metabolism has been considered for cancer therapy , and most efforts have been focused on nicotinamide phosphoribosyltransferase ( NAMPT ) , the rate-limiting enzyme of the NAD+ salvage pathway , whose expression is increased in multiple types of cancer ( Garten et al . , 2015; Lucena-Cacace et al . , 2018; Ohanna et al . , 2018; Pylaeva et al . , 2019 ) . Disappointingly , several clinical trials of NAMPT inhibitors have failed due to low efficacy and high toxicities ( Sampath et al . , 2015 ) , which demands the urgent consideration of an alternative target in the NAD+ metabolic pathway . Nicotinamide mononucleotide adenylyltransferase ( NMNAT ) , the last enzyme in the NAD+ salvage synthetic pathway , has recently emerged as a potential candidate ( Chiarugi et al . , 2012 ) . NMNAT has three isoforms in mammals with distinct subcellular localizations: NMNAT1 , in the nucleus; NMNAT2 , in the cytosol; and NMNAT3 , in the mitochondria ( Berger et al . , 2005 ) . Dysregulations of both NMNAT1 and NMNAT2 have been implicated in cancer . For example , NMNAT1 is considered a poor prognostic marker for renal cancer ( Uhlén et al . , 2015; Uhlen et al . , 2017 ) . Decreased NMNAT1 expression leads to epigenetic silencing of tumor suppressor genes ( Henderson et al . , 2017 ) . Inhibition of NMNAT1 delays DNA repair and increases rRNA transcription ( Song et al . , 2013 ) . In colorectal cancer , NMNAT2 upregulation correlates with the cancer invasive depth and TNM stage ( Cui et al . , 2016; Qi et al . , 2018 ) . In non-small cell lung cancer ( NSCLC ) , NMNAT2 enzymatic activity is upregulated by SIRT3-mediated deacetylation process or p53 signaling ( Li et al . , 2013; Pan et al . , 2014 ) . Moreover , the depletion of NMNAT2 inhibits cell growth indirectly by reducing glucose availability in neuroblastoma cells ( Ryu et al . , 2018 ) . These observations indicate the regulatory link between compartmentalized NAD+ synthesis and cellular metabolism and rapid cancer cell growth , and further underscore the potential of NMNAT as a viable alternative target in NAD+ synthetic pathway , given their aberrant regulation and critical role in cancer metabolism . In this report , to address the knowledge gap regarding the role of NMNAT in glioma , we adapted an in vivo glial neoplasia in Drosophila ( Read et al . , 2009 ) and discovered a genetic requirement for NMNAT in glioma growth . Combined with human glioma cell culture models , we characterized the mechanism of NMNAT in gliomagenesis . Our results identified the upregulation of enzymatically active NMNAT as an essential metabolic regulator for promoting gliomagenesis and revealed that NMNAT-sustained PARylation and deacetylation of p53 results in suppression of apoptosis , a key tumor-inhibitory response .
The Ras/Raf/ERK signaling cascade is one of the most conserved pathways both in Drosophila and human , and a major component of the MAP kinase signaling stress-response network ( Morrison , 2012 ) . RAS mutations are the most commonly found oncogenic alteration in human cancers , most frequently observed in KRAS ( 85% ) , and to a lesser degree in NRAS ( 12% ) and HRAS ( 3% ) ( Simanshu et al . , 2017 ) . Upregulated RAS and mutant RAS have been detected in gliomas ( Arvanitis et al . , 1991; Guha et al . , 1997; Knobbe et al . , 2004; Rajasekhar et al . , 2003 ) , and activation of Ras has been used to model human glioma in Drosophila ( Read , 2011; Read et al . , 2009 ) . Ras oncogene at 85D ( Ras85D ) is the Drosophila orthologue of human RAS . The constitutively active Ras85D mutation ( G12V ) , Rasv12 , has been suggested to be analogous to human oncogenic RAS mutation and used to induce tumor ( Barbacid , 1987; Wu et al . , 2010 ) . We established a Drosophila glial neoplasia model by overexpressing Rasv12 in glial cells , driven by the pan-glial driver repo-GAL4 ( Read et al . , 2009 ) . Green fluorescent protein ( GFP ) was co-expressed as a reporter to mark the Ras expressing cells . Under normal conditions , the Drosophila CNS is wrapped by perineurial , subperineurial , and ensheathing glia ( Freeman , 2015 ) . Powered with high-resolution quantitative brain morphology analysis ( Brazill et al . , 2018b ) , we analyzed glial neoplasia tissue using three criteria , ( i ) tissue double-positive for GFP and endogenous Repo expression; ( ii ) tissue mass consists of multiple layers of glia of at least 400 cells , and ( iii ) tissue mass volume greater than 12 . 4 × 103 μm3 ( Figure 1—figure supplement 1 ) . When Rasv12 was expressed in glia , numerous glial neoplasia tissues marked by GFP and Repo in the brain and ventral nerve cord ( VNC ) were detected as early as 100 hr after egg laying ( AEL ) , and the volumes of glial neoplasia increased with age ( Figure 1A , B and G ) . The brain tumors caused early lethality in pupal stage and greatly reduced survival rate ( Figure 1H ) . Notably , compared with the normal brain ( Figure 1C and E ) , we found significantly increased endogenous NMNAT in glial cells at both 100 and 150 hr AEL . NMNAT was most prominently increased in the nuclear region ( Figure 1D and F ) , suggesting a possible role for NMNAT1 , the nuclear isoform , in Rasv12-induced glial neoplasia formation in Drosophila . To determine whether increased NMNAT is required for glial neoplasia development , we used the RNAi approach to downregulate NMNAT expression in Rasv12-induced glial neoplasia cells ( Brazill et al . , 2018a ) . NMNAT RNAi-mediated knockdown in Rasv12 overexpression cells reduced NMNAT expression level to around 36% of wild-type flies ( Figure 2—figure supplement 1 ) . Interestingly , knocking down Nmnat drastically reduced both the volume and the number of individual Rasv12-expressing glial cells in the brain and VNC at 100 hr AEL ( Figure 2A , C and D ) , demonstrating a strong antitumor effect of NMNAT inhibition in vivo . We further analyzed RNAi-mediated knockdown of NMNAT in normal glial cells ( without Rasv12 expression ) and found no growth inhibition ( Figure 2—figure supplement 2 ) , suggesting NMNAT is not essential for normal cell survival . Next , we tested whether upregulating NMNAT can promote glial neoplasia formation and growth . Drosophila has one Nmnat gene , expressing two protein isoforms through alternative splicing , a nuclear isoform Nmnat-PC and a cytosolic isoform Nmnat-PD . The Nmnat-PC ( nuclear ) and Nmnat-PD ( cytoplasmic ) isoforms share similar enzymatic activity but are differentially regulated under stress conditions ( Ruan et al . , 2015 ) . In Rasv12-induced glial neoplasia , dramatically increased Nmnat is mainly observed in the nuclear region ( Figure 1D and F ) , likely to be the Nmnat-PC ( nuclear ) isoform . To further evaluate the compartment-specific role of NMNAT during glial neoplasia formation , we generated flies expressing Rasv12 together with Nmnat-PC ( nuclear ) or Nmnat-PD ( cytoplasmic ) . Consistent with the previous report , Nmnat-PC ( nuclear ) is highly enriched in the nucleus and colocalizes with the nuclear marker Repo , while Nmnat-PD is predominantly cytoplasmic ( Ruan et al . , 2015 ) . Interestingly , overexpression of Nmnat-PC ( nuclear ) , but not Nmnat-PD ( cytoplasmic ) , significantly increased the total volumes of glial neoplasia ( Figure 2B and C ) , while the number of glial neoplasia showed no significant difference among the groups ( Figure 2D ) . The lethality of the flies ( Figure 2E ) was positively correlated with glial neoplasia size and overexpression of Nmnat-PC ( nuclear ) significantly increased the lethality . To determine whether the enzyme activity of NMNAT is required for glial neoplasia tumorigenesis , we generated flies expressing an enzyme inactive mutant Nmnat-PC ( nuclear ) isoform ( PCWR ) where two key residues for substrate binding were mutated ( Figure 2—figure supplement 3; Zhai et al . , 2006 ) . We found that Nmnat-PCWR ( nuclear ) overexpression did not significantly affect glial neoplasia volumes or numbers or survival outcome when compared to the control ( Figure 2C–E ) . These results suggest that nuclear enzymatically active NMNAT promoted glial neoplasia growth . We next examined the function of NMNAT in human glioma cell proliferation , specifically human NMNAT1 ( nuclear ) and NMNAT2 ( cytoplasmic ) ( Berger et al . , 2005 ) . Since approximately 51% of glioma are mutated for p53 , we included two glioma cell lines with different p53 status , U87MG with wild-type p53 , and T98G with a gain-of-function M237I mutation ( Van Meir et al . , 1994 ) , to dissect common mechanisms of the role of NMNAT in glioma cell growth . We determined NMNAT1 and NMNAT2 protein levels in human glioma cells and normal astroglia cells ( SVG p12 ) . Compared to SVG p12 cells , NMNAT1 and NMNAT2 are increased in both glioma cells T98G and U87MG ( Figure 3—figure supplement 3 ) . Next , we manipulated the expression of NMNAT by siRNA-mediated knockdown and plasmid-mediated overexpression in T98G cells and monitored real-time cell growth using the xCELLigence platform ( Ke et al . , 2011 ) . Interestingly , we found T98G cell proliferation was drastically inhibited when either NMNAT1 or NMNAT2 was knocked down ( Figure 3A ) . This observation was confirmed and extended in an MTT assay ( van Meerloo et al . , 2011 ) , where NMNAT1 or NMNAT2 knockdown reduced cell proliferation ( Figure 3—figure supplement 1 ) . In contrast , overexpressing NMNAT1 or NMNAT2 promoted cell growth ( Figure 3D ) . Moreover , we used a plate colony formation assay to determine clonogenic survival ( Franken et al . , 2006 ) , and found that knockdown of NMNAT1 or NMNAT2 reduced the colony numbers of T98G , while overexpression of NMNAT1 or NMNAT2 increased colony formation ( Figure 3B–F ) . These results are consistent with the genetic dependency on NMNAT observed in the fly glial neoplasia models , suggesting the conservation of NMNAT function in promoting glioma cell growth and proliferation . To further determine whether NMNAT is involved in glioma cell survival , we carried out a flow cytometric apoptosis detection assay through flow cytometry . We transfected siRNA targeting NMNAT into T98G cells and then analyzed Annexin V-FITC/PI by flow cytometric 72 hr post-transfection . Interestingly , we found that knockdown of NMNAT , at the knockdown rate of 40–50% for NMNAT1 or at 20–30% for NMNAT2 , significantly increased the percentage of apoptotic cells , including early apoptotic and late apoptotic cells ( Figure 3G and H ) . We also examined the cell cycle distribution of these cells . The cell cycle assay showed G2/M phase was only slightly increased in T98G cells with NMNAT1 knockdown ( Figure 3—figure supplement 2 ) . These results suggest that NMNAT promotes glioma cell growth mainly through inhibiting cell apoptosis . The cysteine-dependent proteases ( caspases ) are activated by upstream proteins to mediate apoptosis ( Kurokawa and Kornbluth , 2009 ) . Caspase-3 is the main effector protease cleaving a large number of substrates during apoptosis . Previous studies revealed that nuclear translocation and accumulation of caspase-3 play a critical role in the progression of apoptosis ( Prokhorova et al . , 2018 ) . The caspase-mediated pathway is highly conserved in mammalian and Drosophila ( Fuchs and Steller , 2011; Shi , 2001; Figure 4—figure supplement 1A ) . To validate the role of caspase pathway in Drosophila glial neoplasia , we examined tumor growth in flies with downregulation of DCP1 , the homolog of mammalian caspase-3/7 . In these flies , glial neoplasia volume was significantly increased ( Figure 4—figure supplement 1B and C ) , suggesting the important role of caspase-mediated apoptosis in preventing Drosophila glial neoplastic growth . To test whether NMNAT regulates this process , we determined the localization and protein levels of caspase-3 in the glial neoplasms with overexpression of different Nmnat isoforms . We used Repo and DAPI to label the nuclei region and observed a significant decrease of caspase-3 levels in glial neoplasms that overexpress Nmnat-PC ( nuclear ) , compared with those overexpressing lacZ , Nmnat-PCWR ( nuclear ) , or Nmnat-PD ( cytoplasmic ) ( Figure 4A and C ) . In addition , when we knocked down Nmnat in Rasv12-expressing glial cells , we observed significant nuclear enrichment of caspase-3 ( Figure 4B and D ) . These results suggest that NMNAT is a negative regulator of glial neoplastic cell apoptosis in Drosophila . Next , we examined apoptosis and the activation of caspase-3 in human glioma cells . We found that knockdown of NMNAT led to increased nuclear caspase-3 ( Figure 5A and C ) . Western blot analysis showed a specific increase of fully processed P17/19 species of cleaved caspase-3 ( Figure 5D ) , indicating the activation of apoptosis ( Porter and Jänicke , 1999 ) . To examine the effect of overexpressing NMNAT on apoptosis , we employed cisplatin treatment to induce apoptosis as the basal level of apoptosis in T98G glioma cells is low ( Kondo et al . , 1995 ) . Cisplatin significantly increased nuclear caspase-3 levels as expected . Interestingly , overexpression of either NMNAT1 or NMNAT2 reduced nuclear caspase-3 in cisplatin-induced apoptosis ( Figure 5B and E ) , specifically the fully processed cleaved caspase species P17/19 as shown by Western blot analysis ( Figure 5F ) . Taken together , these results suggest that NMNAT promotes glioma growth by inhibiting caspase-mediated apoptosis . DNA instability is one of the hallmarks of cancer . Two common strategies cancer cells use to avoid the triggering cell apoptosis by DNA damage are hyperactivating DNA damage repair , and inactivating cell apoptosis initiation ( Norbury and Zhivotovsky , 2004 ) . Since NAD+ plays important regulatory roles in both DNA damage repair and cell apoptosis , and NAD+ synthase activity is required for glial neoplasia growth ( Figure 2 ) , we next examined the effect of NMNAT on the DNA damage pathway in glioma . We first determined DNA damage by using a phosphor-specific antibody to histone 2A variant ( H2Av ) , a marker for DNA double-strand breaks ( Lake et al . , 2013 ) . We observed a significant elevation of H2Av signal in Nmnat-PC ( nuclear ) overexpressing brains compared to that in Nmnat-PD ( cytoplasmic ) , Nmnat-PCWR ( nuclear ) , or lacZ overexpressing brains ( Figure 6A ) , suggesting DNA damage level is higher in glial neoplasia with Nmnat-PC ( nuclear ) overexpression . We next examined the distribution of endogenous p53 in glial neoplasia and found that while in control glial neoplasia cells ( LacZ group ) , p53 was relatively evenly distributed with ~40% of p53 in the nucleus , a significantly reduced nuclear p53 pool ( ~20% ) was found in Nmnat-PC ( nuclear ) overexpressing glial neoplasia cells ( Figure 6B and D ) . Together with the observation of higher DNA damage levels in Nmnat-PC ( nuclear ) overexpressing glial neoplasia cells , these results indicate that Nmnat-PC ( nuclear ) expression potentially regulates p53 response to DNA damage , presumably to allow higher tolerance to DNA damage . p53 is a key player controlling cell fate in response to DNA damage: initiate DNA repair when there is limited DNA damage , and induce apoptosis when DNA damage is too severe ( Roos and Kaina , 2013 ) . To validate the role of p53 in glial neoplasia development in Drosophila , we examined the effect of a p53 inhibitor: pifithrin-α ( PFT-α ) . PFT-α is reported to inhibit translocation of p53 and affect p53-related transactivation ( Komarov et al . , 1999; Leker et al . , 2004; Murphy et al . , 2004 ) . We analyzed glial neoplasia tissue volume with GFP and DAPI staining in the CNS of flies ( Figure 7A ) . The glial neoplasia volume was significantly increased in PFT-α-treated flies compared to that in DMSO-treated flies ( Figure 7C ) . The increase in glial neoplasia volume was accompanied by a decrease in survival ( Figure 7B ) , and the reduced cleaved caspase-3 intensity ( Figure 7D ) . These results suggest p53 is critical for inhibiting glial neoplastic growth in Drosophila , and p53 inhibition phenocopies NMNAT overexpression in glial neoplasia growth . p53 depletion rescues NMNAT knockdown induced caspase-3 activation . To further assess the role of p53 in NMNAT knockdown-induced apoptosis , we examined the effects of p53 depletion combined with NMNAT knockdown in human glioma cells . We employed two approaches to reduce/deplete p53: siRNA transfection or shRNA lentiviral transduction in both U87MG and T98G cell lines . After p53 depletion , we carried out siNMNAT-mediated knockdown and probed for cleaved caspase-3 to examine the activation of apoptosis in both T98G and U87MG cells . Under all four conditions , two cell types and two modes of p53 depletion , we observed a consistent reduction of siNMNAT-induced apoptosis activation when p53 was depleted . As shown in Figure 8 ( shRNA lentiviral knockdown ) and Figure 8—figure supplement 1 ( siRNA knockdown ) , cleaved caspase-3 expression level in p53 and NMNAT double knockdown cells was reduced compared to those in siNMNAT cells , suggesting that p53 depletion reduced significantly cleaved caspase-3 expression in NMNAT knockdown glioma cells . These results indicate p53 is a key mediator of NMNAT knockdown-induced apoptosis in glioma . Our observations that NMNAT overexpression-induced higher tolerance to DNA damage and altered p53 response is intriguing . Maintaining functional DNA repair is critical for cancer cells to survive during rapid cell proliferation and the accompanying constant need for DNA replication . In response to DNA damage , PARP1 catalyzes NAD+-dependent PARylation of a large number of proteins ( including p53 ) , a process that is one of the largest NAD+ consumers in the nucleus ( Fischbach et al . , 2018; Kim et al . , 2005 ) . It has been shown that PARylated p53 has reduced stability and activity ( Simbulan-Rosenthal et al . , 1999 ) . We hypothesize that NMNAT regulates PARylation in glioma cells . To test this hypothesis , we first examined the level of protein PARylation under NMNAT overexpression , and using dot blot analysis , we found that protein PARylation level was significantly increased in NMNAT1 or NMNAT2 overexpressing cells and significantly reduced with siRNA knockdown ( Figure 9A and B and Figure 9—figure supplement 1 ) . Next , we examined the protein-protein interaction among p53 , NMNAT1 , and PARP1 using immunoprecipitation . Interestingly , we detected PARP1 and NMNAT proteins in the p53-immunoprecipitated fraction ( Figure 9C ) . Furthermore , although total p53 levels were not significantly affected by NMNAT expression , the level of PARP1 immunoprecipitated with p53 was increased with NMNAT overexpression ( Figure 9C ) . We observed consistent results in U87MG cells that NMNAT interacts with p53 ( Figure 9—figure supplement 2 ) . These results suggest the presence of a trimeric p53/ NMNAT/PARP1 complex , and a potential role of NMNAT in promoting the trimeric complex formation . To confirm and extend the biochemical analysis , we carried out immunofluorescent colocalization studies of T98G glioma cells expressing NMNAT1 and detected the colocalization of NMNAT1 with p53 ( Figure 9E1 ) as well as of NMNAT1 with PARP1 ( Figure 9G1 ) . Consistent with Western blot analysis ( Figure 9C ) , p53 protein level is not altered by NMNAT expression as p53 immunofluorescence intensity was similar between NMNAT1 expression cells and neighboring untransfected cells , or DsRed expressing control cells ( Figure 9D2 and quantified in Figure 9H ) . Interestingly , the distribution of p53 changed from diffuse to clustered in NMNAT1-positive hotspots , as visualized by a fluorescence surface plot in Figure 9E2’ . Similarly , PARP1 protein also clustered in NMNAT1-positive hotspots ( Figure 9G2 and G2’ ) , suggesting the close proximity of NMNAT , p53 , and PARP1 . In addition , in NMNAT-expressing cells , PARP1 levels exhibit a small but significant upregulation ( Figure 9I ) . Collectively , these results suggest that NMNAT interacts with PARP1 and promotes PARylation of PARP1-targeting proteins , including p53 , through increasing the local NAD+ availability . In addition to PARylation , p53 is modified by another NAD+-dependent posttranslational modification ( PTM ) , deacetylation . p53 is acetylated by p300/CBP and deacetylated by SIRTs family of NAD+-dependent deacetylases ( Vaziri et al . , 2001 ) . SIRT1 is the major deacetylase regulating p53 activity through deacetylation of p53 at K382 , and hence inhibiting the p53-mediated apoptosis pathway ( Cheng et al . , 2003 ) . NMNAT1 has been reported to interact with SIRT1 directly ( Zhang et al . , 2009 ) . We determined the level of acetyl-p53 in cell extracts and through immunoprecipitating by anti-p53 antibodies from T98G glioma cells with or without NMNAT overexpression , and then probing for acetyl-p53 at K382 . Interestingly , with NMNAT1 or NMNAT2 overexpression , acetyl-p53 was specifically reduced while total p53 levels remained the same ( Figure 10A–C ) , although a stable complex of p53 and SIRT1 was not detected . It is interesting to note that endogenous SIRT1 expression was upregulated in NMNAT overexpressing cells ( Figure 10D ) , suggesting a potential coregulation of NMNAT and SIRT1 . Notably , similar results were observed in U87MG cells ( Figure 10—figure supplement 1 ) , suggesting a common effect of NMNAT on p53 modification . Collectively , these results show that NMNAT upregulation promotes the NAD+-dependent deacetylation of p53 and specifically reduces the pool of acetyl-p53 . To further examine the disease-relevant role of NMNAT in glioma growth , we analyzed patient data from the Cancer Genome Atlas ( TCGA ) to determine how NMNAT expression levels affect survival in glioma and glioblastoma , using the gene expression profiling interactive platform , GEPIA ( http://gepia . cancer- pku . cn/ ) . A strong negative correlation between NMNAT1 expression and survival can be seen in patients with brain lower grade glioma ( LGG ) , with elevated NMNAT1 expression significantly associated with a lower disease-free survival rate , both when comparing survival in the median high-low tumor expression patient groups ( Figure 11A ) and in the highest and lowest 10% expression groups ( Figure 11B ) . In the aggressive form of glioma , glioblastoma multiforme ( GBM ) , from which the T98G and U87MG cell lines are derived , high NMNAT1-expressing tumors ( top 10% ) again showed a significant correlation with more aggressive disease and poorer outcome ( Figure 11C ) . However , NMNAT2 GBM expression levels did not correlate with patient survival ( Figure 11D ) . These brain glioma patient data set results indicate the strong correlation between high NMNAT1 expression with lower survival and poorer clinical outcome . As both PARylation and deacetylation modifications of p53 have been reported to inactivate p53-mediated function and activity ( Juan et al . , 2000; Luo et al . , 2000; Malanga et al . , 1998; Simbulan-Rosenthal et al . , 1999 ) , collectively , our results suggest a model where NMNAT promote glioma growth through facilitating NAD+-dependent PTMs of p53 to ameliorate apoptosis ( Figure 11E ) .
We adapted a glial neoplasia model in Drosophila using the UAS-Ras85Dv12 and repo-GAL4 driver system that induces overgrowth of glial cells to mimic glial neoplasia formation ( Read et al . , 2009 ) . Although RAS alterations in human glioma occur at a lower frequency than some other higher frequency driver alterations ( Brennan et al . , 2013 ) , our rationale for using mutant RAS overexpressing model in Drosophila was to study the effects of NMNAT on the broader common ( rather than Ras-specific ) processes underlying tumorigenic development in a validated glioma model in Drosophila . It will be an important future direction to establish Drosophila models using other high-frequency glioma drivers . Since all Drosophila glia express Repo , we can easily monitor the formation of Rasv12-driven glial neoplasia in the brain by GFP reporter , Repo , and F-actin labeling . In fluorescence imaging , normal brains typically have two to three layers of Repo-positive cells visible in each section ( Figure 1B ) . Therefore , any tissue mass consisting of more than three layers of glia would be atypical and potentially tumor-like . We analyzed glial neoplasia with three key criteria: cell type ( Repo-positive ) , cell number ( more than three layers with at least 400 ) , and tissue size ( volume of at least 12 . 4×103 μm3 ) . Combined with our high-resolution imaging capability , these criteria allow us to distinguish tumor from non-glial neoplasia tissue with high confidence and to analyze glial neoplasia in the most robust and reproducible manner . In Rasv12 expressing flies , we observed glial neoplasia occurred extensively in the brain and VNC . In addition to the morphological phenotypes , we found that glial neoplasia reduced the animal survival rate . Specifically , the total volume of glial neoplasia tissue is positively correlated with the severity of reduced animal survival rate . Such correlation allows the use of high-resolution in vivo morphological imaging as a strong predictor of pathological outcome and a powerful tool to identify genetic modulators of tumorigenesis as we have done in this study , and potential pharmacological modulators for cancer therapy in the future . Our results show that NMNAT expression promotes glioma growth but is likely dispensable for its initiation , as NMNAT overexpression alone did not trigger tumorigenesis . Our results showed that the enzymatic function of NMNAT is required for glioma growth . This finding is not surprising given the fundamental role of NAD+ as a signaling cofactor that regulates cancer metabolism through its coenzymatic function in the redox reactions underlying essential bioenergetic pathways , including glycolysis , the TCA cycle , and oxidative phosphorylation ( Hanahan and Weinberg , 2000 ) . While NAMPT is the rate-limiting enzyme , NMNAT is downstream of NAMPT and directly regulates the level of NAD+ by catalyzing the reversible reaction of NAD+ synthesis . The direction of the reaction , forward ( NAD+ production ) or reverse ( NAD+ breakdown ) , is dependent upon the availability of subtracts . Therefore , NMNAT functions as a cellular metabolic sensor and maintains the homeostasis of NAD+ pools . NAD+ is highly compartmentalized , with each subcellular NAD+ pool differentially regulated and preferentially involved in distinct NAD+-dependent signaling or metabolic events ( Zhu et al . , 2019 ) . Compartment-localized NMNAT isoforms contribute to the maintenance of subcellular NAD+ pools . In mammals , NMNAT1 is nuclear and NMNAT2 is cytoplasm-localized ( Berger et al . , 2005 ) . Our analysis of the public glioma cancer data set GEPIA identified a strong negative correlation between NMNAT1 expression and disease-free survival in patients with brain LGG as well as the progressive GBM . These findings have several implications . First , the critical requirement for nuclear NAD+-consuming events in tumor growth demands a constant supply of nuclear NAD+ pool by nuclear-localized NMNAT . Indeed , as our results show , NAD+-dependent PARylation and deacetylation of proteins including p53 underlies the mechanism of tumorigenesis . Second , the difference in the tumor-promoting effects of nuclear vs . cytoplasmic NMNAT isoforms may inform cellular metabolic needs and genotoxic load . Interestingly , the CBio portal databases show NMNAT1 and NMNAT2 genes appear to be amplified in distinct cancer types ( Figure 11—figure supplements 1 and 2 ) . Future work is required to identify the specific roles of NMNAT1 and NMNAT2 in different cancer types . During the submission of this manuscript , two groups reported the distinct roles of NMNAT1 and NMNAT2 in acute myeloid leukemia and ovarian cancer respectively ( Challa et al . , 2021; Shi et al . , 2021 ) . Shi et al . , 2021 showed that NMNAT1-mediated NAD+ metabolism regulates p53 acetylation and enables acute myeloid leukemia ( AML ) to evade apoptosis ( Shi et al . , 2021 ) . Challa et al . , 2021 showed that NMNAT2-mediated cytosolic NAD+ synthesis regulates ribosome ADP-ribosylation to maintain protein homeostasis in ovarian cancer ( Challa et al . , 2021 ) . It is important to note that in mammalian cells , nuclear and cytoplasmic NMNATs can regulate each other’s activity , likely through feedback from dynamic pool of substrates NMN and ATP , as overexpressing cytoplasmic NMNAT may exhaust the supply of NMN therefore repress nuclear NAD+ synthesis ( Ryu et al . , 2018 ) . Consequently , altering the nuclear NAD+ pool may regulate gene transcription and influence cell differentiation or proliferation state ( Ryu et al . , 2018 ) . Our observation of the specific upregulation of endogenous nuclear NMNAT upon oncogenic RAS-expression further supports the hypothesis that nuclear and cytoplasmic NMNAT react differently in stress conditions and likely be important in different stages of tumor growth . PARylation , phosphorylation , acetylation , and ubiquitination are PTMs that have been shown to regulate the stability and activity of p53 ( Bode and Dong , 2004 ) . Among the most common PTMs of p53 , PARylation and acetylation are both NAD+-consuming processes mediated by NAD+-dependent enzymes , PARPs , and SIRTs ( Lee et al . , 2012; Vaziri et al . , 2001 ) . When PARP1 activity is induced in the DNA damage response process , extensive protein PARylation occurs and many proteins including p53 and DNA repair machinery components are PARylated ( Amé et al . , 2004; Fischbach et al . , 2018 ) . Numerous studies have shown that PARylation of p53 may inhibit p53-mediated function , including cell cycle arrest and apoptosis ( Kanai et al . , 2007; Simbulan-Rosenthal et al . , 1999; Simbulan-Rosenthal et al . , 2001 ) . With abundant NAD+ supply , PARylation is an efficient way to repair DNA damage and ensure cell survival; whereas under conditions of insufficient NAD+ supply , apoptosis is induced ( Herceg and Wang , 2001 ) . In response to DNA damage , the activity of p53 is also modulated by acetylation . Acetyl-p53 is resistant to degradation by ubiquitination and has higher stability , and therefore can exert longer effects of growth arrest , senescence , and apoptosis ( Li et al . , 2002 ) . NAD+-dependent PARylation and acetylation have the opposite effects on p53 activity , where PARylation inhibits p53 activity and acetylation prolongs p53 activity . NAD+ thus plays a critical role in balancing the pro-apoptotic activity of p53 . NMNAT1 regulates functions of NAD+-dependent enzymes such as SIRT1 and PARP1 ( Zhang et al . , 2009; Zhang et al . , 2012 ) . Interestingly , our results identified a trimeric complex of NMNAT-PARP1-p53 and increased PARP1 and SIRT1 , which supports the model that NMNAT recruits NAD+-utilizing enzymes , including PARP1 and SIRT1 , together with protein substrates , and locally supply NAD+ for NAD+-dependent protein modification . Such an NMNAT-PTM modifying enzyme-protein substrate trimeric protein complex will not only sustain the local supply of NAD+ but also facilitate and expedite the modification process . It is important to note that p53 is not the only target for PARylation and deacetylation regulation . The role of NMNAT in PARylation of other target proteins has also been indicated . For example , it has been shown that decreased NMNAT1 expression caused nuclear NAD+ deficiency and subsequently reduced PARylation of multifunctional nuclear protein CCCTC-binding factor , leading to epigenetic silencing of tumor suppressor genes ( Henderson et al . , 2017 ) . As noted above , the most recent study showed increased NMNAT2 mediated cytosolic NAD+ synthesis activity supports mono ( ADP-ribosyl ) ation ( MARylation ) through PARP-16 in ribosome ( Challa et al . , 2021 ) . These reports together with our findings support a specific role of NAD+ in modulating tumorigenesis through regulating PTMs , including PARylation and deacetylation . Our findings in both in vivo and in vitro models highlight NMNAT’s roles in promoting glioma development . Specifically , the direct interaction we identified among p53 , NMNAT , and PARP1 has important implications regarding the utility of NMNAT as a potential target for glioma therapy . Because the protein-protein interaction interface of NMNAT-PARP1-p53 could provide allosteric targeting of NMNAT , in addition to its enzyme pocket , this may open new possibilities for alternative inhibitors of NAD+-dependent pathway with less toxicity . It should be noted that although T98G cells carry a gain-of-function p53 mutation ( M237I ) in the DNA binding domain , this mutation does not affect the sites of PARylation and acetylation ( Yamamoto and Iwakuma , 2018; Yi et al . , 2013 ) . Moreover , prior studies support the ability of cells harboring this p53 mutant to undergo apoptosis , which can be abrogated by p53 inhibition ( Enns et al . , 2004 ) . Our findings that NMNAT similarly affects p53 modification in either wild-type ( U87MG ) or mutant p53 ( T98G ) cells suggest NAD+-dependent PARylation or deacetylation of p53 is independent of the p53 [M237I] mutation . Indeed recent studies have shown that mutant p53 proteins retain the ability to induce apoptosis despite losing tumor-suppressive transactivation functionality ( Timofeev et al . , 2019 ) . Further studies will be required to fully understand the effects of NMNAT on p53 transcription factor function . In conclusion , our studies have identified NMNAT as an NAD+ synthase that plays an essential role in regulating the function and activation of p53 during DNA damage-induced apoptosis in glioma cells . These results support the development of specific NMNAT inhibitors as potentially efficacious therapeutic agents in cancers with upregulated NMNAT levels .
Flies were maintained at 25°C room temperature with standard medium . The following lines were used in this study obtained from the Bloomington Drosophila Stock Center: ( 1 ) The driver used in all experiments: repo-GAL4; ( 2 ) UAS-Rasv12 ( II ) ; ( 3 ) UAS-Rasv12 ( III ) ; ( 4 ) UAS-Nmnat RNAi ( III ) ; ( 5 ) UAS-p35; ( 6 ) UAS-Diap1; ( 7 ) UAS-Dronc RNAi; and ( 8 ) UAS-DCP1 RNAi . UAS-Drosophila melanogaster Nmnat ( UAS-PC , UAS-PCWR , UAS-PD ) were generated in the laboratory . Larvae were collected and treated with 100 μM of Pifithrin-α ( Sigma-Aldrich , P4359 ) with standard medium at 25°C room temperature . T98G and U87MG ( human glioma cells ) cell lines were purchased from the American Type Culture Collection ( ATCC; CRL-1609 ) . SVG p12 cell line was from Dr . Michal Toborek ( University of Miami ) . Cells were maintained in Eagle’s Minimum Essential Medium ( EMEM; Sigma-Aldrich , M0325 ) supplemented with 10% fetal bovine serum ( FBS; ATCC , 30–2020 ) . Cells were cultured at 37°C , 5% CO2 . To induce apoptosis , cells were treated with 50 μM of cisplatin for 8 hr ( Sigma-Aldrich , 232120 ) . The following commercially available antibodies were used: anti-Repo ( 1:250 , DSHB , 8D12 ) , anti-Caspase-3 ( 1:250 for Immunocytochemistry of fly brain , 1:1000 for Western blot analysis , Cell Signaling Technology , 9665 ) , anti-Cleaved Caspase-3 ( 1:1000 , Santa Cruz , 9661 ) , anti-H2AvD ( 1:50 , Rockland , 600-401-914 ) , anti-p53 ( E-5 ) ( 1:50 , Santa Cruz , sc-74573 ) , p53 ( DO-1 ) ( 1:1000 , Santa Cruz , sc-126 ) , anti-Drosophila Nmnat ( 1:3000 ) , anti-NMNAT1 ( 1:1000 , Abcam , ab45548 ) , anti-NMNAT1 ( 1:1000 , Santa Cruz , 271557 ) , anti-NMNAT2 ( 1:500 , Abcam , ab56980 ) , anti-PARP1 ( 1:1000 , Santa Cruz , sc-8007 ) , anti-pADPr ( 1:1000 , Santa Cruz , sc-56198 ) , anti-SIRT1 ( 1:1000 , Cell Signaling Technology , 2492 ) , anti-acetyl-p53 ( 1:1000 , Cell Signaling Technology , 2525 ) , anti-β-actin ( 1:10 , 000 , Sigma-Aldrich , A1978 ) , and anti-tubulin ( 1:300 , Abcam , ab15246 ) . The secondary antibodies conjugated to Alexa 488/546/647 ( 1:250 , Invitrogen ) , or near-infrared ( IR ) dye 700/800 ( 1:5000 , LI-COR Biosciences ) . HRP-anti-mouse and HRP-anti-rabbit ( 1:5000 , Thermo Fisher Scientific ) . Four recombinant plasmids were generated for this study: pDsRed , pDsRed-NMNAT1 , pDsRed-NMNAT2 , NMNAT1 , and NMNAT2 . Small interference RNA sequences targeting human NMNAT were purchased ( GenePharma ) . The siRNA sequences were listed in Supplementary ( Figure 3—figure supplement 2 ) . Stable p53 knockdown was performed using the pLKO lentiviral shRNA system . Lentiviral supernatant production was carried out in HEK 293T cells and transduction of target U87MG and T98G cells with either the shp53 or the control shGFP supernatant was performed as described previously ( Rai et al . , 2009 ) . The following validated shRNA target sequences ( Burton et al . , 2013 ) ; ( Patel et al . , 2015 ) were used: Transduced cells were selected in 2 . 5 µg/ml puromycin-containing culture media for a minimum period of 5–7 days ( corresponding to the time taken for untransduced cells to die completely in selection media ) . The total RNA was extracted by TRIzol reagent ( Invitrogen ) from T98G cells according to the manufacturer’s protocol . cDNA was synthesized from RNA with a cDNA Reverse Transcription Kit ( Applied Biosystems ) . RNA was performed using a Real-Time System and SYBR Green Kit ( Applied Biosystems ) . Relative gene expression was compared to actin as an internal control . The primers used in detection were listed in the Supplementary ( Figure 3—figure supplement 2 ) . Cells for transfection were seeded in a six-well culture vessel ( VWR ) containing EMEM media with 10% FBS for 24 hr . Plasmids or siRNA were transfected with transfection reagent ( jetPRIME ) . Gene expression was measured by Western blot analysis and real-time qPCR after cells were transfected at 48 hr . Cells were grown on 22 mm glass coverslips ( VWR ) . After treatment , cells were rinsed three times with phosphate-buffered saline ( PBS ) , fixed for 15 min in 4% paraformaldehyde , washed three times with PBS , and permeabilized with 0 . 4% Triton X-100 in PBS for 5 min . After three times washing in PBS , blocking was performed by incubation in 5% normal goat serum in PBTX ( PBS with 0 . 1% Triton X-100 ) at 37°C for 30 min . Incubation with primary antibodies was performed in 5% goat serum in PBTX at 37°C for 2 hr . Next , cells were washed three times with PBS and incubated for 1 hr at 37°C with secondary antibodies in 5% goat serum in PBTX . Then , after three times washing with PBS , cells were stained with 4′ , 6-diamidino-2-phenylindole ( DAPI , 1:300 , Invitrogen ) at 37°C for 5 min in PBTX solution . The cells were washed three times with PBS , and the coverslips were mounted on glass slides with VECTASHIELD Antifade Mounting Medium ( Vector Laboratories ) and kept at 4°C before imaging . The larval brains were dissected in PBS ( pH 7 . 4 ) , and fixed in PBS with 4% formaldehyde for 15 min . After the brains were washed in PBS containing 0 . 4% ( v/v ) Triton X-100 ( PBTX ) for 15 min three times , the brains were incubated with primary antibodies diluted in 0 . 4% PBTX with 5% normal goat serum overnight . Then , secondary antibodies were at room temperature for 1 hr , followed by DAPI ( 1:300 , Invitrogen ) staining for 10 min . Brains were mounted on glass slides with VECTASHIELD Antifade Mounting Medium ( Vector Laboratories ) and kept at 4°C before imaging . Confocal microscopy was performed with an Olympus IX81 confocal microscope coupled with ×10 , ×20 air lens or ×40 , ×60 oil immersion objectives , and images were processed using FluoView 10-ASW ( Olympus ) . Specifically , Figure 7B and C were analyzed using the ImageJ interactive 3D surface Plot plugin . Proteins were extracted from cells in RIPA ( radioimmunoprecipitation assay ) buffer 1 mM protease inhibitor cocktail ( Sigma-Aldrich ) . Samples were heated at 100°C for 10 min in a 4× loading buffer . Proteins were separated on a Bis-Tris gel and transferred to nitrocellulose membranes . Then , membranes were blocked with blocking buffer ( Rockland ) for 1 hr at room temperature . Primary antibodies were incubated at 4°C overnight and secondary antibodies were incubated for 1 hr at room temperature . Images were processed on an Odyssey Infrared Imaging System or Amersham Imager 600 and analyzed using Image Studio software or ImageJ . Proteins were extracted from cells in RIPA buffer 1 mM protease inhibitor cocktail ( Sigma-Aldrich ) . Proteins were loaded with same amount on PVDF membranes . Then , membranes were blocked with Casine buffer for 1 hr at room temperature . Primary antibodies were incubated at 4°C overnight and secondary antibodies were incubated for 1 hr at room temperature . Images were processed on an Amersham Imager 600 and analyzed using ImageJ software . Cells were seeded into the E-Plate 96 ( ACEA ) with the same confluence per well . Then , the plate was incubated at 37°C in 5% CO2 for about 100 hr . The instrument was used to monitor the cell growth index . The cell growth curve was drawn with the value of each group from xCELLigence RTCA SP instrument . Cells were seeded with 1000 per well in a six-well plate containing 2 ml medium and replaced medium every 2 days . Cells were washed with 1 ml PBS three times and fixed with 1 ml formaldehyde for 15 min . After washed with PBS , cells were stained in 0 . 1% crystal violet buffer ( Sigma-Aldrich ) for 15 min . Cells were washed with pure water gently , and plates were put at room temperature to dry . Images were processed on an Amersham Imager 600 and analyzed using ImageJ software . Proteins were extracted from cells in RIPA buffer . Proteins were incubated with Protein-A beads ( Thermo Fisher Scientific ) conjugated with anti-p53 antibody or Mouse IgG at 4°C overnight with gentle shaking . After removing the supernatant , the bead pellets were collected and suspended with lysis buffer . Proteins were heated with loading buffer for 10 min at 100°C for loading to gel . Cells were prepared according to cell cycle and cell apoptosis detection kits ( BD Pharmingen ) after knockdown of NMNAT 72 hr . AnnexinV:PI gating was selected and analyzed to divide the data into quadrants , where Q3 was considered as viable , Q4 as early apoptosis , and Q2 as end stage apoptosis and death . For each statistical test , biological sample size ( n ) , and p-value are indicated in the corresponding figure legends . All data in this manuscript are shown as mean ± SD or median ± quartiles ( specified in figure legends ) . t-test was used to compare between two groups , and one-way ANOVA with Bonferroni’s post hoc test was applied to compare among three or more groups . Data were analyzed with Prism ( GraphPad Software ) . Specifically , fly survival data were analyzed by the Chi-square test in R . | One of the most common types of brain cancer , glioma , emerges when harmful mutations take place in the ‘glial’ cells tasked with supporting neurons . When these genetically damaged cells are not fixed or eliminated , they can go on to multiply uncontrollability . A protein known as p53 can help to repress emerging tumors by stopping mutated cells in their tracks . Glioma is a highly deadly cancer , and treatments are often ineffective . Some of these approaches have focused on a protein involved in the creation of the coenzyme NAD+ , which is essential to the life processes of all cells . However , these drugs have had poor outcomes . Instead , Liu et al . focused on NMNAT , the enzyme that participates in the final stage of the creation of NAD+ . NMNAT is known to protect neurons , but it is unclear how it involved in cancer . Experiments in fruit flies which were then validated in human glioma cells showed that increased NMNAT activity allowed glial cells with harmful mutations to survive and multiply . Detailed molecular analysis showed that NMNAT orchestrates chemical modifications that inactivate p53 . It does so by working with other molecular actors to direct NAD+ to add and remove chemical groups that control the activity of p53 . Taken together , these results show how NMNAT can participate in the emergence of brain cancers . They also highlight the need for further research on whether drugs that inhibit this enzyme could help to suppress tumors before they become deadly . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience",
"cancer",
"biology"
] | 2021 | NMNAT promotes glioma growth through regulating post-translational modifications of P53 to inhibit apoptosis |
The receptor tyrosine kinase Her2 , an intensely pursued drug target , differs from other members of the EGFR family in that it does not bind EGF-like ligands , relying instead on heterodimerization with other ( ligand-bound ) EGFR-family receptors for activation . The structural basis for Her2 heterodimerization , however , remains poorly understood . The unexpected recent finding of asymmetric ectodomain dimer structures of Drosophila EGFR ( dEGFR ) suggests a possible structural basis for Her2 heterodimerization , but all available structures for dimers of human EGFR family ectodomains are symmetric . Here , we report results from long-timescale molecular dynamics simulations indicating that a single ligand is necessary and sufficient to stabilize the ectodomain interface of Her2 heterodimers , which assume an asymmetric conformation similar to that of dEGFR dimers . This structural parallelism suggests a dimerization mechanism that has been conserved in the evolution of the EGFR family from Drosophila to human .
Her2 ( also known as Neu or ErbB2 ) , a receptor tyrosine kinase belonging to the human epidermal growth factor receptor ( EGFR ) family that also includes EGFR/Her1 , Her3 , and Her4 , is an important component of cell-signaling networks , and is implicated in the growth of a variety of cancers ( Hynes and Lane , 2005; Riese et al . , 2007; Baselga and Swain , 2009; Lemmon and Schlessinger , 2010 ) . The receptors of the EGFR family activate through dimerization ( Figure 1A ) , which is promoted by the binding of ligands from the EGF family ( Ushiro and Cohen , 1980; Schreiber et al . , 1983; Chung et al . , 2010 ) to the receptors’ extracellular regions ( ‘ectodomains’ ) . This activation process relies on a number of allosteric interactions in the extracellular , transmembrane , and intracellular portions of the receptor ( Endres et al . , 2011 , 2013; Arkhipov et al . , 2013 ) , which lead to the formation of a specific asymmetric active dimer of the intracellular kinase domains ( Zhang et al . , 2006 ) . Her2 is unique in the family in that it does not homodimerize under normal conditions , and its ectodomain does not bind ligands . Instead , it activates through heterodimerization with other members of the family ( EGFR and Her3 , in particular ) when they are bound to ligands ( Citri and Yarden , 2006; Baselga and Swain , 2009 ) . 10 . 7554/eLife . 00708 . 003Figure 1 . Receptors of the human EGFR family and conformations of their ectodomains . ( A ) Left: the four members of the human EGFR family , each consisting of an ectodomain , a single-pass transmembrane helix , and an intracellular module that includes a kinase domain . As shown , Her2 bears a closed ligand binding site and does not bind EGF-like ligands . Her3 is kinase-dead , and its intracellular module is thus colored gray . Right: common homo- and heterodimers of the EGFR family . The Her2 homodimer is rendered semitransparent to indicate its instability in normal cell conditions . ( B ) Schematic of the ligand-free Her2 ectodomain monomer , as observed crystallographically ( PDB entries 1N8Z , 2A91 , 1S78 , 3N85 , and 3MZW ) . Domain II is bent . ( C ) Schematic of the crystal structure of the 2-ligand EGFR ectodomain dimer ( PDB entry 1NJP ) . The dimer is symmetric , and domain II is straight in both subunits . ( D ) Schematic of the crystal structure of the dEGFR ectodomain dimer ( PDB entry 3LTF ) . Although Spitz ligands are bound to both subunits , the structure is asymmetric; domain II is straight in one subunit but bent in the other . Domain V and part of domain IV were not resolved in this crystal structure and are not shown in the schematic . The conformations of the bent and straight domain IIs in ( B ) , ( C ) , and ( D ) are indicated by the black lines . DOI: http://dx . doi . org/10 . 7554/eLife . 00708 . 003 The complexity of signaling underpinned by homo- and heterodimerization of EGFR family members emerged relatively recently in evolution . EGFR families in invertebrates have only one member , but gene duplication gave rise to two-member EGFR families in early vertebrate species , such as fishes . Further gene duplication eventually generated the four members of the mammalian EGFR family ( Stein and Staros , 2000 ) . The increase in the number of EGFR family members , accompanied by increasing diversity of the extracellular ligands they interact with , gave rise to a complex signaling network: homo- and heterodimerization of various EGF receptors , induced by ligand binding , began to generate a variety of unique signaling outputs controlled by the identity of the bound ligands . The unique dimerization properties of Her2 are believed to be essential to its critical role as a potent signal amplifier for the other receptors of the EGFR family . The Her3–Her2 heterodimer is particularly prevalent and potent in signaling . The prominence of Her3–Her2 heterodimers is particularly intriguing , since Her2 itself lacks an activating ligand and Her3 is impaired in its kinase activity ( Figure 1A ) . These receptors act primarily through heterodimerization . The unique partnership of Her2 and Her3 has been aptly dubbed ‘the deaf and the dumb’ ( Citri et al . , 2003 ) . By activating several downstream pathways , including those of MAPK , PI3K , phospholipase C , protein kinase C , and Janus kinase , the Her3–Her2 heterodimer plays a pivotal role in the determination of cell lineage in a variety of tissues in epithelial organs . Experiments have shown that knocking out Her2 and Her3 genes leads to defective development of the heart , mammary gland , and nervous system in mice phenotypes . It is thus not surprising that Her3–Her2 heterodimers are implicated in a number of forms of cancer ( Citri et al . , 2003 ) or that Her2 is an important drug target in cancer therapeutics . The structural mechanism underlying Her2’s dimerization properties remains incompletely understood . In particular , the conformations of Her2 homo- and heterodimers remain obscure , and how these conformations relate to the dimerization affinities is uncertain . Moreover , the manner in which ligands binding to Her2’s heterodimerization partners regulate the conformations and stability of the heterodimers is poorly understood . Although various structural and electrostatic factors that may affect Her2 dimerization have been discussed in the literature ( Cho et al . , 2003; Franklin et al . , 2004; Garrett et al . , 2003; Alvarado et al . , 2009 , 2010; Liu et al . , 2012 ) , further structural elucidation is needed for a better molecular understanding of Her2 activation . Crystal structures of the ectodomains of human EGFR-family receptors ( Lemmon , 2009 ) provide a starting point for elucidation of the structural mechanisms of Her2 heterodimerization . In these structures ( Figure 1B , C ) , the ectodomains ( each consisting of four domains numbered I , II , III , and IV ) are found either as monomers or as symmetric homodimers with ligands bound to both dimer subunits ( Garrett et al . , 2002; Lu et al . , 2010; Ogiso et al . , 2002; Liu et al . , 2012 ) . In the homodimer structures , domain II constitutes most of the ectodomain’s dimer interface ( Figure 1C ) , and its conformation is critical to ectodomain dimerization . A bent domain II is found in all ligand-free , inactive ectodomains of the EGFR-family receptors ( Figure 1B ) , while a straight domain II is associated with the ( ligand-bound ) active dimer ( Figure 1C ) . The domain II of Her2 appears to be constitutively bent ( Cho et al . , 2003; Garrett et al . , 2003; Franklin et al . , 2004; Alvarado et al . , 2009 ) , which is consistent with its poor homodimerization . Unfortunately , no crystal structure is available for human EGFR family heterodimers with only a single ligand bound , as is the case for Her2 heterodimers . Intriguingly , recent crystal structures of Drosophila EGFR ( dEGFR ) ectodomains reveal asymmetric dimers that bear only one fully formed ligand binding site , with the other partially closed , regardless of whether one or both subunits are ligand-bound ( Alvarado et al . , 2009 , 2010 ) . These dEGFR structures hint at a possible structural mechanism for Her2 heterodimerization , but translating the dEGFR results to human Her2 is difficult due to differences between these receptors . First , dEGFR dimers assume an asymmetric conformation even when ligands are bound to both dimer subunits , whereas all existing human EGFR-family dimer structures exhibit a symmetric conformation . Second , the ectodomains of Her2 and other members of the human EGFR family each comprise four domains , whereas the dEGFR ectodomain includes a fifth domain , consisting of 166 residues immediately N-terminal to the transmembrane helix . Here we investigate the dimerization of Her2 using molecular modeling and long-timescale molecular dynamics ( MD ) simulations . We modeled and simulated the EGFR–Her2 and Her3–Her2 heterodimers as well as the Her2 homodimer . The heterodimers were found to be stable when a ligand was bound to the EGFR or Her3 subunit . Our simulations further showed that , when the single bound ligand was removed from a Her2 heterodimer , a substantial gap developed in the dimer interface . A similar gap was also observed in our simulation of the Her2 homodimer , which explains the weak homodimerization of Her2 . Structural analysis shows that such a gap arises from the bending of the domain IIs in both subunits due to the absence of bound ligands . These observations are strikingly similar to the findings from the crystal structures of dEGFR homodimers ( Alvarado et al . , 2009 , 2010 ) : the ligand-free , but not the ligand-bound , dEGFR dimer exhibits a gap in the dimer interface , which explains the reduced dEGFR dimerization in the absence of ligands . In agreement with the recent experimental data showing that a single bound ligand is sufficient to activate an EGFR dimer ( Liu et al . , 2012 ) , our simulations demonstrate that the activation mechanism of the receptors of the human EGFR family and of Her2 in particular conserves Drosophila EGFR’s capacity to form stable asymmetric ectodomain dimers upon the binding of a single ligand .
The general approach of this study was to use long-timescale MD simulations to infer from crystal structures of EGFR ectodomain dimers the structures of dimers with different constituent receptors or with different bound ligands . To examine the validity of this approach , we first applied it to dEGFR , for which the structures of both the ligand-bound and ligand-free ectodomain dimers ( Alvarado et al . , 2009 , 2010 ) have been resolved . Our simulations correctly demonstrated that , once the ligands are removed from the ligand-bound structure , a substantial gap between the two subunits develops , leading to a reduced dimerization interface . Unlike the symmetric human ‘2-ligand’ EGFR dimer ( in which both subunits bear fully formed and occupied ligand binding sites; Figure 1C ) , the 2-ligand dEGFR dimer is asymmetric: although both ectodomain subunits are ligand-bound , only one bears a fully formed ligand binding site and a straight domain II ( Figure 2A ) . The ‘1-ligand’ dEGFR dimer ( in which only one subunit is ligand-bound ) is essentially identical to the 2-ligand dimer , except for the missing ligand . The gap between the two subunits is closed in both dEGFR dimers . Conversely , the ligand-free dEGFR dimer is symmetric , featuring a bent domain II in both subunits , and a gap in the dimer interface . We removed both ligands from the crystal structure of the 2-ligand dEGFR dimer ( PDB entry 3LTF ) and simulated the resulting system . In two independent simulations , a substantial gap emerged in the dimer interface , and the resulting conformation resembled that of ligand-free dEGFR crystal structures ( Figure 2A , B ) . Figure 2C shows a significant reduction in the surface area buried within the dimer interface , consistent with the buried areas in the crystal structures of the ligand-free dEGFR dimers . We note that the dimer conformation drifted away from that of the ligand-free crystal structure after ∼2 µs . This reflected the high flexibility of the simulated dimer , likely a result of the truncation of the domain IVs in the resolved crystal structure from which the simulations were initiated and the lack of inter-subunit contact between the two domains . The fact that our dEGFR simulations were able to reproduce the gap in the dimer interface found in the crystal structures of ligand-free dEGFR dimers lends support to the approach we adopted for the analysis of Her2 dimerization . 10 . 7554/eLife . 00708 . 004Figure 2 . Simulations reproduce the gap in the interface of the dEGFR ectodomain dimer . ( A ) The schematic shows the 2-ligand dEGFR ectodomain dimer on the left . In simulations initiated from this structure with both ligands removed , gap opened in the dimer interface , as indicated by a V-shaped outline in the right diagram . In the ligand-free dimer , domain II is bent in both subunits . The simulation snapshots are shown below the schematic diagrams . ( B ) The simulated dEGFR dimer from ( A ) , at t = 0 . 5 µs , is compared with the crystal structure of the ligand-free dEGFR dimer ( tan ) . Molecular renderings ( A , B ) omit domains IV and V for clarity ( although the crystallographically resolved portion of domain IV was present in simulations ) . ( C ) The surface area buried within the dimer interface ( counting the contributions from domains I , II , and III ) . The results of three independent simulations are shown , two starting from the 2-ligand crystal structures ( after removal of the ligands ) , and one from the ligand-free crystal structure . DOI: http://dx . doi . org/10 . 7554/eLife . 00708 . 004 The 2-ligand EGFR homodimer has been resolved crystallographically ( Garrett et al . , 2002; Ogiso et al . , 2002; Lu et al . , 2010 ) and studied previously using MD simulations ( e . g . , Tynan et al . , 2011; Zhang and Wriggers , 2011 ) . Given the close homology between the EGFR-family receptors ( ‘Materials and methods’ ) , this crystal structure may serve as a starting point for the modeling of Her2 homo- and heterodimers . The first model we constructed was that of the ligand-bound EGFR–Her2 heterodimer , in which the EGFR subunit is ligand-bound and the Her2 subunit is ligand-free . We assumed that a 1-ligand EGFR homodimer is structurally similar to this EGFR–Her2 heterodimer and thus may serve as a template . To obtain a structure of the 1-ligand EGFR homodimer , we removed one ligand from the crystal structure of the 2-ligand EGFR homodimer ( PDB entry 3NJP [Lu et al . , 2010] ) and simulated the remaining complex ( Figure 3A ) . In all three independent simulations performed , the dimer assumed an asymmetric conformation , in which the ligand-free subunit differs from the ligand-bound one . In the former , the space previously occupied by the bound ligand between domains I and III is closed and domain II is bent ( Figure 3B ) , in agreement with a recent MD study ( Tynan et al . , 2011 ) . It is notable that the removal of one of the two ligands from the ectodomain dimer did not lead to substantial decrease in the area of the dimer interfaces ( Figure 3E ) , suggesting that , in agreement with recent experimental findings ( Liu et al . , 2012 ) , a single bound ligand is sufficient to maintain a stable ectodomain dimer of EGFR . By comparison , no significant conformational changes were observed in the control simulations of the 2-ligand EGFR dimer . 10 . 7554/eLife . 00708 . 005Figure 3 . Bending of domain II . ( A ) A schematic showing how the model of the 1-ligand EGFR dimer ( right ) is generated by simulation after the removal of one EGF ligand from the crystal structure of the 2-ligand dimer ( left ) . The resulting structure is asymmetric: domain II of the ligand-free subunit is bent and the binding site is closed , whereas domain II of the ligand-bound subunit is straight and the binding site is open ( as is the case in both subunits of the 2-ligand dimer ) . The angle θ , which characterizes the bending of domain II , is measured between the Cα atoms of EGFR residues 194 , 239 , and 296 . ( B ) Bending of domain II . A simulation snapshot of the ligand-free subunit ( red ) from the 1-ligand EGFR dimer is overlaid with the crystal structures of a subunit from the 2-ligand EGFR dimer ( green ) . EGFR residues 240–309 are used for reference . ( C ) An overlay of the ligand-free subunit of the 1-ligand EGFR dimer , generated by simulation ( red ) , with the crystal structure of Her2 monomer ( cyan ) . In the interest of clarity , domain IV is not shown . ( D ) The angle θ ( illustrated in [A] ) is shown as function of time in the three independent simulations of the 1-ligand EGFR homodimer , for the ligand-free subunit ( middle ) and for the ligand-bound subunit ( right ) . The value of θ in the crystal structure of Her2 monomer is indicated by a straight line . ( E ) The buried surface area at the interfaces of 1-ligand and 2-ligand EGFR dimers in simulation . DOI: http://dx . doi . org/10 . 7554/eLife . 00708 . 005 The conformation of the ligand-free subunit in the simulated 1-ligand EGFR homodimer resembles the crystal structure of Her2 monomer ( Figure 3C ) , especially in the bending of domain II . ( The bent domain II is also observed in the crystal structures of ligand-free EGFR monomers [PDB entries 1NQL and 1YY9; Ferguson et al . , 2003; Li et al . , 2005] . ) The bending in domain II can be characterized by the angle θ formed by the top , middle , and bottom regions of the domain II dimerization interface ( Figure 3A , D ) . In our simulations , this angle was ∼150° to 160° in the ligand-free subunits and ∼170° in the ligand-bound subunits , with the former values being very close to that of the Her2 crystal structure ( 156° ) . The EGFR–Her2 heterodimer was thus modeled by superimposing a crystal structure of the monomeric Her2 ectodomain ( PDB entry 3BE1 [Bostrom et al . , 2009] ) onto the ligand-free EGFR subunit in the 1-ligand EGFR dimer using the portion of domain II directly involved in dimerization ( EGFR residues 240–309; see ‘Materials and methods’ ) . We subsequently modeled the ligand-free EGFR–Her2 heterodimer and the Her2 homodimer using a similar approach . In the former case , the ligand was removed from our model of the EGFR–Her2 heterodimer and the resulting ligand-free heterodimer was then simulated . In the simulations , domains I , II , and III of the EGFR subunit again assumed a ‘Her2-like’conformation . This model of the ligand-free EGFR–Her2 heterodimer was then used as a template for the Her2 homodimer , in the same way that the model of the 1-ligand EGFR homodimer was used as a template for the EGFR–Her2 heterodimer ( ‘Materials and methods’ ) . We subsequently simulated the EGFR–Her2 heterodimer and Her2 homodimer starting from these models . In these simulations , the ligand-bound EGFR–Her2 heterodimer remained stable ( Figure 4A ) , the dimer interface was largely intact , and the ligand wedged between domains I and III of the EGFR subunit , preventing any bending in this subunit’s domain II . In contrast , a large gap opened in the dimer interface of the ligand-free EGFR–Her2 heterodimer ( this was consistently observed in two independent simulations ) between the N-terminal portions of the domain IIs ( Figure 4B; Video 1 ) . This occurred as domain II of the EGFR subunit bent away from the dimer interface upon the removal of its bound ligand . Similarly , a gap at the dimer interface was also observed in two independent simulations of the Her2 homodimer ( Figure 4C ) . Plots of the buried surface area show that the size of the gap fluctuated in the simulations ( Figure 4B ) , but on average , was significantly lower in the ligand-free EGFR–Her2 heterodimer and Her2 homodimer than in the ligand-bound heterodimer . 10 . 7554/eLife . 00708 . 006Figure 4 . Simulations of the EGFR–Her2 heterodimer and the Her2 homodimer . ( A ) The EGFR–Her2 heterodimer with the ligand bound to EGFR ( ‘+EGF’ ) . ( B ) The ligand-free EGFR–Her2 heterodimer or EGFR–EGFR homodimer ( ‘−EGF’ ) . ( C ) The Her2 homodimer . Top: snapshots from the simulations . Middle: plots of the surface area buried within the dimer interface ( counting the contributions from domains I , II , and III ) as a function of simulation time ( for EGFR–Her2 [−EGF] and Her2–Her2 , results of two independent simulations are shown ) . Bottom: schematics illustrating the conformation of domain II and the dimer interface . Conformations of the bent and straight domain II are highlighted by the black lines , and the gap in the dimer interface is indicated by a V-shaped outline when present . For clarity , domain IV is not shown . DOI: http://dx . doi . org/10 . 7554/eLife . 00708 . 00610 . 7554/eLife . 00708 . 007Video 1 . The simulation of the ligand-free EGFR–Her2 ectodomain heterodimer with Her2 colored red and EGFR blue . A gap between the N-terminal portion of each subunit’s domain II develops during the simulation . DOI: http://dx . doi . org/10 . 7554/eLife . 00708 . 007 In terms of their location and potential impact on dimer stability , the gaps observed in our simulations are reminiscent of the one in the crystal structure of ligand-free dEGFR dimer ( Alvarado et al . , 2009 , 2010 ) . The dEGFR gap results in a reduced dimer interface area of ∼2300 Å2 , compared to ∼3400 Å2 in the 1- and 2-ligand dEGFR dimers , where the gap is closed ( Figure 2 ) . This reduction of the dimer interface area is believed to account for the relatively weak dimerization affinity of the ligand-free dEGFR , which is ∼30 times lower than that of ligand-bound dEGFR . Figure 4 shows that the reduction of the dimer interface is similarly observed in our simulations of the ligand-free EGFR–Her2 heterodimer and Her2 homodimer: when the gap is wide open , the surface area buried within the interface is ∼2600 Å2 , compared to ∼3500 Å2 in the ligand-bound heterodimer . We thus suggest that the same structural mechanism that weakens ligand-free dEGFR homodimers ( Alvarado et al . , 2009 , 2010 ) also disfavors ligand-free EGFR–Her2 dimerization and Her2 homodimerization . On the other hand , ligand binding in the EGFR subunit helps to keep domain II straight , which prevents the opening of the gap and stabilizes the dimer interface of EGFR–Her2 heterodimer ( Figure 4 ) . This is reminiscent of the recent finding ( Liu et al . , 2012 ) that one bound ligand is sufficient to activate an EGFR dimer . Among the dimerization partners of Her2 , Her3 is particularly notable in that the Her3–Her2 heterodimers are potent signaling units , and in that the normal and pathogenic signaling through Her2 and Her3 relies strongly on Her3–Her2 dimerization ( Baselga and Swain , 2009; Lemmon , 2009 ) . We thus studied the Her3–Her2 heterodimer in ways similar to our study of EGFR–Her2 heterodimers and the Her2 homodimer . We first modeled the Her3–Her2 heterodimer using the resolved crystal structure of the Her3 monomer ( Cho and Leahy , 2002 ) and our model of the EGFR–Her2 heterodimer as templates ( ‘Materials and methods’ ) . The EGF-like domain of the Her3 ligand heregulin-alpha ( HRG ) was then introduced into the Her3 binding site between domains I and III , in accordance with the pose of EGFR-bound EGF . The resulting model of the Her3–Her2 heterodimer was then simulated , first with HRG bound , in which case the dimer interface remained stable ( Figure 5A ) . After ∼4 µs of simulation the ligand was removed , and the resulting ligand-free heterodimer was then simulated again . Without the ligand , the Her3 subunit underwent conformational changes similar to those described above for the ligand-free EGFR subunit , including a bending of domain II . As a result , a gap again emerged in the dimer interface . In the first series of simulations the gap in the Her3–Her2 heterodimer was often partially closed , but re-opened repeatedly and resulted in a significant decrease in the buried surface area in the dimer interface . In the second series , the gap was steadily open once the ligand was removed ( Figure 5B ) . These observations suggest that Her3–Her2 heterodimerization , similar to that of EGFR–Her2 , is promoted by the stabilization of the dimer interface following ligand binding . 10 . 7554/eLife . 00708 . 008Figure 5 . Her3–Her2 heterodimer . ( A ) Snapshots from the simulations of the Her3–Her2 heterodimer with ( left ) and without ( right ) HRG bound to Her3 . At the end of the simulation with HRG bound to Her3 , HRG was removed , and the resulting system was resolvated and further simulated without the ligand . A gap opened in the dimer interface , as illustrated by the snapshot on the right . For clarity , these images omit domain IV . ( B ) The surface area buried within the dimer interface , counting the contributions only from domains I , II , and III , plotted as a function of time . Two independent sets of two simulations each ( with and without HRG ) are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 00708 . 008 Our simulations suggest that ligand binding promotes the formation of asymmetric dEGFR or EGFR-family dimers by allowing specific favorable interactions between the domain II N-terminal regions of the two subunits . We found that the asymmetric dimers we examined—the EGFR–Her2 and Her3–Her2 heterodimers , the 1-ligand EGFR homodimer , and the 1- and 2-ligand dEGFR homodimers—all shared certain key atomic-level interactions between subunits . In particular , hydrogen bonds between Gln194 ( in EGFR or Her3 numbering ) of the ligand-bound subunit and Cys213 and His215 ( in Her2 numbering; Figure 6A ) of the ligand-free subunit ( Figure 6B ) were particularly stable . The presence of these hydrogen bonds is consistent with the reported crystal structures of the dEGFR asymmetric dimers after a 180° χ2 rotamer flip at the histidine ( which would presumably be consistent with the same electron density ) . In contrast , these hydrogen bonds are not present in the crystal structure or in simulations of the symmetric 2-ligand EGFR dimer , and they are precluded by the gap in the dimer interface of all ligand-free dimers we examined ( including the ligand-free dEGFR crystal structure and our simulations of ligand-free EGFR and Her2 homodimers and EGFR–Her2 heterodimers ) . Our observations suggest an important role for Gln194 of the ligand-bound subunit in stabilizing 1-ligand dimers . Notably , the glutamine residue is conserved in EGFR , Her3 , Her4 , and dEGFR , which bind ligands , but not in Her2 ( Figure 6A ) , which does not . Mutating Gln194 in EGFR and Her3 may thus hinder their heterodimerization with Her2 , and mutating the corresponding residue ( Gln189 ) in dEGFR may hinder its homodimerization . 10 . 7554/eLife . 00708 . 009Figure 6 . Details of the dimer interfaces in Her2 heterodimers . ( A ) The sequence alignments of EGFR , Her2 , Her3 , and dEGFR for the part of domain II corresponding to EGFR residues 167–286 . Three sets of residues ( green , red , and blue , respectively ) , which are involved in three sets of important interactions at the dimer interface , are highlighted . ( B ) Top: the role of EGFR residue Gln194 in the EGFR–Her2 dimer interface . Hydrogen bonds and salt bridges are indicated by red dashed lines . Bottom: the distances between the key atoms involved in these hydrogen bonds are shown as functions of time . ( C ) Left: superimposition of 1-ligand EGFR homodimer and 1-ligand EGFR–Her2 heterodimer , with dimerization arms highlighted by the red box . Middle and right: Her2’s dimerization arm in the EGFR–Her2 heterodimer is conformationally less stable than the dimerization arm of the ligand-free subunit of the 1-ligand EGFR homodimer , likely because the Tyr251–Arg285 cation–π interaction is only present in the latter dimer . Dark blue is used for the EGFR dimerization arm in the EGFR homodimer , and dark gray for the EGFR dimerization arm in the EGFR–Her2 heterodimer . DOI: http://dx . doi . org/10 . 7554/eLife . 00708 . 009 Certain inter-subunit interactions appeared to be unique to Her2 heterodimers . In particular , Arg228 of EGFR interacted stably with Glu243 of Her2 through a salt bridge in our simulations of the ligand-bound EGFR–Her2 heterodimer ( Figure 6B ) . Similarly , Arg228 of Her3 interacted stably with the same glutamate through a water bridge in our simulations of the Her3–Her2 heterodimer . Although this arginine is present in both EGFR and Her3 , it is not present in Her2 , Her4 , or dEGFR . Notably , EGFR and Her3 are the two common dimerization partners of Her2 in the EGFR family , and only Her2 bears a negatively charged residue at the position of the Glu243 ( Figure 6A ) . This is consistent with the notion that Her2 has been fine-tuned for heterodimerization with EGFR and Her3 . Experimentally , heterodimerization between Her2 and ligand-bound EGFR or Her3 ectodomains is much weaker than homodimerization of EGFR , Her4 , or dEGFR ectodomains ( Horan et al . , 1995; Ferguson et al . , 2000 ) . Our simulations suggest that this difference in stability may be due to unique conformational dynamics in the dimerization arm of Her2 . The dimerization arm plays a key role in maintaining dimer stability , and dimerization arm-mediated interactions are very similar in crystal structures of dEGFR and human EGFR dimers ( Alvarado et al . , 2010 ) . In our simulations , however , the dimerization arm of Her2 in the EGFR–Her2 dimer deviated significantly from its initial conformation in the course of the simulation , whereas the dimerization arms of EGFR and dEGFR in homodimers did not ( Figure 6C ) . A cation–π interaction at the tip of the dimerization arm in the 1-ligand EGFR dimer between Tyr251 and Arg285 of the unliganded subunit ( Figure 6C ) appears to be particularly important in stabilizing the dimer . The residue at the position of Tyr251 is conserved as tyrosine or phenylalanine in dEGFR and all members of the human EGFR family . The residue at the position of Arg285 is conserved as arginine , except in Her2 , where it is replaced by leucine ( Leu291; Figure 6A ) . It is tempting to suggest that this missing arginine in Her2 may explain its weak ectodomain heterodimerization compared to the homodimerization of dEGFR , EGFR , and Her4 , and that an L291R mutation in Her2 may promote Her2 heterodimerization and facilitate the crystallization of a Her3–Her2 or EGFR–Her2 dimer .
Given the implication of Her2 in disease and its central role in mediating cell signaling , elucidation of the molecular mechanisms regulating Her2 activation is of great importance . The long-timescale MD simulations reported here support the notion of a simple and evolutionarily conserved structural mechanism controls homo- and heterodimerization of Her2 , and thereby Her2 activation . Our simulations revealed inherent instability in the dimer interfaces of the ligand-free EGFR–Her2 and Her3–Her2 heterodimers and the Her2 homodimer , and indicated that the binding of a single ligand is sufficient to stabilize the heterodimer interfaces . These observations complement the crystallographic analysis , shedding further light on the structural mechanism underlying the preference of Her2 to partner with a ligand-bound EGFR or Her3 , rather than with Her2 or ligand-free EGFR or Her3 ( Lemmon , 2009; Graus-Porta et al . , 1997; Nagy , et al . , 2010 ) . Our simulations showed the conformation of domain II , which is determined by ligand binding , to be critical to the stability of the dimer interface . A ligand bound to EGFR ( or to another member of the family ) acts as a wedge that pushes domains I and III apart , straightening the otherwise bent domain II ( Alvarado et al . , 2009 ) . ( Domain II is bent in the crystal structures of ligand-free EGFR , Her2 , Her3 , Her4 , and dEGFR . ) We simulated the transition from the straight to the bent conformation of domain II upon removal of the bound ligand in EGFR and Her3 and the consequent development of the gap in the dimer interface . Our simulations suggest that , for the gap to open , both domain IIs of the two subunits in a dimer need to adopt a bent conformation . One feature unique to the dimerization and activation of receptors of the EGFR family ( Her2 included ) , but not to other families of receptor tyrosine kinases ( Lemmon and Schlessinger , 2010 ) , is that the dimer interface is mediated by the receptors and does not involve any direct contribution from the bound ligands . Instead of directly participating in the dimerization , a bound ligand regulates the conformation of domain II and potentiates it for dimerization . In this context , it is not surprising that the seemingly minor effect of ligand binding on the conformation of domain II in the heterodimerization partners of Her2 demonstrated by our simulations is of great importance to the regulation of Her2 activity . Our consistent observation of the gap that disrupts the dimer interfaces in the ligand-free EGFR–Her2 and Her3–Her2 heterodimers , and in Her2 and dEGFR homodimers , as well as the fact that domain II is bent in crystal structures of ligand-free EGFR , Her2 , Her3 , Her4 , and dEGFR , suggests that the structural mechanism for the control of dimerization has been preserved from Drosophila to the human EGFR family .
Because EGFR , Her2 , and Her3 are closely related phylogenetically and share a high degree of sequence identity , we used the EGFR ectodomain homodimer as a structural template for the modeling of the Her2 ectodomain homodimer and the EGFR–Her2 and Her3–Her2 ectodomain heterodimers . The sequence identity between the ectodomains of EGFR and Her2 is 40% , while that between the ectodomains of EGFR and Her3 is 41% . In the structural alignments for the modeling , we used the region corresponding to EGFR residues 240–309 , since this is the part of domain II that appears to be essential for dimerization ( Dawson et al . , 2005 ) , and since in our simulations of EGFR the N-terminal portion of domain II exhibited bending around the hinge situated approximately at residue 240 ( Figure 3 ) . In this region , sequence identity to EGFR is 50% and 46% for Her2 and Her3 , respectively ( Figure 6A ) . The Her2 subunit of the EGFR–Her2 heterodimer was modeled based on the Her2 monomer crystal structure from PDB entry 3BE1 ( Bostrom et al . , 2009 ) , and missing residues in domains I–III were filled from PDB entries 1S78 ( Franklin et al . , 2004 ) and 2A91 ( Garrett et al . , 2003 ) . The Her2 monomer was aligned with the ligand-free EGFR subunit from the 1-ligand EGFR dimer using EGFR residues 240–309 ( and the corresponding residues of Her2 ) for reference . The part of domain IV of Her2 that is missing in the crystal structures ( beyond residue 608 ) was modeled based on the EGFR structure . The minor clashes between Her2 and EGFR in the original model were resolved by adjusting the side chains of the amino acids involved . To obtain the Her2 homodimer , the Her2 monomer was used to replace the EGFR subunit in the ligand-free EGFR–Her2 heterodimer . The Her3–Her2 heterodimer was modeled based on our model of the EGFR–Her2 dimer taken at t = 0 ( i . e . , before simulation ) . Individual domains from the crystal structure of the tethered Her3 ectodomain ( PDB entry 1M6B [Cho and Leahy , 2002] ) were aligned on the respective domains of the EGFR template . Because domain II is straight in the EGFR template and bent in the Her3 crystal structure , the alignment was carried out in three separate steps . Specifically , Her3 residues 166–188 , 189–238 , and 239–307 were separately aligned with their EGFR counterparts . The residues at the borders between the parts of Her3 that were aligned as rigid bodies were then adjusted to ensure appropriate connectivity; a few minor clashes between the domains of Her3 and between Her3 and Her2 were eliminated by adjusting amino acid side chains . HRG was taken from PDB entry 1HAE ( Jacobsen et al . , 1996 ) and aligned with the EGF bound to the EGFR template , using EGF’s residues 26–46 for reference ( as this part appears to be most similar structurally between the two ) . Domains IVs of receptors of the EGFR family are generally more flexible than the other ectodomains and are commonly unresolved in crystal structures . Thus , structural analysis of domain IVs starting from homology models may be particularly challenging . Although domain IVs were included in our models and simulations , they were not a main concern of this study . The simulations were performed on the special-purpose supercomputer , Anton ( Shaw et al . , 2009 ) , using the CHARMM22* force field ( MacKerell et al . , 1998; Piana et al . , 2011 ) for proteins and TIP3P ( Jorgensen et al . , 1983 ) as the water model . The simulated systems were solvated in water with NaCl ( Na+ ions were first added to neutralize the net charge of the system , and then equal numbers of Na+ and Cl− were added so that the concentration of Na+ reached 0 . 15 M ) , with residues set to their dominant protonation states at pH 7 . As an equilibration stage , the protein backbone atoms were first restrained to their initial positions , using a harmonic potential with a force constant of 1 kcal mol−1 Å−2 . The force constant was linearly scaled down to zero over 50 ns . Simulations were performed in the NPT ensemble for the equilibration step and in the NVT ensemble afterwards , with T = 310 K , p=1 bar , and Berendsen’s coupling scheme ( Berendsen et al . , 1984 ) with one temperature group . Water molecules and all bond lengths to hydrogen atoms were constrained using M-SHAKE ( Kräutler et al . , 2001 ) . Van der Waals and short-range electrostatic interactions were cut off at 12 . 5 Å . Long-range electrostatic interactions were calculated using the k-space Gaussian split Ewald method ( Shan et al . , 2005 ) with a 64 × 64 × 64 mesh . The simulation time step was 1 fs for the equilibration stage and 2 . 5 fs for production simulations; the r-RESPA integration method was used with long-range electrostatics evaluated every 5 fs ( Tuckerman et al . , 1992 ) . The simulated systems included the EGFR ectodomain dimers with two EGF ligands ( three simulations ) or one ligand ( also three simulations ) ; 1-ligand EGFR–Her2 ectodomain heterodimer ( one simulation ) , ligand-free heterodimer ( two simulations ) , Her2 homodimer ( two simulations ) , 1-ligand Her3–Her2 heterodimer ( two simulations ) , and ligand-free Her3–Her2 heterodimer ( two simulations ) ; and ligand-free dEGFR dimer ( two simulations starting from the 2-ligand crystal structure and one from the ligand-free crystal structure ) . The size of the simulated systems was ∼270 , 000 atoms , in a periodic box of approximately 140 × 140 × 140 Å3 . Modeling , analysis , and visualization were performed using VMD ( Humphrey et al . , 1996 ) . The data for buried area shown in the figures was averaged over a 200-ns window . Repeated simulations of a given system produced largely similar results . In both simulations of the ligand-free EGFR–Her2 dimer , for example , a gap consistently developed at the dimer interface ( Figure 4B ) . Three simulations of the 1-ligand EGFR homodimer ( Figure 3D ) , generated average structures with a root-mean-squared deviation of 1 . 9–4 . 8 Å from one another , as measured based on the Cα atoms of domains I–III . By the same measurement , the two crystal structures of the 2-ligand EGFR homodimer ( PDB entries 1IVO and 1MOX ) differ from one other by 4 . 0 Å . | ErbB proteins are found in most multi-cellular organisms , and are involved in the regulation of a number of important cellular processes , including proliferation , migration , and differentiation . Humans have four ErbB proteins , which span the plasma membrane of cells . These proteins respond to interactions with molecules outside the cell—such as growth factors and hormones—by sending signals along the appropriate signaling pathway within the cell . ErbB proteins have three portions: an ectodomain that extends outside the cell; a single helix that spans the membrane; and a cytoplasmic domain inside the cell . When a signaling ligand molecule outside the cell binds to the ectodomain of an ErbB protein , this protein must then combine with another ErbB protein to form a dimer before a signal can be sent within the cell . These dimers can include two copies of the same ErbB protein or two different ErbB proteins . However , one of the ErbB proteins—Her2—works in a different way . It cannot bind ligands outside the cell , and it can only send a signal within the cell if it first forms a dimer with an ErbB protein of another type , which itself must be bound to an external ligand . The four ErbB proteins diverged from a common ancestor relatively recently , yet they are now diverse enough to play key roles in a variety of complex signaling networks . In particular , the fact that Her2 cannot bind external ligands , and that it must form a dimer with a different ErbB protein before it can send a signal , has led to suggestions that the role of Her2 is to amplify the signals from other ErbB proteins . Since high levels of Her2 are associated with aggressive forms of breast and ovarian cancer , understanding how it is activated could improve our understanding of these cancers . Arkhipov et al . have now used computer simulations to model how Her2 forms dimers with other ErbB proteins in human cells . They based these simulations on crystal structures of human ErbB proteins and dEGFR , a growth-factor receptor found in fruit flies that closely resembles the ErbB proteins found in humans . They found that the dimers were stable as long as one protein within the dimer was bound to a ligand . Removing this ligand , however , distorted the ectodomain of the host protein , creating a gap that weakened the dimer and prevented Her2 from sending a signal within the cell . Similar results were obtained with the fruit fly dEGFR proteins . These simulations suggest that ErbB proteins form dimers and send signals through a mechanism conserved in evolution . Research in this field might help ongoing efforts to develop new treatments for human tumors characterized by high levels of Her2 expression . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2013 | Her2 activation mechanism reflects evolutionary preservation of asymmetric ectodomain dimers in the human EGFR family |
High resolution crystal structures of DNA polymerase intermediates are needed to study the mechanism of DNA synthesis in cells . Here we report five crystal structures of DNA polymerase I that capture new conformations for the polymerase translocation and nucleotide pre-insertion steps in the DNA synthesis pathway . We suggest that these new structures , along with previously solved structures , highlight the dynamic nature of the finger subdomain in the enzyme active site .
DNA polymerase I ( DNAP-I ) has long been viewed as the canonical model for DNA synthesis in cells ( Lehman et al . , 1958 ) . Structural insights into the mechanism of DNA synthesis have been obtained from crystal structures of a thermostable bacterial ( Geobacillus stearothermophilus , Bst ) DNAP-I large fragment that retains catalytic activity inside the crystal lattice ( Johnson et al . , 2003; Kiefer et al . , 1998 ) . The prevailing mechanism invokes the use of a distinct pre-insertion site , observed in the translocated product of in crystallo catalyzed primer-extension reactions where dNTP substrates are soaked into pre-formed crystals of DNAP-I bound to a primer-template duplex ( Figure 1—figure supplement 1 ) ( Johnson et al . , 2003; Kiefer et al . , 1998 ) . The pre-insertion site is a hydrophobic pocket located between the O and O1 helices of the finger subdomain where the n + 1 templating base resides prior to forming the nascent base pair with the incoming dNTP substrate ( Johnson et al . , 2003 ) . However , the pre-insertion site has not been witnessed in polymerases with homologous active sites ( Eom et al . , 1996; Li et al . , 1998; Yin and Steitz , 2002 ) , implying that DNAP-I follows a complex enzymatic pathway that contains numerous intermediates , many of which have not yet been observed in protein crystals . Here we report five crystal structures of DNAP-I that capture new conformations for the polymerase translocation and nucleotide pre-insertion steps in the DNA synthesis pathway . Together , these structures provide new insight into the mechanism of DNA synthesis and highlight the dynamic nature of the finger subdomain in the enzyme active site .
Recognizing that in crystallo and solution catalyzed enzymatic reactions can produce different structural results with potentially different functional interpretations ( Ehrmann et al . , 2017 ) , we chose to investigate the translocated intermediates of DNAP-I using a direct crystallization method that involves solving crystal structures of the enzyme-product complex obtained from primer-extension reactions performed in solution rather than inside the environment of a protein crystal . In these reactions , the starting enzyme-primer-template complex was incubated with solutions of either buffer , dTTP , or dTTP and dATP for 30 min at 37°C . Following primer-extension , the enzyme-product complex was crystallized and cocrystal structures of Bst DNAP-I were solved to resolutions of 1 . 5 – 2 . 0 Å ( Table 1 ) . This approach was used to obtain high resolution structures of DNAP-I for the starting primer-template complex ( n ) and two translocated products obtained for the n + 1 and n + 2 nucleotide addition steps using the same primer-template duplex ( n ) described in previous in crystallo studies ( Figure 1a ) ( Johnson et al . , 2003 ) . Structures of the enzyme-primer-template complex ( n ) before catalysis reflect the initiation step of DNA synthesis . Superposition of the new structure obtained for the initiation step against the previously solved structure reveals that both structures adopt the same active site conformation ( Figure 1—figure supplement 2a ) . This result implies that any structural differences observed between the translocated product of solution and in crystallo catalyzed reactions should be due to the catalysis environment rather than the starting polymerase conformation . To evaluate the elongation step of DNA synthesis , the translocated products obtained from solution and in crystallo catalyzed primer-extension reactions were compared , both globally and locally within the enzyme active site ( Johnson et al . , 2003; Kiefer et al . , 1998 ) . All of the structures adopt the same overall topology commonly observed for A-family DNA polymerases ( Figure 1b ) . However , careful analysis of the enzyme active site did reveal clear conformational differences between structures obtained from solution-catalyzed reactions versus those obtained from in crystallo catalyzed reactions ( Figure 1c , d ) . The in crystallo catalyzed reactions adopt an active site conformation that is nearly identical to the starting conformation , which represents the initiation step of DNA synthesis ( Figure 1—figure supplement 2a ) . However , the solution catalyzed reactions produce a different active site conformation that binds the duplex in a different position and base pair geometry ( Figure 1—figure supplement 2b , c ) . Major structural differences are depicted in the 2D interaction maps , which show that the solution catalyzed reactions produce a translocated product with markedly fewer contacts to the phosphodiester linkage , sugar , and nucleobase moieties of the primer-template duplex as compared to the translocated product obtained by in crystallo catalysis ( Figure 1—figure supplements 3 and 4 , Supplementary file 1a ) . A particularly striking example of conformational disparity is Tyr714 , a critical active site residue involved in the mechanism of DNA synthesis ( Bell et al . , 1997; Carroll et al . , 1991 ) . In the solution catalyzed structures , Tyr714 stabilizes the newly formed base pair by stacking above the primer strand , while this residue stacks above the template strand in the in crystallo catalyzed structures ( Figure 1c , d ) . Importantly , the pre-insertion site is not observed in the solution catalyzed reactions due to a kink in the O-helix , which abrogates the O-O1 loop in the finger subdomain ( Figure 1d ) . Absent a hydrophobic pocket , the n + 1 nucleotide in the template strand stacks against Tyr719 in the O1 helix , which positions the base for a subsequent round of catalysis . The solution catalyzed structures obtained for the n + 1 and n + 2 translocated products adopt identical active site conformations ( Figure 1 – figure supplement 2d ) , which together represent a new intermediate along the DNA replication pathway of Bst DNAP-I . Next , we examined whether a solution catalyzed conformation could be converted to an in crystallo conformation through a round of in crystallo catalysis . Accordingly , dATP was soaked into a crystal of the n + 1 translocated product obtained by crystallization of a solution catalyzed reaction . Following one cycle of in crystallo catalysis , an n + 2 translocated structure was produced that now contained the pre-insertion site and matched the active site conformation of previous in crystallo results ( Figure 1 – figure supplement 2e , f ) . This observation demonstrates that in crystallo catalysis favors an active site conformation that contains the pre-insertion site , as the same active site conformation is obtained from two different starting points . Interestingly , the translocated product obtained from the set of solution catalyzed reactions is similar to known Bst DNAP-I structures solved with duplexes that contain damaged DNA intermediates and active site mutations ( Figure 1—figure supplement 5 , Supplementary file 1b ) . These structures were previously thought to contain a distorted active site conformation due to the position of Tyr714 relative to its conformation in the in crystallo catalysis structures ( Gehrke et al . , 2013; Johnson and Beese , 2004; Wang et al . , 2012 ) . However , given the homology of these structures to the translocated product of solution catalyzed reactions , we postulate that Tyr714 functions as a regulatory checkpoint in the mechanism of DNA synthesis by evaluating the geometry of the newly formed base pair . Next , we wondered whether the mechanism of DNAP-I included the formation of a pre-insertion complex , which is a ternary structure different from the previously discussed pre-insertion site observed in the binary structure of in crystallo catalyzed primer-extension reactions . Previously , Wu and colleagues solved the ternary structure of a mutant version of Bst DNAP-I bound to an incoming dATP substrate ( Miller et al . , 2015 ) . Although that structure was originally described as an open ternary complex , presumably to avoid confusion with the pre-insertion site , it resembles the pre-insertion complex first observed in Klentaq1 ( Li et al . , 1998 ) . The key difference between the open ternary and pre-insertion complex is whether the incoming nucleotide is paired opposite the templating base or an active site residue ( Doublié et al . , 1998; Yin and Steitz , 2004 ) . Since the structure by Wu and colleagues shows the incoming substrate paired opposite Tyr714 , it should be considered a pre-insertion complex . We demonstrated that the wild-type polymerase is also capable of forming a pre-insertion complex by solving the ternary structure of the enzyme bound to the non-hydrolyzable analog , dAMPNPP . The resulting structure ( Figure 2 ) closely resembles the mutant Bst polymerase structure determined by Wu and colleagues and shows Tyr714 paired opposite the incoming nucleotide ( Miller et al . , 2015 ) . Although the phosphate tail shows nearly 100% occupancy , the sugar and nucleobase moieties are flexible , which is consistent with the dynamic properties of the incoming nucleotide in an open polymerase conformation . Nevertheless , the structure shows that the incoming nucleotide is stabilized by polar contacts to the negatively charged triphosphate moiety . These observations demonstrate that Bst DNAP-I adopts a pre-insertion complex similar to other A-family DNA polymerases ( Rothwell and Waksman , 2005 ) , which clarifies an important step in the mechanism of DNA synthesis . Based on the structures reported here , we propose a revised mechanism for DNA synthesis by DNA polymerase I . The catalytic cycle consists of four key steps that derive from high resolution structures of Bst DNAP-I and its homolog T7 RNA polymerase ( Figure 3 ) . Starting from the newly determined post-translocation complex , the polymerase undergoes a conformation change to adopt the pre-insertion complex with an incoming nucleotide paired opposite Tyr714 in the enzyme active site . This conformational change involves release of the n + 1 templating base from its stacking interaction with Tyr719 in the O1 helix and the repositioning of Tyr714 in the enzyme active site . The enzyme then undergoes a more significant conformational change to adopt the closed ternary complex ( Johnson et al . , 2003 ) , which defines the pre-catalytic state of the enzyme . Immediately following phosphodiester bond formation , the enzyme adopts a post-catalytic complex in which the primer has been extended by one nucleotide ( Yin and Steitz , 2004 ) . The enzyme then translocates to the next position on the template to initiate another cycle of nucleotide addition . In summary , we present crystal structures of DNA polymerase I that capture the translocation and nucleotide pre-insertion steps in the DNA synthesis pathway . We suggest that these new structures , along with previously solved structures obtained by in crystallo catalysis , highlight the dynamic nature of the finger subdomain in the enzyme active site . Together , the new and existing structures expand our understanding of the mechanism of DNA synthesis by capturing important intermediates in a complicated reaction pathway .
The Bst ( amino acid residues 299–876 ) gene was PCR amplified from a previously constructed pDEST007-Bst vector generously donated by Prof Thomas Carell using Bst_for ( ATCCATATGGCATTTACGCTTGCTGAC , IDT ) and Bst_rev ( ATGCGGCGGTCTCC TCGAGTCATTATTTCGCATCATACCACG , IDT ) primers containing NdeI and BsaI restriction enzyme sites ( underlined ) , respectively . Purified PCR product and the expression vector , pGDR11 , were digested with NdeI and BsaI restriction enzymes ( NEB ) and ligated and the resulting pGDR11-Bst construct was sequence verified ( Retrogen ) . DH5-α cells ( NEB ) harboring pGDR11-Bst were grown aerobically at 37°C in LB medium containing 100 μg mL−1 ampicillin . At an OD600 of 0 . 8 , expression of a tagless Bst was induced with 1 mM isopropyl β-D-thiogalactoside at 18°C for 16 hr . Cells were harvested by centrifugation for 20 min at 3315 x g at 4°C and lysed in 40 mL lysis buffer ( 50 mM Tris-Cl pH 7 . 5 , 1 mM EDTA , 10 mM BME , 0 . 1 % v/v NP-40 , 0 . 1 % v/v Tween20 , 5 mg egg hen lysozyme ) by sonication . The cell lysate was centrifuged at 23 , 708 x g for 30 min and the clarified supernatant was heat treated for 20 min at 60°C and centrifuged again at 23 , 708 x g for 30 min . The supernatant was loaded onto two 5 mL HiTrap Q HP columns ( GE ) assembled in tandem and washed with low salt buffer ( 50 mM Tris-Cl pH 7 . 5 , 100 mM NaCl , 1 mM EDTA , 10 mM BME ) . Bst was eluted with a high salt buffer ( 50 mM Tris-Cl pH 7 . 5 , 1M NaCl , 0 . 1 mM EDTA , 10 mM BME ) using a linear gradient . Eluted fractions containing Bst were visualized by SDS-PAGE , pooled , and dialyzed against low salt buffer . The dialyzed sample was loaded onto a 5 mL HiTrap Heparin column ( GE ) , washed with low salt buffer , and eluted using a linear gradient of high salt buffer . Eluted fractions containing Bst were visualized using SDS-PAGE and concentrated using a 30 kDa cutoff Amicon centrifugal filter ( Millipore ) . Further purification was achieved by size exclusion chromatography ( Superdex 200 HiLoad 16/600 , GE ) pre-equilibrated with Bst buffer ( 50 mM Tris-Cl pH 7 . 5 , 150 mM NaCl , 1 mM EDTA , 10 mM BME ) . Purified Bst was concentrated to 20 mg mL−1 for crystallization trials using a 30 kDa cutoff Amicon centrifugal filter ( Millipore ) . | DNA molecules consist of two separate strands that spiral around each other to form a structure called the double helix . Each strand contains repeating units , with every unit consisting of a phosphate group and a sugar molecule bound to one of four bases . The two strands are held together by bonds between the bases . When a cell divides , it needs to make a copy of the DNA , so that each new cell will have an exact replica from the old cell . During this process , the helix unwinds and enzymes called polymerases produce new strands ( using the old ones as a template ) . Each strand is copied by adding new bases one at a time . Every time a new base is added , the polymerases must modify their structures several times . If this process becomes faulty , it can lead to various diseases , including cancer . Scientist often use a technique called X-ray crystallography to study intermediate structures of frozen polymerase crystals as the enzyme constructs DNA . Yet , to fully understand the mechanisms of DNA synthesis all intermediate structures need to be identified . Now , Chim , Jackson et al . used a particular method for making frozen polymerase crytals by allowing the enzyme to add new bases in liquid form . The reaction was then frozen and X-ray crystallography was used to take images . This modified method captured different steps in the process and detailed how the enzyme adjusts its structure as it moves along the template strand . The intermediate structures that Chim , Jackson et al . uncovered may help scientists develop new biotechnologies and medicines . Understanding how polymerases modify their form while making DNA copies could lead to better therapies for diseases in which this process has become faulty , like cancer . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2018 | Crystal structures of DNA polymerase I capture novel intermediates in the DNA synthesis pathway |
A broadly known method to stimulate the growth potential of axons is to elevate intracellular levels of cAMP , however the cellular pathway ( s ) that mediate this are not known . Here we identify the Dual Leucine-zipper Kinase ( DLK , Wnd in Drosophila ) as a critical target and effector of cAMP in injured axons . DLK/Wnd is thought to function as an injury ‘sensor’ , as it becomes activated after axonal damage . Our findings in both Drosophila and mammalian neurons indicate that the cAMP effector kinase PKA is a conserved and direct upstream activator of Wnd/DLK . PKA is required for the induction of Wnd signaling in injured axons , and DLK is essential for the regenerative effects of cAMP in mammalian DRG neurons . These findings link two important mediators of responses to axonal injury , DLK/Wnd and cAMP/PKA , into a unified and evolutionarily conserved molecular pathway for stimulating the regenerative potential of injured axons .
Repair of lost axonal connections generally fails to occur after neuronal injury in the adult mammalian central nervous system ( CNS ) . This failure is not only a reflection of the growth inhibitory nature of CNS tissue ( Fawcett et al . , 2012; Filbin , 2003; Silver et al . , 2015 ) , but also due to the lack of intrinsic capacity for neurons in the adult CNS to grow axons ( Liu et al . , 2011; Sun and He , 2010 ) . However , landmark studies by Richardson and Issa have suggested that neurons indeed possess an innate ability to regenerate their axons in the adult mammalian CNS , and that this ability can be unlocked by a ‘conditioning lesion’ ( Richardson and Issa , 1984 ) . In adult DRG neurons , an injury to peripherally projecting axons , i . e . a compression injury to the sciatic nerve , unleashes growth programs within the DRG and allows for regeneration of its centrally projecting axons in the spinal cord ( Neumann and Woolf , 1999; Richardson and Issa , 1984 ) . This growth can be induced by a lesion in the peripheral nervous system ( PNS ) even after the CNS lesion has occurred ( Ylera et al . , 2009 ) , hence it is of great interest from a therapeutic perspective to understand the molecular mechanisms that allow for the unlocking of such regenerative potential . Previous studies have discovered that several signal transduction pathways are activated in DRG neurons upon a conditioning injury , including JAK-STAT3 ( Qiu et al . , 2005 ) , ATF3 ( Fagoe et al . , 2015; Hollis and Zou , 2012 ) , Smad1 ( Zou et al . , 2009 ) , Activin ( Omura et al . , 2015 ) , HIF-1alpha ( Cho et al . , 2015 ) and cAMP ( Qiu et al . , 2002; Neumann et al . , 2002; Cai et al . , 1999 ) . Impressively , ectopic elevation of cAMP alone is sufficient to strongly enhance regeneration ( Xiao et al . , 2015; Qiu et al . , 2002; Neumann et al . , 2002 ) . Since this second messenger is commonly modulated by growth signals and neuronal activity , cAMP modulation has been suggested as a potential therapeutic inroad to stimulate the regenerative potential of neurons ( Xiao et al . , 2015 ) . However , the downstream pathways that are engaged by this broadly utilized second messenger to actually promote axonal regeneration are not known . Much attention has focused upon the cAMP-responsive element binding protein ( CREB ) , since constitutive activation of CREB is sufficient to stimulate axonal regeneration in the presence of CNS myelin in vivo ( Gao et al . , 2004 ) . However , more recent studies indicate that endogenous CREB is not required for cAMP elicited axonal regeneration in vitro ( Ma et al . , 2014 ) . Hence it remains elusive how cAMP elevation activates axonal regrowth programs in neurons . A recent study has identified an essential role for the dual zipper-bearing kinase DLK in the pro-regenerative effect of a conditioning lesion in adult DRG neurons ( Shin et al . , 2012 ) . Similarly , the Drosophila homologue Wallenda ( Wnd ) , mediates protective effects of a conditioning lesion in Drosophila motoneurons ( Brace and DiAntonio , 2016; Xiong and Collins , 2012 ) . This conserved axonal mitogen activated kinase kinase kinase ( MAPKKK ) is thought to function as a sensor of axonal damage , and therefore should become activated upon conditioning injury . In support of this , Wnd/DLK is transported in axons ( Xiong et al . , 2010 ) and is required acutely in injured axons for the generation of signals that are retrogradely transported to the cell body ( Xiong et al . , 2010; Shin et al . , 2012 ) . DLK/Wnd is required for axonal regeneration in many types of neurons , including motoneurons in mammals , flies and worms , and CNS neurons where regeneration is ectopically induced by PTEN mutations ( Yan et al . , 2009; Hammarlund et al . , 2009; Xiong et al . , 2010; Shin et al . , 2012; Watkins et al . , 2013 ) . Conversely , in mammalian CNS neurons that do not regenerate ( eg . retinal ganglion cells , RGCs ) , DLK activation after injury mediates cell death ( Welsbie et al . , 2013; Watkins et al . , 2013 ) . Collectively , these findings support the model that a conserved function of the Wnd/DLK kinase is to ‘sense’ axonal damage . Through a yet unknown mechanism , axonal damage leads to activation of Wnd/DLK’s kinase function . Once activated , downstream signaling mediates both beneficial and deleterious outcomes in neurons , depending upon the context . The high stakes outcomes of regeneration or death , combined with additional findings that DLK mediates cell death in models for nerve growth factor withdrawal ( Huntwork-Rodriguez et al . , 2013; Ghosh et al . , 2011 ) , glaucoma ( Welsbie et al . , 2013 ) , MPTP toxicity ( Mathiasen et al . , 2004 ) and excitotoxicity ( Pozniak et al . , 2013 ) , have inspired much interest in understanding the unknown pathways that lead to the activation of DLK/Wnd in injured axons . Here we identify a direct upstream activator of DLK/Wnd in injured axons , in the form of the cAMP effector kinase PKA . We find that PKA phosphorylates evolutionarily conserved serines within the activation loop of DLK , which is sufficient to activate DLK independently of its downstream signaling mechanisms . In addition , our functional studies in both Drosophila motoneurons and adult mammalian DRG neurons indicate that the ability of cAMP and PKA to promote axonal regeneration depends entirely upon the ability of PKA to activate the DLK/Wnd kinase . These findings present a unified and evolutionarily conserved molecular pathway , from cAMP to PKA to DLK , which plays a central role in stimulating the ability of injured axons to regenerate .
Previous studies in mammalian and C . elegans neurons suggest that cAMP signaling stimulates regenerative axonal growth ( Qiu et al . , 2002; Neumann et al . , 2002; Cai et al . , 1999; Ghosh-Roy et al . , 2010 ) . To study this axon regeneration pathway in Drosophila , we used previously developed axon injury assay in third instar larvae ( Xiong et al . , 2010 ) , and found that knockdown of phosphodiesterase dunce ( dnc ) or activation of PKA by overexpression of the catalytic subunit ( PKACA ) ( Li et al . , 1995 ) led to an enhanced growth response of Drosophila motoneuron axons after nerve crush injury ( Figure 1A ) . The new axonal growth from the injured proximal stump generally assumes a highly branched shape , characterized by a network of small branches and a general thickening of the axon diameter . To assess the injury response , we quantified the total membrane volume within 100 µm of the axonal tip ( indicated by the dash line in Figure 1A ) . In control animals , this total volume increases 3 fold , from 68 . 5 µm3 to 200 µm3 15 hr after injury . PKA activation led to a 1 . 5 fold increase in this volume compared to control ( WT ) axons ( Figure 1B ) . The enhanced sprouting response stimulated by PKA was lost when DLK/Wnd function was inhibited by co-expression of RNAi targeting Wnd ( but not a control RNAi ) ( Figure 1A and B ) . These observations are consistent with the previous finding in C . elegans that DLK is required for the regeneration that is induced by cAMP signaling ( Ghosh-Roy et al . , 2010 ) . In addition , PKA alone is required for Drosophila motoneurons to initiate regenerative sprouting , as RNAi-mediated knockdown of the PKA catalytic subunit inhibited the sprouting response by 50% compared to control axons ( Figure 1A and B ) . cAMP and PKA therefore play an influential role in the regenerative capacity of Drosophila motoneuron axons . 10 . 7554/eLife . 14048 . 003Figure 1 . PKA stimulates and is required for axonal regeneration in Drosophila motoneurons . Single motoneuron axons were labeled by expression of UAS-mCD8-GFP using the m12-Gal4 driver and imaged either 0 hr or 15 hr after nerve crush injury . Representative images are shown in ( A ) , while ( B ) shows quantification of the increased volume in axonal membrane , which is measured within 100 µm of the proximal axon tip , indicated in dotted line . Genotypes used in ( A ) : wild type ( WT ) ( ;;m12-Gal4 , UAS-mCD8GFP/+ ) ; control-RNAi ( UAS-dcr2;; UAS-moody-RNAi/m12-Gal4 , UAS-mCD8GFP ) ; dnc-RNAi ( UAS-dcr2;; UAS-dnc-RNAi/m12-Gal4 , UAS-mCD8GFP ) ; wnd-RNAi ( UAS-dcr2;; UAS-wnd-RNAi/m12-Gal4 , UAS-mCD8GFP ) ; PKACA ( ; UAS-PKACA/+; UAS-mCD8GFP/+ ) ; PKACA , control-RNAi ( UAS-dcr2; UAS-PKACA/+; UAS-moody-RNAi ( VDRC 100674 ) /m12-Gal4 , UAS-mCD8GFP ) ; PKACA , wnd-RNAi ( UAS-dcr2; UAS-PKACA/+; UAS-wnd-RNAi/m12-Gal4 , UAS-mCD8GFP ) ; pka-c1-RNAi ( UAS-dcr2; UAS-pka-c1-RNAi/+; m12-Gal4 , UAS-mCD8GFP [using two different lines , Bloomington 31277 and 35169] ) . All data are represented as mean ± SEM; At least 10 animals ( ≥50 axons ) are examined per genotype; ***p<0 . 001; ‘n . s . ’ indicates non-significant; scale bar , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14048 . 003 While the above and previously described genetic interactions ( Ghosh-Roy et al . , 2010 ) suggest a relationship , whether Wnd/DLK functions downstream of PKA or in a parallel pathway cannot be discerned from genetic epistasis alone . To probe the relationship between PKA and Wnd , we first utilized previously established tools in Drosophila for monitoring the activation of Wnd and downstream nuclear signaling . Wnd signaling induces expression of the c-Jun N-terminal Kinase ( JNK ) phosphatase puckered ( puc ) , which can be measured as lacZ expression using fly lines that contain the puc-lacZ enhancer trap reporter ( Xiong et al . , 2010 ) . Puc-lacZ is expressed at low levels in uninjured motoneurons , however it is induced by axonal injury in a manner that requires both Wnd and JNK kinase function ( Xiong et al . , 2010 ) . We found expression of either dnc-RNAi or PKACA induced the expression of puc-lacZ in motoneurons ( Figure 2A ) . This induction is Wnd dependent , as RNAi knockdown of Wnd ( but not a control RNAi ) rescued the puc-lacZ elevation ( Figure 2A ) . 10 . 7554/eLife . 14048 . 004Figure 2 . PKA modulates the levels of Wnd protein and downstream signaling in Drosophila neurons . ( A ) The puc-lacZ transcriptional reporter for Wnd/JNK signaling indicates that activated PKA stimulates Wnd signaling . A pan-neuronal driver ( BG380-Gal4 ) is used to express UAS-dnc-RNAi , UAS-PKACA or UAS-PKACA together with UAS-wnd-RNAi or a control-RNAi . Example images are shown of cell bodies in the dorsal midline of the ventral nerve cord; ( all but two of these neurons are motoneurons ) . Quantification ( described in methods ) was carried out for 10 animals per genotype . ( B ) Endogenous Wnd protein levels are increased in PKACA expressing neurons . Ventral nerve cords were dissected from third instar larvae ( BG380-Gal4 [WT control] and BG380-Gal4; UAS-PKACA/+ ) and processed for Western blotting with anti-Wnd and anti-tubulin antibodies . The quantification shows Wnd/tubulin ratios ( normalized to WT control ) averaged from 3 independent experiments ( 25 nerve cords per experiment ) . ( C ) PKA increases DLK levels via a posttranscriptional mechanism . GFP-tagged kinase dead Wnd ( GFP-WndKD ) was ectopically expressed using m12-Gal4 driver ( WT control ) or co-expressed with UAS-PKACA and imaged directly after fixation . Example images and quantification of GFP-WndKD intensity in cell bodies and axons within segmental nerves . n>10 animals for each condition . ( D ) PKA-C1 is required for induction of Wnd protein after axonal injury . Example images and quantification of GFP-WndKD in nerve cords and segmental nerves before and after ( 8 hr ) injury . UAS-GFP-WndKD was expressed in motoneurons by OK6-Gal4 . WT or together with UAS-pka-c1-RNAi or UAS-moody-RNAi ( control ) and imaged similarly to Figure 2C . The quantification method for GFP intensity is described in materials and methods . n>10 animals for each condition . All data are represented as mean ± SEM; ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 , ‘n . s . ’ indicates non-significant; scale bars , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14048 . 004 Previous studies in multiple organisms suggest that Wnd/DLK is highly regulated at the level of protein turnover and increased levels of DLK correlate with the activation of downstream signaling ( Xiong et al . , 2010; Huntwork-Rodriguez et al . , 2013; Welsbie et al . , 2013; Collins et al . , 2006; Nakata et al . , 2005; Nihalani et al . , 2000; Hammarlund et al . , 2009 ) . We therefore tested whether PKA activation altered Wnd levels . The total levels of endogenous Wnd within 2nd instar larval brains were significantly elevated ( by 75% ) when PKACA was expressed in neurons ( Figure 2B ) . To test whether the change of Wnd level is due to a posttranscriptional mechanism , we used the Gal4/UAS system to ectopically express a GFP tagged Wnd transgene in Drosophila motoneurons . Since overexpression of WT Wnd can cause lethality , we expressed an inactive ( kinase dead ) version ( GFP-WndKD ) that contains a point mutation in the kinase domain ( Collins et al . , 2006 ) . We found that expression of PKACA induced a 12-fold increase in the levels of GFP-WndKD in motoneuron axons ( Figure 2C ) . In contrast , GFP-WndKD was not significantly altered in cell bodies ( Figure 2C ) . The increase in axonal GFP-WndKD when PKA is induced has remarkable similarity to what occurs in axons after injury ( Xiong et al . , 2010; Huntwork-Rodriguez et al . , 2013 ) . We therefore tested whether PKA is required for the induction of Wnd in proximal axons after nerve crush injury . In control ( WT ) motoneurons , a significant increase in the mean intensity of the GFP-WndKD was observed 7 hr after injury ( Figure 2D ) . This increase is abolished by co-expression of RNAi targeting PKA-C1 , but not a control RNAi ( Figure 2D ) . These observations suggest that PKA is required for the activation of Wnd signaling and the induction of Wnd protein levels downstream of axonal injury . To test whether the regulation of DLK by PKA is conserved in mammalian neurons , we examined endogenous DLK protein in cultured embryonic rat cortical neurons . Treatment with forskolin , which activates PKA via elevation of cAMP , led to a 2-fold increase in the level of endogenous DLK protein ( Figure 3A ) . This effect of cAMP elevation requires PKA , since the increase in DLK levels was abolished by co-treatment with the PKA inhibitor H-89 ( Figure 3A ) . In contrast , treatment with H-89 alone led to a significant reduction in the DLK levels ( Figure 3B ) . Similar results were observed in HEK293 cells co-transfected with Flag-tagged DLK and either a control empty plasmid or PKACA . PKACA induces an approximately two-fold increase in DLK protein levels ( Figure 3E and 4B ) . PKACA also stimulates a phosphatase-sensitive increase in DLK molecular weight , which is most visible when equal amounts of DLK protein are compared ( Figure 3F ) . 10 . 7554/eLife . 14048 . 005Figure 3 . PKA activates DLK via phosphorylation of its activation loop . ( A-B ) Changes in endogenous DLK abundance in response to treatment with forskolin ( 30 µM ) ( A ) or the PKA inhibitor H-89 ( 5 µM ) ( B ) for 6 hr in cultured rat embryonic cortical neurons . Quantification shows relative DLK/Tubulin levels in western blots . ( C ) Alignment of activation loop sequences in different species . ( D ) The anti-pDLKS302 antibodies recognize transfected Flag-DLKWT , but not activation loop mutation Flag-DLKS301A , S305A . Both proteins were transiently expressed in HEK293 cells . Western blots were probed with anti-pDLKS302 antibody , anti-Flag antibody to detect the total DLK expression levels , and anti-Tubulin ( which remains similar in all manipulations ) for normalization . ( E ) PKACA stimulates phosphorylation of DLK S302 in HEK293 cells . HEK293 cells were co-transfected with Flag-tagged DLKWT and an empty control plasmid or PKACA . Cell lysates were probed with anti-DLK antibody , anti-pDLKS302 antibody , anti-PKA C antibody and anti-tubulin antibody . ( F ) PKACA stimulates an increase in DLK molecular weight . Flag-DLK protein was immunoprecipitated from HEK293 cells co-transfected with DLK and either Flag-tagged DLKWT and an empty control plasmid or PKACA . The immunoprecipitated Flag-DLK was then incubated with either glycerol ( control ) or lambda protein phosphatase ( λPP ) . PKACA induced a upward shift in DLK molecular weight , which was lost upon phosphatase treatment . ( G ) The activation loop is required for axonal regeneration in Drosophila neurons . Single axons in Drosophila third instar larva are labeled by mCD8RFP using eve-Gal4 driver . 24 hr after injury , these neurons in animals heterozygyous for wnd ( wnd3/+ ) show robust axonal sprouting . However , sprouting fails to occur in wnd3/wnd3 animals . Expression of UAS-Wnd ( WT ) can restore axonal regeneration in wnd mutant background ( UAS-Wnd , wnd3; wnd3 , eve-Gal4 , UAS-mCD8RFP ) . However , expression of activation loop mutant UAS-WndS301A , S305A failed to rescue the sprouting defect in wnd mutant animals ( UAS-WndS301A , S305A , wnd3; wnd3 , eve-Gal4 , m12-mCD8RFP ) . Quantification of the volume of axonal membrane within 100 µm of the distal ending of the proximal stump . n> 50 axons for each genotype . Data are presented as mean ± SEM for 3 independent experiments; ***p<0 . 001; scale bar , 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14048 . 00510 . 7554/eLife . 14048 . 006Figure 3—figure supplement 1 . PKA can directly phosphorylate DLK at S302 . ( A ) PKA can induce DLK S302 phosphorylation in vitro . Flag-DLK was purified from HEK293 cells by anti-Flag immunoprecipitation and used for an in vitro kinase activity with purified PKA catalytic subunit . 5 µg of purified DLK was incubated with or without 10 , 000 U PKA catalytic subunit . Equal amounts of DLK in both samples ( as demonstrated by probing with anti-DLK antibody ) were analyzed by western blotting with anti-pDLKS302 antibody . ( B ) Quantification shows relative pDLKS302/total DLK levels . DOI: http://dx . doi . org/10 . 7554/eLife . 14048 . 00610 . 7554/eLife . 14048 . 007Figure 3—figure supplement 2 . DLK activation by PKA does not require TORC1 . ( A-B ) The phosphorylation level of DLK S302 is not sensitive to treatments of TORC1 inhibitors . HEK293 cells were either untransfected or transfected with Flag-DLKWT + empty plasmid or Flag-DLKWT + PKACA . Cells were treated with torin1 ( A ) or rapamycin ( B ) for 2 hr . The efficiency of the drugs were demonstrated by probing with anti-phospho-S6K antibody . DOI: http://dx . doi . org/10 . 7554/eLife . 14048 . 00710 . 7554/eLife . 14048 . 008Figure 4 . PKA promotes the stability of DLK independently of DLK downstream signaling . ( A-B ) Activation of PKA promotes DLK stability independently of JNK . HEK293 cells were transiently transfected with Flag-DLKWT , and ( A ) treated with forskolin ( 6 hr , 30 µM ) or ( B ) co-transfected with either PKACA or empty vector ( control ) . In both cases , co-treatment with JNK inhibitor VIII ( 10 µM , 6 hr ) led to a decrease in total Flag-DLK levels . However , both forskolin and PKACA induced an increase in DLK levels even in the presence of JNK inhibitor . Quantification shows average total DLK/Tubulin intensities and average pDLKS302/total DLK ratios ( where total DLK is detected using anti-Flag antibody ) . All data are represented as mean ± SEM; quantifications of relative intensity from Western Blots were averaged from 3 independent experiments; ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 , ‘n . s . ’ indicates non-significant . ( C-D ) Activation of PKA increases the stability of kinase dead DLK mutants , DLKK185A ( C ) and DLKS302A ( D ) . HEK293 cells were transiently transfected with Flag-DLKK185A or Flag-DLKS302A together with PKACA or empty plasmid . Treatment JNK inhibitor VIII ( 10 µM ) for 6 hr had no effect upon the PKACA induced levels of DLKK185A and DLKS302A mutant protein . Quantifications are similar to Figure 4A–B . Western bands intensity were averaged from 4 independent experiments; data are shown as mean ± SEM; ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 , ‘n . s . ’ indicates non-significant . ( E ) Proposed model for the activation and stabilization of DLK by cAMP and PKA . cAMP elevation and PKA activation leads to the phosphorylation of S302 on DLK , which activates its kinase activity . Indicated in the blue arrow , downstream signaling via JNK leads to enhanced DLK stability and phosphorylation of DLK at other sites ( Huntwork-Rodriguez et al . , 2013 ) . PKA also enhances DLK’s stability via an additional mechanism that is independent of S302 ( red arrow ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14048 . 00810 . 7554/eLife . 14048 . 009Figure 4—figure supplement 1 . Summary of predicted PKA phosphorylation sites on DLK/Wnd in different species . The sequence of DLK homologues in different species ( human , mouse , Drosophila and C . elegans ) were analyzed by Group-based Prediction System ( GPS ) to computationally predict PKA phosphorylation sites ( Xue et al . , 2005; 2008 ) . Identified potential PKACA phosphorylation motifs in different species are shown , with the phosphorylation site highlighted in red . Numbering is shown for mouse DLK . Two sites , S302 and S389 , are conserved among all the species . Other sites are conserved in mammals , while fly DLK has a similar number of sites but in distinct locations . Not shown , C . elegans DLK also has 11 additional predicted sites , but at distinct locations from mammalian and fly DLK . DOI: http://dx . doi . org/10 . 7554/eLife . 14048 . 009 Although the mechanism of DLK activation is unknown , phosphorylation of the activation loop is important for activation of other kinases in the mixed lineage kinase family that DLK belongs to ( Durkin et al . , 2004; Leung and Lassam , 2001 ) . Notably , the activation loop contains a conserved consensus sequence for PKA ( Figure 3C ) , and a recent study has demonstrated that the predicted phosphorylation site S302 is required for DLK to activate downstream kinases ( Huntwork-Rodriguez et al . , 2013 ) . To test whether PKA stimulates phosphorylation of DLK’s activation loop ( S302 ) , we have generated phospho-specific antibodies against a phosphorylated peptide corresponding to activation loop of mouse DLK ( KELSDKpSTKMpSFAGTV ) . The phospho-DLK antibodies detected a strong band in HEK293 cells that overexpress WT mouse DLK , but show no reactivity for mutant DLKS302A ( Figure 3D ) . Remarkably , expression of PKACA in HEK293 cells induces a dramatic increase of phospho-S302 DLK ( Figure 3E ) , even when normalized to the levels of total DLK ( Figure 4B ) . Similar results were observed when cells were treated with forskolin for 3 hr ( Figure 4A ) . Since the activation loop contains a conserved consensus sequence for PKA substrates , it should be capable of phosphorylating this site in DLK directly . Indeed , we found that purified PKA can strongly stimulate pS302 reactivity upon purified Flag-DLK in vitro ( Figure 3—figure supplement 1 ) . A recent study in Drosophila has suggested that TORC1 may activate and phosphorylate Wnd ( Wong et al . , 2015 ) , so we considered whether TORC1 plays a role in the activation of DLK by PKA . We used both torin1 and rapamycin to inhibit TORC1 in the presence or absence of PKA in HEK293 cells . However , in both cases we observed no effect upon DLK and pDLK levels ( Figure 3—figure supplement 2 ) . PKA therefore stimulates phosphorylation of DLK’s activation loop independently of TORC1 function . To confirm that S302 is a critical site for Wnd/DLK function in axonal regeneration , we conducted a rescue experiment in Drosophila motoneurons based on the requirement of Wnd for axonal sprouting after injury . We generated UAS-WndS301A , S305A transgenic flies expressing Wnd with mutations in two serines analogous to S298 and S302 in the activation loop of DLK . As shown in Figure 3G , all axons in wnd mutants fail to initiate a sprouting response , which can be rescued by co-expression of WT Wnd , but not WndS301A , S305A . In addition , overexpression of WndS301A , S305A does not give rise to any of the previous described gain-of-function phenotypes similar to WT Wnd , suggesting that S302 is indeed required for DLK function . A previous study has described a positive feedback loop for DLK stabilization that involves the action of DLK’s downstream effector JNK ( Huntwork-Rodriguez et al . , 2013 ) . JNK activation stimulates phosphorylation of DLK at sites outside of its activation loop ( T43 and S533 ) and changes DLK’s sensitivity to degradation via the ubiquitin proteasome system ( UPS ) ( Huntwork-Rodriguez et al . , 2013 ) . This increase in protein stability leads to an increase in total levels of DLK . Since PKA stimulates an increase in ectopically expressed DLK , it most likely increases protein stability . We therefore tested whether this increased stability involves the previously described JNK-dependent feedback mechanism . If this is the case , the effects of PKA and forskolin should depend on the function of JNK . As shown in Figure 4A and B , treatment with JNK inhibitor VIII led to a 30% decrease in total DLK level , as expected for JNK’s previously demonstrated role in promoting DLK stability . However , even in the presence of JNK inhibitor , treatment with forskolin ( Figure 4A ) or transfection with PKACA ( Figure 4B ) increases DLK level by 50% . Moreover , treatment with JNK inhibitor had very little effect upon the fraction of total DLK that is phosphorylated at S302 ( Figure 4A and B ) . These results suggest that PKA stimulates DLK phosphorylation and stabilization independently of downstream JNK activation . Previous biochemical studies suggest that DLK’s activation mechanism is associated with dimerization and autophosphorylation ( Nihalani et al . , 2000; Mata et al . , 1996; Merritt et al . , 1999 ) . We therefore further considered whether PKA could function either upstream or downstream of DLK’s own ability to function as a kinase . To test this , we utilized a kinase-dead version of DLK , DLKK185A , which is unable to activate downstream signaling or undergo autophosphorylation ( Nihalani et al . , 2000; Mata et al . , 1996; Merritt et al . , 1999 ) . Consistent with the previously described feedback mechanism ( Huntwork-Rodriguez et al . , 2013 ) , the DLKK185A mutant protein was less stable , and addition of JNK inhibitor had no further effect upon the levels of kinase dead DLK protein ( Figure 4C ) . However , PKACA stimulated a strong increase of DLKK185A levels and phosphorylation at S302 for DLKK185A . The ability of PKA to increase DLK protein stability and activation loop phosphorylation independently of DLK’s own signaling abilities places PKA firmly upstream of DLK , as an upstream regulator/activator . Since PKA promotes DLK stability and directly phosphorylates S302 , we wondered whether PKA stabilizes DLK by phosphorylation of S302 . Previous work has shown that a decreased stability for DLKS302A mutant protein is linked to the fact that it is inactive for kinase activity , and is therefore unable to activate the downstream stabilization mechanism via JNK ( Huntwork-Rodriguez et al . , 2013 and Figure 4D ) . Consistent with this , the reduced levels of DLKS302A protein are not further reduced in the presence of JNK inhibitors ( Figure 4D ) . However surprisingly , co-transfection with PKACA still caused an increase of the levels of DLKS302A protein ( Figure 4D ) . We interpret that PKA regulates DLK via an additional mechanism , in conjunction with phosphorylation of the critical activation loop S302 . This additional mechanism may involve other sites of phosphorylation , or other modes of regulation ( discussed further below ) . Our finding that PKA activates DLK , taken together with previous findings that DLK promotes axonal regeneration in different neuronal cell types ( Shin et al . , 2012; Xiong et al . , 2010; Hammarlund et al . , 2009; Yan et al . , 2009 ) , led to the hypothesis that DLK is an important downstream mediator of cAMP-stimulated axon regeneration . To test this hypothesis , we employed a recently described replating assay for DRG neurons cultured from adult mice , which allows for a controlled and quantitative measure of the induction of axonal regeneration by in vitro manipulations such as forskolin treatment ( Frey et al . , 2015; Valakh et al . , 2015 ) . In this assay ( depicted in Figure 5A ) , DRG neurons removed from adult mice were first cultured for 4–5 days , which allowed for the regenerative response activated by the dissection to subside . Neurons were then treated with forskolin for 24 hr , and then replated onto a fresh dish . The replating process removes all existing neurites so that the neurites observed within the second culture period can be identified as new growth . As shown in Figure 5B and E , treatment with forskolin stimulates the regenerative response ( Frey et al . , 2015 ) . In addition , co-treatment with H-89 abolished the effect of forskolin on neurite outgrowth ( Figure 5B–D ) , suggesting that PKA is required . 10 . 7554/eLife . 14048 . 010Figure 5 . PKA stimulates axonal regeneration via DLK in adult DRG neurons . ( A-D ) Induction of regeneration by forskolin requires PKA . Experimental design ( A and also see Materials and Methods ) . To demonstrate that forskolin-induced neurite outgrowth is mediated by PKA , we assessed whether PKA signaling was required using the PKA inhibitor H-89 ( PKAi , 5 µM ) . Representative neurons are shown in ( B ) . Neurite outgrowth was assessed by quantifying mean neurite length ( C ) and distribution of longest neurite ( D ) . Data are mean ± SEM for 4 independent experiments . ( E-G ) Induction of regeneration by forskolin requires DLK . WT and DLK KO neurons were treated with DMSO or forskolin ( 30 µM ) as described in ( A ) . Representative neurons are shown in ( E ) . Neurite outgrowth was assessed by quantifying mean neurite length ( F ) and distribution of the longest neurite ( G ) . Data are mean ± SEM for 3 independent experiments . ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 , ‘n . s . ’ indicates non-significant; scale bars , 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14048 . 01010 . 7554/eLife . 14048 . 011Figure 5—source data 1 . Measurements of the longest neurite length for 100 neurons after replating in each condition . The longest neurite length for 100 WT neurons after replating in four different conditions ( DMSO+Vehicle , DMSO+forskolin , PKA inhibitor+Vehicle , PKA inhibitor+forskolin ) are shown in Table ‘PKAi’ . This table contains data from 4 independent experiments . The longest neurite length for 100 DLK KO neurons after replating in two different conditions ( DMSO and forskolin ) are shown in Table ‘DLK KO’ . The table contains data from 3 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 14048 . 011 To determine whether forskolin-induced neurite outgrowth requires DLK , we performed the same experiment in neurons from DLK ( Map3k12 ) conditional knockout ( KO ) mice and littermate controls . The effects of forskolin on neurite outgrowth were abolished in DLK conditional knockout DRG neurons ( Figure 5E–G ) . These findings suggest that DLK and its downstream signaling pathway ( s ) are important mediators of the pro-regenerative effects of cAMP elevation in neurons .
The Wnd/DLK kinase is likely to function as a sensor of axonal damage . Depending upon the context , its activation can lead to either axonal regeneration or cell death and degeneration ( Tedeschi and Bradke , 2013 ) . The factors that determine beneficial versus detrimental outcomes , along with the general cellular mechanisms that lead to the activation of DLK , are poorly understood . In this study , we found that an immediate upstream activator of DLK is the cAMP regulated kinase PKA . Elevation of cAMP signaling , which is activated by pro-regenerative manipulations such as a conditioning lesion , is the most widely known pathway for promoting axonal regeneration ( Hannila and Filbin , 2008 ) . We found that an essential component of this regenerative pathway is the activation of Wnd/DLK by PKA . These findings delineate an evolutionarily conserved mechanism for the activation of the Wnd/DLK kinase . Taken together with previous findings that Wnd/DLK is an essential regulator of responses induced by a conditioning injury ( Shin et al , 2012; Xiong and Collins 2012 ) , the activation of Wnd/DLK by PKA in injured axons presents a unified molecular pathway for activating a regenerative response to axonal damage . In contrast to a merging of cAMP and DLK pathways , some other studies have suggested that these pathways may act separately ( Li et al . , 2015; Chung et al . , 2016 ) . A recent study has noted that in certain sensory neuron types in C . elegans , PKA gain-of-function mutations can induce axonal outgrowth even in dlk mutants ( Chung et al . , 2016 ) . Hence , multiple pathways for axonal regeneration may be inducible by PKA . However DLK is strongly required for cAMP stimulated regeneration in other neuron types in C . elegans ( Ghosh-Roy et al . , 2010 ) , and , importantly , in axonal regeneration induced by a conditioning lesion in the mammalian PNS ( Shin et al . , 2012 ) . Our findings now indicate that DLK is an important molecular target and effector of cAMP-induced regeneration in mammalian neurons . Previous biochemical studies indicate that DLK activation involves dimerization via its leucine zipper domains and autophosphorylation , at locations that are yet undefined ( Nihalani et al . , 2000; Mata et al . , 1996 ) . Because ectopic elevation of DLK/Wnd protein is sufficient to activate its downstream signaling ( Mata et al . , 1996; Nihalani et al . , 2000; Huntwork-Rodriguez et al . , 2013 ) , and DLK is known to be highly regulated at the level of protein turnover ( Collins et al . , 2006; Xiong et al . , 2010; Huntwork-Rodriguez et al . , 2013; Nakata et al . , 2005; Hammarlund et al . , 2009 ) , a plausible mechanism for its regulation is to hold its levels and/or its ability to dimerize in check ( Mata et al . , 1996; Nihalani et al . , 2000 ) . The existence of a direct upstream activator of the kinase was not previously implied , and has thus far been unknown . Here we found that PKA stimulates the phosphorylation of the activation loop of DLK independently of DLK’s kinase activity , and also independently of downstream JNK signaling . This defines PKA as an upstream activator of DLK . A previous study in C . elegans has described a mechanism through which transient elevation of intracellular Ca2+ upon axonal injury leads to the activation of DLK-1 ( Yan and Jin , 2012; Cho et al . , 2013; Ghosh-Roy et al . , 2010; Spira et al . , 2001 ) . In addition , earlier studies have implicated calmodulin-regulated calcineurin in the regulation of mammalian DLK ( Mata et al . , 1996 ) . However , the hexapeptide that mediates activation by Ca2+ in C . elegans is not present in mammalian or Drosophila DLK/Wnd , and mammalian DLK can be activated independently of Ca2+ elevation by cytoskeletal destabilizing agents ( Valakh et al . , 2015 ) . In contrast , the consensus PKA phosphorylation site in the activation loop of Wnd/DLK is conserved in all phyla ( Figure 3C ) , suggesting this pathway as a central ( although not necessarily exclusive ) mechanism for regulating DLK . We note that in conjunction with phosphorylation of DLK’s essential activation loop , PKA enhances DLK’s stability via an additional mechanism ( Figure 4E ) , since the mutated protein DLKS302A can still be stabilized upon PKA activation . This mechanism may involve additional phosphorylation sites on DLK , and indeed , multiple PKA consensus sequences are observed in Wnd/DLK’s sequence ( Figure 4—figure supplement 1 ) . However , it is also possible that PKA regulates DLK’s stability via other mechanisms , such as previously described ubiquitination ( Nakata et al . , 2005; Collins et al . , 2006 ) , palmitoylation modification ( Holland et al . , 2016 ) , changes in DLK’s interacting proteins or subcellular localization . Such aspects of regulation could involve additional molecular targets of PKA , which would provide some potential for context specificity in DLK’s regulation . While PKA has many cellular targets , its specificity can be highly regulated at a subcellular level by interactions with AKAP scaffolding proteins and local changes in cAMP ( Tasken and Aandahl , 2004; Wong and Scott , 2004 ) . It will be interesting to identify the additional players in cAMP and PKA regulation of DLK through future work . It is remarkable that PKA stimulates a specific increase in Wnd levels in axons but not cell bodies ( Figure 2C ) , and inhibition of PKA strongly inhibits the induction of Wnd protein after axonal injury ( Figure 2D ) . Does PKA act locally in axons to stimulate DLK ? Other studies have suggested that Wnd/DLK can regulate retrograde signaling pathways that originate in axons ( Xiong et al . , 2010; Ghosh et al . , 2011; Shin et al . , 2012; Huntwork-Rodriguez et al . , 2013; Watkins et al . , 2013; Yan et al . , 2009; Holland et al . , 2016 ) . Intriguingly , the Hiw/Rpm-1/Phr1 ubiquitin ligase , which is previously known for its role in regulating Wnd/DLK’s levels in axons ( Collins et al . , 2006; Nakata et al . , 2005; Lewcock et al . , 2007; Babetto et al . , 2013 ) , contains a RCC1-like domain that biochemically inhibits adenylate cyclase , and therefore may negatively regulate cAMP signaling ( Pierre et al . , 2004 ) . Hiw/Rpm-1/Phr1 is previously known for its role in regulating synaptic arborization and growth via its regulation of Wnd/DLK ( Nakata et al . , 2005; Collins et al . , 2006; Wang et al . , 2013; Wan et al . , 2000; Shin and DiAntonio , 2011 ) . The addition of cAMP and PKA into this regulatory pathway suggests a mechanism that may be broadly utilized to orchestrate structural changes within presynaptic terminals . We propose that the regulation of DLK by PKA may be generally important for neuronal plasticity as well as responses to axonal damage .
The following fly strains were used in this study: Canton-S ( WT ) , m12-Gal4 ( Ritzenthaler et al . , 2000 ) , BG380-Gal4 ( Budnik et al . , 1996 ) , OK6-Gal4 ( Aberle et al . , 2002 ) , RRa ( eve ) -Gal4 ( Fujioka et al . , 2003 ) , puc-lacZE69 , wnd3 , UAS-Wnd , ( Collins et al . , 2006 ) , UAS-GFP-WndKD ( Xiong et al . , 2010 ) , UAS-PKACA ( Li et al . , 1995 ) . UAS-WndS301A , S305A flies were generated from pUAST-WndS301A , S305A plasmid for this study . UAS-wnd-RNAi ( VDRC 13786 ) and UAS-moody-RNAi ( VDRC 100674 ) were from the Vienna RNAi center ( Dietzl et al . , 2007 ) . UAS-pka-c1-RNAis ( 31277 , 35169 ) and UAS-dnc-RNAi ( 27250 ) were acquired from Bloomington stock center . Peripheral nerve crush assays in 3rd instar larvae were performed according to Xiong et al . ( 2010 ) . Briefly , the segmental nerves of third instar larvae were pinched and crushed by a fine No . 5 forceps while the larvae were anesthetized with CO2 gas . After injury , larvae were transferred to a grape plate and kept in 25°C incubator for specified time periods . Drosophila third instar larva were dissected in ice-cold PBS and fixed in 4% paraformaldehyde for 25 min . Antibodies were used in PBS supplemented with 0 . 3% Triton and 5% normal goat serum . Anti-lacZ ( 40-1a , Developmental Studies Hybridoma Bank ) was diluted 1:100 . Anti-dsRed polyclonal antibody ( 632495 , Clontech ) was diluted 1:1000 . For secondary antibodies , A488- or Cy3-conjucated goat anti-mouse or rabbit ( Invitrogen , Carlsbad , CA ) were used at 1:1000 . Confocal images were collected on an Improvision spinning disk confocal system , consisting of a Yokagawa Nipkow CSU10 scanner , and a Hamamatsu C1900-50 EMCCD camera , mounted on a Zeiss Axio Observer with 40X ( 1 . 3NA ) oil objectives . Similar settings were used for imaging of all compared genotypes and conditions . Volocity software ( Perkin Elmer ) was used for intensity measurements and quantification of all confocal data . A quantification of the sprouting response in injured motoneuron axons was measurement of the change of total volume of regenerating axonal membrane , labeled by mCD8-GFP , within a 100 µm distance from the injured tip . Pixels within the most distal 100 µm of the injured proximal stump were selected based on mCD8-GFP intensity criteria of >3 standard deviation above the mean , and then summed to measure total membrane volume . Figure 1 and 3F report the change in average volume at 15 hr after injury compared to T=0 ( immediately after injury ) . Puc-lacZ expression was quantified by measurement of mean intensity for lacZ staining in the nuclei of motoneurons located along the dorsal midline ( segments A3-A7 ) of the nerve cord of third instar larva . The mean intensity of GFP-WndKD within segmental nerves was quantified by measuring GFP intensity within 100 µm distance of each nerve at the site of exit from the ventral nerve cord . HEK293 cells were cultured in DMEM/F12 ( Gibco ) supplemented with 10% fetal bovine serum ( Gibco ) and 1% penicillin-streptomycin ( Gibco ) . For transient transfection , 1 µg of given vectors were transfected into 3 . 5 cm dish using Lipofectmine 2000 ( Invitrogen ) . 20 hr after transfection , cells were processed for Western blotting . Plasmids used for transfection were Flag-DLK ( Huntwork-Rodriguez et al . , 2013 ) , Flag-DLKS302A ( Huntwork-Rodriguez et al . , 2013 ) , Flag-DLKK185A ( directly generated from Flag-DLK ) and PKACA ( Merrill et al . , 2011 ) . Cortical neurons were dissected from E18 rat embryos . The cortex was digested by incubating with 0 . 5% trypsin-EDTA ( Gibco ) and DNAse I ( Roche ) at 37°C for 10 min . Following digestion , neurons were washed twice in DMEM medium ( Gibco ) containing 10% FBS before resuspension in neuronal growth media which containing neurobasal ( Gibco ) , Glucose ( Sigma ) , Glutamax ( Gibco ) , penicillin-streptomycin and B27 supplement ( Gibco ) . All the plates were coated with 100 µg/mL ploy-D-lysine ( P7886 , Sigma ) for 2 hr . Neurons were then triturated and plated at a final concentration 400 , 000 cells/mL . For adult DRG experiments , DRG neurons were collected from either CD1 ( Charles River ) , or Map3k12 F/F; Advillin-Cre-/- ( WT ) , or Map3k12 F/F; Advillin-Cre+/- ( DLK KO ) mice . WT and DLK KO mice were age-sex matched . Neurons were prepared as previously described ( Frey et al . , 2015 ) . Briefly , DRG were digested for 15 min at 37°C with 0 . 35 mg/mL liberase Blendzyme ( Roche ) , 10 mg/mL bovine serum albumin ( Sigma ) , and 0 . 6 mg/mL DNase ( Sigma ) followed by another 15 min digest at 37°C with 0 . 05% trypsin . DRG were tritterated in culture media ( DMEM containing 10% FBS , 100 µg/mL penicillin and 100 µg/mL streptomycin ) to dissociated cells . Cells were plated on PDL ( 10 mg/mL ) and laminin ( 10 mg/mL ) coated plates . On day in vitro ( DIV ) 1 , half of the media was removed and fresh media containing AraC ( Sigma , 10 nM final ) was added . Drug treatment and replating were performed as described previously ( Frey et al . , 2015 ) . On DIV4 , DMSO or forskolin ( 30 µM ) were added to cells . 24 hr after drug application , drugs were washed out with DMEM . Neurons were lightly trypsinized with 0 . 025% trypsin for 5 min in the incubator ( 37°C , 5%CO2 ) . Trypsin was removed and fresh culture media was added to the cells which were then gently pipetted and transferred to culture slides . 18 hr after replating , neurons were fixed ( 4% PFA ) and stained for βIII-tubulin ( Covance , mouse anti-Tuj1 , 1:500 ) . At least 100 neurons were imaged per group using either Leica DFC310 FX or DFC7000T color fluorescence cameras and longest neurite was traced using NeuronJ plugin for ImageJ . For replating experiments with PKA inhibitor ( H-89 , 5 µM ) , vehicle or inhibitor were added at the same time as DMSO and forskolin . For detection of Wnd protein levels in larva nerve cords , the whole brains were carefully dissected from third instar larva . The two brain lobes were removed before they were frozen in liquid nitrogen and processed for western blotting ( 25 nerve cords per lane ) . For western blots using HEK293 cells or cortical neurons , cells are lysed by incubating on ice for 10 min with RIPA buffer ( BP-115 , Boston BioProducts ) supplemented with Complete protease inhibitor cocktail ( Roche ) and PhosSTOP phosphatase inhibitor ( Roche ) . Protein concentrations were measured by BCA assay kit ( Thermo Scientific ) . Equal amount of protein samples were loaded on each lane of NuPAGE 4–12% Bis-Tris gels ( Invitrogen ) and subject to electrophoresis separation in MOPS buffer ( Invitrogen ) . Blots were visualized with SuperSignal chemiluminescent substrate ( Thermo Scientific ) and exposure to either film or ChemiDoc ( Bio-Rad ) . Bands intensities were determined using software ImageJ ( NIH ) using the gel analysis plug-in . The following antibodies were used for Western blotting: anti-Wnd A3-1 , 2 at 1:700 ( Collins et al . , 2006 ) ; anti-DLK at 1:5000 ( Huntwork-Rodriguez et al . , 2013 ) ; anti-β-tubulin at 1:1000 ( E7; Developmental Studies Hybridoma Bank ) ; anti-Flag at 1:1000 ( F1804 , Sigma ) ; anti-PKA C ( catalytic subunit ) antibody at 1:1000 ( 4782 , Cell signaling ) ; anti-phospho-cJun antibody at 1:1000 ( 3270 , Cell signaling ) ; anti-phospho-S6K antibody at 1:1000 ( 9234 , Cell signaling ) and anti-S6K antibody at 1:1000 ( 2708 , Cell signaling ) . Anti-phospho-DLKS302 antibodies were raised by immunization of rabbits with the peptide KELSDKpSTKMpSFAGTV and affinity purified before use at dilution 1:100 . While the antibodies were raised against a dually phosphorylated peptide ( pS298 , pS302 ) , no difference in reactivity was noticed for DLK S298A mutants ( data not shown ) . Forskolin ( F6886 , Sigma ) was applied to either HEK293 cells or cortical neurons at the final concentration of 30 µM for 6 hr . H-89 ( B1427 , Sigma ) was used at the final concentration of 5 µM for cortical neurons for 6 hr . JNK inhibitor VIII ( 420135 , EMD Millipore ) was used at the final concentration of 10 µM on HEK293 cells for 6 hr . Torin1 ( 4247 , Tocris ) or Rapamycin ( LC laboratories ) was applied to the cells at 1 µM ( final concentration ) for 2 hr before harvest . 24 hr after transfection of the given constructs , HEK293 cells in a 6 cm dish were washed by ice-cold PBS and harvested in ice-cold RIPA buffer supplemented with EDTA-free protease inhibitor ( Roche ) . The cell lysates were incubated on ice for 30 min and centrifuged at 14 , 000 rpm at 4°C for 10 min . The soluble lysate was incubated with 3 µg anti-Flag antibody pre-bound with 15 µl Dynabeads Protein G ( Novex ) for 1 hr at 4°C . After removal of supernatants the beads were washed 3 times with RIPA buffer , and then incubated with 1X PMP buffer ( NEB ) , 1X MnCl2 ( NEB ) and either 1000 U lambda protein phosphatase ( NEB ) or equal amount of glycerol ( control ) in RIPA buffer for 30 min at 30°C . Beads were then removed from the reaction buffer and Flag-DLK was eluted by boiling in SDS sample buffer for 10 min . Equal amount of samples were analyzed by western blot . HEK293 cells transfected with Flag-DLK were washed by ice-cold PBS and harvested using ice-cold cell lysis buffer ( 50 mM Tris-HCl , pH7 . 5 , 150 mM NaCl , 1 mM EDTA , 1% Triton X-100 , Complete protease inhibitors ( Roche ) and 30 µM MG132 ( Sigma ) ) . Flag-tagged DLK was immunoprecipitated from HEK293 cell lysates using anti-Flag M2 antibody ( Sigma ) and Protein G Dynabeads ( Novex ) . The DLK-bound beads were then washed with 3 times with lysis buffer and incubated with 2000 U lambda protein phosphatase ( NEB ) for 30 min at 30°C to remove all the phosphate groups . After incubation , the beads were washed with wash buffer containing cocktail phosphatase inhibitor ( Roche ) and split equally into two tubes containing kinase reaction buffer ( 50 mM Hepes , pH7 . 2 , 10 mM MgCl2 , 0 . 01% Triton X-100 , 2 mM DTT , and 30 µM ATP ) . 10 , 000 U recombinant human full length PKA catalytic subunit alpha ( NED Millipore ) was added to one of the tubes , while the control tube were added with glycerol . Both tubes were incubated at 30°C for 90 min . Flag-DLK was eluted from beads by boiling in the SDS sample buffer . Equal amounts of samples were analyzed by western blotting with anti-DLK and anti-pDLKS302 antibodies . For experiments in flies , we knew from previous work that a sample size of 10 animals per genotype was large enough to detect significant differences among genotypes ( Xiong et al . , 2010; 2012 ) . Therefore , at least 10 animals ( ≥ 50 axons ) were examined and quantified in each genotype . Each experiment was repeated at least 3 times with independent biological replicates . For experiments in mice , we knew from previous work that measuring 100 neurons per genotype and an N=3–4 was sufficient to detect reproducible differences between the experimental groups ( Frey et al . , 2015; Valakh et al . , 2015 ) . Therefore , all the experiments were performed with at least 3 independent biological replicates . One way ANOVA and multiple comparisons were conducted when more than two samples are compared . Tukey post-hoc test was used to correct for multiple comparisons . For binned DRG neurites length distribution , statistical significance was determined by two way ANOVA followed by Bonferroni post-hoc test . p values smaller than 0 . 05 were considered statistically significant . All p values are indicated as *p<0 . 05 , **p<0 . 01 , and ***p<0 . 001 and ****p<0 . 0001 . Data are presented as mean ± SEM . | Adult mammals typically cannot repair damage to the nerve fibers in their brain or spinal cord . This is because these nerve cells cannot generally grow new nerve fibers . However this inability to regenerate nerve fibers is not set in stone . Instead , it can be unlocked by a second injury in nerves elsewhere in the body , the so-called “peripheral nervous system” . This process relies on an enzyme called DLK , which becomes activated in damaged nerve fibers . But how does DLK ‘sense’ damage to nerve fibers ? Injuring the peripheral nervous system causes the levels of a molecule called cAMP to increase in the damaged nerve cells , and the elevated cAMP levels stimulate the nerve fibers to regenerate . However , it was not known if cAMP activates DLK , or if the two act independently of each other . By looking at the regeneration of damaged nerve fibers in fruit fly larvae , Hao et al . now show that the cAMP and DLK signaling pathways are clearly linked . Further experiments with nerve cells from mice and human cells revealed more detail about this link . Together the results showed that another enzyme called PKA activates DLK directly when cAMP levels are high . These findings reveal a unified pathway that is the key to unlocking the regenerative potential of injured nerve fibers , which has been conserved for hundreds of millions of years of evolution . Further work could now ask if the DLK enzyme is involved in the other known roles of cAMP signaling in nerve cells; or if cAMP and PKA activate DLK in other forms of nerve damage , including injuries where nerve fibers normally fail to regenerate . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | An evolutionarily conserved mechanism for cAMP elicited axonal regeneration involves direct activation of the dual leucine zipper kinase DLK |
Dyslexia is a prevalent reading disability whose underlying mechanisms are still disputed . We studied the neural mechanisms underlying dyslexia using a simple frequency-discrimination task . Though participants were asked to compare the two tones in each trial , implicit memory of previous trials affected their responses . We hypothesized that implicit memory decays faster among dyslexics . We tested this by increasing the temporal intervals between consecutive trials , and by measuring the behavioral impact and ERP responses from the auditory cortex . Dyslexics showed a faster decay of implicit memory effects on both measures , with similar time constants . Finally , faster decay of implicit memory also characterized the impact of sound regularities in benefitting dyslexics' oral reading rate . Their benefit decreased faster as a function of the time interval from the previous reading of the same non-word . We propose that dyslexics’ shorter neural adaptation paradoxically accounts for their longer reading times , since it reduces their temporal window of integration of past stimuli , resulting in noisier and less reliable predictions for both simple and complex stimuli . Less reliable predictions limit their acquisition of reading expertise .
Dyslexics are diagnosed on the basis of their persistent difficulties in acquiring peer-level reading skills despite adequate education . Their general reasoning skills are within the normal range ( or above ) , but they consistently show difficulties in some language-related skills such as verbal working memory ( e . g . Torgeson and Goldman , 1977 ) and phonological manipulations ( which typically also load on short-term memory [e . g . Landerl et al . , 1997] ) . Dyslexics also often have higher thresholds in simple perceptual discrimination tasks ( Mcanally and Stein , 1996; Witton et al . , 1998; Hämäläinen et al . , 2013 ) , particularly when administered with serial presentations ( Ben-Yehudah and Ahissar , 2004; discussed in Ramus and Ahissar , 2012 ) . In most of these studies , the responses of participants can be more successful by taking into account the frequency statistics of previous stimuli ( Ahissar et al . , 2006; Oganian and Ahissar , 2012 ) . The putative causes of dyslexics’ difficulties on simple serial tasks have been examined in a series of studies using 2-tone frequency discrimination . Ahissar et al . ( 2006 ) measured the impact of sound regularities on dyslexics’ performance . They assessed a well-documented observation ( Harris , 1948 ) that listeners use a repeated reference tone as an ‘anchor’ , and that their performance was improved in protocols that included this reference tone compared with a no-reference protocol . The benefit that dyslexics obtained from this repetition was smaller than that obtained by 'good readers’ . A similar deficit was found in dyslexics' benefit from repetition of speech sounds . This led to the hypothesis that dyslexics have a deficit in using sound stimuli as perceptual anchors for the formation of sound predictions ( Ahissar et al . , 2006; Ahissar , 2007; Oganian and Ahissar , 2012 ) . Raviv et al . ( 2012 ) extended the protocol-specific account of benefits from stimulus repetition to generate a computational model which takes the experiment’s statistics into account . This model assumes that listeners implicitly infer the mean ( frequency ) even when it is not presented explicitly . The inferred mean of previous stimuli ( prior ) is combined with the representation of the current stimulus , and forms an integrated percept ( posterior ) . The resulting percept is contracted ( biased ) towards that mean ( the ‘contraction bias’; Woodrow , 1933; Preuschhof et al . , 2010 ) . This bias is advantageous when the observation of the current stimulus is noisy , and hence integration with prior knowledge is likely to improve its accuracy . Indeed , in the general population , ‘noisier’ listeners weigh the prior more than ‘less noisy’ listeners when integrating the prior with current representation , resulting in their larger contraction bias . Jaffe-Dax et al . ( 2015 ) found that dyslexics’ bias is smaller than controls’ even though they tend to be ‘noisier’ listeners . In the two-tone frequency discrimination task , the representation of the first tone is contracted towards the prior more than the second tone , because of the noise added to its representation during the encoding and retention in memory through the inter-stimulus time interval . The contraction of the first tone towards the prior can increase the perceived difference between the two tones in the trial , and hence improve discrimination ( Bias+ trials; e . g . , trial t in Figure 1A ) ; alternatively , it could decrease the perceived difference between the two tones and disrupt performance ( Bias-; e . g . , trial t−1 in Figure 1A ) . The difference in performance between Bias+ and Bias- trials reflects the magnitude of the contraction bias ( i . e . context effect ) . Dyslexics’ underweighting of the prior yields a smaller contraction bias , namely a smaller performance difference between these two types of trials ( Jaffe-Dax et al . , 2015 ) . 10 . 7554/eLife . 20557 . 003Figure 1 . The three parameters that additively determine performance in frequency discrimination: frequency difference within a trial ( Δf ) , Global context ( the difference between the current f1 and the global mean ) and Recent context ( the difference between the current f1 and the previous f1 ) . ( A ) A schematic illustration of the Global bias:four trials and the direction of the contraction , which contracts the representation of the first tone in the trial towards the global mean . ( B ) A schematic illustration of the same four trials and the direction of the Recent bias , which pulls the representation of this tone towards the recent f1 frequency . ( C ) The estimated contribution ( β ) of each of the three parameters to the overall performance of each group . Dyslexics differed from controls only in the smaller magnitude of the contribution of their Global predictor ( p<0 . 01 ) . Filled bars denote the mean β values; controls in blue and dyslexics in red . Error bars denote SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 20557 . 003 Using this measure of contraction bias , we studied the dynamics of controls’ and dyslexics’ benefits from the statistics of recent sound stimuli in both simple discriminations and in oral reading . We found that dyslexics are similarly affected ( biased ) by recent stimuli , but less affected by earlier stimuli , as expressed both behaviorally and in the dynamics of the compulsory ERP components ( N1 and P2 ) produced by the auditory cortex . These observations suggest that dyslexics’ automatic integration of previous sounds spans shorter time intervals , and is therefore noisier and produces less reliable predictions . We propose that noisier predictions impede dyslexics’ acquisition of expert-level performance in a range of tasks , including reading . These observations also pave the way to pursuing the deficits underlying dyslexia in non-human animals .
Performance in the 2-tone frequency discrimination task is affected by two basic components ( Raviv et al . , 2012 , 2014 ) . First , the frequency difference between the two tones composing each trial – participants are more accurate when this difference is larger . Second , the context effects; namely , the perceptual bias induced by contracting the first tone towards the estimated mean-frequency in previous trials ( Woodrow , 1933; Preuschhof et al . , 2010; see Introduction and Figure 1A–B ) . This estimated prior can be viewed as a combination of a recent component and a global component . The most recent factor is the frequency of the first tone of the previous trial ( Figure 1B ) , whose impact on the estimated prior is substantially larger than that of any other previous trial ( ‘recency effect’; Schab and Crowder , 1988; Stewart and Brown , 2004; Fischer and Whitney , 2014 ) . The global prior ( illustrated in Figure 1A ) is the mean frequency of all previous trials . As described above , dyslexics’ discrimination behavior is less affected by previous trials ( Jaffe-Dax et al . , 2015 ) . We now aimed to decipher whether dyslexics’ deficit encompasses both the recent and the global context effects . To test this , we designed a novel protocol for the two-tone frequency discrimination task , which allowed us to assess the contribution of recent and earlier trials’ on performance separately . Unlike in most sequences , where the recent and more global contexts tend to have the same direction , and are therefore very difficult to dissociate , in the new sequence , the recent and the global contexts were not correlated ( we ensured that the directions of the local and global effects were not correlated , as described in Materials and methods ) . The overall accuracy of the two groups in this task did not differ ( controls’ mean % correct ± SEM = 75 . 3 ± 1 . 6 , dyslexics’ = 73 ± 1 . 9; z = 1 . 3 , n . s . Mann-Whitney U test ) , as predicted given that the use of experiment’s statistics was not expected to be significantly beneficial in this stimulus series ( replicating Ahissar et al . , 2006 ) . We calculated a GLM model with three predictors ( βs ) for each participant ( n = 60; 30 control and 30 dyslexic participants ) by estimating the magnitude of the contribution of each of the following components to participants’ responses: ( 1 ) frequency difference in the current trial; ( 2 ) contraction bias of the first tone toward the global mean of previous trials; and ( 3 ) contraction bias of the first tone towards the first tone of the most recent trial . Dyslexics’ frequency difference predictor did not differ from controls’ ( Figure 1C; z = 1 . 6 , n . s . ) . Namely , the frequency difference within the trial had a similar impact on the response in the two groups ( similar levels of sensitivity ) . The impact of the context effect of the most recent trial ( ITI ≈ 1 . 5 s ) was also similar in the two groups ( z = 0 . 05 , n . s . ) . However , dyslexics’ bias towards the global mean was significantly smaller than controls’ ( z = 2 . 7 , p<0 . 01 ) . Indeed , the difference in controls between the contributions of the Global ( all previous trials except the most recent one ) and Recent contexts was larger than that found in dyslexics ( z = 2 . 4 , p<0 . 05 . Mann-Whitney U tests ) . Importantly , in both groups , the contribution of Recent context was significantly above zero ( controls: z = 4 . 4 , p<0 . 0001; dyslexics: z = 4 . 7 , p<0 . 00001; Wilcoxon tests ) . Thus , floor effect cannot account for this interaction . The observation that dyslexics assign less weight than controls to earlier trial statistics could result from either of the following two mechanisms . The first is that dyslexics experience larger interference effects , and that intervening trials mask each other . The second is that dyslexics’ memory trace decays faster in time , even without intervening sounds . To dissociate these alternatives , we manipulated the time interval between consecutive trials . We reasoned that if dyslexics’ memory trace decays faster , we might be able to track the neural correlate of this behavioral dynamic . We administered the two-tone frequency discrimination task with four different inter-trial intervals ( ITIs; i . e . the intervals between the second tone of the previous trial and the first tone of the current trial ) , in four separate blocks . We chose ITIs of 1 . 5 , 3 , 6 and 9 s ( roughly; see Materials and methods ) , based on previous reports of cortical adaptation duration ( N1 and P2 components; Hari et al . , 1982; Lu et al . , 1992; Sams et al . , 1993 ) . As shown above , dyslexics benefit less from recent sound statistics . We hypothesized that a similar observation would be found for recent encounters with complex sound stimuli and particularly with novel syllabic combinations . If demonstrated , this correlate would provide support for the relevance of the observations of Experiments 1–2 to the context of reading . To assess the dynamics of the simplest context effect in reading , we administered an oral reading task in which participants were asked to read single simple disyllabic non-words aloud , and we measured repetition effects . The non-words were presented on a screen , one at a time , and participants were asked to read them as quickly ( and accurately ) as they could . Consecutive non-words were presented at an individually paced rate . Non-words were repeated throughout the block at various intervals ( i . e . with a different number of other intervening non-words ) , as schematically illustrated in Figure 5A . 10 . 7554/eLife . 20557 . 007Figure 5 . Dyslexics’ benefit from a previous exposure to the same non-word decayed faster than controls’ . ( A ) Schematic illustration of the reading task . Subjects were asked to read the non-words aloud as quickly as possible . Presentation switched to the next word with the subject’s voice offset . The closest repetition of the same non-word was with one intervening non-word ( i . e . an ITI of <2 s ) . ( B ) Benefit in RTs ( response times from visual word presentation to vocal onset ) as a function of the time interval between the first and second presentation of the same non-word . Improvement was calculated as the difference in RT between the first and second presentation of the same non-word in the block . At very short intervals ( <2 s ) , the benefit was similar for both groups . However , this benefit decayed faster ( at interval >2 s ) among dyslexics ( red ) than among controls ( blue; p<0 . 005 ) . Error bars denote standard error ( SEM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20557 . 007 Participants’ response times ( RTs , measured as vocal response onset ) were shorter in the second presentation of the same non-word . However , the magnitude of this improvement declined as the time interval from the previous presentation of the same non-word increased ( i . e . when there were more intervening words ) . Overall , dyslexics’ RTs were longer than controls’ ( controls — n = 29; mean RT per word ± SEM = 725 ± 28 ms; dyslexics — n = 23; mean RT ± SEM = 929 ± 47 ms; z = 3 . 3 , p<0 . 001 . Mann-Whitney U test ) and they were less accurate ( controls — mean % correct ± SEM = 98 . 6 ± 0 . 3; dyslexics — 96 . 0 ± 1 . 4; z = 3 . 2 , p<0 . 005 . Mann-Whitney U test ) , as expected . Both groups showed large recent context ( repetition ) effects . Specifically , both groups showed substantially faster reading RTs in the second than in the first encounter with the same non-word , when the words were proximal in time ( <2 s interval ) , even though they were never consecutive . The smallest interval between two presentations of the same non-word was with one intervening word . As word presentation rate was based on reading rate , and dyslexics are slower readers , we equated the inter-word time interval in the two groups by considering only trials in which the second presentation occurred within <2 s ITI ( offset of the first presentation to the onset of the second presentation of the same non-word ) . In these trials , dyslexics’ benefit from repetition was similar to controls’ ( Figure 5B; mean benefit ± SEM for controls — 25% ± 2% or 170 ± 19 ms , for dyslexics — 23% ± 3% or 177 ± 20 ms; z = 0 . 1 , n . s . Mann-Whitney U test ) . However , in trials with larger inter-word intervals , dyslexics’ benefits were smaller . Importantly , dyslexics’ benefits were smaller than controls’ , even though their starting point had longer RTs , as explained above ( mean ITI of 42 s , max . ITI of 163 s; mean benefit ± SEM for controls — 12% ± 1% or 90 ± 10 ms , for dyslexics — 6% ± 1% or 52 ± 8 ms , z = 3 , p<0 . 005 . Mann-Whitney U test ) . Indeed , among dyslexics , the difference in benefit from repetitions with small ( <2 s ) compared to larger ( >2 s ) intervals was greater than this difference among controls ( z = 2 . 2 , p<0 . 05; Mann-Whitney U test ) . Note that for short inter-word intervals , dyslexics’ RT benefits do not differ from controls’ in spite of an intervening word . Thus , here too , dyslexics’ faster decay is consistent with a time-based effect rather than with an enhanced interference effect .
We propose that dyslexics’ shorter time constant of adaptation reflects a shorter time constant of implicit sound integration . Dyslexics’ faster decay decreases the time constant of their integration of sound statistics , and consequently reduces the reliability of their implicitly calculated priors . Owing to their noisier priors , attaining the same level of processing reliability requires the extraction of more on-line information , which requires more processing time , leading to the stereotypic ‘slower processing time’ in dyslexia . This account explains dyslexics’ often poorer performance on rapidly presented brief auditory stimuli for both simple tones ( Tallal , 1980; Temple et al . , 2000 ) and speech ( McArthur and Bishop , 2005; Boets et al . , 2007 ) . Moreover , it resolves seemingly inconsistent observations in dyslexics’ pattern of perceptual difficulties . On the one hand , dyslexics’ poor short term memory skills are amply documented ( Beneventi et al . , 2010; Banai and Yifat , 2012 ) and suggest that their performance should improve when memory load is reduced by decreasing task’s retention intervals . On the other hand , dyslexics are known to perform poorly with short stimuli presented at brief intervals ( Tallal , 1980 ) , suggesting that they should gain from longer inter-stimulus intervals . We claim that these seemingly contradictory observations result from the same basic impairment . The key feature is the proportion of on-line computations that need be allocated to solve the task . The more predictable the stimuli are for controls , both locally within session and globally in terms of long-term regularities ( such as in language ) the greater relative slowness ( ‘slow processing speed’ ) dyslexics are expected to show ( consistent with Ahissar et al . , 2006 ) . This slowness stems from their poor implicit memory mechanisms; namely , their faster decay of the neural trace that integrates the statistics of recent stimuli . Dyslexics’ phonological and reading skills reflect the outcome of long-term learning processes which are difficult to track . The use of a simple task of discrimination of pure tones ( which are physically simple but very infrequent environmentally ) enabled us to observe the dynamics of context effects in a relatively novel situation . It served to dissociate the benefits of very short-term ( <2 s; the scale attributed to working memory processes; Baddeley , 2010 ) implicit effects from somewhat longer-term effects . Dyslexics only benefitted less from prior statistics in the latter case . Dyslexics’ noisier priors ( predictions based on past experiences ) may lead to slower long-term learning of characteristic sound regularities , such as those in the native language ( Nicolson et al . , 2010 ) . It follows that dyslexics should benefit less than controls from a given number of experiences , and that group differences should increase with additional specific experiences . Thus , although both good and poor readers are expected to improve with practice , dyslexics’ benefits per exposure are expected to be smaller . Hence , counterintuitively , dyslexics’ relative difficulties are expected to be greater for highly trained stimuli . Our observation of reduced repetition effects in dyslexics’ reading rate is compatible with imaging observations by Pugh et al . ( 2008 ) , who compared the impact of word repetition in reading on BOLD activation in posterior reading-related areas , in dyslexics versus controls . They found that although both populations gained from such repetitions , the impact differed between groups . Whereas controls’ activation was monotonically reduced with repetitions , dyslexics’ activation initially increased . They interpreted this pattern in a manner consistent with ours by suggesting that dyslexics’ pattern seemed similar to that expected after a lesser amount of exposure to these words; i . e . , slower long-term learning of these ( sound ) patterns . Interestingly , dyslexics’ slower learning rate per event ( or exposure ) was also proposed by Nicolson et al . ( 2010 ) , who based their argument on slower learning curves in a different context . They found that dyslexics exhibit slower learning of simple motor tasks , which perhaps tap mechanisms that are partially common to those used in serial comparison tasks . We found that dyslexics’ P2 component had a shorter adaptation duration under both active and passive conditions . We also found that their N1 had a shorter adaptation duration under the active condition when compared to controls . The P2 ( which peaks 200 ms after tone onset ) component was shown to reflect the accumulation of sound statistics ( Tremblay et al . , 2010; Jaffe-Dax et al . , 2015 ) automatically , without explicit attention ( Sheehan et al . , 2005 ) . Moreover , it was shown to be abnormal among dyslexics ( Bishop and McArthur , 2004; Jaffe-Dax et al . , 2015 ) . Indeed , its dynamics matched those of the behavioral context effects measured simultaneously . However , even the earlier N1 component ( 100 ms ) had a shorter time constant here among dyslexics under the active condition . Whether these two components originate from one or two cortical generators is still unclear ( e . g . Mayhew et al . , 2010; Lanting et al . , 2013 ) . Nevertheless , recent studies have found that the magnitudes of adaptation of both are sensitive to the statistics that characterize the experiment ( e . g . Herrmann et al . , 2015 ) . We found a shorter time constant of adaptation in N1 than in P2 . Our observations suggest that the dynamics of adaptation in the auditory cortex of dyslexics differs from that of controls . But they do not directly point to the anatomical source of this abnormality as 100–200 ms ERP signals reflect the combined contribution of various brain sites . Thus , although dyslexics’ processing deficit is likely to result from a structural variation ( which probably has a genetic origin; Giraud and Ramus , 2013 ) , our observations are consistent with several accounts . The abnormally fast decay of N1 and P2 adaptation may reflect a different anatomy of the auditory cortex , as suggested in a recent longitudinal study ( Clark et al . , 2014 ) , or impaired long-range connectivity between posterior and frontal areas ( Boets et al . , 2013; Ramus , 2014 ) , which are not mutually exclusive . With respect to nature of dyslexics’ auditory processing deficits , several accounts are consistent with our current findings . These include both impaired top-down control ( Díaz et al . , 2012 ) , and unreliable ( i . e . more variable ) auditory responses ( Hornickel and Kraus , 2013 ) associated with seemingly noisier sensory systems ( Sperling et al . , 2005 ) , particularly in some frequency bands ( Goswami et al . , 2002; Goswami , 2011 ) . All these observations are expected outcomes of poorer implicit predictions made on the basis of stimuli statistics . Indeed , when dyslexics’ behavioral sensitivity to complex and challenging noise stimuli was assessed , it was found that their impairment is specific for repeated stimuli ( with >2 sec intervals ) . Moreover , their performance tended to be even better than controls’ when stimuli differed from the repeated , and hence implicitly predicted , ones ( Daikhin et al . , 2016; for visual analog see also Jaffe-Dax et al . , 2016 ) . Importantly , the involvement of the auditory cortex does not exclude the contribution of sub-cortical regions . Both the cerebellum and the basal ganglia are involved in the process of integrating sound regularities into improved perceptual performance with reduced reliance on on-line working memory processes and increased reliance on specific sound predictions ( Daikhin and Ahissar , 2015 ) . Thus , the hypothesis linking dyslexics’ reduced rate of automatization ( in reading and in other tasks ) with impaired usage of cerebellar processes ( Nicolson and Fawcett , 1990; Nicolson et al . , 2010 ) is also in line with our findings and interpretation . The attempts to map dyslexics’ processing difficulties to several potentially relevant anatomical abnormalities suggest that there may not be a single anatomical source of difficulty , and that the issue of the core deficit underlying dyslexia should be seen as process-related rather than structure-related . Our observation that the process of dyslexics’ neural adaptation is abnormally short paves the way for future studies of the neural processes that underlie reading difficulties in non-humans . An important insight suggested by our study , which is further supported by recent human studies in speech perception ( e . g . Kleinschmidt and Jaeger , 2016 ) , is that adaptation reflects a mechanism for statistical learning . This interpretation is in line with recent animal studies , suggesting that adaptation reflects a mechanism for making implicit high-resolution stimulus predictions on the basis of an experiment’s statistics ( Khouri and Nelken , 2015 ) . Another studied aspect of neural adaptation is its occurrence at multiple time scales , as observed both in animals ( Ulanovsky et al . , 2004; Khouri and Nelken , 2015 ) and in humans ( reviewed in Lu and Sperling , 2003 ) . Here we propose that dyslexics’ deficit is characterized at time scales >2 s after stimulus presentation . It would be interesting to compare whether individuals with memory deficits at shorter time intervals ( e . g . in iconic/anechoic memory , i . e . <500 ms ) have broader cognitive deficits , as suggested by previous studies ( Lu et al . , 2005; Miller et al . , 2010 ) . About half of our dyslexic participants showed adaptation time scales that were within the range of controls ( Figures 2–4 ) . This partial overlap between the two groups could have stemmed from the low reliability of our measurements . Another potential source for this variability might be different types of dyslexia among our dyslexic participants , which may relate to different underlying neural mechanisms . The standard reading tests that were used for inclusion or exclusion of participants might not have been fine-grained enough to define participants along the dimensions that map to different sources of underlying difficulties ( Zoccolotti and Friedmann , 2010 ) . We only tracked the short-term effects of dyslexics’ faster decay of memory trace . Therefore , we can only speculate about the relationship between these observations and dyslexics’ difficulties in acquiring expert-level proficiency in reading . One putative conceptual link may be provided by the Bayesian framework , when the principle of efficient coding is introduced ( Wei et al . , 2015 ) . This principle predicts that more likely stimuli gradually acquire denser , more reliable representations . The typical gradual enhancement of the representation of more likely stimuli may be reduced in dyslexia , leading to reduced sensitivity to the specific phonological and morphological forms that characterize native language . Assessing this hypothesis experimentally requires long-term tracking of the learning of novel sound statistics .
Sixty native Hebrew speakers ( 30 dyslexics and 30 good readers ) , all of whom were students at the Hebrew University [mean age ( STD ) = 24 . 2 ( 5 . 4 ) years; 36 females] were recruited for this study . Recruitment was based on ads at the Hebrew University . Monetary compensation for participation was according to standard student rates . The study was approved by the Hebrew University Committee for the Use of Human Subjects in Research . All dyslexic participants had been diagnosed as having a specific reading disability by authorized clinicians . Reading-related measures were also assessed in our lab ( detailed in Table 1 ) . Participants with more than 2 years of formal musical education were excluded , so that musical training would not be a major contributor to their pitch sensitivity ( Micheyl et al . , 2006; Parbery-Clark et al . , 2011 ) . Participants with poor Block Design scores ( lower than a normalized score of 7 ) were also excluded from the study . All participants filled in a questionnaire regarding any neurological or psychiatric disorders . None of the participants reported any such disorders . None of them had ever participated in a similar auditory experiment in our lab . 10 . 7554/eLife . 20557 . 008Table 1 . Cognitive scores for the dyslexic and control groups ( mean and standard deviation ) . Dyslexics differed from their good-reader peers in all phonological tasks and in verbal working memory , but not in their general reasoning skills ( Mann-Whitney U tests ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20557 . 008TestControl ( STD ) N = 30 Dyslexic ( STD ) N = 30 Mann-Whitney z valueAge ( years ) 25 . 8 ( 3 . 0 ) 24 . 3 ( 3 . 1 ) 1 . 6General cognitive ( scaled ) Block Design12 . 4 ( 2 . 9 ) 12 . 1 ( 3 . 5 ) 0 . 3Digit Span11 . 1 ( 2 . 8 ) 7 . 8 ( 1 . 7 ) 4 . 7***Phonological speed [items/minute] Pseudo-word reading rate58 . 4 ( 24 . 4 ) 32 . 2 ( 10 . 5 ) 4 . 4***Single-word reading rate96 . 8 ( 32 . 5 ) 68 . 3 ( 25 . 8 ) 3 . 3**Word pattern recognition rate68 . 1 ( 15 . 2 ) 39 . 9 ( 13 . 5 ) 5 . 6***Passage reading rate140 . 4 ( 23 . 8 ) 97 . 8 ( 22 . 3 ) 5 . 7***Spoonerism rate10 . 0 ( 3 . 0 ) 5 . 8 ( 3 . 2 ) 4 . 6***Phonological accuracy [% correct] Pseudo-word reading accuracy89 . 7 ( 11 . 2 ) 62 . 4 ( 18 . 3 ) 5 . 1***Single-word reading accuracy97 . 1 ( 4 . 3 ) 87 . 6 ( 8 . 3 ) 4 . 8***Word pattern recognition accuracy100 ( 0 ) 96 . 2 ( 6 . 4 ) 4 . 3***Passage reading accuracy98 . 6 ( 1 . 4 ) 94 . 8 ( 4 . 5 ) 4 . 9***Spoonerism accuracy92 . 2 ( 6 . 8 ) 77 . 9 ( 18 ) 3 . 2***p<0 . 05; **p<0 . 005; ***p<0 . 0005 . Participants were administered four sessions on four different days . In session one , participants were administered a series of cognitive assessments . Thirty dyslexics and 30 controls were admitted to this session . In session two , participants performed a two-tone frequency discrimination task with a specially designed sequence of trials ( Experiment 1; Figure 1A–B ) . The same 30 dyslexics and 30 controls participated in this session . In session three , ERPs were recorded both passively and while performing the discrimination task . First , participants watched a silent movie while a series of single tones was presented to them in four blocks of four different Inter Stimulus Intervals ( ISI ) of 2 , 3 . 5 , 6 . 5 or 9 . 5 s ( Experiment 2b ) , in a random order . Second , participants actively engaged in the two-tone frequency discrimination task in four blocks with different Inter-Trial-Intervals ( ITI – time interval between the second tone in the trial and first tone in the following trial ) of 1 . 4 , 2 . 9 , 5 . 9 and 8 . 9 s ( Experiment 2a ) . The ISI between the two tones in the trial was 600 ms . Thus , in both passive and active conditions , the Stimulus Onset Asynchronies , i . e . the time intervals between the onset of the first tones in adjacent trials were 2 , 3 . 5 , 6 . 5 and 9 . 5 s . A subgroup of 25 dyslexics and 23 controls participated in this session . In session four , participants performed a fast reading task of visually presented single non-words . Voice response was recorded ( Experiment 3; Figure 4A ) . Both rate and accuracy were obtained . 29 dyslexics and 23 controls participated in this session . All sessions were administered in a sound-attenuated room . Sounds and visual presentations were produced using Matlab ( The Mathworks , Inc . , Natick , MA ) . Tones were presented and voice response was recorded by Psychtoolbox and Psychportaudio ( Kleiner et al . , 2007 ) through a Saffire 6 USB audio interface ( Focusrite Audio Engineering ltd . , High Wycombe , UK ) . General cognitive abilities and phonological skills were assessed using standard tasks: A . Non-verbal reasoning ability . This was measured with the Block Design , a standard test for assessing visuo-spatial reasoning ( WAIS-III; Wechsler , 1997 ) . B . Short-term verbal memory . This was evaluated with the standard Digit Span task ( forward and backward; Hebrew version of WAIS-III; Wechsler , 1997 ) . C . Phonological decoding and single-word reading . Pseudo-word and single word reading were assessed using standard Hebrew lists designed by Deutsch and Bentin ( 1996 ) . D . Word pattern recognition . Subjects were presented with 24 pairs , each composed of a word and a pseudo-homophone , and were asked to point to the word in each pair . E . Fluent reading . Subjects read an academic level passage of 150 words followed by a comprehension question . F . Phonological awareness was assessed using the Spoonerism task ( MacKay , 1970; Möller et al . , 2007 ) . Participants heard ( Hebrew ) word pairs and were asked to switch the first phonemes of the two words and respond vocally ( e . g . : /laila tov/ , ‘good night’ in Hebrew , should be switched to /taila lov/ ) . In all phonological and reading tasks , both accuracy and rate were scored . Participants performed four blocks of 150 trials of the two-tone frequency discrimination task . Each trial contained a tone pair ( 50 ms , 70 dB each tone; 600 ms inter-tone intervals ) . They were asked to indicate which of the two tones had a higher pitch . A short demo of 10 trials preceded the actual experiment . Feedback was provided only in the demo trials . 80% success on the 10 demo ( easy ) trials was a prerequisite for continuing the task . We did not administer feedback during the assessment because we did not want to affect the magnitude of the listeners’ contraction bias . The task was administered with a set of constant stimuli that we designed specifically for this experiment ( available at: https://goo . gl/UnrG1A ) . Its design allowed us to evaluate the contribution to the context effect of the most recent trial separately from that of all previous trials . Assessing these effects separately required a specifically designed sequence , since these effects are typically correlated . Specifically , the direction of the frequency distance between the first tone of the current trial and that of the first tone of the most recent trial , and the direction of the distance from that tone to the average across trials , are typically correlated . In the design of this series , we ensured that they were not correlated . In other words , the sign of the global context ( G ) : G ( t ) =sign ( f1 ( t ) −⟨ f1 ⟩ ) and that of the recent context ( R ) : R ( t ) =sign ( f1 ( t ) −f1 ( t−1 ) ) , in each trial were not correlated . In this sequence , the overall contribution of both the local and global context was positive , yet small . In the active condition ( Experiment 2a ) , participants were administered four blocks of 100 trials each of the two-tone frequency discrimination task . Each trial contained a tone pair ( 50 ms , 70 dB each tone; 600 ms inter-tone intervals ) , and listeners were asked to indicate which of the two tones had a higher pitch . The sequence of trials was randomly drawn for each participant . In each trial , a tone was chosen randomly from 800 Hz to 1250 Hz . The other tone was chosen randomly to have a frequency difference ( plus or minus ) between 1% and 30% from the previously chosen tone . The order of the tones was also randomly chosen . Trial onset asynchrony was fixed for each block at 2 , 3 . 5 , 6 . 5 or 9 . 5 s . Block order was counterbalanced across subjects . In the first , passive , part of the session ( Experiment 2b ) , only the first tone in each pair was presented . We compensated for it by increasing the Inter-Stimulus Intervals ( between the tone’s offset on the previous trial and the onset of the current trial ) in this condition by 0 . 6 s . Consequently , the onset-to-onset intervals between first tones of adjacent events were the same in the two conditions . Subjects watched a silent movie and were asked to ignore the tones . Participants were presented with six blocks of 120 non-words and were asked to read them aloud as fast as they could . Voice onset and offset were acquired . Each non-word was presented 500 ms after the voice offset of the preceding non-word . Presentation remained until the voice offset of the current non-word . Non-words were randomly generated by conjunction of two randomly chosen valid Hebrew syllables ( consonant-vowel and consonant-vowel-consonant , or vice versa ) . Electrophysiological activity was recorded from 32 active Ag-AgCl electrodes mounted on an elastic cap using the BioSemi ActiveTwo tools and recording software ( BioSemi B . V . , Amsterdam , The Netherlands ) . Electrode sites were based on the 10–20 system ( American Electroencephalographic Society , 1991 ) . Two additional electrodes were placed over the left and right mastoids . Horizontal EOG was recorded from two electrodes placed at the outer canthi of both eyes . Vertical EOG was recorded from electrodes on the infraorbital and supraorbital regions of the right eye in line with the pupil . EEG and EOG signals were amplified , filtered with an analogue band-pass filter of 0 . 16–100 Hz , and sampled at 256 Hz . Offline analysis was performed using Brain Vision Analyzer 1 . 05 software ( Brain Products GmbH , Gilching , Germany ) and EEGLAB toolbox for Matlab ( Delorme and Makeig , 2004 ) . The EEG signal was digitally band-pass filtered between 1 Hz and 30 Hz to remove large drifts in signal and high-frequency noise . ICA analysis was trained on the entire length of each block and on all scalp electrodes to identify components that reflect eye-blink- or eye-movement-evoked electrical activity . An eye-related component was identified by its time-correlation with the occurrence of blinks or saccades . This relationship between the identified component and eye-blink activity was verified by confirming that the component's scalp distribution was typical of eye-related electrical activity ( Delorme et al . , 2007 ) . Data were referenced to the nose channel to remove external electrical influence . Artifact rejection was applied to the non-segmented data according to the following criteria: any data point with an EOG or EEG > ± 100 µV was rejected along with the data ± 300 ms around it . In addition , if the difference between the maximum and the minimum amplitudes of two data points within an interval of 50 ms exceeded 100 µV , data ± 200 ms around it were rejected . Finally , if the difference between two adjacent data points was more than 50 µV , the data ± 300 ms around it were rejected . Trials containing rejected data points were omitted from further analysis . Groups did not differ on the number of trials that were analyzed ( active condition: controls — 387 ± 6; dyslexics — 382 ± 5; z = 1 . 6 , n . s . ; passive condition: controls — 394 ± 4; dyslexics — 394 ± 4; z = −0 . 2 , n . s . ; mean number of trials ± SEM , Mann-Whitney U tests ) . For ERP averaging across trials , the EEG was parsed to 2000 ms epochs starting 500 ms before the onset of the first stimulus in each pair , and averaged separately for each electrode . The baseline was adjusted by subtracting the mean amplitude of the pre-stimulus period ( 500–150 ms before the onset of the first stimulus in the trial ) of each ERP from each data point in the epoch . The pre-stimulus baseline period was calculated from this time interval to exclude effects of anticipatory responses that preceded informative anticipated stimuli ( CNV; Walter et al . , 1964 ) . ERP analysis was based on the epochs that were recorded with electrode Cz ( at the vertex of the scalp ) . This electrode measured the most prominent response to the auditory stimuli . Data from each acquisition session were analyzed separately . The magnitude of the ERP components was calculated as the area under the curve between 70 ms and 130 ms after first tone’s onset for N1 and 150–250 ms time range for P2 . We repeated the entire analysis for adjacent electrodes ( Fz , FC1 , FC2 , C3 , C4 , CP1 , CP2 and Pz ) and found similar results . Context effects and ERPs in each of the four different ITIs ( experiment 2 ) were fitted with an exponential decay model for each participant separately . This model was previously used to characterize the decay of context effects both behaviorally and for MEG measurements ( Lu et al . , 1992; Sams et al . , 1993 ) . In a previous study , the effect of context as a function of number of trials back was quantitatively measured and indeed resembled an exponential decay ( Raviv et al . , 2012 ) . Following that quantitative description , we found that the exponential decay of previous trials captures context effects during this task ( Jaffe-Dax et al . , 2015 ) . In the current study , we modeled the impact of previous trials as a function of the temporal interval instead of number of trials . The model , α+βexp ( −t/τ ) , had three parameters: α – asymptote after recovery; i . e . , the value expected when t→∞ , β – the magnitude of adaptation; i . e . , the value expected at t=0 minus α , and τ – time constant of adaptation; i . e . , the time it takes for the measure expected at t=0 to decay to 1/e ( ~37% ) of its initial value . A small τ indicates fast decay and a large τ indicates a slow decay . Formally , we searched for the triplet of parameters that minimizes the squared difference between the data and the model prediction: the units of α and β are those of the fitted measure ( d’ for behavioral bias , μV x msec for ERP ) . In the passive condition , t is the Inter-Stimulus Interval ( ISI ) between the offset of a tone and the onset of the consecutive tone . In the active condition , t is the Inter-Trial Interval ( ITI ) between the offset of the second tone and the onset of the first tone in the consecutive trial . Fitted parameters were estimated by minimizing the squared error of the exponential decay model in a limited range . For the ERP adaptation , limits were from 0 to 15 , 000 for α and from −15 , 000 to 0 for β . For the d’ difference , the limits were from 0 to 100 for α and β . For both measures , τ was limited to be from 0 to 100 s . The groups did not differ on the exponential curve’s Goodness-of-Fit for any of the analyzed measurements ( z < 1 . 8 , n . s . Mann-Whitney U tests ) . The exponential decay model captured the ERP adaptation decay well for both groups ( R2 >0 . 4 ) . The model was less able to account for the behavioral bias decay , especially for controls ( R2 = 0 . 2 ) , suggesting a more complex mechanism than could be well described by a single exponential decay . For purposes of compatibility with our previous studies ( Raviv et al . , 2012; Jaffe-Dax et al . , 2015 ) and to avoid assumptions of normal distribution , we used conservative , non-parametric tests throughout the study . Using standard parametric tests yielded similar statistical significance . | The term “dyslexia” comes from the Greek for “difficulty with words” . People with dyslexia struggle with reading and spelling: they may mix up letters within words and tend to read and write more slowly than others . However , not every symptom of dyslexia is related to literacy . Affected individuals also differ from good readers on simple perceptual tasks , such as distinguishing between tones of different frequencies . In a series of trials involving discrimination between pairs of tones , a person’s performance on each trial will be influenced by the tones presented on previous trials . Both good readers and individuals with dyslexia automatically form a subconscious memory of the tones they hear , and use this memory to guide their performance on subsequent trials . However , people with dyslexia benefit less from this effect than good readers . Jaffe-Dax et al . have now identified the mechanism that underlies this phenomenon , revealing new insights into how dyslexia influences brain activity . By varying the interval between successive pairs of tones , the experiments showed that the memory of previous tones decays faster in people with dyslexia than in good readers . A similar effect occurs when the stimuli are nonsense words . Both good and poor readers manage to read nonsense words more quickly on their second attempt . However , people with dyslexia benefit less from the previous exposure when the gap between repetitions is longer than a couple of seconds . Further studies are needed to determine whether and how the faster decay of memory traces for words is related to impaired reading ability in people with dyslexia . One possibility is that the faster decay of memory traces makes it more difficult to predict future stimuli , which may impair reading . An imaging study is underway to investigate where in the brain this rapid decay of memory traces occurs . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Dyslexics’ faster decay of implicit memory for sounds and words is manifested in their shorter neural adaptation |
Mitochondria move throughout neuronal dendrites and localize to sites of energy demand . The prevailing view of dendritic mitochondria as highly motile organelles whose distribution is continually adjusted by neuronal activity via Ca2+-dependent arrests is based on observations in cultured neurons exposed to artificial stimuli . Here , we analyze the movements of mitochondria in ganglion cell dendrites in the intact retina . We find that whereas during development 30% of mitochondria are motile at any time , as dendrites mature , mitochondria all but stop moving and localize stably to synapses and branch points . Neither spontaneous nor sensory-evoked activity and Ca2+ transients alter motility of dendritic mitochondria; and pathological hyperactivity in a mouse model of retinal degeneration elevates rather than reduces motility . Thus , our findings indicate that dendritic mitochondria reach stable positions during a critical developmental period of high motility , and challenge current views about the role of activity in regulating mitochondrial transport in dendrites .
Mitochondria provide energy in the form of ATP and phosphocreatine , and participate in Ca2+ signaling . In neurons , mitochondrial biosynthesis occurs in the soma , but sites of energy use and Ca2+ influx are dispersed across axonal and dendritic arbors ( Davis and Clayton , 1996 ) . To meet these distributed demands , neuronal mitochondria are transported throughout axons and dendrites along microtubule tracks ( Ehlers , 2013; Lin and Sheng , 2015 ) . A number of recent studies have explored the dynamics and function of axonal mitochondria , including detailed analyses of mitochondrial movements in peripheral and central axons in vivo ( Breckwoldt et al . , 2014; Misgeld et al . , 2007; Plucinska et al . , 2012; Takihara et al . , 2015 ) . By comparison , dendritic mitochondria are less explored and their dynamics have not been examined in intact neural circuits . Most of the energy in dendrites is consumed at synapses ( Attwell and Laughlin , 2001; Howarth et al . , 2012 ) , which are also a primary site of Ca2+ influx ( Augustine et al . , 2003; Grienberger and Konnerth , 2012 ) . Dendritic mitochondria have been shown to localize to synapses in several systems and appear to contribute to their formation and plasticity ( Chang et al . , 2006; Ishihara et al . , 2009; Li et al . , 2010; Li et al . , 2004 ) . Increased ER complexity and Golgi outposts at dendritic branch points support protein and lipid biosynthesis , and secretory trafficking ( Cui-Wang et al . , 2012; Ehlers , 2013; Horton et al . , 2005; Ye et al . , 2007 ) . In addition , branch points are hotspots for Ca2+ signals in some dendrites ( Fitzpatrick et al . , 2009; Larkum et al . , 2003 ) . Mitochondria are required for dendrites to establish normal branching patterns ( Fukumitsu et al . , 2015; Kimura and Murakami , 2014 ) , but whether mitochondria localize to branch points is not clear . Moreover , when associations between mitochondria and synapses or branch points emerge during development has not been examined . Finally , the prevailing view of dendritic mitochondria as being highly motile is based on observations of cultured neurons isolated from embryonic tissue . Whether this view holds true for neurons in their native environments , and how mitochondrial motility in dendrites changes as branching and connectivity patterns stabilize during circuit maturation remains unknown . Neuronal activity is thought to control the movements of dendritic mitochondria . In particular , increases in intracellular Ca2+ were shown to uncouple mitochondria from motor proteins via the Ca2+-dependent adaptor protein Miro1 ( MacAskill et al . , 2009; Wang and Schwarz , 2009 ) . However , the evidence supporting a leading role for activity in dendritic mitochondrial transport was obtained from cultured neurons exposed to artificial stimuli ( e . g . high external K+ ) , raising the question whether physiologically occurring activity patterns in intact circuits exert a similar influence . Here , we analyze how development and neuronal activity shape the distribution and dynamics of mitochondria in dendrites of ganglion cells in the intact retina . We find that mitochondrial density in retinal ganglion cell ( RGC ) dendrites reaches near-mature levels before most synapses are formed . In the adult retina , mitochondria are enriched at synapses and branch points of RGC dendrites . Mitochondria localize to synapses early , whereas their accumulation at branch points occurs gradually during circuit development . The motility of mitochondria changes fundamentally across development . During the period of dendrite growth and synaptogenesis approximately 30% of mitochondria are in motion at any time . By contrast , we observed no movements of dendritic mitochondria in mature circuits . This drastic decline in motility is cargo-specific , as peroxisomes remain motile in mature RGC dendrites . Using simultaneous two-photon imaging of mitochondrial movements and intracellular Ca2+ , we find that although elevation of external K+ reduces motility of dendritic mitochondria in the retina as it does in cultured neurons , neither spontaneous waves of activity during development , nor sensory-evoked activity at maturity alter the motility of dendritic mitochondria in RGCs . Finally , pathologically elevated RGC activity , a common feature of retinal degeneration , restores motility in mature retinas to developmental levels .
To analyze the distribution of dendritic mitochondria relative to sites of high energy demand and Ca2+ signaling , and assess its changes across development , we biolistically labeled RGCs with mitochondrially targeted yellow fluorescent protein ( mtYFP ) , the postsynaptic density protein 95 fused to CFP ( PSD95-CFP ) , and a red cytosolic fluorophore ( tdTomato ) . Others and we previously showed that PSD95-CFP localizes specifically to BC synapses on RGC dendrites ( i . e . excitatory synapses ) ( Jakobs et al . , 2008; Kerschensteiner et al . , 2009; Morgan et al . , 2008 ) . We labeled RGCs at three different ages ( Figure 2A–C ) : P9 ( early synaptogenesis , mid dendritogenesis , spontaneous retinal waves ) , P15 ( mid synaptogenesis , mature dendrites , transition from retinal waves to light-evoked activity ) , and P21 ( mature synapses , mature dendrites , light-evoked activity ) . We found that mitochondria enter dendrites to near-mature levels before most synapses are formed ( Figure 2D , E ) . Similar to other systems , mitochondria in RGC dendrites are found closer to synapses than expected by chance ( Figure 2F ) . This synaptic localization of mitochondria was apparent from the earliest time point examined ( P9 ) . In addition , at maturity , mitochondria are enriched >15-fold at branch points . Unlike their localization to synapses , the enrichment of mitochondria at branch points emerges gradually during development ( Figure 2G ) . Thus , mitochondria enter RGC dendrites during early development , associate with synapses as circuits go through the trial-and-error process of establishing precise connections , and progressively accumulate at branch points even after mature arborization patterns are established . 10 . 7554/eLife . 11583 . 004Figure 2 . Mitochondrial distribution in RGC dendrites across development . ( A–C ) Representative RGCs expressing mtYFP and PSD95-CFP in P9 ( A ) , P15 ( B ) , and P21 ( C ) retinas . Top panels show maximum intensity projections ( MIPs ) through confocal image stacks and bottom panels show MIPs of excerpts of the same stacks at higher magnification . ( D ) Density of mitochondria in RGC dendrites expressed as a volumetric fraction ( see 'Materials and methods' ) across development . ( E ) Density of synapses along RGC dendrites given per length of dendrite based on skeletonization of the respective arbors ( see 'Materials and methods' ) across development . ( F ) Scatter plots comparing the average nearest neighbor distance from synapses to mitochondria ( NND-observed ) to the mean average NND obtained from Monte Carlo simulations in which the positions of synapses along dendrites were randomized ( NND-random ) . Dots represent individual cells and circles ( error bars ) indicate the mean ( ± SEM ) at the different ages examined . ( G ) Normalized mitochondrial density plotted as a function of distance from dendritic branch points at P9 ( left ) , P15 ( middle ) , and P21 ( right ) . Solid lines ( shaded areas ) indicate the mean ( ± SEM ) across a number of RGCs ( P9 n = 6 , P15 , n = 6 , P21 , n = 8 ) . mtYFP , mitochondrially targeted yellow fluorescent protein; RGC , retinal ganglion cell . DOI: http://dx . doi . org/10 . 7554/eLife . 11583 . 00410 . 7554/eLife . 11583 . 005Video 1 . Mitochondrial motility across development . Time-lapse confocal recording of RGC dendrites expressing mtYFP at P9 ( top panel ) , P15 ( center panel ) , and P21 ( bottom panel ) . Scale bar: 5 μm . 0 . 9 Hz time-lapse , playback speed = 15 frames per second . mtYFP , mitochondrially targeted yellow fluorescent protein; RGC , retinal ganglion cell . DOI: http://dx . doi . org/10 . 7554/eLife . 11583 . 005 To study the dynamics of dendritic mitochondria across development , we performed time-lapse imaging experiments ( 0 . 9 frames per second or fps ) on RGCs biolistically or transgenically ( Thy1-mtCFP-P ) labeled with mtYFP and mtCFP , respectively ( Misgeld et al . , 2007; Williams et al . , 2013 ) . Because mitochondria co-localize with synapses even at ages when synapse turnover is high ( Figure 2F ) , we expected mitochondria during development to be motile , allowing them to adjust to changing patterns of connections . Indeed , similar to studies on cultured neurons , we found that at P9 approximately 30% of mitochondria moved at least once per minute ( i . e . motile fraction ) ( Figure 3A , B and Video 1 ) . However , subsequently , the motile fraction of mitochondria in RGC dendrites decreases steeply , falling to approximately 15% by P15 , and at P21 , we observed no mitochondrial movements in dendrites of 7 RGCs imaged in four retinas for 35 min . This drastic decline in motility is specific to mitochondria and does not reflect a general transition in dendritic trafficking , as SKL-GFP-labeled peroxisomes ( Monosov et al . , 1996 ) in RGC dendrites remain motile at P21 ( Figure 3C , D ) and beyond ( data not shown ) . Interestingly , the movement patterns - including duration of runs , speed during uninterrupted motion , and duration of short pauses - of motile mitochondria at P15 were not significantly different from those observed at P9 ( Figure 3E–G ) . Thus , as circuits mature dendritic mitochondria undergo a cargo-specific transition from a partially motile to a stationary phase , in which they preferentially localize to synapses and branch points . 10 . 7554/eLife . 11583 . 006Figure 3 . Motility of dendritic mitochondria and peroxisomes across development . ( A ) Kymographs of representative time-lapse imaging series of mitochondria ( mtYFP , 0 . 9 fps , A ) at P9 ( left ) , P15 ( middle ) , and P21 ( right ) . Top panels show still frames at t = 0s of the branch segments depicted in the kymographs in the bottom panels . ( B ) Summary data of the motile fraction of mitochondria across development ( P9 n = 10 RGCs , P15 n = 9 RGCs , P21 n = 7 RGCs ) . ( C , D ) Analogous to A ( C ) and B ( D ) but for time-lapse imaging of peroxisomes labeled with SKL-GFP ( P9 n = 11 RGCs , P15 n = 10 RGCs , P21 n = 12 RGCs ) . ( E–G ) Bars ( error bars ) indicating the mean ( ± SEM ) duration of uninterrupted runs ( E ) , duration of pauses ( F ) , and speed during uninterrupted motion ( G ) for mitochondria at P9 and P15 ( P9 n = 65 mitochondria , P15 n = 32 mitochondria ) . See also Video 1 . mtYFP , mitochondrially targeted yellow fluorescent protein; RGCs , retinal ganglion cells . DOI: http://dx . doi . org/10 . 7554/eLife . 11583 . 00610 . 7554/eLife . 11583 . 007Video 2 . Mitochondrial motility during spontaneous global Ca2+ transients . Simultaneous time-lapse two-photon recording of GCaMP6 and mtDsRed expression in P9 RGC dendrites . Top panel shows Ca2+ events in a section of dendrite . Center panel shows mtDsRed signal in the same dendritic branch . Bottom panel shows dots overlaid at each mitochondrion position tracked for mitochondria in the center panel . In frames in which a mitochondrion is moving , its dot is green; dots are red for stationary mitochondria . Scale bar: 5 μm . 0 . 9 Hz time-lapse , playback speed = 15 frames per second . RGC , retinal ganglion cell . DOI: http://dx . doi . org/10 . 7554/eLife . 11583 . 007 Neuronal activity has been proposed to control the motility of dendritic mitochondria . In particular , increases in intracellular Ca2+ accompanying neuronal activation are thought to uncouple mitochondria from microtubule motors , causing acute movement arrests ( MacAskill et al . , 2009; Wang and Schwarz , 2009; Yi et al . , 2004 ) . The evidence supporting this model comes from studies of cultured neurons exposed to artificial stimuli , raising the question whether physiologically occurring activity patterns and Ca2+ transients similarly control the movements of dendritic mitochondria in intact circuits . The developing retina spontaneously generates propagating waves of RGC activity . In retinal waves , excitatory input ( P0 – P10: cholinergic , P10 – P15: glutamatergic ) elicits bursts of action potentials in RGCs ( Kerschensteiner , 2013 ) . As waves propagate across the retina , they synchronize the firing of neighboring RGCs . We first recorded ensembles of RGCs at P9 on multielectrode arrays ( MEAs , Figure 4A ) , confirming that biolistically labeled explants generate waves of activity with similar frequency and correlation structure to those observed in unlabeled preparations and in vivo ( Figure 4C , D ) ( Ackman et al . , 2012; Demas et al . , 2003 ) . We then performed two-photon Ca2+ imaging of RGCs biolistically labeled with GCaMP6s . At P9 , nearby RGCs ( <200 μm between cell bodies ) exhibited synchronous global Ca2+ transients across their dendritic arbors . The frequency and correlation of these transients was indistinguishable from that of spike bursts in MEA recordings ( Figure 4B–D ) , indicating that global Ca2+ transients in RGC dendrites accompany retinal waves ( Lohmann et al . , 2002 ) . To test whether this physiologically occurring activity pattern regulates the motility of mitochondria in developing dendrites , we co-labeled RGCs with GCaMP6s and mtDsRed . Simultaneous imaging of dendritic Ca2+ signals and mitochondrial movements ( Figure 4E ) , revealed that wave-associated Ca2+ transients do not alter mitochondrial motility in RGC dendrites: neither the instantaneous motile fraction ( Figure 4H and Video 2 ) nor the average speed of moving mitochondria ( △ speed: 0 . 132 ± 0 . 067 μm/s , n = 66 mitochondria , p>0 . 1 ) change following a dendritic Ca2+ transient . 10 . 7554/eLife . 11583 . 008Figure 4 . Spontaneous neuronal activity , dendritic Ca2+ transients , and mitochondrial motility during development . ( A ) Raster plot of representative spike trains of eight neighboring RGCs recorded in a biolistically labeled retinal explant on an MEA at P9 . ( B ) Representative △F/F traces of GCaMP6s signals recorded from three neighboring RGCs by two-photon imaging . ( C ) Cross-correlations of firing rates ( top panel ) and dendritic Ca2+ transients ( bottom panel ) of neighboring RGCs ( <200 μm between recording sites on MEA , <200 μm between cell bodies in Ca2+ imaging ) ( MEA recordings n = 30 pairs , Ca2+ imaging n = 11 pairs ) . ( D ) Average frequencies of waves of correlated activity on MEA ( black bar ) and global Ca2+ signals in RGC dendrites observed by two-photon imaging ( green bar ) in biolistically labeled retinal explants at P9 ( MEA n = 6 retinas , Ca2+ imaging n = 22 retinas , p>0 . 6 ) . ( E–G ) GCaMP6s and mtDsRed signals during simultaneous fast ( 0 . 9 fps ) time-lapse two-photon imaging . Top: still frames of mtDsRed channel at t = 0s . Left panels: ΔF/F traces aligned with time plotted in kymograph . ΔF/F traces were calculated based on the average GCaMP6s intensity in the dendrite segment shown above the kymograph . Center panels: kymographs of mtDsRed signal . Green overlay indicates the timing Ca2+ transients . Right panels: black lines represent schematized depictions of mitochondrial runs in the kymographs . ( F ) Break between top and bottom panels represents 10 min between pre-treatment ( top ) and 30 mM K+ application ( bottom ) . ( G ) 5 µM DHβE treatment . ( H ) Top: Plot of the instantaneous mitochondrial motility as a function of time relative to a Ca2+ transient ( n = 134 mitochondria , 10 RGCs ) . Bottom: Normalized Ca2+ imaging trace as a function of time ( aligned on t = 0 as the time of the Ca2+ transient ) . Lines ( shaded areas ) represent the mean ( ± SEM ) . ( I , J ) Paired plots of RGCs before and after 30 mM K+ ( I ) or DhβE ( J ) application . In the left panel of I the percentage of time above threshold measured was measured as percentage of frames in which the average GCaMP6s intensity within the frame was more than 2 SD above mean intensity of across the whole recording . The left panel of J show the frequency of Ca2+ transients . Right panels of I and J show changes in mitochondrial motility between pre-treatment and treatment portions of the recordings ( 30 mM K+ n = 8 RGCs , DhβE n = 7 RGCs ) . RGCs , retinal ganglion cells . DOI: http://dx . doi . org/10 . 7554/eLife . 11583 . 00810 . 7554/eLife . 11583 . 009Figure 4—figure supplement 1 . Miro1 expression and effect on mitochondrial motility in RGCs during development . ( A ) Western blot showing Miro1 protein expression levels in whole retinal lysate at P6 – P21 . Bottom panel shows β-actin loading control . ( B ) Bars indicate the intensities of the western blot signal relative to β-actin intensity and normalized to P12 intensity level . ( C ) Kymographs of representative time-lapse imaging series of mitochondria ( mtYFP , 0 . 9 fps ) in P9 RGCs overexpressing Miro1 ( left ) or expressing mtYFP alone ( right ) . Top panels show still frames at t = 0s of the branch segments depicted in the kymographs in the bottom panels . ( D ) Summary data of the motile fraction of mitochondria in RGCs overexpressing Miro1 , a mutant form of Miro1 unable to bind Ca2+ ( MiroKK ) , or mtYFP alone ( control ) ( Miro1 n = 9 RGCs , MiroKK n = 4 , control n = 6 ) . fps , frames per second; mtYFP , mitochondrially targeted yellow fluorescent protein; RGC , retinal ganglion cell . DOI: http://dx . doi . org/10 . 7554/eLife . 11583 . 00910 . 7554/eLife . 11583 . 010Figure 4—figure supplement 2 . Spontaneous , local dendritic Ca2+ transients and mitochondrial motility during development . ( A ) Bars ( error bars ) indicating the mean ( ± SEM ) frequency of global and local Ca2+ events in P9 RGCs ( global n = 32 RGCs , local n = 7 ) . ( B ) Paired plot of average mitochondrial speed in the 5 s before ( Pre ) and 5 s after ( Post ) a local Ca2+ event within a moving mitochondrion’s path ( n = 7 mitochondria ) . ( C ) Representative images of a P9 RGC dendritic branch during global and local signals . Heat maps of ΔF/F intensities are overlaid onto a maximum intensity projection of the GCaMP6 signal in all three panels . The top panel is a representative image of the baseline GCaMP6 intensity; the middle panel is a global Ca2+ event and the bottom panel shows a local Ca2+ transient . Scale bar: 5 μm . RGC , retinal ganglion cell . DOI: http://dx . doi . org/10 . 7554/eLife . 11583 . 010 In cultured neurons , Ca2+ elevations were shown to stop moving mitochondria via the adaptor protein Miro1 , which uncouples mitochondria from kinesin motors upon Ca2+ binding . Miro1 is readily detected in Western blots of the retina from at least P6 onwards and its expression reaches mature levels by P15 ( Figure 4—figure supplement 1 ) . In addition , we found that biolistic delivery of wild-type Miro1 or of a Ca2+-binding-deficient mutant ( Miro1KK ) does not alter the motility of dendritic mitochondria . Thus , a lack of Miro1 expression or function is unlikely to account for the lack of effect of physiologic activity patterns on the movements of mitochondria in RGC dendrites . Several previous studies raised external K+ concentrations to elicit changes in mitochondrial motility . We therefore tested the effects of this manipulation in the developing retina . Elevating external K+ to 30 mM tonically increased the concentration of Ca2+ in RGC dendrites and , similar to observations in cultured neurons , decreased the motile fraction of mitochondria ( Figure 4F , I and Video 3 ) . By contrast , blocking cholinergic retinal waves and the accompanying Ca2+ transients with the nicotinic antagonist DhβE did not affect mitochondrial motility in RGC dendrites ( Figure 4G , J and Video 3 ) . 10 . 7554/eLife . 11583 . 011Video 3 . Spontaneous global Ca2+ transients and pharmacological treatment during retinal waves . ( A ) Time-lapse two-photon recording of GCaMP6 signal in a P9 RGC undergoing spontaneous Ca2+ events while participating in retinal waves . ( B ) A P9 RGC before and during wash-in of 30 mM High K+ mACSF , exhibiting a large sustained increase in GCaMP6 intensity . ( C ) A P9 RGC before and during wash-in of 5 μM DhβE , exhibiting an abolition of spontaneous wave activity . Scale bars: 5 μm . 0 . 9 Hz time-lapse , playback speed = 20 frames per second . RGC , retinal ganglion cell . DOI: http://dx . doi . org/10 . 7554/eLife . 11583 . 01110 . 7554/eLife . 11583 . 012Video 4 . Mitochondrial motility in pathologically hyperactive mature cells . Time-lapse confocal recording of RGC dendrites expressing mtYFP in a P21 WT dendrite ( top panel ) and a P21 Crx-/- dendrite ( bottom panel ) . Scale bars: 5 μm . 0 . 9 Hz time-lapse , playback speed = 15 frames per second . DOI: http://dx . doi . org/10 . 7554/eLife . 11583 . 012 A subset of RGCs ( 21 of 59 cells ) exhibited local dendritic Ca2+ transients at a similar frequency as global signals ( Figure 4—figure supplement 2 , local: 0 . 0102 ± 0 . 0055 events / s , n = 7 cells , global: 0 . 0141 ± 0 . 0022 events / s , n = 32 cells , p>0 . 4 ) . Local transients have previously been observed in RGC dendrites in the developing chick retina , where they are caused by Ca2+ release from internal stores ( Lohmann et al . , 2002 ) . Like global signals , local Ca2+ transients in developing mouse RGC dendrites do not appear to alter mitochondrial motility ( Figure 4—figure supplement 2 ) . Together these results suggest that physiologically occurring activity patterns and Ca2+ transients in the developing retina do not regulate mitochondrial motility in RGC dendrites , irrespective of Miro1 expression , and that tonic elevations of Ca2+ upon application of high K+ solution may engage different mechanisms to physiological transients . As the retina matures , waves of spontaneous activity subside and light-evoked inputs begin to drive RGC activity . We wanted to test whether sensory-evoked activity affects the distribution and/or movements of mitochondria in RGC dendrites at maturity . Because mitochondrial movements are rare at P21 , we acquired z-stacks of RGC dendrites in Thy1-mtCFP-P retinas every 15 min on a two-photon microscope . In the intervals between image acquisitions , retinas were either kept in darkness ( P21 – no stim ) or presented a full-field white noise stimulus with alternating contrast levels ( P21 – stim , see 'Materials and methods' ) . MEA recordings confirmed that this stimulus robustly elevates the firing rates of RGCs ( Figure 5A–C ) . In darkness , mitochondria rarely changed position even across two imaging intervals at P21 , whereas approximately half of them were displaced over the same time at P9 ( Figure 5D , E ) . Light stimulation did not alter the stability of mitochondria at P21 and very few changed their position in the imaging intervals ( Figure 5D , E ) . Thus , similar to our observations for spontaneous activity in developing circuits , sensory-evoked activity does not appear to regulate the motility of dendritic mitochondria in the mature retina . 10 . 7554/eLife . 11583 . 013Figure 5 . Sensory-evoked neuronal activity and mitochondrial motility at maturity . ( A , B ) Spike raster plots of eight representative RGCs recorded in darkness ( A ) and during presentation of a full-field white noise stimulus ( B , see 'Materials and methods' ) . ( C ) Bars ( error bars ) indicate the mean ( ± SEM ) firing rates of RGCs ( n = 334 RGCs , 3 retinas , p<10–26 ) . ( D ) Representative RGCs expressing mtCFP at t = 0 min ( top panels ) , t = 30 min ( middle panels ) after being exposed to white noise stimulus ( right panels ) or kept in darkness ( left and center panels ) . Bottom panels show merged images of t = 0 and t = 30 min . ( E ) Bars ( error bars ) indicating the mean ( ± SEM ) of % mitochondria displaced between t = 0 min and t = 30 min ( P9 – no stim . n = 5 RGCs , P21 – no stim . n = 5 RGCs , P21 – stim n = 5 RGCs ) . RGCs , retinal ganglion cells . DOI: http://dx . doi . org/10 . 7554/eLife . 11583 . 01310 . 7554/eLife . 11583 . 014Figure 5—figure supplement 1 . Antimycin A and Oligomycin and mitochondrial motility . ( A ) Kymographs of representative time-lapse imaging series of mitochondria ( mtDsRed , 0 . 9 fps ) in a P9 RGC before ( left ) and after Oligomycin/Antimycin A application . Top panels show still frames at t = 0s of the branch segments depicted in the kymographs in the bottom panels . ( B ) Paired plots of RGC mitochondrial motility before and after 30 min of 10 µM Oligomycin/ 4 µM Antimycin A application at P9 ( n = 5 RGCs , p>0 . 2 ) . ( C ) Representative images of P21 RGCs expressing mtDsRed at t = 0 min ( top panels ) , t = 30 min ( middle panels ) in the presence ( right panels ) or absence ( left panels ) of Oligomycin/Antimycin A . Bottom panels show merged images of t = 0 and t = 30 min . ( D ) Bars ( error bars ) indicating the mean ( ± SEM ) of % mitochondria displaced between t = 0 min and t = 30 min ( n = 4 RGCs , p>0 . 4 ) . fps , frames per second; RGC , retinal ganglion cell . DOI: http://dx . doi . org/10 . 7554/eLife . 11583 . 014 Toxins that depolarize mitochondria and inhibit their ability to synthetize ATP have been reported to increase , decrease or not change mitochondrial motility in axons and dendrites of different neuron types . To test how toxin-mediated stress affects mitochondrial transport in RGC dendrites , we applied a mixture of Antimycin A ( a complex III inhibitor ) and Oligomycin ( an ATP synthase inhibitor ) to P9 and P21 retinas . Fast ( 0 . 9 fps ) and long-interval ( 15 min ) time-lapse imaging revealed no significant changes in the motility of mitochondria in developing or mature RGC dendrites ( Figure 5—figure supplement 1 ) . Spontaneous oscillatory hyperactivity of RGCs is a common feature of retinal degenerations and originates in presynaptic circuits ( Borowska et al . , 2011; Margolis et al . , 2008; Soto and Kerschensteiner , 2015; Soto et al . , 2012; Stasheff , 2008; Yee et al . , 2012 ) . In Crx-/- mice , a model of Leber congenital amaurosis , retinal waves are preserved . But , from P15 on RGCs exhibit rhythmic hyperactivity as a result of enhanced input from BCs ( Soto et al . , 2012 ) . We previously showed that this elevated activity prolongs and enhances BC-RGC synaptogenesis ( Soto et al . , 2012 ) . Given the correlation between synaptic development and mitochondrial motility in wild-type mice ( Figures 2 and 3 ) , we first confirmed RGC activity was elevated in explants from P21 Crx-/- mice ( Figure 6A–C ) , and then examined mitochondrial transport in RGC dendrites at this age . We found that dendritic mitochondria in Crx-/- retinas remained motile at P21 , at a level intermediate to those at P9 and P15 in wild-type mice ( Figure 6D , E and Video 4 ) . Moreover , the movement patterns of motile mitochondria at P21 in Crx-/- mice except for slightly longer pauses were indistinguishable from those in wild-type at P9 and P15 ( Figure 6F–H ) . 10 . 7554/eLife . 11583 . 015Figure 6 . Pathological hyperactivity and dendritic mitochondria in retinal degeneration . ( A , B ) Spike raster plots of eight representative RGCs recorded from P21 WT ( A ) and from P21 Crx-/- retinas ( B ) . ( C ) Bars ( error bars ) indicating the mean ( ± SEM ) firing rates of WT and Crx-/- RGCs ( WT n = 100 RGCs , 3 retinas , Crx-/- n = 235 RGCs , 3 retinas , p<10–8 ) . ( D ) Kymographs of representative time-lapse series of a P21 WT RGC ( left panels ) and Crx-/- RGC ( right panels ) . ( E ) Bars ( error bars ) indicating the mean ( ± SEM ) motile fraction of mitochondria in P21 WT and Crx-/- dendrites ( P21 WT n = 7 RGCs , Crx-/- n = 6 RGCs ) . ( FH ) Bars ( error bars ) indicating the mean ( ± SEM ) mitochondrial speed during uninterrupted motion ( F ) , duration of uninterrupted runs ( G ) , and duration of pauses ( H ) for mitochondria in P9 – 15 WT and Crx-/- dendrites ( Crx-/- n = 30 mitochondria , pooled WT P9 and P15 n = 97 mitochondria ) . RGC , retinal ganglion cell . DOI: http://dx . doi . org/10 . 7554/eLife . 11583 . 015
The importance of mitochondrial dynamics and distribution to neuronal function is highlighted by the fact that mutations that alter the fission-fusion balance and transport of mitochondria cause neurodegenerative diseases ( Chan , 2006 ) . Among the neurons affected by such diseases are RGCs , which degenerate in autosomal dominant optic atrophy ( ADOA ) , the most common hereditary optic neuropathy ( Olichon et al . , 2006; Yu-Wai-Man et al . , 2011 ) . A majority of ADOA patients harbors mutations in the OPA1 gene , which encodes a GTPase involved in mitochondrial fusion ( Olichon et al . , 2006; Yu-Wai-Man et al . , 2011 ) . Changes in fission-fusion balance alter the transport and distribution of mitochondria in neurons ( Chan , 2006 ) . In axons , mitochondrial transport and distribution have been explored in detail , including recent in vivo imaging studies ( Breckwoldt et al . , 2014; Misgeld et al . , 2007; Plucinska et al . , 2012; Takihara et al . , 2015 ) , and the contributions of mitochondria to axonal branching and synaptic transmission are relatively well understood ( Courchet et al . , 2013; Medler and Gleason , 2002; Spillane et al . , 2013; Verstreken et al . , 2005; Werth and Thayer , 1994 ) . By comparison , dendritic mitochondria have been studied less , and many aspects of their transport , distribution , and function , particularly in intact circuits , remain obscure . We find that during development mitochondria enter RGC dendrites to near-mature levels before branching patterns are established ( Figure 2 ) . Similar observations were recently made for dendrites of cerebellar Purkinje cells ( Fukumitsu et al . , 2015 ) , and together with experiments that block export of mitochondria into dendrites support the notion that local mitochondria are required for the formation and/or maintenance of dendritic branches ( Fukumitsu et al . , 2015; Ishihara et al . , 2009 ) . We find that during development , dendritic mitochondria gradually become enriched ( >15-fold ) near branch points ( Figure 2 ) . These mitochondria likely provide energy for lipid and protein biosynthesis , and secretory trafficking in the complex ER structures and Golgi outposts found at branch points ( Cui-Wang et al . , 2012; Horton et al . , 2005; Ye et al . , 2007 ) . Consistent with this idea , ER structures and Golgi outposts at branch points are required for the formation and maintenance of dendritic branches ( Cui-Wang et al . , 2012; Horton et al . , 2005; Ye et al . , 2007 ) , and the branching phenotypes of neurons lacking dendritic mitochondria can be rescued by exogenous supply of ATP-phosphocreatine ( Fukumitsu et al . , 2015; Li et al . , 2004 ) . Mitochondria in RGC dendrites localize to excitatory synapses ( Figure 2 ) . Similar observations have been made on cultured neurons , although the extent of co-localization varied between studies ( Chang et al . , 2006; Li et al . , 2004 ) . Local mitochondria are thought to participate in the formation of synapses and their plasticity ( Ishihara et al . , 2009; Li et al . , 2010; Li et al . , 2004 ) . Experiments in cultured hippocampal neurons suggested that enhancing the dendritic mitochondria content is sufficient to increase the number of synapses ( Li et al . , 2004 ) . As mitochondrial density in RGCs dendrites reaches near-mature levels before most synapses are formed , it seems unlikely that mitochondria limit the rate of synaptogenesis during development of this circuit . Mitochondrial transport in RGC dendrites is well described by a state diagram recently applied to axonal transport , in which mitochondria exist in a motile or a stationary state ( Obashi and Okabe , 2013 ) . Motile mitochondria alternate between short runs and short pauses ( Figure 3 ) and frequently switch the direction of their movements ( data not shown ) . By contrast , stationary mitochondria remain in place for long periods of time . Unlike transitions between runs and pauses in the motile state , transitions between motile and stationary state are rare ( Obashi and Okabe , 2013 ) . Our results show that as circuits mature , dendritic mitochondria undergo a nearly complete shift to the stationary state . Thus , whereas approximately 30% of mitochondria move at any time in P9 dendrites , we observed no mitochondrial movements at P21 during fast time-lapse imaging ( 0 . 9 fps , Figure 3 ) and displacements were rare even during long imaging intervals ( 15–30 min , Figure 5 ) . These rare displacements may be associated with mitochondrial fission or support mitochondrial fusion or mitophagy ( Chen and Chan , 2009; Maday and Holzbaur , 2014 ) . Disruption of mitochondrial fusion in a mouse model of ADOA ( Opa1+/- mice ) results in shortened mitochondria , which accumulate in proximal RGC dendrites ( Williams et al . , 2012 ) . As Opa1+/- mice age ( >1 yr ) , the number of excitatory synapses and branches of RGC dendrites declines , highlighting the importance of mitochondrial fusion for dendritic and synaptic integrity ( Williams et al . , 2010; Williams et al . , 2012 ) . In isolated cortical neurons , a modest decline in mitochondrial motility was observed with increasing time in culture ( Chang and Reynolds , 2006 ) . In the intact retina , we discover a drastic developmental shift ( motile to stationary ) in dendritic mitochondria . Importantly , this shift is cargo-specific , as peroxisomes keep moving in dendrites of mature RGCs ( Figure 3 ) , and compartment-specific , as mitochondria remain motile in RGC axons ( data not shown ) ( Takihara et al . , 2015 ) . Together with results on dendritic localization ( Figure 2 ) , these observations reveal that mitochondria form stable associations with synapses and branch points in RGC dendrites , and suggest that the prevailing view of mitochondria as highly motile organelles applies to developing but not to mature dendrites . Using simultaneous two-photon imaging of Ca2+ and mitochondria in the intact retina , we find that neither patterned spontaneous activity during development ( Figure 4 ) nor sensory-evoked activity at maturity ( Figure 5 ) regulate the motility of mitochondria in RGC dendrites . These results are in contrast to studies of cultured neurons , which suggest that neuronal activity controls motility of dendritic mitochondria via the adaptor protein Miro1 ( Li et al . , 2004; MacAskill et al . , 2009; Rintoul et al . , 2003 ) . We confirmed that Miro1 is expressed in the developing and mature retina and showed that overexpression of wild-type Miro1 or a mutant unable to bind Ca2+ ( Fransson et al . , 2006 ) do not affect mitochondrial motility in RGC dendrites . Thus , it seems unlikely that differences in Miro1 expression or function account for the discrepant results on the influence of activity . Instead , we suggest that differences between activity patterns physiologically occurring in intact circuits and artificial stimuli used to elicit effects on mitochondrial motility in culture may be a contributing factor . In support of this notion , we find that elevation of external K+ reduces mitochondrial motility in RGC dendrites ( Figure 4 ) as it does in cultured neurons ( Li et al . , 2004 ) , but elicits tonic rises in intracellular Ca2+ not observed during physiologic activity . In a mouse models of retinal degeneration , circuits in the inner retina exhibit spontaneous hyperactivity ( Borowska et al . , 2011; Margolis et al . , 2008; Soto and Kerschensteiner , 2015; Soto et al . , 2012; Stasheff , 2008; Yee et al . , 2012 ) . In Crx-/- mice , we previously showed that this hyperactivity enhances and prolongs synaptogenesis between BCs and RGCs ( Soto et al . , 2012 ) . Here , we find that extended synaptogenesis is matched by an increase in mitochondrial motility at P21 in Crx-/- mice . Together with the parallel decline in synapse turnover and mitochondrial motility across normal development ( Figure 3 ) ( Kerschensteiner et al . , 2009; Morgan et al . , 2008 ) , this suggests that the two processes are linked . In summary , we discover a cargo- and compartment-specific developmental shift in the transport of mitochondria , which during a critical period of high motility localize to synapses and branch points in RGC dendrites and subsequently maintain stable positions . In addition , using simultaneous two-photon imaging of Ca2+ signals and mitochondrial transport , we find that neither spontaneous nor sensory-evoked activity patterns regulate the motility of dendritic mitochondria in the intact retina . Together these results suggest important amendments to our understanding of dendritic mitochondria .
All animals were handled according to a protocol ( # 20140095 ) approved by the Animal Studies Committee of Washington University School of Medicine and performed in compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals . Thy1-mtCFP-P ( Misgeld et al . , 2007 ) and Crx-/- ( Furukawa et al . , 1999 ) mice were backcrossed to C57BL/6J for more than five generations . Eyes were removed from mice deeply anesthetized with CO2 . Retinas were dissected from eye cups and prepared as flat mounts on permeable filter paper ( Millipore , Sigma-Aldrich , Saint Louis , MO ) . Tissue was prepared in ice-cold mouse Artificial Cerebrospinal Fluid ( mACSF ) buffered with HEPES ( biolistic experiments , concentrations in mM: 119 NaCl , 2 . 5 KCl , 2 . 5 CaCl2 , 1 . 3 MgCl2 , 1 NaH2PO4 , 11 Glucose , and 20 HEPES - pH adjusted to 7 . 35–7 . 40 with NaOH ) or sodium bicarbonate ( light stimulation and MEA , concentrations in mM: 125 NaCl , 2 . 5 KCl , 1 MgCl2 , 1 . 25 NaH2PO4 , 2 CaCl2 , 11 Glucose , and 26 NaHCO3 ) . Retinas used for biolistic transfection were incubated at 33°C in an O2-perfused chamber overnight . Retinas used for light stimulus and MEA experiments were dissected under infrared illumination ( excitation wavelength >900 nm ) and incubated at 33°C in a light-tight chamber for approximately 2 hr before experiments . Live retinal tissue was constantly perfused at 1 ml/min with 33°C mACSF bubbled with O2 ( confocal experiments ) or 95% O2 / 5% CO2 ( two-photon experiments ) during light stimulation and imaging . In pharmacologic experiments , 5 µM DhβE ( Tocris Bioscience ) or 4 µM Antimycin A ( Enzo Life Sciences ) and 10 µM Oligomycin ( Sigma-Aldrich ) were added to mACSF . To raise extracellular K+ concentrations to 30 mM , KCl was substituted for NaCl in mACSF ( concentrations in mM: 97 . 5 NaCl , 30 KCl , 1 MgCl2 , 1 . 25 NaH2PO4 , 2 CaCl2 , 11 Glucose , and 26 NaHCO3 ) . A gene gun was used to transfect RGCs in the ex vivo retinal flat-mount preparations as described previously ( Kerschensteiner et al . , 2009; Morgan et al . , 2008 ) . Briefly , plasmids encoding fluorescent proteins were precipitated onto 1 . 6 µm gold particles ( Bio-Rad ) and gold particles loaded into Tefzel tubing ( Bio-Rad ) ( Morgan and Kerschensteiner , 2012 ) . Helium pressure at 40 psi was used to deliver gold particles to flat-mounted retinal tissue ( Morgan and Kerschensteiner , 2011 ) . Multi-electrode array ( MEA ) recordings of RGC action potentials were acquired as described previously ( Pearson and Kerschensteiner , 2015 ) . Briefly , rectangular pieces of isolated dorsal retinal tissue were mounted onto planar arrays of 252 electrodes ( MultiChannelSystems ) . Spike waveforms recorded at each electrode were used to sort activity into trains of RGC action potentials using principal component analysis ( Offline Sorter; Plexon ) . Correlation coefficients normalized for nonstationary firing rates were calculated as described previously ( Kerschensteiner and Wong , 2008 ) . Fixed and live retinas were imaged on Olympus Fv1000 laser scanning confocal and two-photon microscopes using 60x 1 . 35 NA oil-immersion ( fixed retinas ) , 60x 1 . 1 NA water-immersion , and 20x 0 . 95 NA water-immersion objectives ( live retinas ) . Images were processed and analyzed using ImageJ ( NIH ) , Amira ( FEI ) , and software written in MATLAB ( MathWorks ) . To analyze mitochondrial and synaptic distributions , retinas were fixed in 4% paraformaldehyde for 30 min at RT . Synapses were identified and dendrites skeletonized using custom MATLAB software described previously ( Kerschensteiner et al . , 2009; Morgan et al . , 2008 ) . Briefly , the cytosolic tdTomato signal was used to create binary masks ( Amira ) of RGC dendrites . Dendrites were then skeletonized into 1 µm-long linked segments based on these masks . Synapses were identified by iterative thresholding of PSD95-CFP signal within the dendritic mask ( Kerschensteiner et al . , 2009; Morgan et al . , 2008 ) . Voxels belonging to mitochondria were similarly identified by local thresholding algorithms on mtYFP signal within the dendritic mask . The density of mitochondria in dendrites was expressed as a volumetric fraction based on the number of voxels assigned to mitochondria and that contained in the dendritic mask . To determine quantitatively whether mitochondria localize to synapses , the observed average nearest neighbor distance ( synapse to mitochondrion ) for a given RGC was compared to the distribution of average nearest distances obtained in Monte Carlo simulations in which the position synapses of the same cell was randomized along its dendrites . For all cells , the mitochondria were closer to real synapses than their simulated counterparts . Time-lapse recordings of mitochondrial and peroxisomal motility , and Ca2+ waves were acquired at 0 . 9 frames per second ( fps ) . Motile fractions were calculated as ( motile organelles ) / ( motile organelles + stationary organelles ) . We refer to the motile fraction as the fraction moving during 1 min of observation , whereas instantaneous motile fraction is used to denote the fraction moving between consecutive frames in 0 . 9 fps imaging series . Kymographs were generated using the ImageJ Multiple Kymograph plugin . Mitochondrial speeds and peroxisomal speeds and mitochondrial run and pause durations were calculated using custom MATLAB software and organelle positions tracked manually using the ImageJ MTrack2 plugin . Mitochondrial displacement during light stimulation of P21 retinas was computed using custom MATLAB software . mtCFP signal was masked in Amira to include only mitochondria in contiguous dendritic branches . mtCFP intensity was binarized and MATLAB software was used to identify regions of connected pixels as individual mitochondria . Images at t = 30 min were subtracted from images at t = 0 min; mitochondria identified at 0 min that did not have at least 20% overlap with a mitochondrion at 30 min was counted as a displaced mitochondrion . Percent displacement was calculated as 100 x ( displaced mitochondria ) / ( total mitochondria ) . Stimulated retinas were shown light stimulus patterns of alternating high- and low-contrast ( mean intensity 5000 R*/rod/s ) periods lasting 60 s each for a total of 15 min . Image stacks of mtCFP signal throughout RGC dendrites were collected between light stimulations . Unstimulated retinas were kept in darkness for 15 min between image stacks . | Inside the cells of animals and plants , compartments called mitochondria play several important roles including supplying chemical energy for cellular processes . The mitochondria inside nerve cells are produced in the main body of each cell , and must travel down a long nerve fiber called the axon and the branch-like extensions called dendrites to reach the sites where they are most needed . As the nerve cells form , dendritic branches grow and retract as the connections between different nerve cells – known as synapses – form and disappear . Later on , the dendrites and synapses become more stable , but it is not clear if the amount that the mitochondria move also changes . Faits et al . used microscopy to study the movement of mitochondria in the developing dendrites of ganglion cells in the eyes of mice . The experiments show that early on in the development of nerve cells , the mitochondria are very mobile . However , as the synapses become more stable later on , the mitochondria become almost motionless . The movement of another type of cell compartment to the dendrites is unaffected , which suggests that this decline in movement is specific to mitochondria . Next , Faits et al . studied mutant mice that suffer from degeneration of part of the eye called the retina . These mice have ganglion cells that display higher levels of spontaneous activity than normal and their synapses continue to form and disappear later in development . The experiments show that the mitochondria in the ganglion cells remain mobile in the adult mutant mice . Faits et al . ’s findings challenge the prevailing views of mitochondria in dendrites , and suggest that mitochondria reach stable positions during a critical period in the development of the retina . Further studies should reveal how the decline in the movement of mitochondria is regulated , which may help us to understand how differences in the movement of mitochondria can lead to the degeneration of nerve cells in some human diseases , such as dominant optic atrophy . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2016 | Dendritic mitochondria reach stable positions during circuit development |
Bacteria , bacteriophages that prey upon them , and mobile genetic elements ( MGEs ) compete in dynamic environments , evolving strategies to sense the milieu . The first discovered environmental sensing by phages , lysis inhibition , has only been characterized and studied in the limited context of T-even coliphages . Here , we discover lysis inhibition in the etiological agent of the diarrheal disease cholera , Vibrio cholerae , infected by ICP1 , a phage ubiquitous in clinical samples . This work identifies the ICP1-encoded holin , teaA , and antiholin , arrA , that mediate lysis inhibition . Further , we show that an MGE , the defensive phage satellite PLE , collapses lysis inhibition . Through lysis inhibition disruption a conserved PLE protein , LidI , is sufficient to limit the phage produced from infection , bottlenecking ICP1 . These studies link a novel incarnation of the classic lysis inhibition phenomenon with conserved defensive function of a phage satellite in a disease context , highlighting the importance of lysis timing during infection and parasitization .
Following the discovery of bacteriophages ( D’Herelle , 1917; Twort , 1915 ) , Escherichia coli’s T1 through T7 phages were widely accepted as model systems ( Keen , 2015 ) and T-even phages were used to determine mutation manifestation at the molecular level , gene topology , and that fact that nucleic acids are decoded in triplets ( Benzer , 1961; Crick et al . , 1961 ) . Early geneticists observed an interesting phenotype in rapid lysing T-even mutants , called r mutants , which produce plaques with clear edges while plaques of wild type ( WT; all acronyms are expanded in Table 1 ) T-even phages have fuzzy edges ( Hershey , 1946; Paddison et al . , 1998 ) . Edge fuzziness is the consequence of inhibited cell lysis triggered by the adsorption of additional phage after initial infection . This ‘superinfection’ also stabilizes infected cells as measured by optical density ( Doermann , 1948 ) . The phenomenon , termed lysis inhibition ( LIN ) , is significant for two reasons ( Figure 1A and B ) : it allows for prolonged production of progeny phage resulting in larger phage bursts , and it protects progeny phage from adsorbing to infected cells , which are not productive hosts for secondarily adsorbed phages ( Abedon , 1990; Abedon , 2019; Doermann , 1948 ) . Consequently , LIN is considered an important adaptation in environments where host bacteria are scarce but free virions are plentiful ( Abedon , 1990; Abedon , 2019 ) . Despite being discovered over half a century ago , LIN has only been well characterized in T-even coliphages where it is mediated by holins and antiholins ( Chen and Young , 2016; Paddison et al . , 1998; Ramanculov and Young , 2001 ) . Holins are the first step in canonical holin-endolysin-spanin lysis systems in Gram-negative bacteria ( Cahill and Young , 2019; Young , 2014 ) . Holins accumulate in the inner membrane until they trigger , making holes that enable endolysin digestion of the peptidoglycan after which spanins fuse the inner and outer membranes to complete cell lysis . During lysis inhibition , antiholins inhibit holin triggering thereby stopping the progression towards lysis ( Chen and Young , 2016; Paddison et al . , 1998; Ramanculov and Young , 2001 ) . Phages are abundant in natural environments including marine ecosystems ( Middelboe and Brussaard , 2017 ) and gut microbiomes ( Shkoporov and Hill , 2019 ) , potentially making LIN a pertinent state for phages infecting many different bacteria , including Vibrio cholerae . V . cholerae poses a substantial global health burden as the causative agent of the diarrheal disease cholera ( Ali et al . , 2015 ) . In both aquatic reservoirs and stool samples from cholera patients , V . cholerae co-occurs with predatory phages . Several studies have implicated V . cholerae phages in playing a role in modulating cholera outbreaks ( D'Herelle and Malone , 1927; Faruque et al . , 2005a; Faruque et al . , 2005b; Jensen et al . , 2006 ) leading to proposals of phage cocktails as prophylactics to curb cholera transmission ( Yen et al . , 2017 ) . The predominant phage in cholera patient samples is ICP1 , a lytic myovirus ( Seed et al . , 2011 ) that is locked in a dynamic arms race with no clear winner as both V . cholerae and ICP1 continue to be isolated from patients in the cholera endemic region of Bangladesh ( Angermeyer et al . , 2018; McKitterick et al . , 2019a; McKitterick et al . , 2019a; Seed et al . , 2011 ) . Added into the evolutionary fray is a parasitic phage satellite called PLE ( phage-inducible chromosomal island-like element ) found integrated in the chromosome of clinical V . cholerae isolates ( McKitterick et al . , 2019b; O'Hara et al . , 2017; Seed et al . , 2013 ) . Previous analysis of isolates dating back to 1949 revealed a succession of five distinct PLEs ( PLE 1 through PLE 5 of which PLE 1 is the most recently circulating PLE ) in V . cholerae which possess shared genomic architecture . Each PLE provides V . cholerae with a clear fitness benefit in the defense against ICP1 as PLE abolishes phage production ( O'Hara et al . , 2017 ) . Upon infection , PLE excises from the bacterial chromosome , harnessing an ICP1-encoded protein as the trigger ( McKitterick and Seed , 2018 ) , replicates using both PLE and ICP1-encoded products ( Barth et al . , 2020; McKitterick et al . , 2019a ) , and accelerates cell lysis after forming particles hypothesized to be made by hijacking ICP1 structural components to transduce the PLE genome to naïve recipient cells ( O'Hara et al . , 2017 ) . As no infectious ICP1 are produced , PLE defends populations of V . cholerae from ICP1 attack , functioning as an abortive infection system . In the face of this anti-phage element , ICP1 acquired a Type I-F CRISPR-Cas system to target PLE in a sequence-specific manner , restoring ICP1 progeny phage production and overcoming PLE ( McKitterick et al . , 2019b; Seed et al . , 2013 ) . While the full extent of ICP1 , PLE , and V . cholerae interactions are unknown and continuing to evolve , efforts to understand how PLEs restrict ICP1 have yet to identify any single PLE open reading frame necessary for inhibition ( via testing of PLE-encoded repA , which is necessary for PLE replication [Barth et al . , 2020] and int , which is required for PLE excision [McKitterick and Seed , 2018] ) . From the only characterized processes , namely excision and replication , it is clear that PLE requires phage-encoded products , consequently depending on ICP1 for horizontal transmission; however , uncharacteristic of other phage satellites like the well characterized Staphylococcus aureus pathogenicity islands ( SaPIs ) which decrease but do not eliminate the production of progeny virions , PLE completely abolishes ICP1 production – a balancing act that likely requires the exploitation of select products at exact times during the 20 minutes before PLE-mediated accelerated cell lysis . The accelerated cell lysis program in V . cholerae harboring PLE led us to investigate the prolonged infection of ICP1 in strains without PLE where we discovered that ICP1 exhibits lysis inhibition . In this work we report the first mechanistic characterization of archetypal LIN outside of E . coli and we reveal ICP1 LIN mechanisms in V . cholerae by identifying previously uncharacterized ICP1 genes with holin and antiholin activity , termed teaA and arrA respectively . Subsequently , we discovered a single PLE-encoded gene we call lidI for lysis inhibition disruption that is sufficient to collapse ICP1-mediated lysis inhibition . All PLEs encode LidI , highlighting a conserved strategy PLEs may use to antagonize an aspect of the phage lifecycle not previously known to be targeted by parasitic satellites . While we cannot be sure of LidI function in the context of the PLE , it alone is sufficient to decrease the yield of ICP1 from infected V . cholerae and impose an evolutionary bottleneck on phage populations .
After infection at low multiplicity of infection ( MOI; MOI = 0 . 1 ) , ICP1 completes virion production within 20–25 minutes in PLE ( - ) V . cholerae ( O'Hara et al . , 2017 ) . This timeframe is markedly abbreviated with respect to the 90 minutes that pass before visible lysis of the same host infected at a high MOI ( MOI = 5 ) ( O'Hara et al . , 2017 ) . This incongruity prompted us to test ICP1 infections at intermediate multiplicities of infection . When infecting PLE ( - ) V . cholerae with ICP1 at MOI = 1 , we observed an early lysis event 20 minutes post-infection after which the optical density of the culture stabilized ( Figure 1C ) . Such lysis kinetics are consistent with canonical LIN by T-even coliphages wherein a portion of infected cells release progeny phages triggering LIN in the remaining population ( Doermann , 1948 ) . These similarities led us to hypothesize that ICP1 exhibits LIN in V . cholerae . During canonical LIN , superinfection , the secondary adsorption of phages after initial phage infection , stabilizes the optical density of infected E . coli cultures because cells stay intact instead of lysing ( Figure 1B ) . To determine whether superinfection by ICP1 of PLE ( - ) V . cholerae shares this characteristic , we infected cultures with ICP1 ( MOI = 1 ) , let phage adsorb for four minutes , then superinfected the culture with ICP1 ( multiplicity of superinfection; MOSI = 5 ) . As expected of a phage exhibiting LIN , the culture optical density was stabilized , eliminating the early lysis event ( Figure 1C ) . Previously characterized lysis inhibition in E . coli is sensitive to the membrane proton motive force and is disrupted by the addition of energy poisons . One such poison , the ionophore 2 , 4-dinitrophenol ( DNP ) , collapses the proton motive force and subsequently disrupts LIN by T-even phages , causing rapid lysis of infected E . coli ( Abedon , 1992; Heagy , 1950 ) . To test if ICP1 LIN is similarly linked to proton motive force , we exposed PLE ( - ) V . cholerae infected at a high MOI ( MOI = 5 ) to 2 , 4-dinitrophenol ( Figure 1D ) and observed the expected crash in optical density , further supporting the conclusion that ICP1 exhibits LIN in V . cholerae . Although genomes of ICP1 isolates from cholera patient stool are abundant , the genes involved in ICP1-mediated lysis have not been identified . To characterize the mechanism underlying ICP1 lysis inhibition , we endeavored to find ICP1’s holin and antiholin ( Figure 2A ) . Holins are diverse proteins , however , they all include at least one transmembrane domain ( Wang et al . , 2000 ) . Further , we hypothesized that the holin would be conserved in isolates over time because lysis timing is key to phage fitness . As a result we narrowed our search to gene products containing a predicted transmembrane domain that were conserved in previously analyzed ICP1 isolates ( Angermeyer et al . , 2018 ) leaving us with three candidate gene products: Gp137-Gp139 ( Figure 2B; further described in Supplementary file 1 ) . Analysis of these proteins using remote homology and synteny via Phagonaute identified the DUF3154 Pfam ( reclassified as GTA_holin_3TM ( PF11351 ) ; Delattre et al . , 2016; El-Gebali et al . , 2019; Sonnhammer et al . , 1997 ) in ICP1 Gp137 , which we have since named TeaA . In E . coli , canonical holins including T4’s T-holin , accumulate in the membrane until they collapse the proton motive force and form pores or are triggered by an energy poison . Premature holin triggering results in loss of viability and commits the cell to eventual lysis , decreasing the optical density ( Garrett and Young , 1982; Josslin , 1971 ) . To experimentally test TeaA for holin activity , we exogenously expressed teaA in PLE ( - ) V . cholerae in the absence of phage and probed its ability to collapse the proton motive force and be triggered by an energy poison . We measured proton motive force using 3 , 3’-diethloxacarbocyanine iodide ( DiOC2 ( 3 ) ) , a fluorescent green membrane stain that forms red fluorescent aggregates in the presence of intact proton motive force ( Kirchhoff and Cypionka , 2017; Novo et al . , 1999 ) . Upon induction , T-holinT4 and TeaA decreased the red fluorescence , consistent with holin activity , while the empty vector ( EV ) did not ( Figure 2C ) . Holin activity by TeaA was further demonstrated by the rapid decrease in optical density after 2 , 4-dinitrophenol addition which was comparable to the decrease in optical density observed with T-holinT4 expressing V . cholerae after 2 , 4-dinitrophenol treatment ( Figure 2D ) . Interestingly , though endolysin activity is described as necessary in holin-endolysin-spanin systems ( Young , 2014 ) , the holin TT4 or TeaA alone is enough to lyse V . cholerae under laboratory conditions . The similarities between TeaA activity and T4’s T-holin motivated the naming of teaA for the gene’s T-holin-esque activity . Next , we sought to identify an antiholin in ICP1 . Antiholins can interact with holins within the inner membrane ( Moussa et al . , 2012; Young , 2013 ) , periplasm ( Tran et al . , 2005 ) , or cytoplasm ( Chen and Young , 2016 ) . Because antiholins interacting directly with holins within the membrane contain transmembrane domains and previously characterized periplasmic antiholins contain transmembrane domains for tethering to the membrane and subsequent release into the periplasm ( Tran et al . , 2007 ) , we continued to focus on the transmembrane domain-containing proteins in ICP1 . Given that TeaA is conserved in ICP1 isolates , and lysis timing , which is fine-tuned by antiholins , is critical , we expected antiholins would also be conserved . Antiholins are often found near holin genes , in many cases utilizing an alternative start site within the holin sequence or occupying an overlapping open reading frame ( Bläsi and Young , 1996; Graschopf and Bläsi , 1999 ) . With these data in mind , we began to investigate Gp138 , which we subsequently named ArrA , as a potential antiholin ( Figure 2B ) . The T4 antiholins , RIT4 and RIIIT4 were initially identified when mutations in these genes demonstrated a rapid lysis plaque morphology . Phages lacking functional antiholins form plaques with sharply defined edges because lysis inhibition no longer occurs . In an effort to find similar rapid-lysing ICP1 mutants with antiholin modifications , we challenged ICP1 with CRISPR-Cas ( + ) V . cholerae containing a spacer targeting arrA . We observed a mixture of edge phenotypes including clear-edged plaques ( Figure 3A ) and recovered the fuzzy-edged phenotype by expressing ArrA in trans ( Figure 3B ) . We engineered a clean ΔarrA ICP1 strain to further confirm ArrA function . Of note , and consistent with arrA acting as an antiholin , we successfully constructed an arrA knockout demonstrating that , despite its conservation , arrA is not an essential gene . In support of ArrA acting as an antiholin whose absence results in rapid lysis , ΔarrA ICP1 forms plaques that have clear edges ( Figure 3C ) . T4 antiholin mutants demonstrate accelerated lysis kinetics in liquid culture ( Chen and Young , 2016; Paddison et al . , 1998 ) , motivating us to test ICP1 ΔarrA in liquid cultures . Attempts to obtain high titer stocks of ΔarrA ICP1 were unsuccessful ( consistent with decreased phage yields ) hindering tests of infection at high multiplicities of infection . Instead , we infected PLE ( - ) V . cholerae with ΔarrA ICP1 , waited for approximately two cycles of infection to complete , and then , once lysis began , we observed a rapid crash in optical density ( Figure 3D ) . Such kinetics are in stark contrast to the more prolonged decline seen in wild type infections exhibiting LIN ( Figure 1C ) . To be sure this was due to ArrA , we supplied ArrA in trans and recovered lysis kinetics characteristic of LIN in which the optical density was stabilized for 30 additional minutes before the culture cleared ( Figure 3D ) . To further test that this stabilization of optical density by ArrA when supplied in trans was accomplished through LIN , we exposed cultures to 2 , 4-dinitrophenol disrupting the proton motive force and collapsing the culture ( Figure 3D ) . These data support the conclusion that ArrA is an ICP1-encoded antiholin that helps regulate lysis timing . BLASTP analysis of TeaA revealed homologs present throughout marine phage and bacterial genomes ( Figure 3—figure supplement 1 and Supplementary file 2 ) . ArrA yielded fewer homologs than TeaA ( Supplementary file 3 ) , however , using less stringent search parameters , we found that some organisms containing TeaA homologs also contain ArrA homologs , though these were limited to vibriophages ( Figure 3E and Supplementary file 2 and 3 ) . This suggests that there are potential homologous LIN systems - complete with both holin and antiholin - present in phages other than ICP1 . In contrast , the presence of ArrA homologs without TeaA homologs raises the question: what are antiholins doing on their own ? Perhaps they have evolved functionality with holins divergent enough to no longer be considered homologous to TeaA under our search parameters , or they have been coopted for divergent functions much like holins ( Mehner-Breitfeld et al . , 2018; Saier and Reddy , 2015 ) . Thus far , characterization of all ICP1 LIN was done in the absence of PLE , a parasitic phage satellite of ICP1 . Although the mechanisms that PLE deploys to inhibit and hijack ICP1 are not completely understood , V . cholerae lysis kinetics during high multiplicity infections vary depending on the presence of PLE . Consistent with previous experiments at high MOI ( MOI = 5 ) ( O'Hara et al . , 2017 ) , ICP1 infection of PLE ( - ) V . cholerae gradually lyses cultures reaching the lowest optical density ~90 minutes after infection . In contrast , upon infection of PLE 1 V . cholerae , rapid lysis starts 20 minutes post-infection ( Figure 4A ) . This accelerated timescale could result from any combination of processes such as PLE deploying its own lysis machinery , PLE modulating the expression or stability of ICP1’s lysis machinery , or PLE inhibiting or collapsing LIN . To investigate the underpinnings of accelerated lysis in the presence of PLE , we scrutinized the PLE for potential lysis machinery . Initial analysis revealed no transmembrane domains in any of the ~25 predicted open reading frames in each of PLEs 1 , 2 , and 3 . However , the earliest known PLEs , ( PLEs 4 and 5 ) contain two ORFs with predicted transmembrane domains: ORF2 and ORF26 . ORF2PLE 4/5 does not have homologs in PLEs 1 or 2 , suggesting it is not a conserved player mediating accelerated lysis . Consequently , we focused on ORF26PLE 4/5 . Although no homologs were immediately obvious , the synteny between PLEs suggested the presence of previously unannotated open reading frames in PLEs 1 , 2 , and 3 ( namely ORF20 . 1PLE 1 , ORF24 . 1PLE 2 and ORF24 . 1PLE 3 ) , which are homologous to ORF26PLE 4/5 and contain transmembrane domains . We subsequently named these genes lidI and the homologs cluster into two groups: lidIPLE 1 and lidIPLE 2 encode for a 66 amino acid long protein , while lidIPLE 3 , lidIPLE 4 and lidIPLE 5 encode larger proteins at 121 amino acids ( Figure 4—figure supplement 1 ) . These ORFs have no significant homology to other genes or predicted functional domains beyond their shared transmembrane domains . Next , to confirm the expression of the newly discovered lidI genes , we endogenously tagged LidIPLE 1 and evaluated expression during ICP1 infection ( Figure 4B and Figure 4—figure supplement 2 ) . We could not visualize FLAG-LidIPLE 1 in the absence of phage infection; however , when infected by ICP1 at MOI = 2 , FLAG-LidIPLE 1 was detectable by Western blot late in infection – 18 to 20 minutes post phage addition and immediately prior to the sudden decrease in OD characteristic of PLE-mediated accelerated lysis ( Figure 4B and Figure 4—figure supplement 2 ) . After confirming lidIPLE 1 expression during ICP1 infection , we next sought to characterize LidIPLE 1 function in PLE ( - ) V . cholerae . During infection with ICP1 , LidIPLE 1 was sufficient to recapitulate the PLE-mediated accelerated lysis phenotype ( Figure 4A ) . This phenotype is consistent with lidIPLE 1 encoding a holin; however , expression of LidIPLE 1 at the same level of induction in the absence of phage did not alter cellular proton motive force or make cells susceptible to 2 , 4-dinitrophenol induced lysis ( Figure 4C and D ) . These data suggest that LidIPLE 1 does not act as a canonical holin when expressed alone and that its ability to mediate cell lysis is dependent on the presence of ICP1 . Having demonstrated that LidIPLE 1 recapitulates PLE-mediated accelerated lysis , we wanted to determine if it was also necessary for this phenotype . Interestingly , however , when we deleted lidIPLE 1 from PLE 1 V . cholerae the lysis kinetics were unchanged ( Figure 4A ) . To be sure the accelerated lysis is not the consequence of PLE inhibiting the production of phage that are necessary to superinfect cells , exogenous phage was added to PLE 1 and PLE 1 ΔlidI strains both of which still demonstrated accelerated lysis ( Figure 4—figure supplement 3A ) . As these findings were at odds with the conservation of lidI homologs in all the known PLEs , we tested a representative of the other cluster of homologs , LidIPLE 4 , for conserved function . Indeed , LidIPLE 4 is sufficient to cause accelerated lysis in the absence of PLE , however again we found that it is not necessary – PLE 4 ΔlidI V . cholerae strains still exhibit accelerated lysis ( Figure 4E ) . To further investigate the role of individual genes in lysis timing , PLE 1 mutants containing a single knockout of each individual open reading frame were exposed to phage and each demonstrated accelerated lysis ( Figure 4—figure supplement 3B ) . To identify other PLE-encoded factors sufficient to recapitulate PLE-mediated accelerated lysis , we expressed each PLE 1 ORF individually and challenged those strains with phage . This screen did not identify any other single genes sufficient to accelerate lysis , however , it is important to note that cryptic genes could be responsible , other accelerated lysis systems could require multiple ORFs to function , or the timing and expression level of genes expressed outside the context of PLE could obscure gene function . Collectively , these results suggest that PLE mediated accelerated cell lysis is the consequence of the activity of two or more functionally redundant gene products , of which LidI is the only product sufficient to phenocopy the PLE-encoded phenotype . Although redundancy is perhaps not expected for mobile genetic elements ( MGEs ) with restricted genome size , we have additionally observed that no single PLE open reading frame is necessary for inhibition of ICP1 plaque formation ( Figure 4—figure supplement 3C ) , suggesting that multiple strategies act synergistically to eliminate phage production in addition to accelerating lysis . As LidIPLE 1 is sufficient to accelerate lysis but does not phenocopy what we expect of a holin , we wanted to determine the mechanism of LidIPLE 1-mediated accelerated lysis in the absence of PLE . Since ICP1 exhibits LIN during high MOI infections , we hypothesized that LidIPLE 1 causes accelerated lysis by disrupting LIN . LIN occurs when phages outnumber hosts , so changing the multiplicity of infection changes the onset of LIN: at low multiplicities of infection , a small fraction of cells produce a burst of phage which are adsorbed by neighbors and this process can repeat a number of times until the majority of cells are infected and LIN is triggered ( Figure 1A ) . This is evidenced by the stabilization of culture optical density until complete lysis ~90 minutes in PLE ( - ) V . cholerae infected at various multiplicities of infection ( Figure 5A ) . The differential onset of LIN means that accelerated lysis mediated by LidIPLE 1 would also be expected to change in accordance with MOI if it functions by disrupting LIN . Indeed , we observed differential lysis timing dependent on the MOI in cultures expressing LidIPLE 1 with up to a 40 minutes delay at the lowest MOI ( Figure 5A ) , suggesting that LidIPLE 1 functions through lysis inhibition disruption . Congruent with LidIPLE 1 disrupting LIN , lidIPLE 1 expression in PLE ( - ) V . cholerae does not change the efficiency of plaquing ( EOP ) by ICP1 , an experiment that probes the number of successful initial infections at a low multiplicity of infection ( Figure 5B ) . Consistent with this , the phenotypic change in plaque morphology expected of disrupted LIN is the loss of fuzzy plaque edges , which we see in PLE ( - ) V . cholerae expressing lidIPLE 1 in trans ( Figure 5C ) . It is important to note that these data showing that LidI disrupts LIN when expressed alone do not reveal the molecular mechanism underlying this activity or ensure that the gene serves the same function in the context of PLE , even though it successfully phenocopies PLE-induced accelerated lysis . Because LIN in T-even coliphages functions to increase phage burst size ( Doermann , 1948 ) , we hypothesized that LidIPLE 1 collapsing LIN could inhibit ICP1 by decreasing progeny phage yield from infection . To determine if LidIPLE 1 alone can impact the number of phage produced from an infection , expression was induced prior to , at the time of , and at various intervals after infection with ICP1 at a high MOI ( MOI = 5 ) . Induction of lidIPLE 1 at the time of infection or 20 minutes before infection resulted in decreased phage yield by one or two orders of magnitude , respectively , in comparison to strains with the empty vector control ( Figure 6A ) . Not only did we find this to be true of LidIPLE 1 , but we also tested the LidIPLE 4 homolog for conserved inhibition of ICP1 and observed the same decreased phage yield ( Figure 6B ) . Subsequent testing revealed that when infections start out with a low number of phage per cell ( MOI ≤0 . 001 ) , there is still accelerated lysis and decreased progeny phage production when LidIPLE 1 expression is induced ( Figure 6—figure supplement 1 ) . These data reveal LidI as the first PLE-encoded ORF that can singlehandedly negatively impact ICP1 phage yield . From an evolutionary perspective , producing fewer virions equates to fewer diverse phages and would limit the phage’s ability to evolve counterattacks to anti-phage mechanisms or escape through mutation . Hence , we hypothesize that PLE-mediated accelerated lysis decreases the ability of ICP1 to evolve in the face of PLE . However , because our data support a model in which accelerated lysis is redundantly encoded , it is not currently possible to test the impact of delayed lysis on ICP1 evolution in the context of the PLE . We can , however , interrogate how the lidI-mediated collapse of LIN and concomitant decrease in phage production in PLE ( - ) V . cholerae constrains ICP1 evolution . To test if LidIPLE 1-mediated accelerated lysis is enough to impact diversity through the acquisition of random mutations in the progeny phage population , we exposed PLE ( - ) V . cholerae with and without lidIPLE 1 to ICP1 ( MOI = 0 . 1 ) , collected the population of progeny phage , and looked for plaque formation on PLE ( - ) V . cholerae encoding a Type I-E CRISPR-Cas system ( Box et al . , 2016 ) ; an expanded schematic of this experiment is available in Figure 6—figure supplement 2 . These host strains of CRISPR-Cas ( + ) V . cholerae were engineered to harbor various anti-ICP1 spacers that allowed for varying rates of ICP1 escape ( Figure 6C ) . For each spacer , fewer phage progeny from lidIPLE 1 V . cholerae were able to overcome targeting than progeny from infections of strains without lidIPLE 1 ( Figure 6D ) . This defect is due to the LidI-mediated decrease in the population of phages as the frequency of phage escaping stays the same ( e . g . ~2 out of every thousand phages can overcome spacer C , Figure 6D ) . Consequently , because less phage are produced from lidIPLE 1-expressing V . cholerae , an order of magnitude fewer phages can overcome the spacer in the population ( Figure 6—figure supplement 4 ) . While exposure to V . cholerae’s CRISPR-Cas system was meant to probe evolution at the level of individual random mutations and proclivity to overcome targeting , phages also readily use homologous recombination to evolve during co-infection . To test the impact of LidIPLE 1 on this aspect of evolvability , we took advantage of the Type I-F CRISPR-Cas system found in ICP1 by engineering two ICP1 variants with nonfunctional CRISPR-Cas systems: one devoid of spacers against PLE with an inactive Cas1 preventing spacer acquisition ( CRISPR*-Cas ICP1 ) ( McKitterick et al . , 2019b ) and the other lacking Cas2-3 ( CRISPR-Cas* ICP1 ) . We then used these phages to coinfect V . cholerae strains ( MOI = 0 . 01 ) with and without LidIPLE 1 . After lysis , progeny phage were tested for their ability to plaque on PLE 1 V . cholerae , which is only possible if homologous recombination between the two variants restored a functional CRISPR-Cas system able to target PLE 1 ( Figure 6E ) ; an expanded schematic of this experiment is available in Figure 6—figure supplement 3 . Predictably , the presence of LidIPLE 1 decreased the number of progeny phages per infection that recombined to reconstitute the CRISPR-Cas system and overcome PLE ( Figure 6F and Figure 6—figure supplement 4 ) . Although LidI expression does not directly impact ICP1 evolvability within infected cells , these data demonstrate that LidI imposes a bottleneck on ICP1’s population size resulting in fewer diverse phages and limiting ICP1’s potential to escape anti-phage activity .
Here , we observe a previously unknown lysis inhibition ( LIN ) state in the globally relevant pathogen , V . cholerae , in response to ICP1 , the predominant phage isolated from clinical samples , and identify the relevant ICP1-encoded lysis machinery: the holin , TeaA , and the antiholin , ArrA . Consistent with T-even induced LIN , ICP1 LIN results in fuzzy-edged plaques and delayed lysis sensitive to cellular proton motive force . This work subsequently reveals that the disruption of LIN is part of PLE’s anti-phage repertoire . A single open reading frame , lidI , which is conserved through all five PLEs spanning the last 70 years , is sufficient to disrupt LIN and limit progeny phage populations when expressed outside its native context in PLE ( - ) V . cholerae . These two opposing forces , ICP1 LIN and accelerated lysis by PLE through LIN disruption ( which our data shows LidI is capable of doing in isolation and yet other undiscovered PLE-encoded mechanisms redundantly accomplish ) , act in the midst of the ongoing evolutionary arms race between V . cholerae and its parasites . In phage-host interactions , research continues to unveil evolutionary strategies used to sense populations in the environment . Phage-encoded anti-CRISPRs ( the counter adaption to combat CRISPR-Cas systems ) can function in a phage-concentration dependent manner ( Borges et al . , 2018; Landsberger et al . , 2018 ) , small arbitrium peptides can influence lysogeny decisions based on other infections throughout the population ( Erez et al . , 2017 ) , and phages can ‘listen in’ on host quorum sensing as a measure of host availability ( Silpe and Bassler , 2019 ) . All of these environmental signals parallel LIN , one of the first discovered forms of communication during phage infection ( Doermann , 1948; Abedon , 2019; Hershey , 1946 ) . The anti-CRISPR system uses multiple subsequent phage infections , each benefiting from anti-CRISPRs expended during previous infections , to overcome host defenses similar to the subsequent superinfections that trigger LIN to make the most of the infected host . Arbitrium peptide signals formed upon infection give phages a proxy for when the majority of neighboring cells are infected , protecting potential progeny phage from adsorbing to previously infected cells by employing lysogeny – this shielding of progeny phages from infected cells is also accomplished during LIN . Phages eaves dropping on the host quorum sensing system provides a measure of available hosts – a lack of which is communicated by secondarily adsorbed virions during LIN . Consequently , our discovery of ICP1’s LIN in V . cholerae reinforces lysis inhibition as a relevant form of environmental sensing and highlights the importance of phages tuning their infection parameters depending on host availability in the environment . Following ingestion , V . cholerae colonizes and blooms in the small intestine before being shed in stool , further contaminating aquatic reservoirs and promoting subsequent ingestion and infection cycles . If large numbers of ICP1 are co-ingested with small numbers of V . cholerae , it would be theoretically beneficial for ICP1 to demonstrate LIN in infected cells , bide time inside the cell while making more virions , and lyse in the small intestine after uninfected V . cholerae have had time to replicate , providing ample hosts for subsequent rounds of phage predation . This is consistent with outbreak models in which phage become more abundant in the environment towards the end of an epidemic and there is phage amplification within cholera patients ( Faruque et al . , 2005a; Faruque et al . , 2005b; Jensen et al . , 2006 ) . Similarly , if V . cholerae bloomed in the gut followed by a subsequent expansion of the phage population to the extent that phages now outnumber hosts , it would be beneficial for ICP1 progeny phage to stay inside a cell , protected from adsorbing to cells that are already making progeny phage , and instead wait for release into the aquatic environment or ingestion by an uninfected patient via person-to-person transmission where ICP1 progeny may have better chances of finding an uninfected bacterial host to carry out its parasitic lifecycle . Interestingly , this begins to touch on variability of the phage life cycle in different aquatic environments ( Nelson et al . , 2008; Silva-Valenzuela and Camilli , 2019 ) , and the potential benefit for a phage population if a lysis inhibited cell is ingested along with V . cholerae from a patient – after which V . cholerae has been shown to be hyperinfectious ( Merrell et al . , 2002 ) . It is important to consider that within this tripartite system there are two parasites at odds with one another and both encode mechanisms to alter lysis timing . While the benefits of LIN to ICP1 have been explored in the parallel T-even phage/E . coli systems ( Abedon , 1990; Abedon , 2019 ) and discussed above , the impact of a MGE , specifically a parasitic phage satellite like PLE , on lysis inhibition could not have been predicted . One could argue that LIN might provide PLE with more resources for horizontal transmission , or counter that LIN provides time enough for ICP1 to evade PLE-encoded anti-phage mechanisms . Here , we show that the PLE accelerates lysis with LidI as a conserved part of its program – presumably collapsing LIN as a hinderance to ICP1 . Known examples showcase the deleterious effects of accelerated lysis on phage fitness , with holin mutations like the λ phage S-holin mutants which accelerate lysis by 20–25 minutes and lower progeny phage yield by orders of magnitude ( Johnson-Boaz et al . , 1994; Wang , 2006 ) . Similarly , a part of an abortive infection system in Lactococcus lactis , the AbiZ protein , causes cells infected by φ31 to lyse 15 minutes early , decreasing phage titers 100-fold ( Durmaz and Klaenhammer , 2007 ) . Indeed , we find even without other PLE-encoded products , accelerated cell lysis by LidI is sufficient to decrease phage population size and this bottlenecks the phage population , likely reducing ICP1’s ability to overcome PLE . The absolute abolishment of progeny phage accomplished by PLE’s complete anti-phage repertoire ( of which LidI is only a part ) is particularly interesting considering functionally similar SaPIs , which lay dormant in the chromosome much like PLE , until phage infection when SaPIs are induced to parasitize phage components ( Novick et al . , 2010 ) . One of many ways that PLEs and SaPIs differ is that SaPIs allow for some propagation of their helper phage , whereas PLE completely ablates ICP1 production of progeny phage . It is easy to think of SaPIs as selfish elements: they integrate and take advantage of vertical transmission . Once the cells are challenged by a helper phage , they excise , inhibit phage for their own ends , and escape the cell while allowing some progeny phage to escape with them all the while promoting diversity and horizontal gene transfer ( Frígols et al . , 2015; Novick et al . , 2010 ) . This lifecycle ensures horizontal transmission of the SaPI as well as continued activation of SaPIs down the line by available helper phages . In contrast , PLE completely blocks ICP1 production by acting as an abortive infection mechanism . Our evidence that PLE functions through collapsing lysis inhibition supports this angle as lysis inhibition could theoretically , as previously mentioned , allow for more time to produce PLE particles , enabling horizontal transmission of PLE to larger numbers of naïve cells . Surprisingly , this is so important that the means to disrupt LIN and execute lysis for PLE’s own selfish benefit are apparently redundantly encoded within the PLE – LidI can collapse LIN in the absence of PLE , though ΔlidI PLE still shows accelerated lysis . There is also limited evidence that PLE uses the hijacked ICP1 machinery to transduce in nature – in the laboratory , conditions allow four of the five PLEs to integrate in many sites across the superintegron; however , natural isolates only ever have one of those four PLEs integrated in one specific site ( O'Hara et al . , 2017 ) . This pattern is indicative of vertical transmission and infrequent horizontal transduction in the strains sampled from epidemics , which makes it easier to reconcile PLE’s abortive infection activity . With these evolutionary hypotheses in mind , ICP1 acquiring the CRISPR-Cas system changed the game: some single spacers encoded in ICP1 targeting PLE allow ICP1 to form progeny while simultaneously allowing for transduction of PLE ( McKitterick et al . , 2019b ) . If all spacers were singular and created equal , selection could drive PLE to act more like a typical satellite phage , embracing horizontal transfer and allowing ICP1 to slide by producing limited progeny phage . We know , however , that this is not the case; CRISPR systems function by dynamically acquiring spacers ( Barrangou et al . , 2007 ) and multiple spacers can abolish PLE’s ability to transduce while also killing the cell harboring PLE , destroying any chance at horizontal or vertical transmission ( McKitterick et al . , 2019b ) . Considering this complication , it is no surprise PLE would employ products like LidI to collapse LIN , perhaps to limit the ability of ICP1 to pick up new spacers against PLE . The dynamic arms race between ICP1 , PLE , and V . cholerae is ongoing as is research on other coevolving parasite/host systems . Focusing future work on LIN and MGEs is particularly promising given that this work represents a novel incarnation of LIN outside of the T-even coliphages , and we found homologs of LIN machinery outside of the limited contexts that LIN has been previously alluded to ( Gromkova , 1968; Latino et al . , 2019; Schito , 1974 ) . This prevalence suggests that LIN exists outside of characterized systems though the impacts of LIN and its disruption are unknown and largely unexplored . In contrast , the importance of MGEs is widely accepted and anti-phage mechanisms are increasingly found on MGEs , making questions about the interplay between MGEs and the complicating factors outlined here particularly attractive . One recently discovered example of how prevalent these confounding factors are is that of the cyclic-oligonucleotide-based anti-phage signaling system , which is found on MGEs in bacteria including V . cholerae , and that , in E . coli , serves as an abortive infection system upon phage infection and can cause lysis on an accelerated timescale ( Cohen et al . , 2019 ) . As a final note on these intriguing areas of future inquiry , there is increased interest in utilizing phages to combat bacterial infections as a part of phage therapy - the successful application of such approaches will likely depend on understanding all the interactions between phages and bacteria including responses that depend on the environment like lysis inhibition and interplay mediated by MGEs like PLE .
Bacteria were propagated at 37°C via streaking from frozen glycerol stocks on solid LB agar plates and growth in Miller LB ( Fisher Bioreagents ) with aeration . Media was supplemented with chloramphenicol ( 2 . 5 μg/mL for V . cholerae and 25 μg/mL chloramphenicol for E . coli ) , kanamycin ( 75 μg/mL ) , ampicillin ( 100 μg/mL ) , and streptomycin ( 100 μg/mL ) when appropriate . Cell densities were measured at OD600 in tubes ( Biochrom Ultrospec 10; 10 mm pathlength referred to as OD600-tube ) and in 96-well plates ( all reported OD600 measurements in figures have a pathlength equivalent to 150 μL in Costar Clear 96-well plates ( Corning ) ) . To induce chromosomal constructs in both liquid and top agar , plasmid constructs in top agar , and the plasmid Ptac-arrA construct for complementation , 1 mM isopropyl β-D-1-thiogalactopyranoside ( IPTG ) and 1 . 5 mM theophylline were added to cultures while the remaining plasmid constructs were induced with 125 μM IPTG and 187 . 5 μM theophylline in liquid cultures . Plasmid constructs were used for all experiments other than phage infection yield experiments which utilized chromosomal constructs to decrease leaky expression in uninduced strains . Bacteriophages were propagated on PLE ( - ) V . cholerae hosts and prepped via polyethylene glycerol precipitation or concentration and media exchange on Amicon Ultra – 15 ( Millipore ) centrifugal filters ( Bonilla et al . , 2016; Clokie and Kropinski , 2009 ) . Stocks were stored in sodium chloride-tris-ethylenediaminetetraacetic acid buffer ( STE ) , and quantified via the soft agar overlay method ( Clokie and Kropinski , 2009 ) . Briefly , titering was completed by growing V . cholerae to mid-log , infecting with cultures with diluted phage , and allowing adsorption to occur for 7 to 10 minutes before plating on 0 . 5% LB top agar . Subsequently , multiplicity of infection ( MOI ) was determined by calculating the number of plaque forming units and varying that with the number of colony forming units of V . cholerae at a given optical density . This does not take into account virions that adsorb but do not successfully form plaques . Consequently , all reported MOIs do not address multiplicity of adsorption which could vary and impact initial changes in optical densities during experiments . Mechanical lysis of cultures infected with phage was accomplished by mixing chloroform into cultures then letting cultures stand at room temperature for ten minutes before spinning at 5000 x g for 15 minutes at 4°C and removing the supernatant for further analysis . Chromosomal integrations in V . cholerae were accomplished through natural transformation of linear DNA created via splicing by overlap extension PCR ( Dalia et al . , 2014 ) . In the case of deletions , antibiotic resistance cassettes were integrated into the locus of the deleted gene and subsequently flipped out as previously described ( Baba et al . , 2006 ) . Plasmids were constructed with Gibson Assembly and Golden Gate reactions . Phage mutants were selected as previously described ( Box et al . , 2016 ) . Briefly , complementary spacer oligos were annealed and inserted into a plasmid-borne CRISPR array . This plasmid was mated into a strain of V . cholerae engineered to include an inducible Type 1-E CRISPR-Cas system ( CRISPR-Cas ( + ) V . cholerae ) . This system in the host strain was induced for 20 minutes before ICP1 infection for plaque assays on 0 . 5% LB top agar containing antibiotics to maintain the CRISPR array plasmid . Plaques were picked into STE ( 100 mM NaCl , 10 mM Tris-HCl 1 mM EDTA ) , purified on the same host twice , and genomic DNA was prepped for PCRs with the DNeasy Blood and Tissue Kit ( Qiagen ) . Phages were subjected to PCR of the targeted gene and subsequent Sanger sequencing . Clean knockouts were accomplished by adding a repair template of homologous sequence containing the desired deletion and flanking DNA to the plasmid containing the CRISPR array as previously described ( Box et al . , 2016 ) . V . cholerae strains were grown in 2 mL cultures to an OD600-tube=0 . 3 and 150 μL of cultures were added to 96-well plates . This transition often results in a slight decrease in optical density at the beginning of experiments where OD600 is tracked . The underlying cause for this decrease is unknown , but is consistent between controls and experimental conditions . Inducers and phage were pre-aliquoted in plates unless otherwise specified . OD600 within the plate was read for each sample on the SpectraMax i3x ( Molecular Devices ) plate reader every two minutes with one minute of shaking between each read while the machine incubated cultures at 37°C . Assays were interrupted for the addition of phage , inducers , 2 , 4-dinitrophenol ( DNP; 2 μL of 8 . 4 mM DNP dissolved in 80% ethanol for a final concentration of 110 μM DNP ) , or ethanol ( 2 μL of 80% ethanol ) during which samples were removed from the plate reader briefly before measurement was resumed . This enabled superinfection of cultures ( initial cultures were infected with ICP1 MOI = 1 , returned to the plate reader for four minutes , and subsequently superinfected with ICP1 MOSI = 5 ) , addition of DNP in ethanol or ethanol alone to ICP1 MOI = 5 infected cultures 25 minutes post-infection , and DNP addition to induced cultures after 20 minuntes of growth with the inducers . For each strain , three 2 mL cultures of V . cholerae were grown . The first was grown to an OD600-tube=0 . 15 , inducer was added , and the culture was returned to the incubator; this culture served as the pre-induced culture . All cultures were grown for an additional 20 minutes to OD600-tube=0 . 3 at which point ICP1 MOI = 5 was added . Inducer was added to one tube at this time; this culture served as the culture induced at time zero . Tubes were returned to the incubator for 5 minutes for phage adsorption . Cultures were spun at 5000 x g for 3 minutes to pellet cells . Unadsorbed phage was aspirated off and cells were washed once with 1 mL of prewarmed LB with or without inducer . Cells were spun again and resuspended in media with and without inducer at which point OD600-tube was determined . Cultures ( 150 μL ) were moved to 96-well plates with one well designated to the pre-induced culture , one well devoted to the culture induced at time zero , and three wells filled with uninduced culture . The plate was returned to the 37°C incubator to shake at 230 RPM . Inducer was added to two wells of the uninduced cultures 20 and 40 minutes post-infection respectively , leaving one uninduced control . The experiment was ended with mechanical lysis of cultures and subsequent quantification of phage titers . V . cholerae containing the specified plasmids were grown to OD600-tube=0 . 2 . Inducers ( 125 μM IPTG and 187 . 5 μM theophylline ) were added and cells were grown at 37°C with aeration for 30 minutes before 0 . 5 mL were pelleted at 5000 x g for 3 minutes and resuspended in 0 . 1 mL of phosphate buffered saline ( pH 7 . 2; Gibco Life Technologies ) containing 20 μM 3 , 3’-diethloxacarbocyanine iodide ( DiOC2 ( 3 ) ; Sigma-Aldrich ) . Fluorescence measurements were completed in black 96-well half-volume plates ( Corning ) in the SpectraMax i3x ( Molecular Devices ) with 480 ( 508 ) and 488 ( 650 ) excitation ( emission ) wavelength settings . V . cholerae strains were grown to mid-log before plaquing assays . For EOP experiments , plasmid constructs were induced for 20 minutes prior to infecting and plating with antibiotic and inducer in the 0 . 5% LB top agar . For plaque edge analysis , plasmids were maintained with antibiotics but not induced prior to plating on 0 . 5% top agar containing antibiotics and inducer . Plates solidified at room temperature prior to incubation at 37°C . For spot plates , V . cholerae was mixed with 0 . 5% top agar prior to infection , vortexed , and poured onto an LB agar plate to solidify before 3 μL spots of phage dilutions were overlaid on the agar . Plates were allowed to dry prior to incubation at 37°C . To visualize plaques , plates were scanned on the EPSON Perfection V800 Dual Lens scanner . Plasmid empty vector ( EV ) and FLAG-LidIPLE 1 were grown to OD600-tube=0 . 2 then induced with 1 mM IPTG and 1 . 5 mM for 40 minutes before 0 . 5 mL samples were taken . To observe expression during infection , strains were grown to OD600-tube=0 . 3 , infected with ICP1 MOI = 2 with 4 mL samples taken at the labeled timepoints . Samples were prepared and visualized as previously described ( McKitterick and Seed , 2018 ) . Briefly , samples were mixed with one volume of cold methanol , pelleted at 15 , 000 x g at 4°C for 15 minutes , washed with 1 mL cold PBS , and pelleted . Pellets were resuspended in PBS with XT sample buffer and reducing agent ( Bio-Rad ) , vortexed , and boiled for 10 minutes before being run on 4–12% Bis-Tris SDS gels ( Bio-Rad Criterion XT ) . Gels were transferred via the Trans-Blot Turbo ( Bio-Rad ) and visualized with rabbit-α-FLAG ( 1:3 , 000 ) primary and goat-α-rabbit-HRP conjugated secondary antibodies on the ChemiDoc MP Imaging System ( Bio-Rad ) . V . cholerae harboring an empty vector control or LidIPLE 1 plasmid were grown to OD600-tube=0 . 15 before being induced and returned to the incubator to grow with aeration until OD600-tube=0 . 3 . At this point 150 μL of culture were added to 96-well plates and infected with ICP1 MOI = 0 . 1 . After lysis 90 minutes post-infection , any remaining cells were mechanically lysed and the resulting phage population was plaqued on CRISPR-Cas ( + ) V . cholerae as previously described ( Box et al . , 2016 ) . Briefly , Cas ( + ) V . cholerae harboring CRISPR array plasmids were induced 20 minutes before phage populations were titered on each specified V . cholerae strain . EOPs were determined by dividing the number of PFU/mL on V . cholerae containing spacers by the PFU/mL on the CRISPR-Cas ( + ) V . cholerae that did not contain a spacer against the phage ( denoted as spacer ‘none’ ) . Coinfection experiments were completed in the same manner as the CRISPR-Cas targeting of phage populations described above with minor alterations: instead of infection with wild type ICP1 , cultures were coinfected in the plate with CRISPR*-Cas ICP1 and CRISPR-Cas* ICP1 each at an MOI = 0 . 01 , observed for 200 minutes in the plate reader , and , after mechanical lysis , phage populations were plaqued on PLE ( - ) V . cholerae and PLE 1 V . cholerae . Control infections with only one phage never formed plaques on PLE 1 V . cholerae . The proportion of phages that successfully recombined was determined by the dividing the PFU/mL of each phage population on PLE 1 V . cholerae divided by the PFU/mL on PLE ( - ) V . cholerae . Transmembrane domains were predicted by conversion of all predicted open reading frames to amino acid sequence by CLC ( CLC , 2020 ) before analysis with TMHMM Server v . 2 . 0 ( Sonnhammer et al . , 1998 ) . PRALINE was used to create amino acid alignments of LidI ( Simossis and Heringa , 2005 ) . Homologs of TeaA ( 30% identity over 85% of the query ) and ArrA ( 20% identity over 75% of the query ) were identified with BLASTP ( NCBI NIH , 2019 ) and arranged into phylogenetic trees as previously described ( McKitterick et al . , 2019a ) . Briefly , alignments were completed with MUSCLE v3 . 8 . 31 ( Madeira et al . , 2019 ) and a bootstrapped ( n = 100 ) maximum-likelihood phylogenic tree was solved with PhyML 3 . 0 ( Guindon et al . , 2010 ) . Default settings were used for amino acid sequences: automatic model selection with Akaike Information Criterion; SPR tree improvement with n = 10 random starting trees . Trees were visualized with FigTree ( Rambaut et al . , 2019 ) . | Bacteriophages , or phages for short , are viruses that infect bacteria , take over the molecular machinery inside the bacterial cells and use it to make more copies of themselves . The bacteriophages then break open , or “lyse” , the bacterial cell , releasing the viral copies into the environment , ready to infect more bacteria nearby . Hays and Seed set out to understand how the timing of lysis can impact the bacteriophage , using the bacterium Vibrio cholerae – which causes cholera – and its bacteriophage called ICP1 . This analysis revealed that the ICP1 phage uses a gene called teaA as the first step in the lysis of bacterial cells . The ICP1 phage can also delay that lysis with a second gene called arrA . This “lysis inhibition” gives the bacteriophages more time to make copies of themselves inside the bacterium , so even more are released when the cell finally breaks open . Hays and Seed also found that the Vibrio cholerae cells can defend themselves against lysis inhibition using a single gene called lidI . This gene is part of a system that defends against bacteriophage attack called the PLE , which consists of several genes of previously unknown function . Hays and Seed saw that the lidI gene disrupts lysis inhibition , speeding up the bursting of infected bacterial cells , which in turn decreases the number of bacteriophages produced from each infected cell . Lysis inhibition had previously only been observed in the bacterium Escherichia coli . Now that researchers know that ICP1 bacteriophages also delay lysis in Vibrio cholerae , this might lead to more studies exploring this process in samples from cholera patients . Further studies could test to see if the phenomenon of lysis inhibition may also exist in yet more bacterial species . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2020 | Dominant Vibrio cholerae phage exhibits lysis inhibition sensitive to disruption by a defensive phage satellite |
Construction of motile cilia/flagella requires cytoplasmic preassembly of axonemal dyneins before transport into cilia . Axonemal dyneins have various subtypes , but the roles of each dynein subtype and their assembly processes remain elusive in vertebrates . The PIH protein family , consisting of four members , has been implicated in the assembly of different dynein subtypes , although evidence for this idea is sparse . Here , we established zebrafish mutants of all four PIH-protein genes: pih1d1 , pih1d2 , ktu , and twister , and analyzed the structures of axonemal dyneins in mutant spermatozoa by cryo-electron tomography . Mutations caused the loss of specific dynein subtypes , which was correlated with abnormal sperm motility . We also found organ-specific compositions of dynein subtypes , which could explain the severe motility defects of mutant Kupffer’s vesicle cilia . Our data demonstrate that all vertebrate PIH proteins are differently required for cilia/flagella motions and the assembly of axonemal dyneins , assigning specific dynein subtypes to each PIH protein .
Motile cilia/flagella are hair-like organelles that project from various types of eukaryotic cells . In humans , malfunctions of motile cilia often cause primary ciliary dyskinesia ( PCD ) , a syndrome characterized by recurrent respiratory infections , male infertility , hydrocephalus , and inversion of visceral laterality ( Knowles et al . , 2013; Brown and Witman , 2014 ) . Motile cilia have a microtubule-based structure called an axoneme , consisting of nine peripheral doublet microtubules ( DMTs ) with or without central-pair microtubules ( so called 9 + 2 , 9 + 0 , respectively ) . Ciliary motility is driven by axonemal dyneins , which have multiple subtypes such as outer arm dyneins ( OADs ) and seven different types of inner arm dyneins ( IADs; IAD a to g; Kagami and Kamiya , 1992 ) . Biochemical analyses of green algae Chlamydomonas revealed that each axonemal dynein consists of multiple subunits ( Hom et al . , 2011; Sakato and King , 2004 ) ; OAD is composed of three heavy chains ( α- , β- , and γ-HC ) , two intermediate chains ( IC1 and IC2 ) , and ten light chains . Six types of IADs ( IAD a , b , c , d , e , and g ) have single HCs with several light chains such as p28 , centrin , and actin . IAD f has two heavy chains ( f α- and f β-HC ) , four intermediate chains , and five light chains . In the process of ciliary construction , axonemal dyneins and all other large ciliary molecules are synthesized in the cytoplasm and undergo gated entry into the ciliary compartment ( Takao and Verhey , 2016 ) . In the cytoplasm , the components of OADs and IADs are detected as preassembled complexes , rather than individual components ( Fok et al . , 1994; Fowkes and Mitchell , 1998; Viswanadha et al . , 2014 ) . This cytoplasmic preassembly of axonemal dyneins requires various proteins collectively called dynein axonemal assembly factors ( DNAAFs; Kobayashi and Takeda , 2012; Mitchison et al . , 2012 ) . As cilia/flagella require multiple types of axonemal dyneins for their motions ( Kamiya , 1995 ) , proper assembly of each dynein complex is essential for ciliary motility . However , the assembly processes of each dynein subtype and their roles in cilia/flagella motions remain elusive in vertebrates . The PIH protein family has been implicated in the preassembly of different subsets of axonemal dyneins . The PIH protein , which contains a PIH1-domain , was first identified in budding yeast ( Saccharomyces cerevisiae ) as an interactor of HSP90 and named as Pih1 ( Protein Interacting with HSP90; also known as Nop17; Zhao et al . , 2005; Gonzales et al . , 2005 ) . Yeast Pih1 is required for the assembly of various multi-subunit protein complexes but is not involved in the assembly of axonemal dyneins , as yeast do not have a cilium . In vertebrates , there are four PIH proteins: PIH1D1 , PIH1D2 , KTU/DNAAF2 , and PIH1D3/TWISTER , and PIH1D1 is the orthologue of yeast Pih1 . Similar to yeast , the human PIH1D1 is a subunit of the R2TP complex ( RUVBL1 , RUVBL2 , RPAP3/Tah1 , and PIH1D1 ) , which interacts with HSP90 to promote assembly of various protein complexes for cellular activities such as box C/D snoRNP and RNA polymerase II ( Kakihara and Houry , 2012 ) . KTU/DNAAF2 is the first protein that was identified as a DNAAF ( Omran et al . , 2008 ) . Genetic and biochemical analyses of KTU/DNAAF2 in medaka ( Japanese killifish ) , human , and Chlamydomonas revealed that KTU/DNAAF2 is required for the assembly of OAD and a subset of IADs ( Omran et al . , 2008 ) . Subsequently , the function of Chlamydomonas MOT48 , a possible orthologue of vertebrate PIH1D1 , was reported; MOT48 is one of three PIH proteins in Chlamydomonas ( MOT48 , PF13/KTU , and TWI1 ) and is involved in the assembly of another subset of IDAs ( Yamamoto et al . , 2010 ) . These pioneering studies proposed that preassembly of different subsets of axonemal dyneins is mediated by distinct PIH proteins . However , evidence for this hypothesis is sparse , due to a lack of systematic studies of all PIH proteins . Although PIH1D3/TWISTER recently turned out to be one of the DNAAFs ( Dong et al . , 2014; Paff et al . , 2017; Olcese et al . , 2017 ) , the function of vertebrate PIH1D1 and PIH1D2 has not been addressed in terms of ciliogenesis and ciliary motility . Furthermore , two Pih1d3 paralogues in mice , Pih1d3 and Twister2 , are differently expressed in ciliary/flagellar organs ( Pih1d3 for testis , while Twister2 for both testis and the others; Dong et al . , 2014 ) , suggesting a divergence of their functions . In this study , we performed systematic and functional analyses of all four PIH genes ( genes encoding PIH proteins ) by generating zebrafish ( Danio rerio ) mutants of each PIH gene . We compared the functions of all PIH proteins in one platform , zebrafish , because functional divergence of PIH proteins among organisms is highly probable . Although zebrafish is often used to analyze the functions of PCD related genes , the detailed structure of their cilia/flagella has not been studied so far . We applied cryo-electron tomography ( cryo-ET ) for the first time to zebrafish sperm , which enabled us to observe the detailed structure of wild-type and mutant axonemal dyneins . Mutations of each PIH gene caused defects of different subtypes of axonemal dyneins , which was correlated with abnormal sperm motility . Interestingly , some mutants showed different phenotypes of ciliary motility between sperm flagella and Kupffer’s vesicle cilia . Together with different expression patterns of various DNAH ( dynein axonemal heavy chain ) genes , we also discuss the organ-specific compositions of axonemal dyneins assembled by PIH proteins . This is the first report that shows all vertebrate PIH proteins as DNAAFs , assigning their functions to specific types of axonemal dyneins and to cilia/flagella motions . Our data provide evidence for the above-mentioned hypothesis that the cytoplasmic assembly of different dynein subtypes is mediated by distinct PIH proteins .
To find all PIH proteins encoded in the zebrafish genome , we performed BLASTp search with the consensus sequence of PIH proteins as a query . Although teleost fish are known to have undergone an additional genome duplication ( Kasahara et al . , 2007 ) , only four hits were obtained , similar to the human genome and consistent with a previous report ( Yamamoto et al . , 2010 ) . We thus conclude that zebrafish has four PIH proteins: Pih1d1 , Pih1d2 , Ktu , and Twister . Their domain structures are well conserved among vertebrates ( Figure 1A; Figure 1—figure supplement 1A , B ) . Transcripts of these four PIH genes were all detected in ciliated organs such as Kupffer’s vesicle , floor plate , otic vesicle and pronephric duct ( Figure 1D–G ) , suggesting their involvement in ciliary functions . The transcript of pih1d1 was also detected in the whole body of 12 hpf ( hours post-fertilization ) embryos ( Figure 1D; black asterisk ) , which is consistent with the reported cellular functions of human PIH1D1 ( Kakihara and Houry , 2012 ) . ktu was also expressed in brain rudiments at 32 hpf ( Figure 1D; black arrowhead ) . In mouse , KTU function in brain was reported for ciliated ependymal cells ( Matsuo et al . , 2013 ) . These results suggest that pih1d2 , ktu , and twister have cilia-specific functions , while pih1d1 has ubiquitous cellular functions in addition to ciliary function . Since DNAAFs including PIH proteins are known to be localized to the cytoplasm ( Kobayashi and Takeda , 2012 ) , we examined the subcellular localizations of zebrafish PIH proteins by immunoblot analysis , using specific polyclonal antibodies made in this study . All PIH proteins were detected in both testis lysate and sperm lysate . However , when spermatozoa were fractionated into sperm heads and flagella , PIH proteins were detected only in the sperm head fraction , indicating that like other DNAAFs , zebrafish PIH proteins are indeed specifically present in the cytoplasm , but not in the flagellar compartment ( Figure 1C; asterisks ) . To analyze the functions of the four PIH proteins in zebrafish , we generated mutant alleles by genome-editing with TALEN ( pih1d1 and pih1d2 ) or CRISPR/Cas9 ( ktu and twister ) ( Figure 1B; Figure 1—figure supplement 1C , D ) . Immunoblot analysis confirmed that all mutations resulted in a null of each PIH protein ( Figure 1—figure supplement 1E ) . Since homozygous mutants of each PIH gene were viable , we established homozygous mutant lines . These were used in the following experiments . Ciliary functions of PIH genes were first examined by observing the motility of mutant spermatozoa using a high-speed camera . Spermatozoa whose heads were attached to a coverslip were selected and subjected to the analyses of beating frequencies and waveforms ( Video 1 ) . In pih1d1-/- , spermatozoa showed a slight reduction of beating frequency ( Figure 2G ) , and propagation of flagellar bending was disturbed , as slopes of shear angle curves changed between traces ( Figure 2B’; asterisk ) . In twister-/- , almost all spermatozoa were immotile , but a few were found to be motile with decreased beating frequencies and severely disturbed waveforms ( Figure 2E ) . In pih1d2-/- and ktu-/- , a significant difference was not observed in either beating frequencies or waveforms ( Figure 2C , D ) . We suspected functional compensation of these two genes , and thus generated double mutants of pih1d2-/-;ktu-/- . Double mutant spermatozoa exhibited abnormal waveforms; motile in the proximal half , while immotile in the distal half ( Figure 2F ) . In the proximal region , beating frequency was about twice as high as that of wild type ( Figure 2G ) . This could be caused by the reduction of sliding distance of DMTs , rather than the change of the sliding velocity of DMTs , because the slopes of shear angle curves was decreased ( Figure 2F’; dotted line ) , indicating that the bending of the proximal flagella was smaller than that of wild type . We also analyzed the length of sperm flagella , but did not find any significant differences between wild type and PIH gene mutants ( Figure 2—figure supplement 1C ) . We then examined the motility of free swimming spermatozoa by CASA ( computer-assisted sperm analysis ) modified for zebrafish ( Wilson-Leedy and Ingermann , 2007 ) . In this analysis , traced paths of swimming sperm heads were used to calculate sperm motility ( Figure 2—figure supplement 1B; Video 2 ) . In all mutants , the ratios of motile ( locomotive ) spermatozoa were significantly decreased , and no sperm in twister-/- showed significant locomotion ( Figure 2H ) . For motile spermatozoa , swimming velocity did not change in pih1d1-/- , pih1d2-/- , and ktu-/- , but significantly decreased in pih1d2-/-;ktu-/- ( Figure 2I ) . The beating frequencies of sperm heads were decreased in pih1d1-/- , but increased in pih1d2-/-;ktu-/- ( Figure 2J ) , which is consistent with Figure 2G . To observe the ultrastructure of zebrafish axoneme , we applied cryo-ET to zebrafish spermatozoa . The axoneme of zebrafish sperm had the characteristic 9 + 2 arrangement of DMTs surrounding central-pair microtubules ( Figure 3—figure supplement 1A ) . To analyze the structure of DMTs in more detail , subtomographic averaging was applied using the 96 nm repeat of DMTs assuming nine-fold rotational symmetry of the axoneme , since we did not detect any obvious heterogeneity of nine DMTs in zebrafish sperm unlike Chlamydomonas flagella and sea urchin sperm ( Hoops and Witman , 1983; Bui et al . , 2012; Lin et al . , 2012 ) . The averaged structure of zebrafish DMT exhibited overall similarity to that of other organisms ( Figure 3C , D ) . Thus , based on the well-studied structure of the Chlamydomonas axoneme ( Bui et al . , 2012 ) , we assigned the structures of OADs , seven types of IADs , radial spokes ( RSs ) , and nexin-dynein regulatory complex ( N-DRC ) in the zebrafish axoneme ( Figure 3A , B; Video 3 ) . To address the evolutionary conservation and diversity of cilia/flagella , we compared the ultrastructure of zebrafish axoneme to that of Chlamydomonas and human axonemes in more detail . Compared with Chlamydomonas , zebrafish axoneme does not have OAD α-HCs , but has longer RS3 ( Figure 3A , D ) . A linker between N-DRC and OAD is not observed in the zebrafish axoneme unlike Chlamydomonas ( Figure 3D; red arrowhead ) . These features of the zebrafish axoneme are also found in human respiratory cilia ( Figure 3C; Lin et al . , 2014 ) . The same features were also reported in mouse respiratory cilia ( Ueno et al . , 2012 ) and found in sea urchin spermatozoa ( Lin et al . , 2012 ) , indicating that these features are common among metazoans . To gain structural insights into abnormal motility of mutant spermatozoa , we observed the structure of mutant axonemes by cryo-ET and subtomographic averaging . Compared to wild type , mutant axonemes exhibited structural defects of various types of axonemal dyneins ( later summarized in Figure 7A ) . IAD c was missing in pih1d1-/- ( Figure 3F; Video 4 ) , while no significant difference was observed in pih1d2-/- ( Figure 3G; Video 5 ) . In ktu-/- , smaller IAD c density was observed , suggesting that IAD c is partially missing in ktu-/- spermatozoa ( Figure 3H; Video 6 ) . In twister-/- , reflecting severe motility defects , OADs and IAD c were missing and smaller IAD g and d were observed ( Figure 3I; Video 7 ) . Intriguingly , in pih1d2-/-;ktu-/- , averaging of all DMT particles did not converge into one structure , thus tomograms were classified as follows . We noticed that out of nine tomograms of axonemes , four axonemes had OADs but five lacked OADs . Using this difference , we divided axonemes into two classes ( +OAD and -OAD ) and averaged , respectively . The +OAD class possessed a full set of axonemal dyneins , except for a smaller IAD c , like the ktu-/- axoneme ( Figure 3J; Video 8 ) . By contrast , the -OAD class lost not only OADs , but also IAD b , c , and e ( Figure 3K; Video 9 ) . However , note that the –OAD class showed faint densities of these IADs in the subtomographic slice ( Figure 3—figure supplement 1C ) , which suggests that IAD b , c , and e were retained partially in the –OAD class axonemes . Although we found structural defects of axonemal dyneins , no significant defect was observed in other DMT structures , such as RSs , in all mutants we examined . To correlate the structural defects of mutants with biochemical data , we performed immunoblot analysis of axonemal dynein components ( Figure 4A ) . We made specific antibodies against zebrafish Dnah8 ( OAD γ-HC ) and Dnah2 ( IAD f β-HC ) . Dnai1 is a component of OADs and is also known as IC1 . Dnali1 is the orthologue of Chlamydomonas p28 , which is the subunit of three types of IADs: IAD a , c , and d ( Piperno et al . , 1990; Hom et al . , 2011 ) . Consistent with the above structural analysis , Dnah8 and Dnai1 were missing from the axoneme of twister-/- ( Figure 4A; asterisks ) . In pih1d2-/-;ktu-/- , the amount of Dnah8 and Dnai1 was decreased , possibly reflecting the presence of the two types of DMT structures ( +OAD and -OAD ) . Dnah2 was not affected in any mutants , and so was the case of IAD f in the structural analysis . Dnali1 was slightly decreased in pih1d1-/- , ktu-/- , and pih1d2-/-;ktu-/- ( Figure 4A; filled circles ) , confirming the loss of IAD c ( one of three IADs containing p28 in Chlamydomonas ) in these mutants . In twister-/- , the structural analysis revealed the loss of IAD c and d ( two of three IADs containing p28 ) , and the amount of Dnali1 was strongly reduced . Interestingly , shifted bands of Dnai1 were observed in pih1d1-/- and pih1d2-/- ( Figure 4A; open circles ) , indicating abnormal construction of OADs in these mutants . However , the structure of OADs in these mutants appeared normal as far as our structural analysis showed at the current resolution . Taken together , all biochemical results are largely consistent with our structural data . From these results , we conclude that all PIH proteins are responsible for the assembly of specific subtypes of axonemal dyneins . Together with their specific cytoplasmic localizations , we identified all vertebrate PIH proteins ( not only Ktu and Twister , but also Pih1d1 and Pih1d2 ) as DNAAFs . Abnormal sperm motility observed in pih1d1-/- , twister-/- and pih1d2-/-;ktu-/- can be explained by the loss of specific subtypes of axonemal dyneins . On the other hand , spermatozoa of pih1d2-/- and ktu-/- appeared to have normal motility , although the structural or biochemical analyses revealed abnormal axonemal dyneins in these mutants . A likely explanation for this discrepancy is that the defects of axonemal dyneins in pih1d2-/- or ktu-/- spermatozoa are so subtle that other normal axonemal dyneins can compensate their loss of function . However , it is worth noting that the affected axonemal dyneins were different between pih1d2-/- ( OAD Dnai1 ) and ktu-/- ( IAD c ) , which indicates distinct functions of Pih1d2 and Ktu , although functional compensation of these two genes was also revealed by pih1d2-/-;ktu-/- . The two types of DMT structures in pih1d2-/-;ktu-/- , ( Figure 3J , K;+OAD and -OAD classes ) led us to examine their distribution in the mutant axoneme . For this , we stained mutant spermatozoa with the anti-Dnah8 ( OAD γ-HC ) antibody ( Figure 4B ) . In wild type , Dnah8 was localized along the entire length of the flagellum . However , in pih1d2-/-;ktu-/- , Dnah8 was consistently absent in the distal region , while it remained in the proximal ( Figure 4B; white arrowhead ) . Thus , the +OAD class structure was localized in the proximal region , while the -OAD class was in the distal . We also analyzed the localization of Dnah8 in other mutants . Consistent with our structural analysis , in pih1d1-/- , pih1d2-/- , and ktu-/- spermatozoa , Dnah8 was normally distributed along the entire length of their flagella , while twister-/- spermatozoa completely lost Dnah8 . Different structural defects of IADs were also observed between the +OAD and -OAD classes . To assess the distribution of IADs in pih1d2-/-;ktu-/- , we analyzed the structure of proximal and distal axoneme directly . Among many cryo-prepared pih1d2-/-;ktu-/- axonemes , we found one axoneme suitable for observing both proximal and distal regions by cryo-ET ( Figure 4—figure supplement 1A ) . Although the obtained subtomograms are noisy due to a smaller number of averaged particles , the DMT structure of a proximal subtomogram possessed OADs and the densities of all IADs , consistent with the structure of +OAD class ( Figure 4—figure supplement 1B , D ) . On the other hand , the DMT structure of a distal subtomogram lost OADs , IAD b , c , and e , which corresponds to -OAD class ( Figure 4—figure supplement 1C , E ) . Therefore , the distribution of not only OADs but also IADs is different between the proximal and distal regions in pih1d2-/-;ktu-/- spermatozoa . In humans , organ-specific compositions of OAD HCs were reported between sperm flagella and respiratory cilia ( Figure 7E; Fliegauf et al . , 2005; Dougherty et al . , 2016 ) , and mouse Pih1d3 was reported as a testis-specific gene ( Dong et al . , 2014 ) . To assess organ-specific functions of zebrafish PIH proteins , we focused on a second ciliated organ , Kupffer’s vesicle , which is orthologous to the mammalian embryonic node . In Kupffer’s vesicle , epithelial cells project mono-cilia that have rotational motility to produce leftward fluid flow in the organ ( Figure 5A , B ) . Like in the mouse node , this leftward flow is required for the determination of visceral asymmetry , and thus defects of Kupffer’s vesicle cilia cause abnormal left-right patterning of the fish ( Essner et al . , 2005 ) . Mutations of each PIH gene caused abnormal motility of Kupffer’s vesicle cilia . To describe ciliary motility , we categorized motion patterns into three classes: rotating , irregular , and immotile ( Figure 5C , D; Video 10 ) . Rotational frequencies were measured from rotating class cilia ( Figure 5E ) . The resulting left-right patterning of embryos was assessed by observing the direction of heart looping ( normally rightward; Figure 5F , G ) . In pih1d1-/- , rotational frequencies of cilia were significantly reduced , but almost all cilia were motile and the ratio of heart-looping reversal was not largely affected . By contrast , in twister-/- and pih1d2-/-;ktu-/- , all cilia were immotile , leading to complete randomization of their left-right patterning . In pih1d2-/- and ktu-/- , the proportions of rotating class cilia were decreased to ~40% and~15% , respectively , with reduced rotational frequencies , resulting in significant levels of heart-looping defects ( Figure 5D ) . Regarding the structure and localization of axonemal dyneins , due to technical difficulties , we were unable to apply cryo-ET and immunohistochemistry with anti-dynein antibodies to the axonemes of Kupffer’s vesicle cilia . Together with sperm analyses , we conclude that all PIH genes of zebrafish are essential for normal motility of both sperm flagella and Kupffer’s vesicle cilia . Intriguingly , however , in pih1d2-/- and ktu-/- , only Kupffer‘s vesicle cilia showed motility defects , while sperm flagella beat normally , indicating that Pih1d2 and Ktu have organ-specific functions . The organ-specific phenotypes of PIH mutants could reflect organ-specific compositions of axonemal dyneins . To address this , we performed whole-mount in situ hybridization of various DNAH genes . Zebrafish have three OAD β-HC genes: dnah9 , dnah9l , and dnah11 , and two OAD γ-HC genes: dnah5 and dnah8 . As for IAD , dnah2 is an IAD f β-HC gene , and dnah3 and dnah7l are other IAD HC genes . The gene correspondence of dynein heavy chains among zebrafish , human , and Chlamydomonas are summarized in Table 1 , based on the comprehensive analysis of dynein phylogeny by Kollmar ( 2016 ) . Zebrafish embryos and testes showed distinct expression patterns of DNAH genes ( Figure 6A ) . When comparing Kupffer’s vesicle and testis , dnah11 expression was specifically detected in Kupffer’s vesicle , while dnah8 and dnah3 were specifically detected in testis ( Figure 6B ) . At the embryonic stages , dnah9l and dnah8 were detected only in the otic vesicle and the pronephric duct , respectively , which also suggested specific combinations of DNAH genes in these organs . These results indicate that components of axonemal dyneins are indeed organ-specific . Intriguingly , however , Kupffer’s vesicle and floor plate , whose cilia exhibit similar rotational motility ( Kramer-Zucker et al . , 2005 ) , showed the same expression patterns of DNAH genes . It is likely that the same transcriptional regulation is required in Kupffer’s vesicle and floor plate to construct similar types of cilia .
Our cryo-ET and biochemical analyses revealed that the PIH proteins are required for the assemblies of specific subsets of axonemal dyneins ( Figure 7C ) : Pih1d1 for OAD ( Dnai1 construction ) and IAD c; Pih1d2 and Ktu for OAD , IAD b , c , and e; and Twister for OAD , IAD c , g , and d . In Chlamydomonas , the mutation of KTU/PF13 affected the assembly of OAD and IAD c , while the mutation of MOT48 affected the assembly of OAD , IAD b , c , d , and e ( Yamamoto et al . , 2010 ) . Although the affected subtypes of axonemal dyneins are not the same , the mutations of PIH genes resulted in the loss of specific subsets of axonemal dyneins in both organisms . Remarkably , OAD and IAD c are most sensitive to the mutations of various PIH genes in both zebrafish and Chlamydomonas . Consistent with this , Dong et al . ( 2014 ) suggested that the assembly of OAD proceeds in a stepwise manner mediated by different PIH proteins . This might also be the case for IAD c , although no experimental evidence has been obtained . In our research , IAD a or f were not affected in any PIH gene mutants ( Figure 7C ) , which was essentially the same as in Chlamydomonas mutants of KTU/PF13 and MOT48 . This suggests that these axonemal dyneins are assembled independently of PIH proteins . Alternatively , multiple PIH proteins redundantly participate in their assembly . IAD a or f may not be constructed automatically , because a defect of DYX1C1 ( a known DNAAF other than PIH family proteins; Tarkar et al . , 2013 ) affects the normal assembly of all types of IADs in Chlamydomonas ( Yamamoto et al . , 2017 ) . Double , triple or quadruple mutants of PIH genes will be needed to answer this question . The phenotypes of zebrafish PIH gene mutants are summarized in Figure 7A , B . Although PIH1D1 has been known to serve as a component of R2TP complex , which has various important cellular functions ( Kakihara and Houry , 2012 ) , the role of PIH1D1 in vertebrate development remains poorly understood . Zebrafish pih1d1-/- mutants were viable and exhibited only ciliary defects as far as we observed . Thus , Pih1d1 could be mostly redundant in cellular functions . In fact , yeast PIH1-deletion cells were also reported to be viable ( Gonzales et al . , 2005 ) , like zebrafish pih1d1-/- . Accumulated knowledge about DNAAFs has suggested the involvement of R2TP-like complex in the process of axonemal dynein assembly ( Li et al . , 2017 ) . We identified Pih1d1 as a novel DNAAF , which strongly support this idea . Since KTU and PIH1D3 are also suggested to participate in R2TP-like complexes ( Tarkar et al . , 2013; Olcese et al . , 2017 ) , each of PIH proteins may serve as a component of R2TP-like complexes . Intriguingly , however , our expression analysis of PIH genes suggested that pih1d2 , ktu , and twister have cilia-specific functions , while pih1d1 has ubiquitous cellular functions in addition to ciliary function . Further analysis of binding partners of PIH proteins can provide us the mechanism of how PIH1D1 promote the assembly of various types of protein complexes and how distinct PIH proteins modulate the assembly of different types of axonemal dyneins . It was surprising that the spermatozoa of zebrafish ktu-/- showed normal motility , since abnormal sperm motility was reported in both human KTU/DNAAF2-/- patients and medaka ktu mutants ( Omran et al . , 2008 ) . This suggests that the function of Ktu and other PIH proteins in zebrafish could have diverged during evolution . Intriguingly , double mutants of pih1d2-/-;ktu-/- showed phenotypes similar to those of the medaka ktu mutant , in terms of complete loss of ciliary motility in Kupffer’s vesicle ( Figure 5D ) and the expansion of pronephric ducts ( Figure 7—figure supplement 1F ) . Furthermore , the waveform of pih1d2-/-;ktu-/- spermatozoa highly resembles that of medaka ktu mutant spermatozoa ( bends do not propagate to the tip of the sperm tail; Omran et al . , 2008 ) . The function of medaka Ktu is thus partially shared by Ktu and Pih1d2 in zebrafish . Such functional divergence of PIH proteins was also reported in human and mouse; human PIH1D3 has functions in various ciliated organs ( Paff et al . , 2017; Olcese et al . , 2017 ) , while mouse Pih1d3 is a testis-specific gene ( Dong et al . , 2014 ) . Twister2 ( paralogue of mouse Pih1d3 ) is expressed in various ciliary organs including testis in mice but is not able to rescue the loss of Pih1d3 in testis . Therefore , PIH proteins tend to be functionally diverse and sometimes interchangeable , even though they stay in the category of DNAAFs . In pih1d2-/-;ktu-/- spermatozoa , OAD , IAD b , c , and e were missing only from the distal region of the flagella . One possible explanation for this phenotype is that the lack of both PIH1d2 and Ktu causes decreased efficiency of axonemal dynein assembly , leading to a shortage of dyneins to be loaded in the distal axoneme . Actually , the axoneme is known to continue elongating by adding flagellar components to its distal end during ciliogenesis ( Johnson and Rosenbaum , 1992 ) , and in mature spermatozoa , the transport of flagellar components is highly improbable , because IFT components disappear as spermatozoa mature ( San Agustin et al . , 2015 ) . Alternatively , the distal and proximal region of zebrafish sperm could differ in the composition of axonemal dyneins . Indeed , human respiratory cilia are known to have two types of OADs , that is DNAH11/DNAH5-containing OADs in the proximal and DNAH9/DNAH5-containing OADs in the distal parts ( Figure 7E; Fliegauf et al . , 2005; Dougherty et al . , 2016 ) . Intriguingly , a mutation in the human KTU/DNAAF2 gene strongly affects the assembly of only distal OADs in respiratory cilia ( Omran et al . , 2008 ) . However , at the moment , we do not have any evidence for the distal-specific dynein composition in zebrafish spermatozoa . Although testis showed the expression of two OAD γ-HC genes: dnah5 and dnah8 , Dnah8 is present along the entire length of sperm flagella ( Figure 4B ) and the distribution of Dnah5 is not known . As for OAD β-HC gene , only dnah9 was detectable in testis ( Figure 6 ) . Further analyses with pih1d2-/-;ktu-/- spermatozoa could shed light on the structural and functional difference between distal and proximal regions of vertebrate spermatozoa . Expression analysis of DNAH genes suggested that the composition of axonemal dyneins differed between sperm flagella and Kupffer’s vesicle cilia . In sperm flagella , Dnah9/Dnah8-containing OADs could be majority as discussed above , while in Kupffer’s vesicle cilia , the axonemes seem to be constituted of Dnah9/Dnah5-containing OADs and/or Dnah11/Dnah5-containing OADs ( Figure 7D ) . We also found that dnah3 ( IAD HC gene ) are differently expressed between testis and Kupffer’s vesicle ( Figure 6 ) . Given that all PIH proteins are expressed in the two organs , phenotypic differences between pih1d2-/- and ktu-/- can be accounted for by the different compositions of axonemal dyneins . It is tempting to speculate that the dynein compositions vary depending on the pattern of ciliary beating , as cilia and flagella of the two organs exhibit the different mode of movement , planar oscillation for sperm flagella and rotation for Kupffer’s vesicle cilia .
Zebrafish were maintained at 28 . 5°C on a 13 . 5/10 . 5 hr light/dark cycle . Embryos and larvae were raised at the same temperature in 1/3 Ringer’s solution ( 39 mM NaCl , 0 . 97 mM KCl , 1 . 8 mM CaCl2 , and 1 . 7 mM HEPES , pH 7 . 2 ) . Developmental stages of embryos and larvae are described according to hpf at 28 . 5°C and the morphological criteria by Kimmel et al . ( 1995 ) . For embryos used in whole-mount in situ hybridization , 200 μM 1-phenyl-2-thiourea was added to 1/3 Ringer’s solution to delay pigmentation . PIH1 domain is registered as PF08190 in the Pfam database . To find all PIH proteins in the zebrafish genome , BLASTp search was performed with the consensus sequence of the PIH1 domain ( https://www . ncbi . nlm . nih . gov/Structure/cdd/cddsrv . cgi ? uid=pfam08190 ) . Four proteins were identified as a match: Pih1d1 ( NP_001153400 . 1 , E value = 2 . 66e-27 ) , Pih1d2 ( NP_001008629 . 1 , E value = 2 . 08e-9 ) , Ktu ( NP_001028272 . 1 , E value = 2 . 37e-31 ) , and Twister ( also known as Pih1d3; NP_001002309 . 1 , E value = 4 . 16e-5 ) . Paralogues of each match were also checked , since teleost fish are known to have undergone an additional genome duplication . BLASTn and tBLASTn search were performed using each PIH sequence as a query; however , only the proteins containing the query sequence were a hit in E value <10 . Therefore , zebrafish have four PIH proteins: Pih1d1 , Pih1d2 , Ktu , and Twister . Zebrafish genome-editing was performed according to previous reports of TALEN ( Bedell et al . , 2012 ) or CRISPR/Cas9 ( Gagnon et al . , 2014 ) . Target sites of our genome-editing are as follows: pih1d1 ( TALEN left ) , GTTGAACACGAGCAGAAACAA; pih1d1 ( TALEN right ) , TGAAGCAGAAGTTGTTGGTA; pih1d2 ( TALEN left ) , TACAGGAGCTTCATTCAG; pih1d2 ( TALEN right ) , TGAGTGAAACTCGGCTCCC; ktu ( CRIPSR gRNA ) , GGAGATCCGGCCACAGCTGG; twister ( CRISPR gRNA ) , GGATAATGATGAGGAAGAAG . Genomic DNA was extracted from the developing embryos and target loci were amplified to check the mutations by sanger-sequencing . After identifying founder fish , each mutant line underwent back-cross twice to remove the effect of possible off-target mutations . Zebrafish sperm was expelled by gently squeezing the sides of the fish , and collected in Hank’s buffer ( 137 mM NaCl , 5 . 4 mM KCl , 0 . 25 mM Na2HPO4 , 0 . 44 mM KH2PO4 , 1 . 3 mM CaCl2 , 1 . 0 mM MgSO4 , and 4 . 2 mM NaHCO3 ) . For the fractionation of sperm head and flagella , spermatozoa were passed through a 26-gauge needle 20 times in Hank’s buffer with 2 mg/ml BSA , and the separated heads and flagella were collected by centrifugation ( head: 400 g , 3 min; flagella: 9000 g , 3 min ) . For purification of sperm axonemes , sperm heads and membranes were removed by adding 2% Nonidet P-40 to Hank’s buffer , and demembranated axonemes were collected by centrifugation ( 10 , 000 g , 3 min ) , then resuspended in HMDEKAc buffer ( 30 mM HEPES at pH 7 . 2 , 5 mM MgSO4 , 1 mM dithiothreitol , 1 mM EGTA , and 50 mM CH3COOK ) . To observe proper motility , spermatozoa were kept on wet ice until analyzed and used within 1 hr of sperm collection . Zebrafish spermatozoa were inactive in Hank’s buffer , but were activated by adding abundant amount of 1/5 × Hank’s buffer . Sperm motilities were observed under bright-field conditions using an inverted microscope ( DMI6000B; Leica ) and a high-speed camera ( HAS-L1; Detect ) . For waveform analysis , spermatozoa whose heads were attached to the coverslip were selected and waveforms of flagella were filmed at 1000 fps . On the other hand , for CASA , 2 mg/ml of BSA was added to buffers to prevent sperm from attaching to the glass and free swimming spermatozoa were filmed at 200 fps . CASA modified for zebrafish was performed as previously reported ( Wilson-Leedy and Ingermann , 2007 ) . Spermatozoa were prepared on glass slides with 10 μm spacers ( 200A10; Kyodo giken chemical ) , and covered with coverslips to provide a consistent fluid depth . Eight independent experiments with two times of 1 s observations were performed to obtain 16 technical replicates of CASA . Purified sperm axoneme were incubated with anti-α-tubulin antibody ( 1:10000 dilution; T9026; Sigma-Aldrich ) for 15 min at 4°C in HMDEKAc buffer , and then with anti-mouse antibody conjugated with 15 nm colloidal gold ( final 1:50 dilution; EM . GMHL15; BBInternational ) and 15 nm colloidal gold conjugated with BSA ( final 1:5 dilution; 215 . 133; Aurion ) were added . Holey carbon grids were glow discharged before use to make them hydrophilic . 5 μl of axoneme solution was loaded onto the grid , and then excess liquid was blotted away with filter paper to make a thin film of the solution . Immediately after , the grid was plunged into liquid ethane at −180°C for a rapid freeze of the solution . Blotting and freezing were automatically performed by an automated plunge-freezing device ( EM GP; Leica ) . Cryo-prepared grids were stored in liquid nitrogen until observation using the electron microscope . Cryo-prepared grids were transferred into a transmission electron microscope ( JEM-3100FEF; JEOL ) with a high-tilt liquid nitrogen cryotransfer holder ( 914; Gatan ) , and kept at −180°C . Images of single axis tilt series were collected using a 4096 × 4096–pixel CMOS camera ( TemCam-F416; TVIPS ) and automated acquisition software ( EM-TOOLs; TVIPS ) . Tilt series were acquired by a stepwise rotation of the sample from −60 to 60° in 2 . 0° increments . The total electron dose was limited to approximately 100 e/Å2 for an individual tilt series to avoid radiation damage of the sample . Images were recorded at 300 keV , with 8 . 8 µm defocus , at a magnification of 30 , 000 × and a pixel size of 7 . 2 Å . An in-column Ω energy filter was used to enhance image contrast in the zero-loss mode with a slit width of 20 eV . The tilt series images were aligned and reconstructed into 3D tomograms using IMOD software ( Kremer et al . , 1996 ) . Alignment and averaging of subtomograms were conducted by custom Ruby-Helix scripts ( Metlagel et al . , 2007 ) and PEET ( Particle Estimation for Electron Tomography ) software suite ( Nicastro et al . , 2006 ) , using a 96 nm repeat of DMT as one particle assuming a nine-fold rotational symmetry of the axoneme . Effective resolutions were determined by Fourier shell correlation with a cutoff value of 0 . 143 . For the visualization of tomographic slices or 3D structures , 3dmod program ( IMOD software ) or isosurface rendering of UCSF Chimera package ( Pettersen et al . , 2004 ) were used , respectively . Spermatozoa in Hank’s buffer were attached to the wells of an eight-well glass slide ( TF0808; Matsunami ) . After washing out excess sperm , attached spermatozoa were briefly demembranated using 1% Nonidet P-40 for 2 min . Specimens were fixed with 2% paraformaldehyde/Hank’s buffer for 10 min at room temperature , followed by treatment with cold acetone and methanol ( −20°C ) . After rehydration with PBST ( phosphate buffered saline containing 0 . 1% Tween-20 ) , specimens were treated with blocking buffer ( 2% normal goat serum , 1% cold fish gelatin in PBST ) . Immunostaining was performed with anti-acetylated tubulin antibody ( 1:500 dilution; T6793; Sigma-Aldrich ) and anti-Dnah8 antibody ( 1:50 dilution; generated in this study ) as primary antibodies . Alexa Fluor 488 Donkey anti-mouse IgG ( 1:250 dilution; ab150105; Abcam ) and Alexa Fluor 555 Donkey anti-rabbit IgG ( 1:250 dilution; ab150074; Abcam ) were used as secondary antibodies with 2 . 5 μg/ml DAPI ( Wako ) for nuclear staining . Specimens were mounted with Fluoro-KEEPER Antifade Reagent ( Nacalai tesque ) and observed with a fluorescence microscope ( BX60; Olympus ) and a CCD camera ( ORCA-R2; Hamamatsu ) . Embryos developing Kupffer’s vesicle were selected at 12 hpf and dechrionated before observations . To align the orientations , embryos were embedded in 0 . 8% of low gelling temperature agarose ( Sigma-Aldrich ) with 1/3 Ringer’s solution . Motility of Kupffer’s vesicle cilia were observed under the bright-field conditions using an inverted microscope ( DMI6000B; Leica ) and a high-speed camera ( HAS-L1; Detect ) at 1000 fps . The sequences of zebrafish PIH genes and DNAH genes were subcloned into pCRII-TOPO plasmid ( Invitrogen ) . From the constructed plasmids , RNA probes were synthesized using SP6 or T7 RNA polymerase ( Roche ) with DIG RNA Labeling Mix ( Roche ) . RNAs were purified using RNeasy Mini Kit ( Qiagen ) . Sequences of primers used in the construction of plasmids are summarized in Table 2 . Dechorionated embryos or dissected testes were fixed with 4% paraformaldehyde ( PFA ) in PBST , and then stored in methanol at −20°C . After rehydration with PBST , specimens were treated with proteinase K and re-fixed with 4% PFA/PBST solution . Hybridization was performed overnight at 63°C in hybridization buffer ( 750 mM NaCl , 75 mM trisodium citrate , 500 μg/ml torula tRNA , 50 μg/ml Heparin , 50% formamide , and 0 . 1% Tween-20 ) with digoxigenin-labeled RNA probes . Hybridized specimens were washed with 50% formamide/2 × SSCT ( saline sodium citrate containing 0 . 1% Tween-20 ) followed by 2 × SSCT and 0 . 2 × SSCT , then treated with AP-conjugated anti-digoxigenin Fab fragments ( 1:4000 dilution; Roche ) in blocking solution ( 150 mM NaCl , 100 mM maleic acid at pH 7 . 5 , 5% blocking reagent ( Roche ) , 5% normal goat serum , and 0 . 1% Tween-20 ) at 4°C overnight . After washing with MABT ( 150 mM NaCl , 100 mM maleic acid at pH 7 . 5 , and 0 . 1% Tween-20 ) , signals were developed using BM-purple ( Roche ) . When desired intensities of staining were obtained , reactions were stopped by stopping solution ( PBST containing 1 mM EDTA ) and 4% PFA/PBST . Before observations , specimens were transferred into 80% glycerol/PBS to make them transparent . For some embryos , to show stained organs clearly , posterior regions were flat-mounted or yolk was removed . Images were taken by a stereoscopic microscope ( MVX10; Olympus ) and a CCD camera ( DP73; Olympus ) . Sequences encoding full length of zebrafish PIH proteins ( Pih1d1 , Pih1d2 , Ktu , and Twister ) were subcloned into the pColdI plasmid vector ( Takara ) . Sequences encoding zebrafish Dnah8 ( amino acid 895–1402 ) and Dnah2 ( amino acid 802–1378 ) were subcloned into the pGEX-6P-2 plasmid vector ( GE Healthcare ) . Recombinant polypeptides were purified from transformed E . coli lysate using Ni-NTA Agarose ( Qiagen ) for PIH proteins or Glutathione Sepharose 4B ( GE Healthcare ) for Dnah8 and Dnah2 . Polyclonal antibodies against each purified polypeptide were raised by immunization of rabbits . Antibodies were affinity purified from serum by the antigens before use . Sequences of the primers used in the antigen production are summarized in Table 2 . Other antibodies used are as follows: anti-Dnai1 antibody ( GTX109719; GeneTex ) , anti-Dnali1 antibody ( anti-p28 antibody; LeDizet and Piperno , 1995 ) , anti-α-tubulin antibody ( T9026; Sigma-Aldrich ) , and anti-acetylated tubulin antibody ( T6793; Sigma-Aldrich ) . Proteins were separated by SDS-PAGE in 5–20% gradient polyacrylamide gels ( Nacalai Tesque ) and transferred onto polyvinylidene difluoride ( PVDF ) membranes ( Millipore ) . After blocking with 5% skim milk ( Nacalai Tesque ) in TBST ( Tris-buffered saline containing 0 . 1% Tween-20 ) , membranes were incubated with primary antibodies , followed by several washes and incubation with secondary antibodies ( goat anti rabbit/mouse IgG antibody peroxidase conjugated; Sigma-Aldrich ) . Protein signals were visualized by ECL Select Western Blotting Detection Reagent ( GE Healthcare ) and observed using luminescent image analyzer ( ImageQuant LAS4000mini; GE Healthcare ) . Data with biological/technical replicates are shown with mean ( bar graphs ) ±SD ( error bars ) . Statistical significances between WT and each mutant were tested by a two-tailed Dunnett’s test , and p value < 0 . 05 was considered to indicate a significant difference . The averaged subtomograms of zebrafish DMTs in this study are available at the EMDataBank ( http://www . emdatabank . org/ ) under the following accession numbers: WT , EMD-6954; pih1d1-/- , EMD-6955; pih1d2-/- , EMD-6956; ktu-/- , EMD-6957; twister-/- , EMD-6958; pih1d2-/-;ktu-/- ( +OAD class ) , EMD-6959; and pih1d2-/-;ktu-/- ( -OAD class ) , EMD-6960 . | Many cells have long , thin structures called cilia on their surface , some types of which can beat back and forth . This beating motion has many roles; for example , cilia on the cells that line the lungs help to sweep out debris , and the tails of sperm beat to move them forward . A structure called the axonemal dynein complex at the core of the cilia generates the beating motion . When the cell makes new cilia , it assembles the complexes in the main body of the cell and then transports them to the right place , like erecting a prefabricated building . Various proteins help to assemble the complexes , of which there are more than eight types . However , the identities of all of these proteins , and their roles in constructing specific axonemal dynein complexes , is not fully known . Studies in algae have suggested that a family of proteins known as PIH ( short for protein interacting with Hsp90 ) helps to construct axonemal dynein complexes . Zebrafish – which share many of the same protein-encoding genes as humans – produce four PIH family proteins . To investigate the roles that each of these proteins play , Yamaguchi et al . used genetic engineering to create four zebrafish mutants that were each unable to produce a different PIH protein . A technique called cryo-electron microscopy enabled the axonemal dynein complexes in the tails of the sperm produced by the zebrafish to be visualized . The sperm from each mutant lacked specific axonemal dynein complexes , revealing that each PIH protein assembles different complexes . The sperm also had difficulties moving . Yamaguchi et al . examined this movement to deduce how specific complexes affect the ability of the sperm to beat their tails . Further work on how PIH proteins interact with the axonemal dynein complexes will help us to understand how cells make cilia , and what happens when this process goes wrong . This could ultimately help us to treat genetic disorders known as ciliopathies , which arise when cilia do not develop normally . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2018 | Systematic studies of all PIH proteins in zebrafish reveal their distinct roles in axonemal dynein assembly |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.